Overview

Literature searches through July 28, 2010 for the current report identified 1987 unduplicated citations. We received dossiers from 6 pharmaceutical manufacturers: Takeda Pharmaceuticals, GlaxoSmithKline, Bristol-Meyers Squibb, Amylin Pharmaceuticals, Novo Nordisk, and Merck. Twenty-two additional references were identified through hand searches of systematic reviews and other sources, and 11 additional articles were identified from the dossiers, 4 from the pioglitazone (Actos®) dossier (Takeda Pharmaceuticals), 1 from the exenatide (Byetta®) dossier , 1 from the liraglutide (Victoza®) dossier (Novo Nordisk), 4 from the sitagliptin dossier (Merck), and 1 from the saxagliptin (Onglyza®) dossier (Bristol-Meyers Squibb). We also retrieved 240 excluded references from the reference database of the Fixed Dose Combination Drug Products for the Treatment of Type 2 Diabetes and Hyperlipidemia DERP report17 in order to review these publications using new inclusion criteria. From all of these sources, we had a total of 2260 references. In addition to these, we carried forward 209 of the included studies from 3 previous DERP reports: Newer Drugs for the Treatment of Diabetes Mellitus,8 Fixed Dose Combination Drug Products for the Treatment of Type 2 Diabetes and Hyperlipidemia,17 and the Drug Class Review on Thiazolidinediones.18

By applying the eligibility and exclusion criteria to titles and abstracts of all identified citations, we obtained full-text copies of 857 citations. After re-applying the criteria for inclusion, we ultimately included 107 new publications from our recent literature searches and other sources, plus the 209 includes from previous reports. See Appendix D for a list of excluded studies and reasons for exclusion at the full text stage. Figure 1 shows the flow of study selection. Among the 107 includes from our recent searches, 79 were trials, 19 were systematic reviews, and 9 were observational studies. Among these, 20 were rated good quality, 63 were fair quality, and 24 were poor quality. Poor-quality studies are listed in Appendix D.

Figure 1. Results of literature search.

Figure 1

Results of literature search.

Key Question 1. What is the comparative efficacy and effectiveness of newer diabetes medications, TZDs, and drug combinations (administered as fixed dose combination products or dual therapy) for children and adults with diabetes mellitus?

I. Newer Drugs for the Treatment of Diabetes Mellitus: Amylin Agonists, DPP-4 Inhibitors, and Incretin Mimetics

Summary of Findings for Amylin Agonists

Pramlintide for type 1 diabetes
Evidence in children
  • No data on children were reported, although people as young as 16 years were eligible for study enrollment in 2 included trials (% of children enrolled was not reported) 19, 20 (insufficient strength of evidence).
Evidence in adults
  • HbA1c was either slightly improved or no different with the addition of pramlintide 30 or 60 mcg/meal to a flexible-dose insulin regimen compared with placebo plus flexible-dose insulin regimen over 29 weeks20 (between-group difference: 0.0%) and 52 weeks19 (between-group difference: 0.27%, P value, not reported) of treatment (low strength of evidence).
  • Greater reduction in HbA1c when pramlintide 60 mcg 3 or 4 times a day was added to fixed-dose insulin therapy (decreased from baseline by 0.29% to 0.34%, P<0.01) than when placebo was added to fixed-dose insulin (decrease by 0.04%, not statistically significant) at 52 weeks21 (low strength of evidence).
  • Slight weight loss with pramlintide in addition to insulin (range: −0.4 to −1.3 kg) compared with slight weight gain with placebo plus insulin in a fixed- or flexible-dose setting (range: +0.8 to +1.2 kg) over 29 and 52 weeks (moderate strength of evidence).
Pramlintide for type 2 diabetes
Evidence in children
  • Children and adolescents ≤ 18 years were not enrolled in any of the included studies (insufficient strength of evidence).
Evidence in adults
  • No included studies focused on health outcomes as the primary outcomes. One study reported some health outcomes among the adverse events.22 Overall evidence was insufficient to determine how pramlintide compares with other treatments for their impact on health outcomes.
  • Greater reduction in HbA1c with pramlintide doses from 75 mcg to 120 mcg given 2 or 3 times daily added to fixed- or stable doses of insulin compared with placebo and insulin (range 0.13% to 0.4% at 52 weeks, moderate strength of evidence).
  • Greater reduction in weight with pramlintide doses from 75 mcg to 120 mcg given 2 or 3 times daily added to fixed- or stable-doses of insulin compared with placebo and insulin (range 1.1 kg to 1.85 kg, placebo-corrected differences at 52 weeks, moderate strength of evidence).
  • No statistically significant difference for reduction in HbA1c between the addition of pramlintide 120 mcg at meals to glargine or detemir compared with rapid acting insulin analog at 24 weeks (1.1% compared with 1.3%, P=0.46, low strength of evidence).
  • No change in weight reported with the addition of pramlintide 120 mcg at meals to glargine or detemir, compared with a 4.7 kg weight gain with rapid acting insulin analog at 24 weeks (+4.7 kg between group difference, P<0.0001, low strength of evidence).

Detailed Assessment for Pramlintide in Type 1 Diabetes

We found no active-control trials. We found 3 placebo-controlled trials. Characteristics of these trials are presented in Table 5 and results for HbA1c and weight are presented in Table 6. All 3 studies were fair-quality and were conducted in a double-blind manner with pramlintide or placebo added to their insulin regimen. None of these trials were similar enough for efficacy data to be pooled.

Table 5. Characteristics of pramlintide placebo-controlled trials in adults with type 1 diabetes.

Table 5

Characteristics of pramlintide placebo-controlled trials in adults with type 1 diabetes.

Table 6. Efficacy outcomes of placebo-controlled trials of Pramlintide in type 1 diabetes.

Table 6

Efficacy outcomes of placebo-controlled trials of Pramlintide in type 1 diabetes.

Efficacy and effectiveness
Flexible-dose insulin

In a fair-quality trial the addition of pramlintide 30 mcg or 60 mcg 3 or 4 times a day with meals to a flexible-dose insulin regimen did not significantly improve HbA1c (−0.5% compared with −0.5%; Table 6) compared to patients receiving a combination of short- and long-acting insulin plus placebo adjusted to achieve specified glycemic targets over 29 weeks.20 Pramlintide-treated patients lost slightly more weight than insulin-only patients (−1.3 kg compared with +1.2 kg).

All patients received stable doses (±10% change from baseline) of intensive insulin therapy using multiple daily injections or continuous insulin infusion before enrolling in the study. Patients were mainly middle-aged and white and had long-standing type 1 diabetes. Mean baseline HbA1c was 8.1%. A 30% to 50% reduction in mealtime insulin was recommended before starting pramlintide to avoid hypoglycemic events.

In a second trial using flexible insulin dosing,19 the addition of pramlintide 30 mcg or 60 mcg 4 times a day to insulin with each meal was slightly more effective than insulin plus placebo in lowering HbA1c and weight (Table 6). The change in HbA1c at week 52 was −0.39% with pramlintide plus insulin and −0.12% with insulin plus placebo (between-group difference: 0.27%, P value, not reported).

This trial was rated fair quality, but there are some aspects of the design and reporting that limit the validity of the results: only 71% of patients completed the 52 weeks of therapy and data from only completers were examined. The total withdrawal rates of 28% to 29% were similar between the treatments, however, more pramlintide-treated patients discontinued due to adverse events than placebo-treated patients during the study (12.8% compared with 8.0%). Nausea was the most common reason for withdrawal. In addition, the authors reported no further details on insulin dose adjustments than that they were made according to “good medical practices.”

Stable insulin dosing

The addition of pramlintide 60 mcg 3 or 4 times a day with meals to fixed or stable background insulin therapy improved HbA1c by 0.29% and 0.34% compared with 0.04% improvement in the insulin plus placebo group over 52 weeks of therapy.21 Pramlintide-treated subjects also demonstrated nominal weight loss from baseline (−0.4 kg at 52 weeks, P<0.05), which was not seen with placebo (+0.8 kg at 52 weeks, P>0.05). This trial was rated fair quality, but there are some aspects of the design and reporting that limit the validity of the results, including high withdrawal rates (>35% in all treatment arms). However a greater proportion of pramlintide-treated patients discontinued due to adverse events (primarily nausea) compared with those in the placebo plus insulin arm (14% to 20% compared with 3% for adverse events).

This trial began with a 90 mcg dose arm, which was removed from efficacy analysis when another trial (identified as study #137–117 in US Food and Drug Administration reviews) revealed an adverse tolerability profile associated with this 90 mcg dose. Specific reasons for “intolerability” with the 90 mcg dose could not be found in either study #137–117 in the US Food and Drug Administration documents or from this trial by Ratner and colleagues. Only general statements were made by Ratner and colleagues: there was 2-fold increase in nausea, vomiting, anorexia and 4-fold increase in severe hypoglycemia event rates associated with pramlintide across the doses compared with placebo. Study #137–117 could not be found in a peer-reviewed publication.

Detailed Assessment of Pramlintide in Type 2 Diabetes

Three placebo-controlled and one active-control trials were found. One post hoc analysis23 of a placebo-controlled trial24 addressing cardiovascular markers was identified. Characteristics of the included trials are presented in Table 7 and results for HbA1c and weight in Table 8. All of the studies were rated fair-quality. Three studies included patients who were not achieving glycemic goals on insulin with or without oral agents.24–26 One study included patients not achieving glycemic goals regardless of insulin or oral agent use.22 None of the trials were pooled due to significant heterogeneity.

Table 7. Characteristics of pramlintide trials (placebo and active controlled) in adults with type 2 diabetes.

Table 7

Characteristics of pramlintide trials (placebo and active controlled) in adults with type 2 diabetes.

Table 8. Efficacy outcomes of placebo and active-control trials of pramlintide in type 2 diabetes.

Table 8

Efficacy outcomes of placebo and active-control trials of pramlintide in type 2 diabetes.

Dose-ranging study

The addition of pramlintide 75 mcg/meal or 150 mcg/meal to fixed-dose insulin, with or without oral hypoglycemic agents (metformin or sulfonylureas), improved HbA1c by 0.3% to 0.4% and weight loss by 1.5 to 2.4 kg (placebo-corrected values) in a population with poorly controlled (HbA1c 9.0% to 9.3%) type 2 diabetes over 52 weeks.26 No significant differences in HbA1c were observed between 2 pramlintide doses at the end of the trial: pramlintide 75 mcg (−0.5%) compared with 150 mcg (−0.6%). The largest reductions in HbA1c (almost 1%) occurred early on at week 13 for those on the 150 mcg dose.

This trial was rated fair quality, but there are some aspects of the design and reporting that limit the validity of the results. These include high withdrawal rates (~30%) which were similar for placebo, pramlintide 30 mcg and 75 mcg groups. Those randomized to pramlintide 150 mcg dose exhibited largest rates of total withdrawal and withdrawal due to adverse events (37.5% and 18%).

Stable insulin dosing

The addition of pramlintide 90 mcg or 120 mcg to fixed or stable doses of insulin with or without oral hypoglycemic agents (metformin or sulfonylureas) gave slightly larger improvements in HbA1c and weight at 52 weeks than patients randomized to placebo plus fixed-dose insulin (placebo-corrected values for HbA1c: 90 mcg: −0.13%, 120 mcg: −0.4% and for weight: 90 mcg: −1.1 kg; 120 mcg: −1.85 kg).25 Effect on HbA1c was greatest at 26 weeks for both pramlintide groups (P<0.05 compared with placebo) and persisted only with the 120 mcg arm at 52 weeks (change in HbA1c from baseline −0.62%, P<0.05). Approximately 20% to 27% of all randomized patients were taking oral hypoglycemic agents at baseline.

During the course of this one fair-quality trial, 25 results from another study (identified as study #137-123 in the US Food and Drug Administration reviews) found that pramlintide 60 mcg was less effective than compared with higher doses. As a result, efficacy and safety information from the 60 mcg arm were not reported by this trial.

Flexible insulin dosing

In contrast to the previous study, another short-term fair-quality trial 24 evaluated pramlintide as a pre-meal medication in conjunction with glargine (without prandial insulin) with or without oral hypoglycemic agents (metformin, sulfonylureas, and/or thiazolidinediones). The comparison group was patients on flexible-dose glargine plus placebo. At 16 weeks, the addition of pramlintide to glargine reduced HbA1c by 0.36% and induced weight loss of 2.3 kg (placebo-corrected values) relative to placebo plus glargine.

Glargine, a basal insulin without pronounced peak effects, was allowed to be adjusted during the study to achieve prespecified fasting glucose targets once pramlintide doses were stabilized. Patients had diabetes for 10 to 11 years. At baseline HbA1c was moderately elevated at 8.5%, and patients were using insulin glargine 48 to 54 units per day, with 50% of patients concomitantly taking ≥2 oral hypoglycemic agents and 89% taking at least 1 oral agent.

Another fair-quality trial 22 compared pramlintide with rapid acting insulin analog (RAIA; lispro, aspart, or glulisine) in addition to basal insulin glargine or detemir). Both basal and RAIA were allowed to be titrated at the investigators discretion, however basal insulin was titrated once or twice weekly to fasting glucose 70–100 mg/dL and RAIA could be titrated only after 4 weeks of basal titration to avoid hypoglycemia. RAIA was increased by 1–2 units every 3–7 days per the investigator based on glucose readings prior to the next meal. RAIA resulted in a non-statistically significant greater HbA1c reduction over pramlintide by 0.2% (P=0.46). No change in weight was noted in the pramlintide group, however patients randomized to RAIA did experience significantly more weight gain, mean change from baseline 4.7 kg (P<0.0001).

Baseline HbA1c was similarly elevated to previously described studies at 8.2% to 8.3% and approximately 50% of patients were taking oral agents. Of the patients in the pramlintide group, 27% were using basal insulin at doses averaging 20–24 units per day, as were 24% of patients in the placebo group.

Summary of Findings for DPP-IV Inhibitors

Eighteen randomized controlled trials for sitagliptin and 5 randomized controlled trials for saxagliptin fulfilled inclusion criteria. Four of the sitagliptin randomized controlled trials were identified through dossier submission, 2 of which were extensions of other studies included. Two systematic reviews including sitagliptin also met inclusion criteria. No comparative cohort or case-control studies were identified reporting either long-term benefits or adverse events. In the US Food and Drug Administration Medical and Statistical Reviews we identified 10 relevant trials for sitagliptin, of which 7 were published in peer-reviewed journals. One of the trials27 identified from the US Food and Drug Administration Reviews was not included because it did not meet inclusion criteria; the 3 remaining trials (study #P10X1, P014, and P014X1) could not be found in the medical literature. Details of included studies are found in Tables 917; their quality assessments are in Evidence Table 6 (Evidence Tables are published in a separate document).

Table 9. Characteristics of sitagliptin active-control trials (with or without placebo study arms) in adults with type 2 diabetes.

Table 9

Characteristics of sitagliptin active-control trials (with or without placebo study arms) in adults with type 2 diabetes.

Table 10. Efficacy outcomes of sitagliptin monotherapy compared with an active agent.

Table 10

Efficacy outcomes of sitagliptin monotherapy compared with an active agent.

Table 11. Efficacy outcomes of Sitagliptin compared with an active agent added to another oral hypoglycemic agent.

Table 11

Efficacy outcomes of Sitagliptin compared with an active agent added to another oral hypoglycemic agent.

Table 12. Characteristics of sitagliptin monotherapy placebo-controlled trials in adults with type 2 diabetes.

Table 12

Characteristics of sitagliptin monotherapy placebo-controlled trials in adults with type 2 diabetes.

Table 13. Characteristics of sitagliptin add-on therapy placebo-controlled trials in adults with type 2 diabetes.

Table 13

Characteristics of sitagliptin add-on therapy placebo-controlled trials in adults with type 2 diabetes.

Table 14. Efficacy outcomes of sitagliptin monotherapy compared with placebo.

Table 14

Efficacy outcomes of sitagliptin monotherapy compared with placebo.

Table 15. Results of meta-analyses for mean change in HbA1c and weight for sitagliptin 100 mg compared with placebo.

Table 15

Results of meta-analyses for mean change in HbA1c and weight for sitagliptin 100 mg compared with placebo.

Table 16. Efficacy outcomes of sitagliptin or placebo added to one oral hypoglycemic agent.

Table 16

Efficacy outcomes of sitagliptin or placebo added to one oral hypoglycemic agent.

Table 17. Efficacy outcomes of sitagliptin or placebo added to 2 oral hypoglycemic agents.

Table 17

Efficacy outcomes of sitagliptin or placebo added to 2 oral hypoglycemic agents.

Summary of Findings for Sitagliptin

Efficacy/Effectiveness
Sitagliptin compared with Saxagliptin
  • We found no head-to-head studies of sitagliptin and saxagliptin meeting inclusion criteria (insufficient strength of evidence).
Evidence in children
  • Children and adolescents ≤ 18 years were not included in any of the published studies on effectiveness or efficacy (insufficient strength of evidence).
Evidence in adults
  • All studies focused on intermediate outcomes with none focusing on health outcomes as primary outcomes. Some studies reported some health outcomes such as all-cause mortality or number of people with macrovascular disease among secondary outcomes or adverse events. Overall evidence was insufficient to determine how sitagliptin compares with other treatments for their impact on health outcomes.
  • No studies provided data on efficacy/effectiveness for follow up beyond 2 years.
  • Sitagliptin monotherapy resulted in slightly less HbA1c reduction than either metformin monotherapy over 54 weeks (between group difference −0.16 for metformin 1000 and −0.47 for metformin 2000 mg/d) or glipizide monotherapy over 12 weeks (between group difference −0.22%) (low strength of evidence for both comparisons).
  • Sitagliptin monotherapy resulted in slight weight gain, compared with slight weight loss for those treated with metformin monotherapy over 54 weeks (between group difference −1.6 to −2.1 at 54 weeks, low strength of evidence).
  • Sitagliptin monotherapy resulted in slightly less weight gain compared with glipizide monotherapy over 12 weeks (+0.4 kg compared with +0.9 kg, low strength of evidence).
  • Greater reduction in HbA1c with liraglutide 1.2 mg and 1.8 mg daily than with sitagliptin 100 mg daily in one trial (−1.24% to −1.5% compared with −0.6%; P<0.0001, low strength of evidence). Weight loss was significantly greater with both doses of liraglutide compared to sitagliptin.
  • Greater reduction in HbA1c with sitagliptin 100 mg/d monotherapy than with placebo (weighted mean difference −0.79%, 95% CI −0.93 to −0.66) in patients inadequately controlled on diet and exercise over 12–24 weeks (moderate strength of evidence).
  • Less weight loss with sitagliptin 100 mg/d monotherapy than with placebo (weighted mean difference 0.66, 95% CI 0.43 to 0.89, moderate strength of evidence).
  • Studies comparing add-on of sitagliptin to other hypoglycemic agents (metformin, pioglitazone, or glimepiride) found sitagliptin-treated subjects to have either more weight gain, less weight loss, or similar changes in weight compared to placebo-treated subjects (low strength of evidence).
  • Overall, in patients with inadequate glycemic control on one (metformin, pioglitazone, or glimepiride) or 2 hypoglycemic agents, the addition of sitagliptin resulted in greater reduction in HbA1c than the addition of placebo (between group difference −0.5 to −1.0, moderate strength of evidence)
  • No significant difference in reduction in HbA1c between rosiglitazone and sitagliptin when added to metformin therapy in two randomized controlled trials (moderate strength of evidence).

Detailed Assessment for Sitagliptin

Systematic reviews

Amori and colleagues28 published a good-quality systematic review of US Food and Drug Administration approved and unapproved GLP-1 analogues (exenatide, linaclotide) and DPP-4 inhibitors (sitagliptin [8 studies] and vildagliptin [12 studies]). Sitagliptin and vildagliptin were examined together, rather than individually. Thus, we do not report results of that systematic review here because vildagliptin is not a medication of interest for this report. The Cochrane Collaboration published one good-quality systematic review of DPP-4 inhibitors vildagliptin and sitagliptin.29 In contrast to Amori and colleagues, this review presented results separately by drug. Two studies compared sitagliptin with another single hypoglycemic agent and found less HbA1c lowering with sitagliptin (weighted mean difference 0.33 (95% CI 0.18 to 0.48). In contrast, when sitagliptin was used in combination with another hypoglycemic agent compared to another hypoglycemic combination (6 studies), sitagliptin combination resulted in slightly greater HbA1c lowering (weighted mean difference −0.40, 95% CI −0.47 to −0.33). Similarly, 6 studies were pooled that compared sitagliptin to placebo as monotherapy and found sitagliptin to reduce HbA1c to a greater extent than placebo (weighted mean difference −0.77, 95% CI −0.85 to −0.65). When changes in weight were examined sitagliptin resulted in less weight loss than either placebo (3 studies) or another single hypoglycemic agent (1 study). Our analysis found similar results in change in HbA1c and weight to the above systematic review.

Randomized controlled trials

Eighteen unique randomized controlled trials were identified, with three extension trials, all of fair-quality. We first address active controlled trials and then placebo-controlled trials. The placebo-controlled section is organized by whether sitagliptin was used as monotherapy or as add-on therapy.

Active-control trials

Seven fair-quality trials (10 articles) compared various doses of sitagliptin to active treatment arms of glipizide or metformin (Table 9). 30–39 Four of these trials included comparisons of sitagliptin monotherapy with glipizide or metformin monotherapy.30–33, 37, 38 The others compared sitagliptin with metformin, rosiglitazone, liraglutide, or glipizide as add-on therapy to active background therapy of metformin or pioglitazone.34–36, 39–41

Monotherapy: Sitagliptin compared with an active agent

In 4 fair-quality trials, various doses of sitagliptin were compared to active treatment arms of glipizide 5–20 mg/d or metformin 1000–2000 mg/d (Table 9).30–33, 37, 38 Patients had baseline HbA1c of 7.2% to 8.9%. The trials ranged from 12–104 weeks and showed overall, patients on glipizide and metformin 1–2 g/d monotherapy had numerically larger reductions in HbA1c compared with sitagliptin monotherapy (Table 10). Pooled analysis was not conducted due to small number of studies with significant heterogeneity.

One study compared sitagliptin 100 mg/day with metformin 1000 mg/day and 2000 mg/day monotherapy (3 other arms included placebo, sitagliptin/metformin 1000 mg/day and sitagliptin 2000 mg/day discussed separately).31–33 Initial results reported after 24 weeks showed greater HbA1c reduction with both doses of metformin when compared to sitagliptin (P value not reported). After 30 additional weeks (54 weeks total) slightly more HbA1c lowering was seen in each group however, the reduction remained greater in the metformin groups. After 104 weeks total, HbA1c changes were similar across all 3 groups. Weight loss was also greater after 24, 54, and 104 weeks in the metformin groups.

Another study compared sitagliptin 100 mg/day with metformin 2000 mg/day (titrated up over 5 weeks).38 Similar to the previous trial discussed, metformin resulted in greater HbA1c lowering (difference in LS mean change 0.14 (95% CI 0.06 to 0.21) and greater weight loss than sitagliptin (P<0.001).

A 12 week dose-response study compared various doses of sitagliptin, all divided twice daily, to glipizide titrated according to the study’s protocol.30 Slightly less HbA1c reduction was seen in the sitagliptin 100 mg/day group than the glipizide group. However, patients randomized to sitagliptin gained less weight than those in the glipizide group.

Another study that stratified patients to sitagliptin 25 mg/day or 50 mg/ day based on their renal function, compared sitagliptin to placebo for the first 12 weeks and then glipizide 5–20 mg/day for the remaining 42 weeks.37 Patients in this study all had chronic renal insufficiency and were allowed to continue insulin therapy if on it prior to randomization. After 54 weeks, the placebo/glipizide group had slightly greater HbA1c reduction than sitagliptin. There was minimal change in HbA1c from 12 weeks to 54 weeks with sitagliptin. After 54 weeks, the sitagliptin group had greater weight loss than the placebo/glipizide group however after 12 weeks the placebo group had greater weight loss (Table 10) Results were not stratified by whether or not patients were taking insulin, however 7 patients in the sitagliptin group (11%) and 2 patients in the placebo group (8%) were on insulin at baseline.

Add-on therapy: Sitagliptin compared with active control (other oral hypoglycemic agent) added to metformin

Four fair-quality trials (4 articles) compared the addition of sitagliptin with the addition of another oral hypoglycemic agent to ongoing metformin therapy.34–36, 40 Pratley, 2010 #5847, 41 Pooled analysis was not conducted due to small number of studies with significant heterogeneity.

One fair-quality trial compared the effects of adding either sitagliptin 100 mg/d or glipizide 5–20 mg/d in patients with inadequate glycemic control on metformin.34, 35 Glycemic control was considered inadequate if the metformin dose was ≥ 1500 mg/d with baseline HbA1c 6.5% to 10% at initial screening or after several weeks of stabilizing the metformin dose prior to a 2-week single-blind, placebo run-in period before randomization.

Over the initial 52 weeks the 2 study groups showed no significant differences in treatment effects for HbA1c (between-group difference 0.04%, 95% CI −0.04, 0.13) (Table 11). There was a statistically significant difference between treatment groups in the change in weight. Sitagliptin-treated subjects experienced slight weight loss) compared with a small weight gain seen in glipizide-treated subjects (between-group difference −2.5kg, 95% CI −3.1, −2.0). Most patients had low baseline HbA1c (mean 7.5%) and more than 70% of patients were on oral monotherapy while approximately 30% were on 2 oral agents at baseline. Results were similar in the 104 week extension.35 Minimal changes were seen in HbA1c change from 52 week results and 104 week results, which is in contrast to a previously discussed trial and extensions investigating sitagliptin as monotherapy. 31–33

Another fair quality trial36 assessed the effects of sitagliptin, rosiglitazone, or placebo added to metformin monotherapy over 18 weeks. Prior to randomization patients had to have inadequate glycemic control (HbA1c 7% to 11%) and had to be taking metformin at stable doses ≥ 1500 mg/d for at least 10 weeks before entering a 2-week run-in period. The mean baseline HbA1c was 7.7%.

In these patients, the addition of sitagliptin or rosiglitazone to metformin was significantly more effective than the addition of placebo to metformin at lowering HbA1c (P≤0.001). The placebo-corrected mean change from baseline was −0.51% (95% CI, −0.70 to −0.32) for sitagliptin, and was −0.57% (95% CI, −0.76 to −0.37) for rosiglitazone. Also, comparisons between sitagliptin and rosiglitazone were conducted and showed no statistically significant differences in lowering HbA1c (between-group difference: −0.06%, 95% CI −0.25 to 0.14). Patients randomized to sitagliptin or placebo exhibited slight weight loss from baseline (sitagliptin, −0.4 kg, 95% CI −0.8 to 0.0 compared with placebo, −0.8 kg, 95% CI −1.2 to −0.4) while patients on rosiglitazone gained weight (from baseline: +1.5 kg, 95% CI 1.0 to 1.9) over 18 weeks of therapy (Table 11).

Another trial compared the addition of colesevelan, rosiglitazone, or sitagliptin to ongoing metformin.40 The trial found no statistically significant difference between the rosiglitazone- and sitagliptin-treated subjects.40

An additional 26 week fair quality active-control trial compared liraglutide (1.2 or 1.8 mg daily) to sitagliptin 100 mg daily.41 All study participants were on metformin ≥ 1500 mg daily as background therapy. The study found a greater improvement in HbA1c with both doses of liraglutide compared to sitagliptin (change in HbA1c: liraglutide 1.2 mg −1.24%; liraglutide 1.8 mg −1.5%; sitagliptin −0.6%; P<0.0001 for both doses of liraglutide compared to sitagliptin). Weight loss was significantly greater with both doses of liraglutide compared to sitagliptin (change in weight: liraglutide 1.2 mg −2.86 kg; liraglutide 1.8 mg −3.38 kg; sitagliptin −0.96 kg; P<0.0001 for both doses of liraglutide compared to sitagliptin). Treatment satisfaction as measured by the Diabetes Treatment Satisfaction Questionnaire (DTSQ) improved both with liraglutide and sitagliptin, but increased significantly more in the liraglutide 1.8 mg arm of the study than in the sitagliptin 100 mg arm of the study.41

Add-on therapy: Sitagliptin compared with active control (other oral hypoglycemic agent) added to pioglitazone

One fair-quality study was identified that compared sitagliptin 100mg daily to metformin 850mg twice daily as add-on therapy to pioglitazone.39 All patients included were not controlled on pioglitazone 30mg daily as monotherapy. Patients were randomized to have sitagliptin added to pioglitazone 30mg daily or metformin added to pioglitazone 15mg daily. The rational for different doses of pioglitazone in the two groups was not addressed in the publication.

After 52 weeks of treatment, there was no difference in HbA1c reduction between the two treatment groups. Both groups had similar reduction in HbA1c (between group difference 0.1%, P >0.05). Patients taking metformin in addition to pioglitazone experienced more weight loss the patients taking sitagliptin in addition to pioglitazone (between group difference 1.1kg, P<0.05) (Table 11).

Placebo-controlled trials

Thirteen fair-quality trials compared various doses of sitagliptin to placebo (Tables 12 and 13).30, 36, 37, 42–51 Six of these trials included comparisons of sitagliptin monotherapy with placebo (Table 12).30, 42–46 The others compared add-on therapy with sitagliptin or placebo to a variety of ongoing treatments (Table 13).36, 37, 47–51

Monotherapy: Sitagliptin compared with placebo

Seven fair-quality trials ranging from 12–24 weeks in duration compared sitagliptin 100 mg/d to placebo (Table 12).30, 31, 42, 43, 45, 46, 52 Approximately 50% to 60% of subjects were on 1 or more oral hypoglycemic agents at screening. These agents were discontinued before diet and exercise run-in periods. Patients not responding to diet and exercise were eligible for study inclusion but were required to participate in a 2-week single-blind, placebo run-in period prior to randomization. Three trials allowed use of prespecified rescue medications based on certain glycemic criteria. Mean baseline HbA1c was 7.6% to 8.8%.

Patients randomized to receive sitagliptin 100 mg/d showed significant reductions in HbA1c (weighted mean difference −0.79%, 95% CI, −0.93% to −0.66%, see Table 15), while placebo-treated patients generally showed worsening glycemic control (Table 14). One dose-ranging study 46 found similar HbA1c lowering across sitagliptin 50 mg daily, 100 mg daily, and 50 mg twice daily (−0.43% to +0.44%), however 25 mg daily resulted in less reduction (−0.25%).

Change in weight varied across the trials, generally decreasing in both treatment arms (range for change from baseline: sitagliptin −0.1 to −0.8 kg compared with placebo −0.5 to −1.1 kg). However, one trial45 found weight gain in the sitagliptin arm (mean change from baseline, 0.6 kg) and no change in weight in the placebo arm. Overall, however, subjects randomized to sitagliptin lost slightly less weight than subjects randomized to placebo (weighted mean difference: 0.661, 95% CI 0.43 to 0.892; see Table 15).

Add-on therapy: Sitagliptin or placebo added to one oral hypoglycemic agent

A total of 6 fair-quality trials compared the addition of sitagliptin or placebo to another oral hypoglycemic agent.36, 47–51 Three trials assessed the effects of sitagliptin compared to placebo in patients who were considered to have “failed” therapy with metformin,36, 47, 50 2 studies assessed sitagliptin compared to placebo in patients who were considered to have “failed” therapy with pioglitazone or glimepiride,48, 49 and 1 study assessed sitagliptin compared to placebo in patients who were inadequately controlled on metformin and insulin >15 units daily.51

Mean baseline HbA1c ranged from 7.7% to 9.2%. Approximately 60% of patients were on more than 1 oral hypoglycemic agent, while 30% were on more than 2 oral agents (Table 13). Patients were considered to have “failed” therapy with metformin, pioglitazone, or glimepiride at screening or after 10–19 weeks of dose stabilization and if HbA1c was between 7% and 10% or 7.5% and 10.5%. Patients also entered 2-week single-blind, placebo run-in periods prior to randomization.

The addition of sitagliptin to metformin, pioglitazone, or glimepiride appears to show larger reductions in HbA1c and compared with the addition of placebo over 18 to 30 weeks (Table 16). Subjects who received placebo plus glimepiride showed worsening glycemic control, while subjects on placebo plus metformin or placebo plus pioglitazone had slight improvements or no change in HbA1c from baseline. Weight gain was generally seen in patients taking sitagliptin in combination with pioglitazone or glimepiride to a similar extent of those taking pioglitazone alone, however no weight gain was seen in those taking glimepiride alone. Patients randomized to add sitagliptin or placebo to metformin lost weight by 0.4 kg to 0.8 kg compared with baseline (Table 16). Pooled analysis was not conducted due to small number of studies and significant heterogeneity.

One fair quality randomized trial50 studied the effects of sitagliptin or placebo added to ongoing metformin therapy. Unlike the other studies47–49, this trial evaluated the effects of sitagliptin in patients with worse glycemic control (baseline HbA1c between 8% and 11%). These patients were on metformin and diet and exercise for 6 weeks, had baseline HbA1c between 8% and 11%, and had ≥85% adherence to their regimens during a 2-week, placebo run-in period. No patients were naïve to oral hypoglycemic agents and approximately 50% were already taking metformin monotherapy or combination oral therapy at baseline. The addition of sitagliptin to ongoing metformin therapy was more effective than placebo plus metformin at lowering HbA1c (placebo-corrected difference: −1.0%, 95% CI −1.4 to −0.6) over 30 weeks. Both treatment groups exhibited weight loss of −0.5 kg over 30 weeks.

One study was unique in that it included patients who were inadequately controlled on insulin and/or metformin therapy.51 Patients were randomized to sitagliptin 100 mg or placebo in addition to their pre-study doses insulin and metformin (if they were taking). Approximately 70% of patients in both groups were taking metformin at baseline. Doses of insulin and metformin were not increased, however insulin could be decreased if hypoglycemia occurred. Similar results were seen in this study as others, with greater HbA1c lowering seen in patients randomized to sitagliptin than placebo (difference in LS mean change −0.6, 95% CI −0.7 to −0.4). Authors reported no difference in HbA1c lowering in patients on metformin or not on metformin (p=0.44). No difference was noted in weight change from baseline between the two groups, P value NR (Table 16).

Add-on therapy: Sitagliptin or placebo added to 2 existing oral hypoglycemic agents

One fair-quality trial evaluated the addition of sitagliptin or placebo in patients whose glycemia was inadequately controlled on glimepiride 4–8 mg/d alone or glimepiride plus metformin 1500–3000 mg/d.49 Results of sitagliptin or placebo added to glimepiride alone have already been reviewed. In this trial, mean baseline HbA1c was 8.3%, and more than 95% of patients were also taking combination oral hypoglycemic agents at baseline and were considered to have failed this regimen either at screening or after several weeks of dose-stabilization of glimepiride and metformin before participating in a 2-week placebo run-in phase prior to randomization.

In patients already on glimepiride plus metformin, the addition of sitagliptin improved HbA1c over 24 weeks of treatment whereas the addition of placebo showed worsening glycemic control (difference in LS mean change −0.89%, 95%CI −1.1 to −0.68). Weight, however, increased slightly (+0.4 kg, 95% CI −0.1 to 0.9) with sitagliptin relative to placebo; whereas, placebo-treated patients showed weight loss (−0.7 kg, 95% CI −1.4 to −0.1) (Table 17).

Summary of Findings for Saxagliptin

Evidence in children
  • We found no studies including children and adolescents ≤ 18 years
Evidence in adults
  • All studies focused on intermediate outcomes with none focusing on health outcomes as primary outcomes. Some studies reported some health outcomes such as all-cause mortality or cardiac death among secondary outcomes or adverse events. Overall evidence was insufficient to determine how saxagliptin compares with other treatments for their impact on health outcomes.
  • No studies provided data on efficacy or effectiveness for follow up beyond 24 weeks.
  • We found no active-control studies meeting inclusion/exclusion criteria for saxagliptin.
  • Greater reduction in HbA1c with saxagliptin monotherapy compared to placebo (between group difference −0.45 to −0.65%, moderate strength of evidence); reduction was greater with saxagliptin 5 mg than with saxagliptin 2.5 mg.
  • Saxagliptin added on to either metformin, a thiazolidinedione, or glyburide resulted in greater HbA1c reduction than placebo added on to metformin, a thiazolidinedione, or glyburide (between group difference ranges were −0.72 to−0.82%, −0.36 to −0.64, and −0.62 to −0.72, respectively; one study was identified for each comparison, low strength of evidence for each comparison; moderate strength of evidence overall for saxagliptin add-on therapy compared with placebo).
  • Weight loss was greater with placebo than with saxagliptin monotherapy and greater weight loss was seen with saxagliptin 2.5 mg than with 5 mg. (between group difference −0.09 to −0.2 kg for placebo compared with saxagliptin 2.5; −0.8 to −1.3 kg compared with saxagliptin 5, moderate strength of evidence).

Detailed Assessment for Saxagliptin

Randomized controlled trials

We found 5 fair-quality randomized placebo-controlled trials meeting our inclusion/exclusion criteria. This section is organized by how saxagliptin was used (monotherapy or add-on therapy). There were no active control studies identified that met inclusion criteria. Characteristics of included studies are shown in Table 18.

Table 18. Characteristics of saxagliptin placebo-controlled trials in adults with type 2 diabetes.

Table 18

Characteristics of saxagliptin placebo-controlled trials in adults with type 2 diabetes.

Monotherapy: Saxagliptin compared with placebo

In 2 fair-quality randomized controlled trials carried out over 12–24 weeks, a wide variety of doses were compared to placebo.53, 54 Data abstracted was for approved doses in the United States, Saxagliptin 2.5 mg and 5 mg although other doses were studied in identified trials. All patients included in the trials were treatment naïve and mean baseline HbA1c for participants ranged from 7.7–8.0 (Table 18).

Overall, reduction in HbA1c was greater with saxagliptin compared to placebo and slightly greater with saxagliptin 5 mg compared to 2.5 mg (Table 19). With saxagliptin, HbA1c reduction ranged from −0.43 to −0.9% and placebo ranged from −0.27 to +0.19%. There was a numerically greater HbA1c reduction with saxagliptin in the 12 week trial53 compared to the 24 week trial,54 however the placebo corrected change was similar between the two trials. Patients were similar between the 2 trials except patients in the 24 week trial had diabetes for longer than those in the 12 week trial (mean duration 2.3–3.1 years compared to 0.8–1.8 years).

Table 19. Efficacy outcomes for saxagliptin monotherapy compared with placebo.

Table 19

Efficacy outcomes for saxagliptin monotherapy compared with placebo.

Weight loss was seen across all groups, however more weight loss was seen in the placebo group than in either saxagliptin 2.5 mg or 5 mg (−1.03 to −1.4 kg compared to −0.1 to −1.2 kg, Table 19). Patients randomized to saxagliptin 5 mg had less weight loss than those randomized to saxagliptin 2.5 mg.

Add-on therapy: Saxagliptin or placebo added to one oral hypoglycemic agent

Three fair-quality trials were identified that compared saxagliptin to placebo as add-on therapy in patients not achieving adequate glycemic control on either metformin, a thiazolidinedione, or glyburide.55–57 Mean baseline HbA1c ranged from 8.0% to 8.4% and trials were all carried out over 24 weeks. Patients were deemed to have inadequate glycemic control if their HbA1c was ≥7% to ≤10% or ≥7% to ≤10.5% on their current therapy for the previous 8–12 weeks prior to screening.

In general, the addition of saxagliptin to metformin, a thiazolidinedione, or glyburide appears to show larger reductions in HbA1c compared with the addition of placebo over 24 weeks (Table 20). Results were not stratified by which thiazolidinedione patients were taking. Varying results were seen in regards to change in weight. The addition of placebo to glyburide or a thiazolidinedione resulted in less weight gain than the addition of saxagliptin to glyburide or a thiazolidinedione. Slightly more weight loss was seen with the addition of saxagliptin 2.5 mg to metformin than with the addition of saxagliptin 5 mg or placebo. No statistical testing was done to determine the statistical significance of these differences.

Table 20. Efficacy outcomes for saxagliptin or placebo added to one oral hypoglycemic agent.

Table 20

Efficacy outcomes for saxagliptin or placebo added to one oral hypoglycemic agent.

One study randomized patients who had inadequate glycemic control on submaximal doses of sulfonylurea therapy.57 Patients were switched from their current sulfonylurea to open label glyburide 7.5 mg/day. After a 4 week single blind run-in period, patients continued their open label glyburide and were randomized to either saxagliptin 2.5 mg/day, saxagliptin 5 mg/day or placebo + blinded glyburide 2.5 mg/day. Therefore patients randomized to placebo had a total daily dose of glyburide 10 mg daily as compared with glyburide 7.5 mg in the saxagliptin groups.

Summary of Findings for GLP-1 Agonists

Efficacy and effectiveness
Exenatide compared with liraglutide
  • In the one included head-to-head trial (N=464), liraglutide 1.8 mg once daily reduced mean HbA1c significantly more than exenatide 10 mcg twice daily (−1.12% compared with −0.79%; estimated treatment difference −0.33; 95% CI −0.47 to −0.18, low strength of evidence).
  • Exenatide and liraglutide resulted in similar weight loss (−2.87 compared with −3.24 kg, respectively; estimated treatment difference −0.38 kg; 95% CI −0.99 to 0.23, low strength of evidence)
Exenatide
Evidence in children
  • No included study examined children or adolescents with type 2 diabetes.
Evidence in adults
  • Except for one study reporting quality of life, no included studies examined the impact of treatment with exenatide on health outcomes (such as MI, death, stroke, or renal failure) (insufficient strength of evidence). The longest duration of an included study was 52 weeks.
  • Four active-control trials compared exenatide to insulin, with both groups also receiving oral diabetes agents, and all found no significant difference between groups for reduction in HbA1c (range for exenatide 10 mcg twice daily −1.0% to −1.4%; range for insulin −0.9% to −1.4%, moderate strength of evidence). In one of the trials, the substitution of exenatide for insulin did not improve HbA1c compared to continuing insulin.
  • Active-control studies demonstrated significant weight loss in exenatide groups compared to weight gain with insulin (treatment difference range 4.1 kg to 5.4 kg, moderate strength of evidence).
  • One active-control trial found no significant difference in improvement in HbA1c between exenatide and glibenclamide (−1.5% compared with −1.8%, P>0.05, low strength of evidence). Weight loss in the exenatide arm of the study was significantly greater than in the glibenclamide arm of the study (−8.0 kg compared with +4.3 kg, P<0.001, low strength of evidence).
  • One trial comparing exenatide to rosiglitazone with all participants on background metformin therapy, found no significant difference in improvement in HbA1c (−0.9% vs. −1.0%, P=0.720), but greater weight loss in the exenatide arm of the study (−2.8 kg vs. +1.5, P<0.001).
  • Greater reduction in HbA1c with exenatide than with placebo, both when added to various oral agents and as monotherapy. For exenatide 5 mcg twice daily compared with placebo (5 studies) weighted mean difference in HbA1c −0.72%, 95% CI −0.99% to −0.45% (moderate strength of evidence); for exenatide 10 mcg twice daily compared with placebo (8 studies) weighted mean difference in HbA1c −0.90%, 95% CI −1.08% to −0.73% (high strength of evidence).
  • For change in weight, pooled analysis (5 studies) found no statistically significant difference between exenatide 5 mcg twice daily and placebo (weighted mean difference −0.61 kg, 95% CI −1.28 to 0.06). However, statistical heterogeneity was high for the pooled analysis (I2=74%), and a sensitivity analysis removing a single study resulted in significant weight loss for exenatide 5mcg compared to placebo (weighted mean difference −0.87, 95% CI −1.35 to −0.40, P<0.001, I2=33%) (low strength of evidence).
  • For change in weight, exenatide 10 mcg twice daily resulted in significant weight loss compared to placebo (weighted mean difference −1.25 kg, 95% CI −1.60 to −0.90, high strength of evidence).
  • Quality of life was examined in only one study of exenatide 10 mcg twice a day. No significant differences were seen between exenatide and insulin glargine (low strength of evidence).
Liraglutide
Evidence in children
  • No study examined children or adolescents with type 2 diabetes
Evidence in adults
  • No included studies focused on health outcomes as the primary outcomes. Several studies reported a health outcome among other secondary outcomes or in the adverse events section. Overall evidence was insufficient to determine how liraglutide compares with other treatments for their impact on health outcomes.
  • The longest duration of an included study was 52 weeks.
  • Three active-control trials comparing liraglutide to glimepiride demonstrated improvement in HbA1c in both treatment groups. Results indicate either no significant difference between treatment groups (2 trials) with liraglutide 0.6 mg daily and glimepiride 1 to 4 mg daily58 and between liraglutide 1.2 mg and 1.8 mg daily and glimepiride 4 mg daily59 or greater improvement in HbA1c with liraglutide (1.2 mg and 1.8 mg daily) than with glimepiride 8 mg daily (insufficient strength of evidence).60
  • Liraglutide 1.2mg and 1.8 mg daily leads to weight loss whereas glimepiride causes weight gain (moderate strength of evidence).
  • Greater reduction in HbA1c in one good quality active-control trial comparing liraglutide 1.8 mg daily to open-label insulin glargine (−1.33% compared with −1.09%; P=0.0015, low strength of evidence)
  • Weight loss with liraglutide compared with weight gain with insulin glargine in the same study (treatment difference −3.43 kg; P<0.0001, low strength of evidence)
  • One trial comparing the addition of rosiglitazone with the addition of liraglutide (to ongoing glimepiride treatment) reported greater reduction in HbA1c with liraglutide (−1.1 compared with −0.4%, P<0.0001, low strength of evidence) and greater weight gain in the rosiglitazone arm compared to all doses of liraglutide (change in weight: liraglutide 0.6mg +0.7kg; liraglutide 1.2mg +0.3 kg; liraglutide 1.8mg −0.2 kg; rosiglitazone 4 mg +2.1 kg; P<0.0001 for all doses of liraglutide compared to rosiglitazone).
  • Greater reduction in HbA1c with liraglutide 1.2 mg and 1.8 mg daily than with sitagliptin 100 mg daily in one trial (−1.24% to −1.5% compared with −0.6%; P<0.0001, low strength of evidence).41
  • Greater weight loss with liraglutide 1.2 mg and 1.8 mg daily than with sitagliptin in the same study (−2.86 kg to −3.38 kg compared with −0.96 kg; P<0.0001, low strength of evidence).41
  • Greater reduction in HbA1c with liraglutide than with placebo (moderate strength of evidence), both when added to various oral agents and as monotherapy (liraglutide 0.6 to 0.65 mg daily weighted mean difference −1.10, 95% CI −1.45 to −0.75; liraglutide 1.2 to 1.25 mg daily weighted mean difference −1.28, 95% CI −1.56 to −1.00; liraglutide 1.8 to 1.9 mg daily weighted mean difference −1.26, 95% CI −1.50 to −1.03). Although all of the individual studies showed that liraglutide significantly decreased HbA1c compared to placebo, statistical heterogeneity in pooled analyses was substantial (I2 71% to 82%).
  • When compared with placebo, liraglutide 1.8 mg to 1.9 mg daily produced a significant decrease in weight (weighted mean difference −1.43 kg, 95% CI −2.33 to −0.56, moderate strength of evidence).
  • There was no statistically significant weight loss for liraglutide 0.6 to 0.65 mg compared with placebo (moderate strength of evidence).
  • Liraglutide 1.2 to 1.25 mg resulted in significant weight loss compared to placebo in all studies except for in the 1 included study in which all participants were on background sulfonylurea therapy; meta-analyses of 3 trials using the 1.2 to 1.25 dose indicated no statistically significant difference in weight change between liraglutide and placebo (weighted mean difference −0.83 kg, 95% CI −1.85 to 0.19) but there was substantial statistical heterogeneity (I2 76%); removing 1 trial where subjects were all on background sulfonylureas resulted in a finding of greater weight loss with liraglutide than with placebo (weighted mean difference −1.31 kg, 95% CI −1.85 to −0.77) (low strength of evidence).

Detailed Assessment of Exenatide Compared with Liraglutide

We found one fair quality randomized controlled trial comparing liraglutide to exenatide.61 In this 26-week open-label study, 464 participants were randomized to liraglutide 1.8 mg once daily or exenatide 10 mcg twice daily. Participants were continued on their background oral antidiabetic therapy which was either metformin, a sulfonylurea, or both.

In this study, liraglutide reduced mean HbA1c significantly more than exenatide (−1.12% [SE 0.08] compared with −0.79% [SE 0.08]; estimated treatment difference −0.33; 95% CI −0.47 to −0.18; P<0.0001). Both liraglutide and exenatide resulted in similar weight loss (liraglutide −3.24 kg [0.33] compared with exenatide −2.87 [0.33]; estimated treatment difference −0.38 kg; 95% CI −0.99 to 0.23; P=0.2235).

Detailed Assessment for Exenatide

Active-control trials

Four open label studies compared exenatide 10 mcg twice a day to insulin therapy (various regimens). All studies used concurrent sulfonylurea and/or metformin in addition to the study treatment regimens (Table 21, Evidence Table 3). Three of these trials were fair-quality noninferiority studies,62–64 and 1 was a fair-quality exploratory substitution study.65 The outcomes in these 4 trials were too heterogeneous to pool in meta-analyses. In addition to the four trials comparing exenatide to insulin, we also identified one trial comparing exenatide to glibenclamide,66 and one trial comparing exenatide to rosiglitazone.67

Table 21. Characteristics of exenatide active-control trials in adults with type 2 diabetes.

Table 21

Characteristics of exenatide active-control trials in adults with type 2 diabetes.

Efficacy and effectiveness

Heine and colleagues63 compared once-daily insulin glargine to exenatide twice daily over 26 weeks of follow-up in a noninferiority study, with both groups receiving metformin and a sulfonylurea. Reductions in HbA1c were 1.11% in both groups (between-group difference 0.017%, 95% CI −0.123 to 0.157%). Weight increased in the insulin glargine group throughout the trial, with progressive reduction in the exenatide group (weight change −2.3 kg with exenatide, +1.8 kg with insulin glargine; between-group difference −4.1 kg, 95% CI −4.6 to −3.5 kg).

Quality of life was assessed in this trial.63, 68 A per protocol analysis of 455 of 549 original trial patients revealed no significant differences between the 2 treatments for measures of symptoms, quality of life, vitality, and treatment satisfaction. These similar outcomes occurred despite an additional injection daily and gastrointestinal adverse events with exenatide.

Another noninferiority study62 also compared exenatide 10 mcg twice daily to insulin glargine, with both groups continuing pre-study single oral agents. Change in HbA1c at 16 weeks was identical in the 2 treatment arms (−1.36%, SE 0.09%, within group P<0.001). Both exenatide and insulin glargine reduced HbA1c by a similar amount in patients with baseline HbA1c ≥ 9% (approximate change −1.8%) and < 9% (change −0.9%).62

A third non-inferiority study64 compared exenatide twice daily with biphasic insulin aspart in patients poorly controlled on sulfonylurea and metformin. The change in HbA1c was similar between groups (change with exenatide −1.04%, change with insulin aspart −0.89%; between group difference −0.15%, 95% CI −0.32 to 0.01). Exenatide patients lost weight while insulin-treated patients gained weight (between-group difference −5.4 kg, 95% CI −5.9 to −5.0 kg).

The fourth active-control trial65 examined persons with type 2 diabetes who were already using insulin and sulfonylurea and/or metformin. In this small (N=51), exploratory randomized controlled trial, exenatide 5 and then 10 mcg twice daily was substituted for insulin, while oral agents were continued. Specific glycemic goals were not set. HbA1c did not change significantly in either group (P>0.05) and there was no significant between-group difference in HbA1c at 12-week follow-up. Exenatide patients noted a decrease in weight (mean weight change −4.2 kg, SD 3.0 kg, P<0.001), in contrast to the insulin group (mean weight change +0.5 kg, SD 1.7, P<0.001).

In addition to the four trials described above comparing exenatide to insulin, we also identified one trial comparing exenatide to glibenclamide, and one trial comparing exenatide to rosiglitazone. In the 12-month trial comparing exenatide to glibenclamide, all participants in the study continued on metformin.66 There was no significant difference in improvement in HbA1c between those treated with exenatide and those treated with glibenclamide (change with exenatide −1.5%, change with glibenclamide −1.8%, P>0.05.) Weight loss in the exenatide arm of the study was significantly greater than in the glibenclamide arm of the study (change with exenatide −8.0 kg, change with glibenclamide +4.3 kg, P<0.001).

In the 20-week study comparing exenatide to rosiglitazone with all participants on background metformin therapy, there was no significant difference in improvement in HbA1c between the exenatide and rosiglitazone arms (change with exenatide −0.9%, change with rosiglitazone −1.0%, P=0.720).67 Weight loss in the exenatide arm of the study was significantly greater than in the rosiglitazone arm of the study (change with exenatide −2.8 kg, change with rosiglitazone +1.5, P<0.001.)

Placebo-controlled trials

We identified 9 fair-to-good-quality placebo-controlled trials69–77 of exenatide (Table 22, Evidence Table 3). Overall, study subjects were fairly homogeneous. Subjects were similar in age (mean 53 to 62 years) and sex (37 to 75% male) with some variation in race and ethnicity. Mean baseline HbA1c ranged from 7.1% to 8.6% and mean duration of diabetes from 2 to 14.8 years.

Table 22. Characteristics of exenatide placebo-controlled trials in adults with type 2 diabetes.

Table 22

Characteristics of exenatide placebo-controlled trials in adults with type 2 diabetes.

Efficacy and effectiveness

We included 9 trials comparing exenatide to placebo (Table 22). All found statistically significant weight loss with exenatide compared to placebo. All but one of the trials found statistically significant reduction in HbA1c with exenatide compared to placebo. The one trial that did not find statistically significant reduction in HbA1c compared to placebo was different from the other trials in that participants in the study had relatively well controlled diabetes at baseline (HbA1c 7.1–7.5 at baseline).77 In this section, we first describe the nine placebo-control trials, and then present the results of our meta-analyses for HbA1c and weight.

Three similar studies compared exenatide to placebo, with both treatment groups taking oral hypoglycemic agents.69–71 Kendall and colleagues71 randomized patients to exenatide 5 mcg or 10 mcg or placebo twice daily over 30 weeks. Patients continued their pre-study metformin and a sulfonylurea. Hemoglobin A1c decreased in the exenatide arms and steadily increased with placebo (placebo-adjusted change in HbA1c for exenatide 5 mcg, −0.8%; 10 mcg, −1.0%; P<0.001 for both treatment groups compared with placebo). Weight decreased progressively in both exenatide arms, more so than in the placebo arm (weight change −1.6 kg, SE 0.2 kg in both exenatide groups; −0.9 kg, SE 0.2 kg with placebo).

In a similarly designed study Buse and colleagues69 compared exenatide to placebo in patients taking a sulfonylurea. Hemoglobin A1c improved in both treatment groups (HbA1c change with exenatide 5 mcg, −0.46%; 10 mcg, −0.86%) while increasing slightly in the placebo group (between-group P≤ 0.0002). Weight decreased more in the exenatide groups (weight change −1.6 kg, SE 0.3) than in the placebo group (weight change −0.6 kg, SE 0.3 kg). DeFronzo and colleagues70 performed a similar study except that all subjects were taking metformin. The researchers noted very similar improvements in HbA1c with exenatide 10 mcg (HbA1c change −0.78%, SE 0.1%) compared with placebo (HbA1c change 0.08%, SE 0.10%) and also a similar decrease in weight with exenatide.

In a fourth placebo-controlled trial, subjects who were inadequately controlled with a thiazolidinedione (with or without metformin), were randomized to exenatide 10 mcg twice daily or placebo.72 Exenatide improved HbA1c (mean between-group difference −0.98, 95% CI −1.21 to −0.74). Exenatide reduced weight but placebo did not (between-group difference −1.51 kg, 95% CI −2.15 to −0.88).

Three additional placebo-controlled trials of subjects inadequately controlled with oral antidiabetic agents found that HbA1c improved and weight was reduced with exenatide treatment compared with placebo, when subjects were continued on oral antidiabetic agents.74–76 One study randomized subjects on metformin with or without a sulfonylurea to exenatide 10 mcg twice daily or placebo.74 At 16 weeks, HbA1c reduction from baseline was significantly greater in the exenatide treatment group than with placebo (−1.2% compared with −0.4%; P<0.001). Weight reduction was also greater with exenatide than placebo (−1.2 kg compared with −0.1 kg; P<0.001). In a similarly designed study, Kadowaki and colleagues75 randomized subjects with suboptimally controlled diabetes on oral antidiabetic agents (a sulfonylurea with or without biguanide or a thiazolidinedione) to exenatide 2.5 mcg twice daily, exenatide 5 mcg twice daily, exenatide 10 mcg twice daily, or placebo. This study found a dose-dependent effect on glycemic control with exenatide compared to placebo (HbA1c change with exenatide 2.5 mcg −0.9%; exenatide 5 mcg −1.2%; exenatide 10 mcg −1.4%; placebo +0.02%; all P<0.001 compared with placebo). This study did not find a significant weight reduction in the exenatide treatment groups compared with placebo. The third study randomized subjects on oral antidiabetic agents (metformin or a sulfonylurea) to exenatide or placebo.76 All subjects continued on oral antidiabetic therapy, and started an intensive lifestyle modification program. The exenatide arm of the study showed greater improvement in HbA1c (−1.21% compared to −0.73%, P<0.0001), and greater weight loss (−6.16 kg compared to −3.97 kg, P=0.003).

One placebo-controlled trial of exenatide did not find improvement in HbA1c with exenatide compared to placebo, but subjects in this study at baseline had relatively well controlled diabetes on background therapy with metformin and/or a thiazolidinedione.77 Exenatide did result in a statistically significant reduction in weight compared to placebo (weight change exenatide −1.8 kg, placebo 0.3 kg, P<0.05).

One placebo-controlled trial evaluated exenatide monotherapy in patients with type 2 diabetes naive to antidiabetic agents.73 Subjects were randomized to exenatide 5 mcg, exenatide 10 mcg, or placebo, and were on no oral hypoglycemic agents. At 24 weeks, HbA1c reduction from baseline was significantly greater in both exenatide treatment groups than with placebo (HbA1c change with exenatide 5 mcg −0.7%; 10 mcg −0.9%; placebo −0.2%; P<0.01 for both treatment groups compared with placebo). Weight reduction was also greater with exenatide than with placebo (weight change with exenatide 5 mcg −2.8 kg; 10 mcg −3.1 kg; placebo −1.4 kg; P<0.01 for both treatment groups compared with placebo).

In several placebo-controlled trials of exenatide combined with oral agents, patients with a baseline HbA1c more than 9.0% achieved greater reductions in HbA1c than subjects with baseline less than 9.0%.69, 71, 78 Weight reductions were greater in persons who had higher body mass index at baseline.79, 80

These studies were sufficiently homogeneous to obtain pooled estimates of effect (Table 23). When compared with placebo, exenatide 5 mcg twice daily produced a significant decrease in HbA1c (weighted mean difference −0.72, 95% CI −0.99 to −0.45, P<0.001, Appendix E).

Table 23. Placebo-control trials of exenatide: Summary of meta-analyses.

Table 23

Placebo-control trials of exenatide: Summary of meta-analyses.

A larger improvement in HbA1c was noted with exenatide 10 mcg twice daily (weighted mean difference −0.90, 95% CI −1.08 to −0.73, P<0.001). There was considerable statistical heterogeneity between the studies in these analyses (I 2=76% for exenatide 10 mcg, I2=78% for exenatide 5 mcg). Because of the considerable heterogeneity, we repeated these meta-analyses without the study by Kadowaki et al. (Table 23). After removing it from the analysis, the statistical heterogeneity was reduced (I2=57% for exenatide 10 mcg, I2=1% for exenatide 5 mcg) and the magnitude of effect size from our pooled estimates was almost the same, but was slightly decreased (exenatide 10 mcg twice daily compared with placebo weighted mean difference −0.84, 95% CI −0.97 to −0.70, P<0.001; exenatide 5 mcg twice daily compared with placebo weighted mean difference −0.60, 95% CI −0.74 to −0.46, P<0.001). We hypothesize that the high heterogeneity when including Kadowaki et al. is due to the study being conducted in a different population (all Japanese participants) and having a small sample size.

When compared with placebo, exenatide 10 mcg twice daily produced a statistically significant decrease in weight (weighted mean difference −1.25 kg, 95% CI −1.60 to −0.90, P<0.001). The decrease for exenatide 5 mcg twice daily was not statistically significant in the meta-analysis including Kadowaki et al. (weighted mean difference −0.61 kg, 95% CI −1.28 to 0.06, P=0.074). There was substantial statistical heterogeneity prior to removing Kadowaki et al. from this analysis (I2=74%). After removing Kadowaki et al, the heterogeneity was not statistically significant, and pooled estimates of effect were increased (exenatide 10 mcg twice daily compared with placebo weighted mean difference −1.34, 95% CI −1.71 to −0.97, P<0.001; exenatide 5 mcg twice daily compared with placebo weighted mean difference −0.87, 95% CI −1.35 to −0.40, P<0.001).

Systematic reviews

Three systematic reviews which included exenatide met our inclusion criteria and were rated fair or good quality.28, 81, 82 In 2007, Amori and colleagues28 published a review of published and unpublished English-language studies of US Food and Drug Administration-approved and unapproved DPP-4 inhibitors (sitagliptin and vildagliptin) and GLP-1 agonists including exenatide. These reviewers derived the following pooled estimates of change from baseline for exenatide compared with placebo (both groups combined with various oral diabetes agents): HbA1c −1.01% (95% CI −1.18% to −0.84%) and weight −1.44 kg (95% CI −2.13 to −0.75 kg). When exenatide was compared with various insulin regimens, the following pooled estimates of change from baseline for exenatide compared with insulin were noted: HbA1c −0.06% (95% CI −0.22% to 0.10%) and weight −4.8 kg (95% CI −6.0 to −3.5 kg). Weight loss was dose-dependent and progressive, with no apparent plateau by week 30.

A second systematic review, published by Pinelli and colleagues in 2008, also compared exenatide to placebo and insulin and in terms of glycemic control and weight loss.81 In a meta-analysis of the 3 included studies, exenatide improved HbA1c compared to placebo (weighted mean difference −0.97% (95% CI −1.11 to −0.83), and also showed a slight improvement in HbA1c compared to insulin in 2 included studies (weighted mean difference −0.08%, 95% CI −.23 to 0.07). A meta-analysis of all 5 included studies on exenatide found significant weight loss with exenatide compared to placebo or insulin therapy (weighted mean difference −2.74 kg, 95% CI −4.85 to 0.64 kg).

Another systematic review of GLP-1 receptor agonists, including exenatide and liraglutide, was also included.82 This study combined trials of both exenatide and liraglutide into one meta-analysis for HbA1c and one meta-analysis for weight loss. Combining the included trials of exenatide and liraglutide derived the following pooled estimates of GLP-1 agonists compared to placebo: HbA1c −1.0% (95% CI −1.1% to −0.8%). Similar results were obtained with separate analyses of exenatide and liraglutide compared to placebo. In our meta-analyses, we separated pooled estimates by dose of liraglutide, and found greater reduction in HbA1c at the higher doses of liraglutide. Monami et al. found significant weight loss with exenatide compared to placebo, and a nonsignificant trend toward weight loss with liraglutide compared to placebo. For our liraglutide analyses (described in the following section), we separated pooled estimates by dose of liraglutide and did find significantly greater weight loss with liraglutide compared to placebo at the higher dose of liraglutide (1.8mg daily).

Detailed Assessment for Liraglutide

Active-control trials

We found 6 fair or good quality active-control trials. Three fair quality active-control trials with a similar design compared liraglutide to glimepiride in terms of HbA1c reduction and weight loss.58–60 In 2 of these studies, subjects were on no other antidiabetic agents.58, 60 In one study, all subjects were taking metformin 1 g twice daily in addition to the study treatment regimes. We did not attempt to pool data for these 6 trials due to heterogeneity of study designs, outcome reporting and comparisons.

One good quality active-control trial compared liraglutide to open-label insulin glargine, with all subjects on combination therapy with metformin and glimepiride.83 One fair quality active-control trial compared liraglutide to rosiglitazone.84 An additional fair quality active-control trial compared liraglutide to sitagliptin.41 These studies are summarized in Table 24, Evidence Table 3.

Table 24. Characteristics of liraglutide active-control trials in adults with type 2 diabetes.

Table 24

Characteristics of liraglutide active-control trials in adults with type 2 diabetes.

Efficacy and effectiveness

Three fair quality-studies compared the efficacy of liraglutide to glimepiride. 58–60 In a phase 2, dose-finding study Madsbad and colleagues58 compared 5 fixed dosage groups of liraglutide (0.045 mg, 0.225 mg, 0.45 mg, 0.60 mg, and 0.75 mg daily) to glimepiride 1–4 mg daily and to placebo. Liraglutide 0.60 mg daily was the only approved dose of liraglutide in this study; we will focus on the outcomes for this arm only. After 12 weeks of therapy, there was a significant reduction in HbA1c compared to placebo for the liraglutide 0.6 mg arm, and the glimepiride arm (HbA1c change: liraglutide 0.60 mg compared to placebo −0.70%, glimepiride compared to placebo −0.74%). Treatment with liraglutide 0.6 mg daily or glimepiride did not significantly increase or decrease body weight in this study. According to the prescribing information for liraglutide, liraglutide 0.6 mg is a dose intended to be used for reduction of gastrointestinal side effects during the initial titration, and should not be used for glycemic control.

Two later studies compared the efficacy of liraglutide to glimepiride with higher doses of liraglutide.59, 60 Nauck and colleagues, as part of the LEAD-2 study, randomized subjects to liraglutide 0.6 mg, liraglutide 1.2 mg, liraglutide 1.8 mg daily, glimepiride 4 mg daily, or placebo. All subjects were also on metformin 1 g twice daily. At 26 weeks, all of the treatment arms showed improvement in HbA1c (change in HbA1c: liraglutide 0.6 mg −0.7%; liraglutide 1.2 mg −1.0%; liraglutide 1.8 mg −1.0%, glimepiride 4 mg −1.0%.) Improvement in HbA1c in the liraglutide 1.2 mg and 1.8 mg arms was noninferior to treatment with glimepiride. There was a statistically significant difference between the weight loss in all of the liraglutide treatment groups and the weight gain in the glimepiride group (weight change liraglutide 0.6 mg −1.8 kg; liraglutide 1.2 mg −2.6 kg; liraglutide 1.8 mg −2.8 kg, glimepiride +1.0 kg; P<0.0001.)

Garber and colleagues, as part of the LEAD-3 Mono (Liraglutide Effect and Action in Diabetes-3 Mono) study, randomized subjects to liraglutide 1.2 mg daily, liraglutide 1.8 mg daily, or glimepiride 8 mg daily. At 52 weeks, all of the treatment arms showed improvement in HbA1c (change in HbA1c: liraglutide 1.2 mg −0.84%; liraglutide 1.8 mg −1.14%, glimepiride 8 mg −0.51%). Reduction in HbA1c was significantly greater in both liraglutide arms than in the glimepiride arm (P<0.01 for both comparisons). There was a statistically significant difference between the weight loss in the liraglutide arms and the weight gain in the glimepiride arm (P<0.0001, exact values of weight change not reported).

Patient-reported outcomes were also followed as part of the LEAD-3 Mono study.85 This study used a survey to assess a composite health-related quality of life score, and found that this score improved more favorably with treatment with liraglutide 1.8 mg compared to glimepiride (P= 0.004). There was no statistical difference for this scale between liraglutide 1.2 mg and glimepiride.

In summary, Garber and colleagues found statistically significantly greater improvement in HbA1c with liraglutide 1.2 mg and 1.8 mg daily compared to glimepiride 8 mg daily with subjects on no other antidiabetic therapy, and Nauck and colleagues showed noninferiority of liraglutide 1.2 mg and 1.8 mg daily compared with glimepiride. Both studies showed significantly greater weight loss with liraglutide compared to glimepiride.

One good quality active-control trial compared liraglutide 1.8 mg daily to open-label insulin glargine, with all subjects on combination therapy with metformin and glimepiride. 83 Liraglutide reduced HbA1c significantly compared to glargine (−1.33% compared with −1.09%; P=0.0015). The study also found greater weight loss with liraglutide compared with insulin glargine (treatment difference −3.43 kg; P<0.0001).

One fair quality 26 week active-control trial compared liraglutide (0.6, 1.2, or 1.8 mg daily) to rosiglitazone. All subjects were on glimepiride 2 to 4 mg daily.84. The study found a greater improvement in HbA1c with the two higher doses of liraglutide compared to rosiglitazone 4 mg daily (change in HbA1c: liraglutide 1.2 mg −1.1%; liraglutide 1.8 mg −1.1%; rosiglitazone −0.4%). The difference was statistically significant (P<0.0001 for liraglutide 1.2 and 1.8 mg daily compared to placebo). The study also found significant weight gain in the rosiglitazone arm compared to all doses of liraglutide (change in weight: liraglutide 0.6mg +0.7kg; liraglutide 1.2mg +0.3 kg; liraglutide 1.8mg −0.2 kg; rosiglitazone 4 mg +2.1 kg; P<0.0001 for all doses of liraglutide compared to rosiglitazone).

One 26 week fair quality active-control trial compared liraglutide (1.2 or 1.8 mg daily) to sitagliptin 100 mg daily.41 All study participants were on metformin ≥ 1500 mg daily as background therapy. The study found a greater improvement in HbA1c with both doses of liraglutide compared to sitagliptin (change in HbA1c: liraglutide 1.2 mg −1.24%; liraglutide 1.8 mg −1.5%; sitagliptin −0.6%; P<0.0001 for both doses of liraglutide compared to sitagliptin). Weight loss was significantly greater with both doses of liraglutide compared to sitagliptin (change in weight: liraglutide 1.2 mg −2.86 kg; liraglutide 1.8 mg −3.38 kg; sitagliptin −0.96 kg; P<0.0001 for both doses of liraglutide compared to sitagliptin. Treatment satisfaction as measured by the Diabetes Treatment Satisfaction Questionnaire (DTSQ) improved both with liraglutide and sitagliptin, but increased significantly more in the liraglutide 1.8 mg arm of the study than in the sitagliptin 100 mg arm of the study.41 None of the other active-control trials examined treatment satisfaction or quality of life.

Placebo-controlled trials

We found 7 fair or good quality liraglutide placebo-control trials (Table 25).58, 59, 83, 84, 86–88 Four of these included an active-control arm in addition to a placebo arm and are described above in the previous section.58, 59, 83, 84 Overall, study subjects were fairly homogeneous. Subjects were similar in age (mean 53 to 60 years). Race was not reported in 5 of the 7 studies, and was 81% to 89% white when it was reported. Sex ranged from 45% to 85% male. Mean baseline HbA1c ranged from 7.1% to 8.6% and mean duration of diabetes from 3 to 10 years.

Table 25. Characteristics of liraglutide placebo-controlled trials in adults with type 2 diabetes.

Table 25

Characteristics of liraglutide placebo-controlled trials in adults with type 2 diabetes.

In 3 of the studies, study participants were on no other antidiabetic therapy.58, 87, 88 In the other 4 studies, participants were on combination therapy with metformin,59 metformin and glimepiride,83 metformin and rosiglitazone,86 and glimepiride. 84

Efficacy and effectiveness

All of the studies showed that liraglutide therapy resulted in a significant decrease in HbA1c compared to placebo. Pooled estimates of effect were obtained through meta-analyses for three doses of liraglutide. (Table 26, Appendix E) There was substantial statistical heterogeneity (I2 71% to 82%) for all of the doses of liraglutide, most likely secondary to differences in background therapy between studies. However, all of the studies found a significant decrease in HbA1c compared to placebo. When compared with placebo, liraglutide 0.6 to 0.65 mg daily produced a significant decrease in HbA1c (weighted mean difference −1.10, 95% CI −1.45 to −0.75, P<0.001, Table 26). A similar improvement in HbA1c was noted with liraglutide 1.2 to 1.25 mg daily (weighted mean difference −1.28, 95% CI −1.56 to −1.00, P<0.001) and 1.8 to 1.9 mg daily (weighted mean difference −1.26, 95% CI −1.50 to −1.03, P<0.001).

Table 26. Placebo-controlled trials of liraglutide: Summary of meta-analyses.

Table 26

Placebo-controlled trials of liraglutide: Summary of meta-analyses.

When compared with placebo, liraglutide 1.8 mg to 1.9 mg daily produced a significant decrease in weight (liraglutide 1.8 mg to 1.9 mg weighted mean difference −1.43 kg, 95% CI −2.33 to −0.53, P=0.002). There was no statistically significant weight loss for liraglutide 0.6 to 0.65 mg or liraglutide 1.2 mg to 1.25 mg daily compared with placebo, although there was considerable heterogeneity in the meta-analysis of liraglutide 1.2 mg to 1.25 mg daily (Table 26). In reviewing the results, this considerable heterogeneity was largely secondary to the inclusion of the LEAD-1 SU study by Marre et al.84 In this study, participants in all arms were on background therapy with glimepiride, and participants in the liraglutide 0.6 mg and 1.2 mg arms of the study gained, rather than lost, weight. Because of this difference, we ran the meta-analyses for weight both including and excluding Marre et al. (Table 26). With the exclusion of Marre et al., there was significant weight loss with liraglutide 1.2 mg compared to placebo (weighted mean difference −1.31 kg, 95% CI −1.85 to −0.77, P<0.001). This suggests that the 1.2 mg dose of liraglutide may lead to weight loss as monotherapy or combined with metformin, but not in combination with a sulfonylurea.

II. Thiazolidinediones (TZDs)

Summary of Findings for Thiazolidinediones (TZDs)

Evidence in children
  • No data on children were reported.
Evidence in adults
  • Pioglitazone compared with rosiglitazone. Meta-analysis of 8 head-to-head randomized controlled trials found no statistically significant difference between pioglitazone and rosiglitazone for their ability to improve glycemic control (for change in HbA1c, weighted mean difference −0.09, 95% CI −0.23, 0.05, I2 0.0%) (moderate strength of evidence). Prior systematic reviews found both drugs appear to have similar effects on HbA1c, producing a decrease of approximately 1%, similar to the change produced with other oral agents (including metformin, glibenclamide, or glimepiride). Effect of both pioglitazone and rosiglitazone appears to be similar when used in either monotherapy or combination therapy.
  • Pioglitazone compared with rosiglitazone. None of the included head-to-head trials reported comparative efficacy/effectiveness of health outcomes or utilization outcomes.
  • Overall, no difference in reduction in HbA1c between pioglitazone and sulfonylureas (moderate strength of evidence). We included 10 trials, 7 finding no statistically significant difference, 2 favoring pioglitazone by 0.19 to 0.32%, and one favoring glimepiride by 0.63%.
  • No significant difference in 7 trials for reduction in HbA1c between pioglitazone and metformin (high strength of evidence).
  • No significant difference in reduction in HbA1c between rosiglitazone and sulfonylureas (moderate strength of evidence). We included 9 trials, 7 finding no statistically significant difference, one favoring rosiglitazone by 0.42%, and one favoring the sulfonylurea group by 0.4%.
  • No significant difference in reduction in HbA1c between rosiglitazone and metformin (moderate strength of evidence). We included 4 trials, 3 finding no statistically significant difference and one favoring rosiglitazone by 0.13%.
  • For reduction in HbA1c, consistent with our findings, prior systematic reviews reported no between-group differences between thiazolidinediones and metformin or second- generation sulfonylureas.
  • Thiazolidinedione plus metformin compared with a second-generation sulfonylurea plus metformin (4 randomized controlled trials) did not show a consistent effect favoring 1 of the combinations, nor did a randomized controlled trial comparing thiazolidinediones with repaglinide.
  • No significant difference in reduction in HbA1c between rosiglitazone and sitagliptin in two randomized controlled trials (moderate strength of evidence).
  • One trial comparing the addition of rosiglitazone with the addition of liraglutide (to ongoing glimepiride treatment) reported greater reduction in HbA1c with liraglutide (−1.1 compared with −0.4%, P<0.0001, low strength of evidence)
  • One trial comparing exenatide to rosiglitazone with all participants on background metformin therapy, found no significant difference in improvement in HbA1c (−0.9% vs. −1.0%, P=0.720).
  • Data were not sufficient to determine the comparative effectiveness of pioglitazone and rosiglitazoneon microvascular or macrovascular complications of diabetes; there were no head-to-head data (insufficient strength of evidence).

Detailed Assessment for TZDs

Systematic reviews: Pioglitazone compared with rosiglitazone

In a report for the Agency for Healthcare Research and Quality report,89 Bolen and colleagues examined 4 head-to-head studies comparing pioglitazone with rosiglitazone and did not find a significant difference for HbA1c between these 2 drugs.

Head-to-head trials: Pioglitazone compared with rosiglitazone

Eight fair-quality, head-to-head, randomized controlled trials (in 14 publications) were identified (Table 27 and Evidence Table 4).90–102

Table 27. Head-to-head trials comparing pioglitazone with rosiglitazone in persons with type 2 diabetes.

Table 27

Head-to-head trials comparing pioglitazone with rosiglitazone in persons with type 2 diabetes.

Details of the trials comparing pioglitazone with rosiglitazone are presented in Table 27 and Evidence Table 4. Some trials compared monotherapy with either medication,92, 93, 99, 101, 102 while others compared adding pioglitazone or rosiglitazone to existing treatment.90, 91, 94–98, 100 All trials reporting improvement in HbA1c (%) for subjects treated with either pioglitazone or rosiglitazone found no statistically significant difference between groups. The range of improvement (change from baseline in HbA1c[%]) with either treatment was from a 0.6 to a 1.4.

Our meta-analysis including 7 of these trials found no statistically significant difference between pioglitazone and rosiglitazone (weighted mean difference −0.09, 95% CI −0.23, 0.05, I2 0.0%, Appendix E). One of the trials did not report sufficient outcome data to be included.93

Systematic reviews: Active- and placebo-controlled trials with TZDs
Original report

For the original Drug Effectiveness Review Project drug class report on TZDs, 10 reviews reporting comprehensive searches were identified (Evidence Tables 1 and 2 from that report).18 Six of the reviews were rated poor quality, as they lacked 1 or more of the following: explicit inclusion criteria, specification of the search strategy, quality assessment of individual studies, or sufficient detail on the individual studies.105, 106 107–110 Details of the 4 fair- to good-quality systematic reviews are provided in Evidence Table 1 from the 2008 TZD report.18

Three systematic reviews examined both pioglitazone and rosiglitazone.111–113 Boucher and colleagues111 compared the 2 thiazolidinediones to other antidiabetic drugs; they did not directly compare pioglitazone and rosiglitazone. They concluded that as monotherapy these 2 drugs have effects on HbA1c similar to the other antidiabetic drugs, and when added to one of those drugs significantly improved HbA1c compared with the original treatment regimen.

Chiquette and coauthors112 reviewed placebo-controlled trials of pioglitazone and rosiglitazone and noted the need for head-to-head studies. They concluded that both drugs decreased HbA1c and increased weight to a similar degree.

In a systematic review for the Health Technology Assessment Programme of the National Health Service,113 Czoski-Murray and colleagues also noted that both pioglitazone and rosiglitazone produced similar improvements in HbA1c (approximately 1.0%). They did not identify any randomized controlled trials comparing the 2 drugs and noted that there were no peer-reviewed data on long-term effects.

Updated report

For the 2008 update of the Drug Effectiveness Review Project drug class review on TZDs, an additional 11 systematic reviews were identified (Evidence Table 1 for 2008 TZD Report).89, 114–123

In these reviews both pioglitazone and rosiglitazone reduced HbA1c by approximately 1.0 absolute percentage point, similar to the change produced with other oral agents, including metformin, glibenclamide, and glimepiride.89, 114, 117, 119, 124 This reduction was also similar to the changes noted in placebo-controlled trials in this report. These reviews did not provide additional direct head-to-head data for HbA1c change for pioglitazone and rosiglitazone. In placebo-controlled trials, Phatak and Yin117 noted a weighted mean change in HbA1c from baseline of −1.03% (standard deviation 0.19) for pioglitazone and −0.98% (standard deviation 0.18) for rosiglitazone. Head-to-head studies were not examined and indirect comparisons were not performed.

Detailed Assessment for TZDs Compared With Active Controls

For the original Drug Effectiveness Review Project TZD report, active-control studies for the outcome of HbA1c were not included. These were, however, included for examination of effectiveness outcomes and for examination of patient subgroups.

For the updated report (2008), active-control studies for both pioglitazone and rosiglitazone were included for the outcome of HbA1c in order to update the Agency for Healthcare Research and Quality report on oral hypoglycemic agents whose search ended January 2006.89 Bolen and colleagues concluded that there were no between-group differences between thiazolidinediones and metformin (7 randomized controlled trials) or second generation sulfonylureas (13 randomized controlled trials). Thiazolidinedione plus metformin compared with a second-generation sulfonylurea plus metformin (2 randomized controlled trials) did not show a consistent effect favoring 1 of the combinations, nor did 2 randomized controlled trials comparing thiazolidinediones compared with repaglinide. One trial comparing pioglitazone to acarbose favored pioglitazone for HbA1c reduction.

In the sections below, we include the active-control good- and fair-quality TZD studies included in the 2008 Drug Effectiveness Review Project drug class review on TZDs (searches through Nov 2007), as well as new good- and fair-quality studies identified since that time (searches through July 28, 2010).

Pioglitazone compared with an active control
Characteristics of studies

We included 16 trials comparing pioglitazone with an active control (Tables 28 and 29).125–140 Seven of these are new to this section in this report.128, 129, 136–140 Seven monotherapy trials compared pioglitazone to a sulfonylurea126, 130, 133, 135 or to metformin. 135, 137, 139, 140 Trials examining combination therapy compared pioglitazone to a sulfonylurea with both groups receiving various oral hypoglycemic agents or insulin125, 127–129, 131 or metformin. 134 Pioglitazone was compared to metformin as add-on to other diabetic therapy in 3 trials.132, 136, 138 Drug dosing across studies was fairly consistent, with most study populations 50–60 years of age. Studies ranged between 3 and 18 months; 5 trials had follow-up of greater than 6 months.127, 128, 130, 135, 137

Table 28. Characteristics of pioglitazone active-control trials with sulfonylureas in adults with type 2 diabetes.

Table 28

Characteristics of pioglitazone active-control trials with sulfonylureas in adults with type 2 diabetes.

Table 29. Characteristics of pioglitazone active-control trials with metformin in adults with type 2 diabetes.

Table 29

Characteristics of pioglitazone active-control trials with metformin in adults with type 2 diabetes.

Efficacy results

HbA1c results for active-control trials of pioglitazone are presented in Tables 30 and 31. Effects on HbA1c were similar between treatment groups, with no statistically significant difference noted between groups in 13 of the 16 trials. The 3 trials reporting a statistically significant difference compared pioglitazone to a sulfonylurea and reported small between-group differences in HbA1c (0.19% to 0.63%).127, 128, 133 None of the trials comparing pioglitazone to metformin reported a statistically significant difference. In a small (N=92), monotherapy study in Japan,133 HbA1c decreased more with glibenclamide (change in HbA1c −1.43%) than with pioglitazone (change in HbA1c −0.80%, between-group P<0.05) at 24 weeks follow-up. In an 18-month trial of glibenclamide compared with pioglitazone in newly-diagnosed diabetic subjects taking a variety of concurrent hypoglycemic agents including insulin,127 HbA1c improved in both groups to a similar degree to week 32, then the improvement was maintained with pioglitazone but not with glimepiride. At the final follow-up (week 72), the between-group difference (in favor of pioglitazone) was −0.32% (95% CI −0.52 to −0.12). In the PERISCOPE trial (N=543), greater improvement in HbA1c was reported for subjects treated with pioglitazone (−0.55%) than for those treated with glimepiride (−0.36%, between group P=0.03) at 18 months.128

Table 30. Change in HbA1c for pioglitazone compared with sulfonylureas in adults with type 2 diabetes.

Table 30

Change in HbA1c for pioglitazone compared with sulfonylureas in adults with type 2 diabetes.

Table 31. Change in HbA1c for pioglitazone compared with metformin in adults with type 2 diabetes.

Table 31

Change in HbA1c for pioglitazone compared with metformin in adults with type 2 diabetes.

Rosiglitazone compared with an active control
Characteristics of studies

We included 14 active-control trials comparing rosiglitazone with an active control (Tables 32 and 33).141–154 Six of these are new to this section in this report. 144, 148, 149, 152–154 There were 4 monotherapy trials comparing rosiglitazone to metformin147, 148 or rosiglitazone to a sulfonylurea.145, 147, 149 The combined therapy trials compared rosiglitazone to a sulfonylurea with both groups receiving metformin or insulin141–144, 152, 154 or compared rosiglitazone to metformin with both groups receiving sulfonylureas151 or various hypoglycemic agents.146 Raskin and colleagues150 compared rosiglitazone to repaglinide and to the combination of the 2 drugs. Kadoglou and colleagues153 compared the addition of rosiglitazone with increasing the dose of metformin for people with inadequately controlled diabetes while taking metformin 850mg daily.

Table 32. Characteristics of rosiglitazone active-control trials with sulfonylurea in adults with type 2 diabetes.

Table 32

Characteristics of rosiglitazone active-control trials with sulfonylurea in adults with type 2 diabetes.

Table 33. Characteristics of rosiglitazone active-control trials with metformin or other in adults with type 2 diabetes.

Table 33

Characteristics of rosiglitazone active-control trials with metformin or other in adults with type 2 diabetes.

Across active-control studies, rosiglitazone dosing was either 4 or 8 mg daily. Follow-up intervals ranged from 24 weeks to 4 years,147 with 7 trials having follow-up of 1 year or more.142, 144–148, 154 Mean age of study subjects was mid-50s for most studies, with 4 studies enrolling older subjects, with mean ages between 60 and 65 years.141, 151, 153, 154

Efficacy results

HbA1c results for active-control trials of rosiglitazone are presented in Tables 34 and 35. One of the 14 trials, A Diabetes Outcomes Progression Trial (ADOPT),147 reported a statistically significantly greater improvement in HbA1c for subjects treated with rosiglitazone than those treated with active controls and one143 reported greater improvement for the active control than for rosiglitazone. The other 12 trials reported no statistically significant difference between groups. ADOPT was a large (N=4360), multicenter, double-blind, randomized controlled trial designed to evaluate monotherapy with rosiglitazone, metformin, or glyburide. The trial reported greater improvement in HbA1c at 4 years for subjects treated with rosiglitazone than for those treated with metformin (treatment difference −0.13%, 95% CI −0.22 to −0.05) and those treated with glyburide (treatment difference −0.42%, 95% CI −0.50 to −0.33). Garber and colleagues reported greater improvement in glycemic control for subjects treated with a combination of glibenclamide 5 mg/metformin 1000 mg (once or twice daily) than for those treated with rosiglitazone 4–8 mg daily combined with metformin 1500–2000 mg daily (between-group difference in HbA1c 0.4%, P<0.001).143

Table 34. Change in HbA1c in rosiglitazone active-control trials with sulfonylurea in adults with type 2 diabetes.

Table 34

Change in HbA1c in rosiglitazone active-control trials with sulfonylurea in adults with type 2 diabetes.

Table 35. Change in HbA1c in rosiglitazone active-control trials with metformin or other in adults with type 2 diabetes.

Table 35

Change in HbA1c in rosiglitazone active-control trials with metformin or other in adults with type 2 diabetes.

Among the monotherapy trials, ADOPT (N=4360) was designed to evaluate monotherapy with rosiglitazone, metformin, or glyburide among subjects recently diagnosed (within 3 years) with type 2 diabetes and who had failed lifestyle therapy but had not started on oral hypoglycemic agents.147 The primary outcome was monotherapy failure defined as fasting plasma glucose level of >180 mg/dL. Median duration of treatment with rosiglitazone was 4 years. The cumulative incidence of monotherapy failure at 5 years was 15% with rosiglitazone, 21% with metformin, and 34% with glyburide (P<0.001 for both rosiglitazone comparisons).

The results of 2 smaller rosiglitazone monotherapy trials were similar to the results from ADOPT when the appropriate follow-up intervals were compared. Hanefeld and colleagues found no significant difference between glibenclamide and rosiglitazone at 52-week follow-up145 and Pop-Busui and colleagues found no significant difference between glyburide and rosiglitazone at 26 weeks.149 Likewise, Kiyici and colleagues reported similar changes from baseline in HbA1c for subjects treated with rosiglitazone and those treated with metformin.148

Among the combination therapy trials where rosiglitazone was added to ongoing metformin therapy compared with adding various sulfonylureas to ongoing metformin, 4 trials did not show significant differences between rosiglitazone and active comparators.141, 142, 144, 152 On the other hand, Garber and colleagues143 reported greater improvement in HbA1c for the fixed combination of glibenclamide 5 mg/metformin 1000 mg (once or twice daily) than for rosiglitazone 4–8 mg daily combined with metformin 1500–2000 mg daily (between-group difference in HbA1c 0.4%, P<0.001).

Combination therapy studies comparing rosiglitazone to metformin with both groups receiving other oral agents did not show significant differences between treatment groups.146, 151 A combination of rosiglitazone and repaglinide150 demonstrated superiority for the combination product over rosiglitazone monotherapy. Rosiglitazone was superior to repaglinide (each as monotherapy; no statistics provided).

In the large RECORD trial158 (discussed further in Key Question 2), subjects who were already taking a sulfonylurea were randomized to add-on rosiglitazone 4 mg daily (titrated up to 8 mg daily) or metformin (titrated up to 2550 mg daily). Subjects taking metformin at study entry were randomized to add-on sulfonylurea. If adequate glycemic control (HbA1c ≤ 8.5%) was not obtained on maximal dosage dual therapy, a third drug was added (either a sulfonylurea or metformin to rosiglitazone subjects and insulin in the control group). HbA1c decreased by approximately 0.5% at 18 months follow-up146 in all 4 treatment groups, with no statistically significant difference between rosiglitazone and other drugs in the background metformin and background sulfonylurea groups.

TZDs compared with newer diabetes drugs
Characteristics of studies

We found 4 trials comparing rosiglitazone with a newer diabetes drug of primary interest to this report (Table 36).36, 40, 67, 84 One compared the addition of rosiglitazone with the addition of sitagliptin to ongoing metformin;36 one compared the addition of rosiglitazone with the addition of liraglutide to ongoing glimepiride;84 one compared the addition of colesevelan, rosiglitazone, or sitagliptin to ongoing metformin;40 and one compared the addition of exenatide, rosiglitazone, or exenatide and rosiglitazone to ongoing metformin.67

Table 36. Characteristics of TZD interclass head-to-head trials in adults with type 2 diabetes.

Table 36

Characteristics of TZD interclass head-to-head trials in adults with type 2 diabetes.

Efficacy results

HbA1c results for these are presented in Table 37. One 18 week trial (N=273) found no significant difference in reduction of HbA1c between those treated with the addition of rosiglitazone and those treated with the addition of sitagliptin.36 The trial that randomized subjects to add-on either rosiglitazone, sitagliptin, or colesevelam to ongoing metformin found no statistically significant difference between the rosiglitazone- and sitagliptin-treated subjects.40

Table 37. Change in HbA1c in TZD interclass head-to-head trials in adults with type 2 diabetes.

Table 37

Change in HbA1c in TZD interclass head-to-head trials in adults with type 2 diabetes.

One 26 week trial (N=1040) comparing the addition of rosiglitazone with the addition of liraglutide (to ongoing glimepiride treatment) reported greater reduction in HbA1c with liraglutide (−1.1 compared with −0.4%, P<0.0001).84

One 20-week trial comparing exenatide to rosiglitazone with all participants on background metformin therapy, found no significant difference in improvement in HbA1c between the exenatide and rosiglitazone arms (−0.9% compared with −1.0%, P=0.720).67

Detailed Assessment of TZDs Compared with Placebo

Placebo-controlled trials of pioglitazone

For this report, we did not update the comparisons of pioglitazone or rosiglitazone compared with placebo. This information was included in the 2008 Drug Effectiveness Review Project drug class review on TZDs. We briefly summarize the findings of that report here.18

In the original report, 16 trials comparing pioglitazone to placebo in at least 1 study arm were identified. All but 1 of these trials had sufficient data to permit a meta-analysis; a study by Saad and colleagues160 did not provide a measure of dispersion. In the updated review 4 new placebo-controlled trials were identified, 2 of combination therapy161, 162 and 2 of monotherapy,163, 164 along with a no-treatment comparison165 study.

The mean difference between groups for all good- and fair-quality studies comparing pioglitazone with placebo ranged from −3.0% to −0.5% and the pooled weighted mean difference was −0.95 (95% CI −1.24 to −0.67) (95% CI −1.27 to −0.84) (Table 38). In other words, overall, pioglitazone improved HbA1c about 1.0% compared with placebo. Results were somewhat more pronounced when pioglitazone monotherapy was compared with placebo than when combined therapy (the addition of pioglitazone to another hypoglycemic drug) was compared with placebo added to the other hypoglycemic drug, although the differences between monotherapy and combined therapy were not significant (Table 38).

Table 38. Meta-analysis results for HbA1c from 2008 Drug Effectiveness Review Project TZDs report.

Table 38

Meta-analysis results for HbA1c from 2008 Drug Effectiveness Review Project TZDs report.

Placebo-controlled trials of rosiglitazone

In the original report, 25 trials compared the efficacy or effectiveness of rosiglitazone to placebo. Four rosiglitazone studies did not provide adequate information for inclusion in the meta-analysis: Honisett et al.166 did not provide a measure of dispersion; the units for HbA1c in a paper by Raskin and colleagues167 were difficult to interpret; Wang et al. 168 provided graphical data only; and Nolan and colleagues169 provided a measure of fasting glucose but not HbA1c. In the updated review of placebo-controlled trials of rosiglitazone, 8 new studies were identified,162, 170–176 including 3 poor-quality studies.162, 173, 174 All but 1 study162 were combination therapy studies.

Mean differences are presented in Table 38 above. Results are similar to those noted for pioglitazone, with a mean change in HbA1c for all good and fair-quality studies of −0.92 (95% CI −1.15 to −0.68). Again, heterogeneity was significant among studies and there were no significant differences between monotherapy and combined therapy. Adjusted indirect comparisons of pioglitazone and rosiglitazone revealed no significant differences between the 2 drugs for HbA1c.

Using meta-regression, the 2008 Drug Effectiveness Review Project TZDs report examined placebo-controlled trials of either pioglitazone or rosiglitazone and found no significant relationships between change in HbA1c and follow-up interval or funder (industry or other). When studies using combination therapy (either thiazolidinedione combined with insulin, sulfonylurea, or metformin) were examined, there were no significant differences among the various treatment combinations for change in HbA1c.

Detailed Assessment of Health outcomes (microvascular and macrovascular disease, lower extremity ulcers, all-cause mortality, and quality of life) for TZDs

None of the head-to-head studies identified in the original or updated review examined macro- or microvascular outcomes. Three placebo-controlled or no-treatment comparison studies identified in the original review examined cardiovascular outcomes; all examined patients with known macrovascular disease and type 2 diabetes,168, 177,178 including the PROACTIVE trial.177 These 3 trials did not provide sufficient data to determine comparative effectiveness of pioglitazone and rosiglitazone on microvascular or macrovascular complications of diabetes. In the updated review several additional trials provided evidence on macrovascular outcomes and on mortality, with 5 trials providing additional evidence on pioglitazone. Here we summarize the information related to health outcomes and TZDs. Of note, we address adverse events (including congestive heart failure and cardiovascular adverse events) in the Key Question 2 section of this report, rather than in this section.

In the PROACTIVE trial,177 a good-quality, European, multicenter, randomized, placebo-controlled trial of 5238 patients with type 2 diabetes and evidence of macrovascular disease, treatment patients received pioglitazone titrated from 15 mg up to 45 mg daily. Ninety-six percent of patients were taking other glucose-lowering agents, including insulin. The average follow-up period was 34.5 months. The primary endpoint was the composite of all-cause mortality, non-fatal myocardial infarction (including silent myocardial infarction), stroke, acute coronary syndrome, endovascular or surgical intervention in the coronary or leg arteries, and amputation above the ankle. The hazard ratio for this endpoint was 0.90 (95% CI 0.80 to 1.02). Congestive heart failure was not included in this composite endpoint, although congestive heart failure was examined as an adverse event. When examined individually (as secondary endpoints), none of the components of the primary endpoint changed significantly (P>0.05). The hazard ratio of the main secondary endpoint (a composite of all-cause mortality, myocardial infarction [excluding silent myocardial infarction], and stroke) was 0.84 (95% CI 0.72 to 0.98).

Wang and colleagues168 performed a randomized controlled trial comparing rosiglitazone 4 mg daily to no treatment (N=70) over 6 months. Included patients were aged 50 to 73 years, had a diagnosis of coronary artery disease (>50% stenosis as proven on angiography), had established type 2 diabetes, and had undergone a percutaneous coronary intervention (Evidence Table 9 from 2008 DERP TZD Report Update 1). Forty-one percent took other anti-diabetic medications. At 6-month follow-up the incidence of coronary events was decreased in the rosiglitazone group (between-group P<0.05 for the composite endpoint), with 4 events in the rosiglitazone group (recurrent angina179 and coronary artery bypass grafting [1]) and 12 in the control group (recurrent angina [5], repeated angioplasty,179 and coronary artery bypass grafting179).

A single-center poor-quality study examined the preventive effects of rosiglitazone on restenosis after coronary stent implantation among 95 persons with type 2 diabetes.178 In this open-label, randomized controlled trial, the treatment group was placed on rosiglitazone 8 mg before undergoing catheterization and 4 mg daily thereafter, combined with conventional antidiabetic therapy using a variety of agents (details of concurrent therapy were not provided). The comparison group received conventional therapy only. The rate of restenosis was 18% in the rosiglitazone group and 38% in the control group (between-group P=0.03). There was also a significant difference in stenosis diameter between groups at 6 months (P=0.004) in favor of the rosiglitazone group.

The available data provided no information on the comparative effectiveness of pioglitazone and rosiglitazone on macro- and microvascular outcomes when used as monotherapy or when added to or substituted for other oral hypoglycemic agents. Dormandy and colleagues177 addressed the question of combined therapy as pioglitazone was added to other anti-diabetic therapy in 96% of patients. In the study by Wang and coauthors168 monotherapy and combined therapy patients were aggregated, so conclusions cannot be drawn about each of these 2 approaches.

In the updated review several additional trials provided evidence on macrovascular outcomes and on mortality, with 5 trials providing additional evidence on pioglitazone.

The CHICAGO trial127 was a multicenter study of pioglitazone 15 to 45 mg per day compared with glimepiride 1 to 4 mg per day in 462 adults who were newly diagnosed with type 2 diabetes. The primary endpoint was the change in carotid artery intima-media thickness after 72 weeks. Secondary endpoints included the composite of cardiovascular mortality, non-fatal myocardial infarction, or nonfatal stroke, and the composite of these outcomes plus coronary revascularization, carotid endarterectomy/carotid stenting, hospitalization for unstable angina, or hospitalization for heart failure. There were few events reported, and no cardiovascular deaths. There were 2 instances of the first composite endpoint in the glimepiride group and none in the pioglitazone group. On the second composite endpoint, there were 10 events in the glimepiride group (8 of which were coronary revascularization) and 4 in the pioglitazone group (3 coronary revascularization).

PERISCOPE was another trial of pioglitazone compared to glimepiride designed to measure progression of atherosclerosis in patients with type 2 diabetes.128 After 18 months of follow-up, there was no difference between groups in the occurrence of clinical endpoints, including the composite of cardiovascular death, nonfatal myocardial infarction, or nonfatal stroke (2.2% for glimepiride compared with 1.9% for pioglitazone; P=0.78), the composite of cardiovascular death, nonfatal myocardial infarction, nonfatal stroke, hospitalization for unstable angina, or congestive heart failure 4.8% for glimepiride compared with 4.1% for pioglitazone; P=0.70) or any components of the composite outcomes. There were 3 cardiovascular deaths in the pioglitazone group and 1 in the glimepiride group (P=0.37).

In a small, fair-quality, randomized controlled trial (N=47), patients with impaired glucose tolerance or type 2 diabetes (combined in the analysis) in addition to nonalcoholic steatohepatitis, received either pioglitazone 45 mg daily or placebo, in addition to a weight loss intervention.180 Glycemic control improved with pioglitazone compared with placebo (P<0.001), with a decrease in weight and body mass index with treatment compared with placebo (P=0.003 and 0.005, respectively). Plasma aspartate and alanine aminotransferase levels and hepatic fat content all decreased with treatment compared with placebo (P<0.05) and liver aminotransferase levels normalized with pioglitazone. Histologic changes in the liver also improved significantly with pioglitazone.

In another small trial,181 patients with acute coronary syndrome received pioglitazone or no additional treatment starting 2 weeks after percutaneous, bare metal stent placement. At 6-months follow-up these researchers demonstrated that late luminal loss was less in the pioglitazone group than in the control group (P=0.0008); the same was found for restenosis rate (between-group P=0.0052; both assessed with quantitative angiography). Major cardiac events (myocardial infarction or revascularization of the target lesion) were significantly decreased in the pioglitazone group at 6 months compared with the control group (7.7% compared with 60.7%, P<0.0001). There were no deaths in either group.

Takagi and colleagues compared pioglitazone with placebo in 44 patients with type 2 diabetes who had undergone coronary stent implantation.182 After 6 months of follow-up, angiographic in-stent restenosis (19% compared with 46%; P=0.0994) and target lesion revascularization (12% compared with 38%; P=0.0835) were less frequent in the pioglitazone group, but the differences were not statistically significant. There was no difference in HbA1c levels at follow-up in this study (See Key Question 1).

The updated search identified several important recent trials of rosiglitazone reporting vascular or mortality outcomes: the RECORD trial146, 157 and ADOPT.147 The RECORD trial was an open-label, multicenter, noninferiority, randomized controlled trial (N=4447). Subjects who were already taking metformin or a sulfonylurea were randomized to add-on rosiglitazone 4 mg daily (titrated up to 8 mg daily) or to metformin (titrated up to 2550 mg daily) plus a sulfonylurea (glyburide, gliclazide or glimepiride, depending on physician preference). If adequate glycemic control (HbA1c ≤8.5%) was not obtained on maximal dosage dual therapy, a third drug was added (either a sulfonylurea or metformin for rosiglitazone subjects and insulin in the control group).

The primary outcome for the RECORD study was time to first occurrence of cardiovascular hospitalization or cardiovascular death. 321 people in the rosiglitazone group and 323 in the active control group experienced the primary outcome during a mean 5.5 year follow up. The hazard ratio for rosiglitazone (plus metformin or a sulfonylurea) compared with metformin plus a sulfonylurea was 0.99 (95% CI 0.85 to 1.16), meeting the criterion of non-inferiority. Heart failure causing admission to the hospital or death occurred in 61 people in the rosiglitazone group and 29 in the active control group(hazard ratio 2.10, 95% CI 1.35 to 3.27).

The large ADOPT147, discussed above for the outcome of monotherapy failure, compared rosiglitazone, glyburide, and metformin in subjects newly diagnosed with type 2 diabetes. Subjects with significant renal or hepatic disease, unstable or severe angina, or congestive heart failure of any New York Heart Association class were excluded. Approximately half of subjects had hypertension, 81% had metabolic syndrome, and 45% were smokers.156 The number of deaths from all causes was similar across the 3 groups, but more cardiovascular events were reported in the rosiglitazone group (4.3%) than in the metformin (4.0%) or glyburide groups (2.8%; no significant differences among groups). Congestive heart failure events were higher with rosiglitazone than with glyburide (further details are presented in Key Question 8). The lower rates of cardiovascular events in the glyburide group were primarily due to lower rates of nonfatal myocardial infarction and congestive heart failure in this group.

Several additional, smaller rosiglitazone trials were also identified in the updated search.174, 176 In a very small (N=16), poor-quality, randomized controlled trial, subjects with coronary stent implantation were randomized to rosiglitazone 4–8 mg daily or placebo for 6 months. Rosiglitazone did not reduce in-stent restenosis and there were no differences in cardiac events between the groups.174

In a study of older adults with type 2 diabetes, Rosenstock and colleagues176 noted no significant difference between rosiglitazone and placebo (both groups received glipizide) in SF-36 component scores, although the rosiglitazone group had greater improvement on the Diabetes Treatment Satisfaction Questionnaire (DTSQ) than the glipizide only group (1.15 point increase compared with 1.61 point decrease, P<0.001).

III. Fixed-dose Combination Products (FDCPs) or Dual Therapy

Summary of findings for FDCPs or Dual Therapy

Evidence in children
  • We did not find any evidence meeting inclusion/exclusion criteria for children (insufficient strength of evidence).
Evidence in adults
  • We found no studies that focused on health outcomes as the primary outcomes for any available FDCP. Two studies reported health outcomes among other secondary outcomes or in the adverse events section.183, 184 Overall evidence was insufficient to determine how FDCPs compare with other treatments for their impact on health outcomes.
  • We found no head to head trials that compared HbA1c control between any 2 FDCPs (insufficient strength of evidence).
  • We found no trials that evaluated the following FDCPs: Duetact®, Janumet®(insufficient strength of evidence).
  • Therapy with Avandamet,® Avandaryl,® or Actoplus Met produced statistically significantly greater reductions in HbA1c compared to monotherapy with any of their respective components.
  • The magnitudes of the differences in HbA1c reductions between the FDCPs and their respective monotherapy components ranged from 0.13% to 0.7% for Avandamet®, 0.6% to 0.8% for Avandaryl®, and 0.2% to 0.9% for Actoplus Met®.
  • Greater reduction in HbA1c with Avandamet® or dual therapy with metformin and rosiglitazone than with component monotherapy in trials of 24 to 32 weeks (reduction in the intervention arms ranged from 0.13% to 0.7%, moderate strength of evidence)
  • Greater reduction in HbA1c with Avandaryl® or dual therapy with rosiglitazone and glimepiride than with component monotherapy in trials from 20 to 28 weeks (reduction in the intervention arms ranged from 0.6% to 0.8%, moderate strength of evidence)
  • Greater reduction in HbA1c with Actoplus Met® or dual therapy with pioglitazone and metformin than with component monotherapy in trials of 24 weeks and 15 months (reduction in the intervention arms ranged from 0.2% to 0.9%, moderate strength of evidence)
  • Greater reduction in HbA1c with dual therapy with metformin and sitagliptin than with component monotherapy in a 24 week trial with additional 30 and 52 week extensions (range 0.4% to 1.2%, moderate strength of evidence).

Detailed Assessment for FDCPs and Dual Therapy

We identified studies that have been conducted specifically using fixed-dose combination tablets comprised of rosiglitazone/metformin (Avandamet®),183, 185 rosiglitazone/glimepiride (Avandaryl®),186 and pioglitazone/metformin (Actoplus Met®).139 Two of these were new since the 2007 Drug Effectiveness Review Project report on FDCPs.139, 183 We found no head-to-head studies comparing FDCPs.

We also included studies using dual therapy of rosiglitazone plus metformin,184 rosiglitazone plus glimepiride,187, pioglitazone plus metformin,188 and sitagliptin plus metformin.31–33 All of these were new for this report. For this report, dual therapy was defined as using the individual components of a FDCP in separate pills/tablets. Studies were required to randomize subjects to the components of a FDCP or to monotherapy with one of the components of the FDCP to be eligible for this report. Studies continuing a ‘background’ therapy (e.g., with metformin) and randomizing subjects to add-on one medication (e.g., rosiglitazone or pioglitazone) or to add-on placebo were classified as comparing that medication (e.g., rosiglitazone or pioglitazone) with placebo.

No studies were identified that used the fixed-dose combination tablets comprised of pioglitazone/glimepiride (Duetact®)189 or sitagliptin/metformin (Janumet®).190 The efficacy and safety of Duetact® and Janumet® have been established based on trials using the co-administration of their separate components.

The majority of the trials were 4- to 6-month evaluations of glycemic control and general adverse events with FDCPs or dual therapy compared to component monotherapy when used as initial treatment for patients with type 2 diabetes. Studies that compared type 2 diabetes combination tablet products to co-administration of their components were few, nonrandomized, and limited to analyses based on refill data from pharmacy claims databases.191–193

We found no evidence to address the effectiveness of combination tablet products in improving long-term health.

Throughout this section, meta-analyses were not performed due to an insufficient number of studies or heterogeneity of study populations, outcomes, and designs.

Avandamet® or dual therapy with metformin plus rosiglitazone

Three randomized controlled trials including either Avandamet® or dual therapy with metformin and rosiglitazone met inclusion criteria. No comparative cohort studies, case-control studies or systematic reviews were identified reporting long-term benefits.

Head-to-head trials

We found no head-to-head trials of Avandamet® or dual therapy with metformin and rosiglitazone comparing them with other FDCPs that met inclusion criteria.

Trials comparing Avandamet® or dual therapy with component monotherapy

Three fair-quality trials compared Avandamet® (2 trials) or dual therapy (one trial) with metformin and rosiglitazone to monotherapy with metformin or rosiglitazone. (Table 39) Two trials compared Avandamet® with metformin monotherapy; one of them also compared Avandamet® with rosiglitazone monotherapy. The dual therapy trial compared concurrent use of metformin and rosiglitazone with metformin monotherapy.

Table 39. Characteristics of Avandamet® (metformin/rosiglitazone) and rosiglitazone plus metformin dual therapy trials in adults with type 2 diabetes.

Table 39

Characteristics of Avandamet® (metformin/rosiglitazone) and rosiglitazone plus metformin dual therapy trials in adults with type 2 diabetes.

Overall, both Avandamet® and dual therapy with metformin and rosiglitazone were associated with greater reductions in HbA1c values, compared with monotherapy (Table 40).

Table 40. Change in HbA1c in Avandamet® (metformin/rosiglitazone) or rosiglitazone plus metformin trials in adults with type 2 diabetes.

Table 40

Change in HbA1c in Avandamet® (metformin/rosiglitazone) or rosiglitazone plus metformin trials in adults with type 2 diabetes.

Both trials comparing Avandamet® with metformin monotherapy found that Avandamet® reduced HbA1c levels by a greater amount than metformin alone. In one of the trials,185 HbA1c was reduced by a mean of 2.3% after 32 weeks in subjects on Avandamet® (mean daily dose = 7.2 mg rosiglitazone + 1,799 mg metformin) compared with 1.8% for subjects on metformin (mean daily dose = 1,847 mg) (P=0.0008). In the other,183 mean HbA1c reduction in the Avandamet® group (mean daily dose = 6.8 mg rosiglitazone + 1,812 mg metformin) was 0.51% over 32 weeks, compared with a 0.38% reduction in the metformin monotherapy group (mean daily dose 2,628 mg) (P=0.0357).

In the 32-week trial comparing Avandamet® with rosiglitazone monotherapy,185 rosiglitazone (mean daily dose = 7.7 mg) reduced HbA1c by a mean of 1.6% - an amount smaller than the 2.3% reduction with Avandamet® (P<0.001).

A 24-week trial of dual therapy (metformin plus rosiglitazone) compared with metformin monotherapy reported that dual therapy (8 mg rosiglitazone + 1,000 mg metformin daily) reduced HbA1c by a mean of 0.93% compared with a mean reduction in the metformin monotherapy group (2,000 mg daily) of 0.71% (P value for the between-group difference was NR; 95% CI indicated statistical significance between the arms (mean change from baseline −0.36, −0.04).

Avandaryl® or dual therapy with rosiglitazone plus glimepiride

Two randomized controlled trials including either Avandaryl® or dual therapy with rosiglitazone and glimepiride met inclusion criteria. No comparative cohort studies, case-control studies or systematic reviews were identified reporting long-term benefits.

Head-to-head trials

We found no head-to-head trials of Avandaryl® or dual therapy with rosiglitazone and glimepiride comparing them with other FDCPs that met inclusion criteria.

Trials comparing Avandaryl® or dual therapy with component monotherapy

Two trials compared Avandaryl® or dual therapy with rosiglitazone plus glimepiride to monotherapy with rosiglitazone or glimepiride. (Table 41) One good-quality trial compared 2 dosages of Avandaryl® with glimepiride monotherapy and with rosiglitazone monotherapy.186 One fair-quality dual therapy trial compared concurrent use of rosiglitazone and glimepiride with rosiglitazone monotherapy.187

Table 41. Characteristics of Avandaryl® (rosiglitazone/glimepiride) and rosiglitazone plus glimepiride dual therapy trials in adults with type 2 diabetes.

Table 41

Characteristics of Avandaryl® (rosiglitazone/glimepiride) and rosiglitazone plus glimepiride dual therapy trials in adults with type 2 diabetes.

In both trials, Avandaryl® or dual therapy with rosiglitazone and glimepiride were associated with greater reductions in HbA1c values, compared with monotherapy (Table 42). The trial comparing 2 dosages of Avandaryl® with glimepiride or rosiglitazone monotherapy186 found that Avandaryl® (4 mg/1 mg daily titrated to 4 mg/4 mg )reduced HbA1c levels by a greater amount (mean reduction 2.41%) than glimepiride monotherapy (1.72%; P<0.0001) or rosiglitazone monotherapy (1.75%; P<0.0001) after 28 weeks. In the 4 mg/1 mg daily titrated to 8 mg/4 mg daily formulation of Avandaryl®, mean HbA1c reduction was 2.52%. This was also a significantly greater reduction compared with glimepiride and rosiglitazone monotherapies (P<0.0001).

Table 42. Change in HbA1c in Avandaryl® (rosiglitazone/glimepiride) or rosiglitazone plus glimepiride trials in adults with type 2 diabetes.

Table 42

Change in HbA1c in Avandaryl® (rosiglitazone/glimepiride) or rosiglitazone plus glimepiride trials in adults with type 2 diabetes.

In the other trial, dual therapy with rosiglitazone and glimepiride also resulted in greater improvement in HbA1c than monotherapy.187 Dual therapy with 8 mg of rosiglitazone and 8 mg of glimepiride daily (titrated up from 8 mg/1 mg daily) was associated with a mean HbA1c reduction of 1.2% after 20 weeks. This was a significantly larger decrease than was found with glimepiride (4 mg titrated to 8 mg daily) monotherapy (mean reduction 0.3%; P<0.001).

Actoplus Met® or dual therapy with pioglitazone plus metformin

We found one study including Actoplus Met® and one controlled trial including dual therapy with pioglitazone and metformin that met inclusion criteria. No comparative cohort studies, case-control studies or systematic reviews were identified reporting long-term benefits.

Head-to-head trials

We found no head-to-head trials of Actoplus Met® or dual therapy with pioglitazone and metformin comparing them with other FDCPs that met inclusion criteria.

Trials comparing Actoplus Met® or dual therapy with component monotherapy

One fair-quality trial compared Actoplus Met® with pioglitazone and metformin monotherapies (Table 43). One good-quality trial compared dual therapy with pioglitazone and metformin to monotherapy with each component.

Table 43. Characteristics of Actoplus Met® (pioglitazone/metformin) or pioglitazone plus metformin dual therapy active-control trials in adults with type 2 diabetes.

Table 43

Characteristics of Actoplus Met® (pioglitazone/metformin) or pioglitazone plus metformin dual therapy active-control trials in adults with type 2 diabetes.

In the active-control FDCP trial, Actoplus Met® (30 mg/1,700 mg daily) was associated with a greater reduction in HbA1c value, compared with either monotherapy (Table 44). At the end of this 24-week RCT, the mean HbA1c reduction in the Actoplus Met® group was 1.83%. Mean reductions in the pioglitazone and metformin monotherapy groups were 0.96% and 0.99%, respectively. The P value of the between-group difference for both Actoplus Met® comparisons was <0.0001.

Table 44. Change in HbA1c in Actoplus Met® (pioglitazone/metformin) or pioglitazone plus metformin trials in adults with type 2 diabetes.

Table 44

Change in HbA1c in Actoplus Met® (pioglitazone/metformin) or pioglitazone plus metformin trials in adults with type 2 diabetes.

In the active-control dual therapy trial, treatment with both pioglitazone (45 mg daily) and metformin (2,550 mg daily) was associated with greater reductions in HbA1c values, compared with monotherapy (Table 44). After 12 months of treatment, the dual therapy group achieved a mean HbA1c reduction of 0.9%, a change significantly greater than the decreases achieved with 45mg daily pioglitazone monotherapy (mean reduction = 0.6%; P<0.01) or 3,000 mg daily metformin monotherapy (mean reduction = 0.7%; P<0.05).

Janumet® or dual therapy with sitagliptin plus metformin

No studies including Janumet® were found that met inclusion criteria. One randomized controlled trial including dual therapy with sitagliptin and metformin met inclusion criteria. This trial resulted in 3 publications; one reporting results after 24 weeks,31 one reporting results after 54 weeks,32 and the other after 104 weeks33 No comparative cohort studies, case-control studies or systematic reviews were identified reporting long-term benefits.

Head-to-head trials

We found no head-to-head trials of Janumet® or dual therapy with sitagliptin plus metformin comparing them with other FDCPs that met inclusion criteria.

Trials comparing Janumet® or dual therapy with component monotherapy

One 24 week trial31 with an optional additional 30 weeks32 and further additional 50 weeks33 (Table 45) compared initial dual therapy of sitagliptin plus metformin to sitagliptin monotherapy and metformin monotherapy in subjects who were inadequately controlled only on diet and exercise. Patients in this study were taken off prior oral hypoglycemic agents and put through a diet and exercise run-in phase in addition to a 2-week single-blind placebo run-in period before enrollment. Approximately 50% of patients were taking oral hypoglycemic agents at baseline, implying that the remainder were medication naive. Mean HbA1c was close to 9% and duration of diabetes was less than 5 years. In all treatment arms metformin was titrated to increase tolerability. Patients were followed initially for 24 weeks, and then had the option to continue for 30 additional weeks and then an additional 50 weeks. Patients originally randomized to placebo were automatically put in the metformin 1000 mg twice daily group for the additional 30 weeks. Since the study was designed to examine the potential benefit of a fixed-dose combination tablet of these 2 agents, sitagliptin was up titrated when metformin was up titrated as it would be with the use of a fixed-dose combination tablet (50 mg daily increased after 1 week to the stable study dose of 50 mg twice daily).

Table 45. Characteristics of metformin/sitagliptin dual therapy active-control trials in adults with type 2 diabetes.

Table 45

Characteristics of metformin/sitagliptin dual therapy active-control trials in adults with type 2 diabetes.

The use of sitagliptin 100 mg/d plus metformin 2000 mg/d or sitagliptin 100 mg/d plus metformin 1000 mg/d significantly improved HbA1c compared with sitagliptin monotherapy or metformin monotherapy over 24 weeks (Table 46). For the subjects continuing for the additional 30 weeks, subjects on sitagliptin and metformin combination therapy maintained HbA1c levels without much change; those on metformin and sitagliptin monotherapy continued to have minimal HbA1c improvement (between group P=NR). Magnitude of benefit remained greater in the combination groups, but statistical significance was not reported. Similar results were seen in patients who continued for an additional 50 weeks (total of 104 week treatment).

Table 46. Change in HbA1c in metformin plus sitagliptin dual therapy trials in adults with type 2 diabetes.

Table 46

Change in HbA1c in metformin plus sitagliptin dual therapy trials in adults with type 2 diabetes.

Key Question 2. What is the comparative tolerability and frequency of adverse events for newer diabetes medications, TZDs, and drug combinations (administered as fixed dose combination products or dual therapy) for children and adults with diabetes mellitus?

I. Newer Drugs for the Treatment of Diabetes Mellitus: Amylin Agonists, DPP-4 Inhibitors, and GLP-1 Agonists

Summary of Findings for Amylin Agonists: Harms

Pramlintide for type 1 diabetes
Evidence in children
  • No data on children were reported, although people as young as 16 years were eligible for study enrollment in 2 included trials.19, 20
Evidence in adults
  • Greater withdrawals due to adverse effects for pramlintide-treated subjects than for insulin-treated subjects (ranges across trials were 5% to 20% compared with 2% to 8%, respectively, moderate strength of evidence).
  • Gastrointestinal adverse events including nausea, vomiting, and anorexia were more commonly reported with the use of pramlintide plus insulin than with placebo plus insulin (moderate strength of evidence).
  • Severe hypoglycemia occurred more frequently with pramlintide plus insulin during the first 4 weeks of treatment compared with placebo plus insulin (moderate strength of evidence). Rates of severe hypoglycemia declined once pramlintide doses stabilized but continued to remain slightly higher than with placebo plus insulin at up to 52 weeks of follow-up (moderate strength of evidence).
  • Studies beyond 52 weeks are lacking.
Pramlintide for type 2 diabetes
Evidence in children
  • Children and adolescents ≤18 years were not included in any of the published studies on efficacy or effectiveness.
Evidence in adults
  • Both pramlintide- and placebo-treated subjects exhibited similar rates of withdrawal and withdrawal due to adverse events.
  • The most commonly reported adverse event was nausea, which occurred more frequently with pramlintide plus insulin than with placebo plus insulin especially during the first 4 weeks of treatment, but declined thereafter (moderate strength of evidence).
  • Severe hypoglycemia occurred more frequently with pramlintide compared with placebo (moderate strength of evidence).
  • Hypoglycemia occurred less frequently in subjects taking pramlintide than those taking rapid acting insulin analogs (RAIA) in one 24 week study (low strength of evidence).

Detailed Assessment of Pramlintide in Type 1 Diabetes: Harms

We found no active-control trials. We found 3 placebo-controlled trials.19–21 Details of these trials are presented in Table 5 in the corresponding section in Key Question 1.

Patients receiving pramlintide in addition to insulin had greater rates of withdrawal due to all causes and withdrawal due to adverse events than patients receiving placebo plus insulin. This was found with both fixed- and flexible-dose insulin (see Evidence Table 7). No included trial reported deaths or listed rare adverse events. There were no significant cardiac, hepatic, renal, or drug-related idiosyncratic adverse events observed in any treatment arm. Adverse events reported in the included studies are summarized in Table 47.

Table 47. Adverse events from placebo-controlled trials of pramlintide in type 1 diabetes.

Table 47

Adverse events from placebo-controlled trials of pramlintide in type 1 diabetes.

Hypoglycemia

During the first 4 weeks of treatment severe hypoglycemia occurred more frequently with pramlintide plus insulin than with insulin plus placebo, with both fixed and flexible insulin regimens. The rate of severe hypoglycemia declined once pramlintide doses were stabilized and not being titrated; however, at weeks 26–5219, 21 and weeks 0–2920 the rate of severe hypoglycemia associated with pramlintide was still slightly higher than placebo (event rates 0.42 to 1.10 compared with 0.30 to 0.52) (Table 47). Only 1 trial20 reported that a 30% to 50% reduction in prandial insulin was allowed before the use of pramlintide. Even in this study, pramlintide-treated patients exhibited slightly higher rates of severe hypoglycemia compared with insulin plus placebo-treated patients (Table 47). No trials reported the overall incidence of mild to moderate hypoglycemic episodes. All 3 trials predefined the term “severe hypoglycemia” to mean: those requiring either assistance of another person, the administration of glucagon, or the administration of intravenous glucose.

Nausea and vomiting

A significant proportion of pramlintide-treated patients experienced nausea during the trials: Across trials overall rates of nausea for pramlintide groups ranged from 46% to 95%; for placebo groups, 12% to 36%. Specifically, patients who did not tolerate pramlintide 60 mcg also frequently experienced nausea with the 30 mcg dose, and the highest reported rates of nausea (95%) were in subjects who received 30 mcg 3 times a day.20 Higher rates of nausea were reported with pramlintide 90 mcg 3 times a day21 than with lower dosages in the same trial.

Severe nausea was much less common than nausea overall, ranging between 5.8% and 8.5% for pramlintide plus insulin and 0.7% to 1.7% for placebo plus insulin across studies.19–21

More than 10% of patients randomized to pramlintide plus insulin experienced vomiting, compared with rates of up to 8.0% with placebo plus insulin. Severe vomiting occurred in up to 2% of patients taking pramlintide compared with 0.4% to 0.7% taking placebo.19–21

Of note, 2 of 3 placebo-controlled trials19, 21 reported that most cases of nausea and vomiting tended to occur within 2–4 weeks of treatment but no data were provided to verify these statements.

Anorexia or reduced appetite

Rate of anorexia was significantly more frequent with pramlintide plus insulin (11% to 18% across trials) than with placebo plus insulin (approximately 2%). Severe anorexia occurred in <2% of pramlintide patients and no placebo patients.19, 21

Other adverse events

One trial reported sinusitis at a rate of 14.0% with pramlintide and 8.8% with placebo (P>0.05).20 Two non-comparative observational studies195, 196 were also evaluated for rare adverse events and neither reported any additional information.

Detailed Assessment of Pramlintide in Type 2 Diabetes: Harms

Pramlintide-plus-insulin and placebo-plus-insulin groups had similar rates of withdrawal due to all causes and withdrawal due to adverse events (see Evidence Table 7). There was no evidence of cardiac, hepatic, renal, or drug-related idiosyncratic adverse events in patients in any treatment arm of the 4 randomized controlled trials identified for this review and no deaths were reported. Adverse effects are summarized in Table 48.

Table 48. Adverse effects reported in placebo and active-control trials of pramlintide in type 2 diabetes.

Table 48

Adverse effects reported in placebo and active-control trials of pramlintide in type 2 diabetes.

Hypoglycemia

Pramlintide-plus-insulin and placebo-plus-insulin groups experienced similar rates of mild-to-moderate hypoglycemia,24, 26 but pramlintide-treated patients experienced more episodes of severe hypoglycemia. Severe hypoglycemia occurred most with pramlintide 120 mcg during the first 4 weeks of therapy (0.9 events/patient-year compared with 0.3 events/patient-year with placebo).25 The incidence of severe symptoms declined with continued use of pramlintide, and rates were similar to placebo for weeks 4–26 and 26–52.25 Compared with RAIA, pramlintide had a lower incidence of hypoglycemia.22 All trials predefined the term “severe hypoglycemia” to mean: those requiring either assistance of another person, the administration of glucagon, or the administration of intravenous glucose.

Nausea

The incidence of mild-to-moderate and severe nausea was significantly higher with pramlintide 75, 90, 120, and 150 mcg than with placebo plus insulin. Two trials reported data showing that most events occurred within the first 4 weeks of treatment.22, 25 When metformin use was stratified in 1 trial, its addition to pramlintide plus insulin appeared to have no significant effect on nausea compared with the larger study population.25 These trials did not report vomiting or anorexia.

Headache

In one trial, higher rates of headache were reported with pramlintide (15% and 17%) than with placebo (8%).25 In another trial 26 rate of headache was similar among treatment groups, ranging from 13.2% in the placebo-plus-insulin group to 19.1% with pramlintide 75 mcg 3 times a day plus insulin. None of the studies provided enough information to determine whether there were any correlations between the incidence of headaches and hypoglycemic events.

Other adverse events

No trials reported any treatment-emergent adverse events occurring with a frequency of more than 2% to 5%. Overall adverse events occurring with a frequency of ≥10% with a minimum 5 percentage point difference between pramlintide- and placebo-treated patients comprised sinusitis, retinal disorder, inflicted injury, and injection site reactions (Table 48).25, 26 Higher incidence of retinal disorder was reported with pramlintide 150 mcg than with lower pramlintide doses and placebo.26 The authors performed detailed medical reviews of these patients with reported retinal disorder and concluded that the increased incidence was likely attributable to preexisting conditions that were not documented at the time of screening.

One post-hock analysis specifically looked at markers of cardiovascular risks.23. After 16 weeks, it was found that pramlintide treated patients had favorable decreases in triglycerides when compared to placebo treated patients (Table 48). No significant changes from baseline in LDL, HDL, or total cholesterol were seen

Summary of Findings for DPP-IV Inhibitors: Harms

Sitagliptin compared with saxagliptin
  • We found no head-to-head evidence.

Summary of Findings for Sitagliptin: Harms

  • The most commonly reported adverse events across treatment groups were hypoglycemia, nausea, vomiting, diarrhea, and abdominal pain.
  • The rates for total withdrawal were slightly lower with sitagliptin than with placebo (pooled relative risk 0.63 95% CI 0.52 to 0.76) and rates of withdrawal due to adverse events were not significantly different between sitagliptin and placebo (pooled relative risk 0.88, 95% CI 0.54 to 1.43, moderate strength of evidence).
  • Hypoglycemia was generally more frequent with glipizide than with sitagliptin (17.1–34.1% compared with 1.6–5.3%) and was more common when sitagliptin was used in combination with other hypoglycemic agents than when used as monotherapy (moderate strength of evidence).
  • Hypoglycemia was not significantly different in subjects taking sitagliptin 100 mg and those taking placebo (pooled relative risk 1.26, 95% CI 0.49 to 3.25, low strength of evidence).
  • Rates of gastrointestinal side effects were higher with metformin than with sitagliptin (moderate strength of evidence).
  • Gastrointestinal side effects were not significantly different between sitagliptin and placebo treated subjects (nausea pooled relative risk 1.4, 95% CI 0.5 to 3.96; vomiting pooled relative risk 0.77, 95% CI 0.20 to 2.88, low strength of evidence).
  • Upper respiratory infections and urinary tract infections were not significantly different between patients taking placebo and those taking sitagliptin (pooled relative risk 1.06, 95% CI 0.66, 1.7, low strength of evidence)
  • Subjects treated with sitagliptin had similar changes or greater improvements in triglycerides than subjects treated with placebo (low strength of evidence); changes in other lipid parameters were not significantly different between sitagliptin and placebo (moderate strength of evidence).

Detailed Assessment of Sitagliptin: Harms

In 7 trials with data suitable for meta-analysis, total withdrawals were slightly lower among patients randomized to sitagliptin monotherapy than patients receiving only placebo (relative risk for total withdrawals 0.63, 95% CI 0.52 to 0.76); there was no significant difference for withdrawals due to adverse events (relative risk for withdrawal due to adverse events 0.88, 95% CI 0.54 to 1.43). Patients on sitagliptin monotherapy had lower rates of total withdrawal relative to patients on glipizide, who experienced more hypoglycemic events and higher rates of total withdrawal relative to patients on metformin. The rate of total withdrawals was also higher in patients whose add-on therapy was sitagliptin than in patients using monotherapy with metformin, pioglitazone, or glimepiride.

The most commonly reported adverse events were hypoglycemia, abdominal pain, nausea, vomiting, and diarrhea.

A total of 20 deaths were reported in 4 trials over 24–104 weeks. None was considered to be related to any study intervention; 8 were sudden cardiac deaths or myocardial infarctions, 2 were secondary to trauma, 1 was related to sepsis, 6 were due to cancer, 1 suicide, 1 was related to chronic obstructive pulmonary disease and interstitial lung disease, and 1 cause of death was unknown.

Rare adverse events

Sixteen randomized controlled trials reported adverse events. In those trials adverse events occurring in at least 4% of study subjects included: upper respiratory tract infections, headache, influenza, nasopharyngitis, and urinary tract infection. Incidence of adverse effects between sitagliptin and active comparator agents is summarized in Tables 4950, and incidence of adverse effects between sitagliptin and placebo is summarized in Tables 5253. Pooled relative risk for upper respiratory and urinary tract infections showed no significant difference between sitagliptin and placebo (relative risk 1.06, 95% CI 0.66 to 1.69) (Table 51).42, 43, 45 Four studies34, 42, 47, 49 reported small increases (≤10% from baseline) in mean white blood cell count, mainly an increase in absolute neutrophil count, in regimens with sitagliptin compared to regimens without. These increases appeared early and remained stable throughout the duration of the studies. No other trials provided data on changes in white blood cell count with sitagliptin. Edema was only reported for 1 study and the incidence was 5% in the rosiglitazone group and 1% in both placebo and sitagliptin groups.36

Table 49. Adverse events of sitagliptin compared with oral hypoglycemic agents.

Table 49

Adverse events of sitagliptin compared with oral hypoglycemic agents.

Table 50. Adverse events of sitagliptin compared with oral hypoglycemic agents (continued).

Table 50

Adverse events of sitagliptin compared with oral hypoglycemic agents (continued).

Table 52. Adverse events of sitagliptin compared with placebo.

Table 52

Adverse events of sitagliptin compared with placebo.

Table 53. Adverse events of sitagliptin compared with placebo (continued).

Table 53

Adverse events of sitagliptin compared with placebo (continued).

Table 51. Meta-analysis comparing adverse events of sitagliptin 100 mg to placebo.

Table 51

Meta-analysis comparing adverse events of sitagliptin 100 mg to placebo.

Hypoglycemia

In general, hypoglycemia was more common in patients treated with comparator agents as opposed to sitagliptin. Pioglitazone was the only comparator that had lower incidence of hypoglycemia. Patients taking sitagliptin in addition to glimepiride experienced more hypoglycemia than those taking glimepiride alone. Similarly, patients taking sitagliptin in addition to insulin and metformin experienced more hypoglycemia than those taking insulin and metformin alone.

There was no statistically significant difference in the overall risk of mild to moderate hypoglycemia between sitagliptin and placebo (pooled relative risk 1.26, 95% CI 0.48 to 3.25) (Table 51).30, 31, 42–46 The rate of mild-to-moderate hypoglycemia increased slightly when sitagliptin was added to glimepiride (7.6% compared with 2.8%) or pioglitazone (1.1% compared with 0%).

Abdominal pain, nausea, vomiting, and diarrhea

Compared with metformin monotherapy, sitagliptin was associated with lower incidence of abdominal pain, nausea, vomiting, and diarrhea (Tables 4950). Combination therapy of sitagliptin plus glimepiride, metformin, or pioglitazone had <6% incidence of abdominal pain, nausea, vomiting, and diarrhea; these results were not significantly different from their comparisons (Tables 4950).

There were no statistically significant differences between sitagliptin monotherapy and placebo in the risk of nausea (pooled relative risk 1.4, 95% CI 0.49 to 3.96) (Table 51). 31, 42, 43, 45 and vomiting (pooled relative risk 0.76, 95% CI 0.20 to 2.87) (Table 51).31, 42, 43 However, based on the elevated relative risks, there appears to be a trend for greater risk of experiencing abdominal pain, and nausea with sitagliptin monotherapy compared with placebo.

Lipids

Six publications reported changes in lipid parameters in patients taking sitagliptin compared to placebo, rosiglitazone, pioglitazone, glipizide, and metformin (Tables 5456).30, 32, 36, 46–48 The data for the remaining 9 publications was received from the manufacturer. In 12 trials, patients taking sitagliptin had either less elevation or greater reduction in triglycerides than those in the comparator groups. Changes in all other lipid parameters were less significant and more variable across studies. The results of our meta-analyses comparing sitagliptin with placebo for lipid parameters are summarized in Table 51. For these analyses, we assumed a pre-post correlation of 0.5 and conducted sensitivity analyses using correlations of 0.3 and 0.7. There was no statistically significant difference for total cholesterol, HDL, or LDL in our main analyses or any of the related sensitivity analyses (Appendix E). For triglycerides, the main analysis favored sitagliptin (WMD −9.97, 95% CI −19.4 to −0.49), but sensitivity analyses found no statistically significant difference between sitagliptin and placebo.

Table 54. Changes in lipid parameters (mean change from baseline, mg/dL).

Table 54

Changes in lipid parameters (mean change from baseline, mg/dL).

Table 55. Changes in lipid parameters (mean change from baseline, mg/dL) (continued).

Table 55

Changes in lipid parameters (mean change from baseline, mg/dL) (continued).

Table 56. Changes in lipid parameters (mean change from baseline, mg/dL) (continued).

Table 56

Changes in lipid parameters (mean change from baseline, mg/dL) (continued).

Summary of Findings for Saxagliptin: Harms

  • The most commonly reported adverse effects were nasopharyngitis, upper respiratory infections, headache, and urinary tract infections.
  • Rates for total withdrawal were lower with saxagliptin 2.5 and 5 mg compared with placebo used as monotherapy or as add-on therapy (2.5 mg relative risk 0.66, 95% CI 0.57 to 0.79; 5 mg relative risk 0.79, 95% CI 0.66 to 0.95, moderate strength of evidence).
  • Rates of withdrawal due to adverse events were not significantly different with saxagliptin 2.5 mg used as monotherapy or as add-on therapy compared with placebo (pooled relative risk 0.85, 95% CI 0.29 to 2.53), however rates were higher in patients taking saxagliptin 5 mg than for those taking placebo (pooled relative risk 2.09, 95% CI 1.07 to 4.10, moderate strength of evidence).
  • The incidence of hypoglycemia was not significantly different with saxagliptin 2.5 mg or 5 mg used as monotherapy or as add-on therapy compared with placebo (2.5 mg: pooled relative risk 2.01, 95% CI 0.63 to 6.39; 5 mg: pooled relative risk 1.04, 95% CI 0.28 to 3.81, low strength of evidence).
  • There were no significant differences in infections between saxagliptin and placebo (low strength of evidence).

Detailed Assessment for Saxagliptin: Harms

In the 5 identified placebo-controlled trials (see Key Question 1 saxagliptin section for study characteristics), total withdrawals were higher in the placebo groups compared to either the saxagliptin 2.5 mg or 5 mg/day groups. Withdrawals due to adverse effects were similar between placebo and saxagliptin 2.5 mg/day, however higher in saxagliptin 5 mg/day. Similar rates of withdrawal due to adverse effects were seen regardless of saxagliptin being used as add-on therapy or monotherapy. Results of our meta-analyses are summarized in Table 57 and results from individual saxagliptin trials for adverse events are summarized in Table 58 and in Evidence Table 8.

Table 57. Meta-Analysis results comparing saxagliptin to placebo as both monotherapy and add-on therapy.

Table 57

Meta-Analysis results comparing saxagliptin to placebo as both monotherapy and add-on therapy.

Table 58. Adverse events in trials of saxagliptin.

Table 58

Adverse events in trials of saxagliptin.

The most common adverse effects seen were headache, upper respiratory infections, nasopharyngitis, and urinary tract infections. Gastrointestinal adverse effects were rarely reported and were most commonly seen when saxagliptin was used in combination with metformin.

Hypoglycemia

Hypoglycemia was reported in all 5 trials. The incidence of confirmed hypoglycemia (≤50 mg/dL) was low ranging from 0 to 2.4% in saxagliptin treated patients, with 2 trials reporting zero incidence in both saxagliptin and placebo treated patients.53, 54 The trials that reported any confirmed hypoglycemia were those using saxagliptin in combination with either glyburide, metformin, or a TZD.55–57 Overall, there was no difference in the incidence of confirmed hypoglycemia in saxagliptin 2.5 mg/day (pooled relative risk 2.01, 95% CI 0.63 to 6.39) or saxagliptin 5 mg/day (pooled relative risk 1.04, 95% CI 0.28 to 3.81) compared to placebo.

Infections

Infection related adverse events were reported in all 5 trials. Pooled relative risk showed no significant difference between saxagliptin 2.5 mg daily and placebo in incidence of upper respiratory tract infections (relative risk 0.95, 95% CI 0.65 to 1.37), nasopharyngitis (relative risk 0.86, 95% CI 0.58 to 1.27), or urinary tract infections (relative risk 1.16, 95% CI 0.81 to 1.68). Similarly, no difference was seen between saxagliptin 5 mg daily and placebo in incidence of upper respiratory tract infections (relative risk 0.99, 95% CI 0.69 to 1.42), nasopharyngitis (relative risk 0.84, 95% CI 0.57 to 1.24), or urinary tract infections (relative risk 1.2, 95% 0.84 to 1.73). Three of the 5 studies reported small numerical decreases in absolute lymphocyte counts in higher dose of saxagliptin (≥10 mg daily), however minimal to nodecrease in either saxagliptin 2.5 mg or 5 mg daily.53–55

Lipids

Changes in lipid parameters were onlyreported in 1 trial. 56 When compared to placebo in addition to a TZD, there was a numerically greater increase in LDL cholesterol in subjects treated with saxagliptin in addition to a TZD (compared with a small decrease in LDL cholesterol with placebo), however there were no statistical comparisons reported. In addition, placebo-treated subjects total cholesterol decreased more than saxagliptin 2.5-treated subjects ( 4.3 compared with 3.1, respectively); subjects treated with saxagliptin 5 demonstrated a small increase (+0.8). There was a greater numerical reduction seen in triglycerides in patients receiving saxagliptin as add-on therapy compared to placebo (P=NR) (Table 59).

Table 59. Changes in lipid parameters (mean change from baseline, mg/dL).

Table 59

Changes in lipid parameters (mean change from baseline, mg/dL).

Summary of Findings for GLP-1 Agonists: Harms

Exenatide compared with liraglutide
  • In the 1 head-to-head randomized-control trial, withdrawal rates were similar between groups. The incidence of nausea was similar between the groups initially, but was more persistent over time in the exenatide group. The proportion of patients who reported minor hypoglycemia was less in the liraglutide group than the exenatide group (26% compared with 34%, 1.93 compared with 2.60 events per patient per year, rate ratio 0.55, CI 0.34 to 0.88; P=0.0131). There was no significant difference in change in total cholesterol, LDL cholesterol, or HDL cholesterol between the exenatide and the liraglutide treatment arms. Reduction in triglycerides was significantly greater in the liraglutide group than the exenatide group ( 15.8 mg/dL (3.9) compared with 8.9 mg/dL (3.9) estimated treatment difference 6.9 mg/dL, CI 14.3 to 0.0; P=0.0485) (low strength of evidence).
Exenatide
  • The longest duration of an included study was 52 weeks.
  • In the active-control trials of exenatide compared to insulin, total withdrawals and withdrawals due to adverse events were higher in the exenatide groups than the insulin groups (moderate strength of evidence).
  • Nausea and vomiting were the most frequent adverse events among exenatide-treated patients, and rates of these symptoms were significantly higher in the exenatide group than insulin and placebo groups. Nausea declined after the first 8 weeks of therapy (moderate strength of evidence).
  • Studies of exenatide did not report an association with pancreatitis, although the US Food and Drug Administration has received reports of acute pancreatitis in patients who received exenatide. A majority of affected patients (90%) in those reports had other risk factors for pancreatitis.
  • Rates of hypoglycemia were similar between insulin and exenatide groups (moderate strength of evidence).
  • In the one trial comparing exenatide to glibenclamide, total withdrawals were higher in the glibenclamide group due to higher rates of hypoglycemia (low strength of evidence).
  • There was no significant difference in total withdrawals between exenatide 5 mcg or 10 mcg daily and placebo (moderate strength of evidence).
  • Withdrawal rates due to adverse events were higher with exenatide 10 mcg twice a day than with placebo (relative risk 3.18, CI 1.70 to 5.93); there was not a statistically significant difference between treatment groups at the 5 mcg twice daily dosing (relative risk 1.76, CI 0.98 to 3.19) (moderate strength of evidence).
  • Nausea, vomiting, and diarrhea rates were significantly higher in subjects treated with exenatide (either dose) compared with those treated with placebo (moderate strength of evidence).
  • The incidence of hypoglycemia was elevated with exenatide 5 and 10 mcg twice a day compared with placebo in all 4 studies of patients on background sulfonylurea therapy (moderate strength of evidence).
  • There was no evidence of cardiovascular, pulmonary, hepatic, or renal adverse effects across studies, and rates of serious events were similar between treatment groups (low strength of evidence).
  • There was no significant difference in lipid profiles between patients on exenatide compared with placebo in the 1 study that examined this outcome (low strength of evidence).
Liraglutide
  • The longest duration of an included study was 52 weeks.
  • Total withdrawal rates were similar between liraglutide- and glimepiride-treated subjects, but withdrawals due to adverse events were slightly higher for liraglutide than glimepiride (low strength of evidence).
  • Rates of gastrointestinal side effects were higher with liraglutide than glimepiride (high strength of evidence).
  • Hypoglycemia rates were lower with liraglutide than glimepiride (moderate strength of evidence).
  • Pancreatitis: In clinical trials, there were more cases of pancreatitis among those treated with liraglutide than among those treated with other medications or placebo. Studies comparing liraglutide with glimepiride could not exclude a weak association between treatment with liraglutide and the development of pancreatitis (1 case compared with 1 case in LEAD-2 study; 2 cases compared with 0 in LEAD-3); there were no reports of pancreatitis in the active-control trial with insulin glargine; only 1 of the included placebo-controlled trials reported any cases of pancreatitis (1 case compared with 1 case) (insufficient strength of evidence).
  • Rates of gastrointestinal side effects were higher with liraglutide than with insulin glargine (1 study) (low strength of evidence).
  • Rates of minor hypoglycemia were similar between liraglutide and insulin glargine (1 study), but more patients treated with liraglutide had major hypoglycemic events (5 compared with 0) (low strength of evidence).
  • In the active-control trial comparing liraglutide to rosiglitazone, the incidence of serious adverse events was similar between treatment arms.84 Nausea was more common in the liraglutide groups compared to rosiglitazone (low strength of evidence).
  • In the active-control trial comparing liraglutide to sitagliptin, the incidence of serious adverse events was similar between treatment arms.41 Gastrointestinal complaints, particularly nausea, were more common in the liraglutide arms of the study than in the sitagliptin arm (low strength of evidence).
  • Total withdrawal rates were lower for liraglutide (0.6 mg daily, 1.2 mg daily, and 1.8 mg daily) than placebo (relative risk range 0.37 to 0.62) (moderate strength of evidence).
  • There was no significant difference in the risk of withdrawal due to adverse events with liraglutide 0.6 mg, 1.2 mg, or 1.8 mg daily compared with placebo (moderate strength of evidence).
  • The incidence of hypoglycemia was elevated with liraglutide 1.8 mg daily compared with placebo (relative risk 1.66, CI 1.18 to 2.34). Rates of hypoglycemia were not significantly different between liraglutide 0.6 mg daily and liraglutide 1.2 mg daily, and placebo (moderate strength of evidence).
  • The rates of gastrointestinal side effects were higher in the liraglutide-treated groups than in the placebo group. The risk increased with higher doses (relative risk 1.76 for 0.6 mg; relative risk 2.33 for 1.2 mg; relative risk 3.14 for 1.8 mg), but generally waned over time (high strength of evidence).
  • In the 2 studies that examined lipid parameters in liraglutide compared to placebo, liraglutide improved triglycerides compared to placebo in both studies, and improved LDL cholesterol levels compared with placebo in 1 study (low strength of evidence).
  • One study compared lipid parameters in liraglutide-treated and sitagliptin-treated subjects and found no significant difference with the exception of a slightly larger decrease in total cholesterol with liraglutide 1.8 mg (−6.6 mg/dL versus −0.8 mg/dL, P=0.0332) (low strength of evidence).

Detailed Assessment of GLP-1 Agonists: Harms

Characteristics of trials included in this section were described in the Key Question 1 section for GLP-1 agonists. In this section, we focus on the results of those trials related to harms. For observational studies, we provide a table summarizing study characteristics (Table 61).

Table 61. Characteristics of exenatide observational studies in adults with type 2 diabetes.

Table 61

Characteristics of exenatide observational studies in adults with type 2 diabetes.

Detailed Assessment of Exenatide Compared with Liraglutide: Harms

In the one 26-week randomized-controlled trial (N=464) of liraglutide compared with exenatide, withdrawal rates were not significantly different between groups.61 In the liraglutide arm of the study, 14% of participants withdrew from the study, and 10% withdrew due to adverse events. In the exenatide arm of the study, 19% of participants withdrew from the study, and 13% withdrew due to adverse events.

Overall, participants in the liraglutide group reported fewer adverse events than in the exenatide group (74.9% compared with 78.9%), but reported more serious and severe adverse events (12.3% compared with 7.3%) Only 1 serious or severe adverse event was judged to be related to study medication (severe hypoglycemia in the exenatide group).

The incidence of nausea was similar between the groups initially, but was more persistent over time in the exenatide group. Otherwise, the distribution of adverse events was similar between the study arms.

There were 2 major episodes of hypoglycemia in patients in the exenatide arm of the study who were also on a sulfonylurea. No major episodes of hypoglycemia occurred in the liraglutide arm of the study. The proportion of patients who reported minor hypoglycemia was significantly less in the liraglutide group than the exenatide group (26% compared with 34%, rate ratio 0.55, CI 0.34 to 0.88; P=0.0131).

There was no significant difference in change in total cholesterol, LDL cholesterol, or HDL cholesterol between the exenatide and the liraglutide treatment arms. Reduction in triglycerides was significantly greater in the liraglutide group than the exenatide group ( 15.8 mg/dL (3.9) compared with 8.9 mg/dL (3.9) estimated treatment difference 6.9 mg/dL, CI 14.3 to 0.0; P=0.0485)

Detailed Assessment of Exenatide: Harms

Active-control trials
Adverse effects

Total withdrawals in the exenatide group ranged from 12.0% to 21.3% and in the comparison group from 0% to 10.1% in the 4 active-control trials comparing exenatide to insulin.62–65 Withdrawals due to adverse events for the exenatide group ranged from 8% to 15% and were less than 1% in the comparison groups. Nausea and vomiting were the most frequent adverse events among exenatide-treated subjects, and rates of these symptoms were significantly higher in the exenatide group than in groups using insulin glargine62, 63 or other insulin routines, 65,64 with rates of nausea ranging from 33% to 57% in the exenatide groups compared with <1 to 9% with the comparison group receiving insulin.

Overall hypoglycemia rates were similar between groups treated with insulin and with exenatide. 62–64 Hypoglycemia was particularly common when exenatide (39%) or insulin (38%) was combined with sulfonylurea and/or metformin;65 79% of hypoglycemia cases were associated with sulfonylurea. In a study comparing exenatide and titrated insulin glargine,62 the overall rate of hypoglycemia with exenatide (14.7%) was not statistically different than that with insulin glargine (25.2%). In subgroup analysis of this study, however, the rate of hypoglycemia in patients who received metformin and exenatide was 2.6% as compared with 17.4% in those receiving insulin glargine (P=0.010), whereas the rates of hypoglycemia in patients taking sulfonylureas was similar with exenatide (30.0%) and insulin glargine (34.5%).

In the one trial comparing exenatide to glibenclamide, total withdrawals were higher in the glibenclamide group due to higher rates of hypoglycemia.(total withdrawals exenatide 6%, total withdrawals glibenclamide 12%; hypoglycemia exenatide 0%, glibenclamide 5%).66

In the one trial comparing exenatide to rosiglitazone, total withdrawals were similar between the treatment arms (exenatide 27%, rosiglitazone 24%).67 Nausea, vomiting, and diarrhea were more frequently reported in the exenatide arm than in the rosiglitazone arm of the study (nausea: exenatide 47%, rosiglitazone 4%; vomiting: exenatide 22%, rosiglitazone 0%; diarrhea exenatide 7%, rosiglitazone 4%). Symptomatic hypoglycemia occurred in 4% of participants in the exenatide arm of the study, and none of the participants in the rosiglitazone arm of the study.

Placebo-control trials
Adverse effects

The placebo-controlled trials were sufficiently homogenous to obtain pooled estimates for adverse effects. Studies were only included for each meta-analysis if they reported sufficient information for the adverse effect under study. For example, only studies that reported numbers of subjects with the adverse effect of headache were included in the meta-analysis for that adverse effect. Results of our meta-analyses are summarized below in Table 60.

Table 60. Placebo-control trials of exenatide: Summary of meta-analyses.

Table 60

Placebo-control trials of exenatide: Summary of meta-analyses.

Based on pooled estimates across the placebo-controlled trials, there was no significant difference in withdrawals from the study between placebo and exenatide 5 mcg twice daily (relative risk 0.77, 95% CI 0.56 to 1.058, P=0.106) or exenatide 10 mcg twice daily (relative risk 1.14, 95% CI 0.83 to 1.56, P=0.415). Among the 9 included placebo-controlled trials of exenatide 10 mcg daily, withdrawals due to adverse events were greater with exenatide 10 mcg twice daily than with placebo. There was no statistically significant difference in withdrawals due to adverse events between exenatide 5 mcg twice daily and placebo (Table 60).

Nausea, vomiting, and diarrhea were significantly more frequent with treatment at both dosages of exenatide than in the placebo group (Table 60). There was considerable statistical heterogeneity in the meta-analysis for nausea for exenatide 10 mcg bid (I2=76%) due to variation among studies in the magnitude of the effect, but all studies consistently did report more nausea among those treated with exenatide compared to placebo. Nausea declined after 8 weeks of treatment, although the statistical significance of the trend was not reported.69–72, 74 There was no correlation between change in body weight and duration70, 71 or severity 79 of nausea. When the incidence of nausea remained stable, body weight continued to decrease.197

Hypoglycemia was more frequent in the exenatide study groups than in the placebo study groups in all 4 studies in which participants were on background sulfonylurea therapy.69, 71, 73, 74 The risk of hypoglycemia was not increased compared with placebo when all subjects received a thiazolidinedioneor metformin 70, 72, 77 or no background therapy.73

There was no evidence of cardiovascular, pulmonary, hepatic, or renal adverse effects across studies. Serious adverse events were rare, and reported to be unrelated to the study drug. In 1 study72 2 treatment-group patients had serious adverse events (chest pain and allergic alveolitis) which did not necessitate study withdrawal. Kadowaki et al. reported 1 serious adverse event (tendon rupture) in the exenatide 2.5 mcg arm of the study. Gao et al. reported serious adverse events among 3 patients in the exenatide arm of the study (chronic cholecystitis, Bartholin’s cyst, and hypoglycemia) and 4 patients in the placebo arm of the study. Only hypoglycemia was judged to be related to the study drug.

None of these studies included in this report noted cases of acute pancreatitis, however, from the date of the drug’s approval through December 2006, the US Food and Drug Administration received 30 domestic reports of acute pancreatitis in patients who received exenatide.198 Median age of patients was 60 years and daily doses ranged from 10–20 mcg. The median time to onset of the symptoms was 34 days (range 4 to 300 days). Median amylase value was 384 IU/L and median lipase value 545 IU/L. Seventy percent of patients required hospitalization. A majority of affected patients (90%) had other risk factors for pancreatitis, including alcohol use or hypertriglyceridemia.

Moretto was the only study reporting changes in lipid profiles among participants. In this 24-week study of exenatide monotherapy, changes in fasting total cholesterol, HDL cholesterol, and LDL cholesterol from baseline to end point were not significantly different with exenatide 5 mcg and 10 mcg treatment compared with placebo.73

Observational studies

We examined adverse events in cohort studies of exenatide and identified 6 single-arm open-label extension studies,78–80, 197, 199, 200 1 single-arm retrospective cohort201 study, and 1 2-arm retrospective cohort study202 (Table 61). All of the open label extension studies assessed exenatide 10 mcg twice daily. In these studies, investigators included only subjects who had previously completed a prior study and several studies79, 80, 197 excluded patients who had received placebo.

An open-label extension study of 3 of the placebo-controlled primarytrials 69–71 included in this report was published in multiple publications with overlapping or identical populations.78–80, 197, 200 These publications represented a pooled synthesis of patients continuing in an open-label extension beyond the original 30-week trial comparing exenatide 5 mcg or 10 mcg twice daily to placebo. Subjects from both the placebo and treatment groups were invited to continue on 10 mcg twice daily along with their existing metformin and/or sulfonylurea regimens for a 2-year78 and then 3-year200 period. Mild-to-moderate nausea was the most frequently reported adverse event, and 3% of subjects withdrew over the extension period (30 weeks to 2 years) because of nausea. Eight percent of subjects continued to complain of nausea after 2-years of follow-up. Hypoglycemia (of any severity) occurred at a rate of 1 case in 1010 person-years of exenatide treatment. There were no cardiovascular, pulmonary, hepatic, or renal effects attributed to treatment.

Adverse events in subjects completing 3-year follow-up of the open label extension of these 3 placebo-controlled trials200 included mild-to-moderate nausea (59%) (5% of subjects withdrew due to nausea over the 3 years), and hypoglycemia (40%) with 2 of 527 subjects withdrawing because of hypoglycemia. This study population was a select group: only approximately half (46%) of subjects originally enrolled in the 3 primary trials enrolled in the open-label extension. Of subjects enrolled, only 54% completed the 2-year follow-up and 41% the 3-year follow-up.

An unrelated open-label, extension study199 (“Study B”) of a 28-day trial reported that nausea and vomiting were the most common adverse effects with exenatide 10 mcg twice daily for 26 weeks, but incidence rates were not reported. Approximately ¾ of subjects also received metformin; the other ¼ received diet and exercise only. A retrospective chart review201 of 200 patients who had used exenatide noted that 13% discontinued treatment due to side effects, including nausea (8%), urticaria (2%), and hypoglycemia (0.5%).

One fair quality observational study examined hypoglycemia in patients newly initiated on exenatide or insulin glargine.202 This study found that the probability of a hypoglycemic events was significantly lower for exenatide than for insulin glargine (total adjusted annualized hypoglycemia event rate for insulin glargine 0.117 +/− 0.007 compared with exenatide 0.065 +/− 0.011, P<0.001) Of note, background use of a sulfonylurea was higher in the insulin glargine group, although this was accounted for in the multivariate analysis.

Systematic reviews

Three systematic reviews of exenatide met our inclusion criteria and were rated either fair or good qaulity.28, 81, 82 Amori and colleagues28 published a systematic review of published and unpublished English-language studies of US Food and Drug Administration-approved and unapproved DPP-4 inhibitors (sitagliptin and vildagliptin) and GLP-1 agonists including exenatide. Severe hypoglycemia was rare (5/2781 patients who used exenatide) and occurred only when combined with sulfonylurea use. The risk ratio for mild to moderate hypoglycemia with exenatide compared with placebo was 2.3 (95% CI 1.1 to 4.9). Dose-dependent nausea and vomiting were the most frequently reported adverse events with exenatide (risk ratio nausea compared with any other treatment 2.9 (95% CI 2.0 to 4.2). Withdrawal rates due to gastrointestinal effects were higher with exenatide (4%) than with placebo.

A second systematic review, published by Pinelli and colleagues in 2008, also compared exenatide to placebo and insulin in terms of adverse events.81 In a meta-analysis of the 5 included studies on exenatide, the pooled odds ratios for nausea was 9.02 (95% CI 3.66 to 22.23), for vomiting was 4.56 (95% CI 3.13 to 6.65), and for diarrhea was 2.96 (95% CI 2.05 to 4.26), The risk of hypoglycemia was not significantly greater in the pooled analysis of exenatide versus comparators (pooled odds ratio 3.53; 95% CI 0.92 to 13.61).

Another systematic review of GLP-1 receptor agonists, including exenatide and liraglutide, was also included.82 This study combined trials of both exenatide and liraglutide into one meta-analysis and found, similar to our results, an increased risk of gastrointestinal side effects with exenatide and liraglutide. Monami et al. found that exenatide increased the risk of hypoglycemia compared to placebo, but only when exenatide was combined with a sulfonylurea. This is similar to the findings of our meta-analyses.

Detailed Assessment of Liraglutide: Harms

Active-control trials

Among the 3 studies comparing liraglutide to glimepiride, total withdrawals in the liraglutide group ranged from 7% to 35% compared with 0% to 39% in the glimepiride group. 58–60 Withdrawals due to adverse events for the liraglutide group ranged from 5% to 12% and were 0% to 6% in the comparison glimepiride group. Nausea, vomiting, and diarrhea were frequent adverse events among liraglutide-treated subjects. Rates of these symptoms were higher in the liraglutide group than in groups using glimepiride. In the LEAD-2 study, 35% to 44% of participants on liraglutide reported nausea, vomiting, or diarrhea, compared to 17% of participants on glimepiride. Vomiting was reported in 5% to 7% of participants on liraglutide, compared to 1% on glimepiride. Diarrhea was reported in 8% to 15% of participants on liraglutide, compared to 4% on glimepiride. There was a trend toward increase side gastrointestinal side effects with the increased dose of liraglutide.59 In the LEAD-3 study, 27% to 29% of participants on liraglutide reported nausea, compared to 8% of participants on glimepiride. Vomiting was reported in 9% to 12% in the liraglutide group, compared to 4% in the glimepiride group. Diarrhea was reported in 16% to 19% in the liraglutide group, compared to 9% in the glimepiride group.60 The majority of the symptoms of nausea occurred in the first weeks of therapy. In both the LEAD-2 and the LEAD-3 study groups, by week 4 less than 10% of subjects in the liraglutide groups reported nausea.

Overall hypoglycemia rates were lower in the liraglutide groups than in the glimepiride group.58–60 Minor hypoglycemia occurred in 3% to 12% of the participants in the liraglutide groups, and 15% to 24% of the participants in the glimepiride groups. There were no reports of major hypoglycemic events in any of the participants in these studies.

Two participants in the LEAD-2 study developed pancreatitis; 1 was in the liraglutide 1.2 mg arm and 1 was in the glimepiride arm. Two participants in the LEAD-3 study developed pancreatitis; both were in the liraglutide arms of the study. A weak association between development of pancreatitis and treatment with liraglutide could not be excluded.59, 60

In the 1 active-control trial comparing liraglutide 1.8 mg daily to open-label insulin glargine, rates of adverse events were higher in the liraglutide arm than the insulin glargine arm. This was in large part secondary to a higher incidence of nausea, dyspepsia, vomiting, and diarrhea in the liraglutide arm of the study (nausea: liraglutide 14% compared with insulin glargine 1%; dyspepsia: liraglutide 7% compared with insulin glargine 2%; vomiting: liraglutide 7% compared with insulin glargine 0.4%; diarrhea: liraglutide 10% compared with insulin glargine 1%.) The incidence of gastrointestinal symptoms was highest in the early weeks of the study and waned over time.83 Rates of minor hypoglycemia were similar between the 2 groups (liraglutide 27.4% compared with insulin glargine 28.9%). Five patients in the liraglutide arm (2.2%) did report major hypoglycemic events in this study, compared to no patients with major hypoglycemic events in the insulin glargine arm of the study. There were no reports of pancreatitis in this study.

In the active-control trial comparing liraglutide to rosiglitazone, the incidence of serious adverse events was similar between treatment arms (rosiglitazone 3%, liraglutide 3–5%).84 Nausea was more common in the liraglutide groups compared to rosiglitazone, although the occurrence of nausea decreased over time in the liraglutide treatment arms. Minor hypoglycemia was experienced by 4.3% of participants in the rosiglitazone arm, and by 5.2% to 9.2% of participants in the liraglutide arms of the study. One participant in the liraglutide 0.6 mg arm developed pancreatitis. No other cases of pancreatitis were reported in the study.

In the active-control trial comparing liraglutide to sitagliptin, the incidence of serious adverse events was similar between treatment arms (3% to 4%).41 Gastrointestinal complaints, particularly nausea, was more common in the liraglutide arms of the study than in the sitagliptin arm (incidence of nausea: liraglutide 21–27%, sitagliptin 5%). The median duration of nausea was 8 days with liraglutide 1.8 mg, and 13 days with liraglutide 1.2 mg.

This study also compared changes in fasting lipid profile over the course of the study between the liraglutide and the sitagliptin study arms, and found no significant difference between the two drugs in terms of fasting lipid profile changes with the exception of a slightly larger decrease in total cholesterol with liraglutide 1.8 mg compared to sitagliptin (−6.6 mg/dL versus −0.8 mg/dL, P=0.0332).41 None of the other active-control trials measured changes in lipid profile with liraglutide.

Placebo-controlled trials

Based on pooled estimates across the placebo-controlled trials included in this systematic review for the 3 dosing ranges of liraglutide (0.6 mg to 0.65 mg daily, 1.2 mg to 1.25 mg daily, and 1.8 mg to 1.9 mg daily) there was significantly lower risk of withdrawal in the liraglutide arms than the placebo arms, for all of the doses of liraglutide (Table 62). Withdrawals due to adverse events, however, were not significantly different between liraglutide and placebo for all 3 dosing ranges.

Table 62. Liraglutide compared with placebo: Summary of meta-analyses.

Table 62

Liraglutide compared with placebo: Summary of meta-analyses.

Gastrointestinal side effects were more frequent with liraglutide than with placebo at all of the liraglutide dosages included in this review (Table 62). There was an increasing risk of gastrointestinal side effects at higher doses of liraglutide (pooled effect liraglutide 0.6 mg once daily relative risk 1.76, 95% CI 1.16 to 2.67, P=0.0196; liraglutide 1.2 mg once daily relative risk 2.33, 95% CI 1.78 to 3.04, P<0.001; liraglutide 1.8 mg once daily relative risk 3.14, 95% CI 2.49 to 3.94, P<0.001). In general, the gastrointestinal side effects were mild in severity, and decreased over time.59, 86

There was no significant difference in the risk of hypoglycemia between liraglutide 0.6 mg daily or liraglutide 1.2 mg daily, and placebo (Table 62). There was an increased risk of hypoglycemia with liraglutide 1.8 mg daily compared with placebo (pooled effect liraglutide 1.8 mg once daily relative risk 1.66, 95% CI 1.18 to 2.34, P=0.004) Russell-Jones et al was the only study reporting any major hypoglycemic events.83 In this study, 5 patients in the liraglutide 1.8 mg daily arm and none in the placebo arm reported a major hypoglycemic event. Only 1 of these events required medical assistance.

There was no evidence of cardiovascular, pulmonary, hepatic, or renal adverse effects across studies. Serious adverse events were rare. Nauck et al reported 1 case of pancreatitis in the liraglutide 1.2 mg arm of the study and 1 case of pancreatitis in the glimepiride arm of the study which required study withdrawal. Other than these cases, there were no reports of pancreatitis in the included studies.

Two studies evaluated lipid parameters. Vilsboll et al found no significant difference between liraglutide and placebo in changes in total cholesterol, LDL cholesterol, or HDL cholesterol, but did find that liraglutide significantly reduced triglyceride levels compared to placebo (liraglutide 1.90 mg compared with placebo: −22%, 95% CI −35 to −6, P=0.0110; liraglutide 1.25 mg compared with placebo: −15%, 95% CI −30 to 2, P=0.0854; liraglutide 0.65 mg compared with placebo: −19%, 95% CI −33 to −2, P=0.0304).87 Zinman et al found no significant difference in total cholesterol or HDL cholesterol, but found a significant reduction in LDL cholesterol and triglyceride levels in participants treated with liraglutide 1.2 mg daily compared to placebo (reduction in LDL liraglutide 1.2 mg daily compared with placebo −10.81 mg/dL compared with −3.86 mg/dL, P<0.05; reduction in triglycerides liraglutide 1.2 mg daily compared with placebo −33.62 mg/dL compared with −11.5, P<0.05). There was no significant difference for any of the lipid parameters tested for liraglutide 1.8 mg compared with placebo.86

Liraglutide causes dose-dependent and treatment-duration dependent thyroid C-cell tumors in rats and mice. It is unknown whether liraglutide causes thyroid C-cell tumors in humans, but because of the association in rats and mice prescribing information for liraglutide indicates that liraglutide is contraindicated in patients with a personal or family history of medullary thyroid cancer or with a history of Multiple Endocrine Neoplasia syndrome type 2.

II. Thiazolidinediones (TZDs)

Summary of Findings for TZDs: Harms

  • In September 2010, the US Food and Drug Administration restricted access for rosiglitazone and combination products that contain rosiglitazone due to an increased risk of cardiovascular adverse events.
  • We found no evidence of increased all-cause mortality or cardiovascular mortality with pioglitazone; some studies suggest reduced risk of all-cause and cardiovascular mortality with pioglitazone (low strength of evidence).
  • Evidence from systematic reviews, randomized controlled trials, and observational studies indicate that both pioglitazone and rosiglitazone increase the risk of heart failure (odds ratios range from 1.32 to 2.18 in various meta-analyses, high strength of evidence).
  • Evidence from systematic reviews, randomized controlled trials, and observational studies indicate that both pioglitazone and rosiglitazone increase the risk of edema (odds ratios range from 2.26 to 4.62 in various meta-analyses, high strength of evidence).
  • The risk of hypoglycemia is reduced with TZDs when compared with sulfonylureas; the risk is similar to the risk with metformin (high strength of evidence).
  • Both TZDs resulted in a similar weight increase. The increase is similar to that with sulfonylureas (moderate strength of evidence).
  • Risk of fractures is increased among patients exposed to TZDs (odds ratio 1.45, 95% CI 1.18 to 1.79, from meta-analysis of 10 randomized controlled trials involving 13,715 participants, moderate strength of evidence). This risk appears to be increased among women (odds ratio 2.23, 95% CI 1.65 to 3.10) but not among men (odds ratio 1.00, 95% CI 0.73 to 1.39). These findings are consistent with the results of the ADOPT trial.
  • Adverse events occurring with pioglitazone and rosiglitazone were similar in head-to-head trials (low strength of evidence).

Detailed Assessment of TZDs: Harms

Restricted access for rosiglitazone

In September 2010, the US Food and Drug Administration announced that GlaxoSmithKline must develop a restricted access program for its drug, rosiglitazone (Avandia®) and combination products that contain rosiglitazone (Avandaryl®, and Avandamet®). These new restrictions are in response to data that suggest an elevated risk of cardiovascular events, such as heart attack and stroke, in patients treated with Avandia. The restrictions limit the use of rosiglitazone to patients with type 2 diabetes who cannot control their glucose levels with other medications and cannot take pioglitazone. Doctors will have to document their patients’ eligibility and patients will have to review information and acknowledge that they understand the risks. Patients who are currently using rosiglitazone and benefiting from it may continue using the medication if they choose.203

The US Food and Drug Administration also ordered GSK to convene an independent group of scientists to review key aspects of the company’s RECORD trial, which studied the cardiovascular safety of Avandia compared to standard diabetes drugs. During the course of the US Food and Drug Administration’s review of the RECORD study, important questions arose about potential bias in the identification of cardiovascular events. In addition, the US Food and Drug Administration halted the GSK’s TIDE trial, comparing Avandia to Actos and to standard diabetes drugs.203

Systematic reviews of active-control and placebo-controlled trials of TZDs

A number of systematic reviews examined adverse effects in the previous Drug Effectiveness Review Project TZDs reports89, 106, 109, 111–116, 119–123(See Evidence Table 1 for 2008 TZD Report). We identified 8 new systematic reviews meeting inclusion criteria for this report (Table 63 and Evidence Tables 17 and 18).81, 204–210 Five were assessed as good quality 81, 204, 207, 208, 210 and 3 were fair quality.205, 206, 209 One review focused on fractures,204 1 focused on cardiovascular outcomes,210, 1 on the risk of myocardial infarction and other major adverse cardiac events,209 and 1 on myocardial infarction and chronic heart failure.206 Mannucci et al 205 reported all-cause mortality in addition to adverse cardiovascular events, and both Pinelli et al and Phung et al reported efficacy and safety outcomes.81, 207

Table 63. Recent systematic reviews reporting adverse events with thiazolidinediones.

Table 63

Recent systematic reviews reporting adverse events with thiazolidinediones.

Mortality

Few reviews examined mortality (total or cardiovascular).89, 115, 121, 122 Eurich and colleagues 115 examined the use of various antidiabetic agents in patients with heart failure and diabetes and identified 3409 thiazolidinedione-treated subjects. Pooled odds ratios for thiazolidinediones compared with other hypoglycemic agents for all-cause mortality was 0.83 (95% CI 0.71 to 0.97, P=0.02) when 4 studies of varying designs (3 were observational studies) were pooled (I2= 52%, P=0.10). Pioglitazone and rosiglitazone were combined in the studies contributing to these pooled effects. These authors note that the finding of lower all-cause mortality with thiazolidinediones should be interpreted with caution, as 3 of the 4 studies contributing to this estimate were observational in design, and subjects receiving these drugs may have been at lower risk for heart failure due to the commonly perceived risk of using them among persons with higher risk of cardiovascular events and congestive heart failure.

In contrast to Eurich and colleagues,115 Singh, Loke, and Furberg122 found no difference in all-cause mortality when they examined only rosiglitazone. In 4 trials, the relative risk for all-cause mortality was 0.99 (95%, 0.80 to 1.23; P=0.92). Cardiovascular mortality rates were similar to all-cause rates (relative risk 0.90 [95% CI 0.63 to 1.26], P=0.53).

The Singh, Loke, and Furberg122 review differed from that of Eurich and colleagues115 as the former review included subjects with either type 2 diabetes or prediabetes, included randomized controlled trials only, and was not restricted to subjects with heart failure. Both of the reviews included active drug and placebo comparisons, and only the randomized controlled trial by Dargie171 was included in both the reviews.

In a systematic review of thiazolidinedione use in subjects who underwent coronary stent implantation, at 6-month follow-up mortality rate was 2/259, a death in each the control and rosiglitazone arms.121 Bolen and colleagues89 did not identify sufficient studies examining mortality to permit calculation of a pooled estimate.

Mannucci et al205 included 94 published and unpublished randomized control trials with over 20,000 subjects with type 2 diabetes to assess whether pioglitazone is associated with increased cardiovascular risk. They found a reduced risk of all-cause mortality (odds ratio 0.30, 95% CI 0.14 to 0.63; P<0.05) in an analysis of trials that reported at least 1 death and excluded the PROACTIVE study.177 PROACTIVE was excluded because it enrolled subjects at very high cardiovascular risk and was considered to not be representative of subjects receiving pioglitazone in the actual world. When all studies, including the PROACTIVE study, were included in the analysis, there was no significant reduction in mortality associated with pioglitazone. An analysis of studies comparing pioglitazone to rosiglitazone showed no significant difference in all-cause mortality between the 2 drugs.

The study by Mannucci and colleagues 205 reported a statistically significant reduced risk of cardiovascular death with pioglitazone use when all studies that reported cause of death were analyzed (relative risk 0.35, 95% CI 0.14 to 0.85). When the authors considered only studies that reported at least 1 cardiovascular death, the results were not statistically significant (Mantel-Haenszel odds ratio 0.49, 95% CI 0.21 to 1.15 [P=0.10]).

Cardiovascular morbidity

Several reviews examined the effects of thiazolidinediones on cardiovascular events; 3 focused on rosiglitazone, 119, 122, 206, 1 focused on pioglitazone,205 and another on both thiazolidinediones.89 Richter and colleagues only identified data from ADOPT147(discussed below). Singh, Loke, and Furburg122 identified 3 randomized controlled trials in type 2 diabetes,147, 158, 171 all of which were included in this update. Pooled estimates were obtained for these 3 randomized controlled trials and the DREAM trial of persons with prediabetes.211 These studies compared various drugs at a variety of follow-up intervals, although statistical tests for heterogeneity were not significant by usual criteria. The relative risk for myocardial infarction of rosiglitazone compared with other drugs was 1.42 (95% CI 1.06 to 1.91); as noted above, the relative risk for cardiovascular mortality was not increased.

Bolen89 stratified studies by the drug used for comparison and did not obtain pooled estimates because of clinical and methodological diversity. Three randomized controlled trials comparing thiazolidinediones and metformin and 2 randomized controlled trials comparing thiazolidinediones and sulfonylureas reported similar rates of nonfatal myocardial infarction or coronary heart disease between the thiazolidinedione and the comparison drug. Five short-duration, placebo-controlled studies also found similar rates of cardiovascular disease events and the PROACTIVE placebo-controlled trial also demonstrated no significant difference.177 Three randomized controlled trials examining restenosis rates noted fewer cardiovascular disease events with thiazolidinediones than with placebo in patients at high risk.

Mannucci and colleagues205 reported no statistically significant difference in the risk of non-fatal coronary events associated with pioglitazone use based on an analysis of 40 randomized control trials (8,248 patients) of pioglitazone compared with any other treatment (relative risk 0.82, 95% CI 0.55 to 1.23). When limited to studies that reported at least 1 non-fatal coronary event, the results still did not meet statistical significance. These analyses did not include the PROACTIVE study.177 When PROACTIVE was included in the analysis, the results reached statistical significance in favor of pioglitazone (results shown in graph only).

In a meta-analysis of 86 trials (30,003 patients) of rosiglitazone compared with any other treatment, Monami and colleagues 206 examined the association of the risk of chronic heart failure (discussed below) and the risk of myocardial infarction with specific baseline characteristics such as HbA1c, body mass index, lipid levels, duration of diabetes, and insulin use. They aimed to identify moderators of the effect of rosiglitazone on the risk of myocardial infarction and chronic heart failure in type 2 diabetic patients. The authors used data from the studies that reported at least 1 myocardial infarction to calculate a Mantel-Haenszel (M-H) odds ratio for myocardial infarction. The observed increased risk of myocardial infarction with rosiglitazone use was not statistically significant (M-H odds ratio 1.18, 0.91 to 1.53). Trials enrolling patients with a higher HbA1c at baseline reported a lower risk of myocardial infarction. This correlation remained significant after adjusting for duration of the trials (r −0.24, P=0.03). Other significant correlations after adjusting for trial duration included lower triglycerides, higher body mass index, and more patients treated with insulin were associated with a higher risk of myocardial infarction.

Congestive heart failure

In a review of persons with diabetes or prediabetes using rosiglitazone,122 the relative risk of heart failure for rosiglitazone compared with various other antidiabetic drugs was 2.09 (1.52 – 2.88), corresponding to a number needed to harm of 383 per year if baseline risk was 0.24% per year (low risk, from the ADOPT trial).147

Singh and colleagues123 also examined onset of congestive heart failure in both pioglitazone and rosiglitazone compared with placebo in 3 randomized controlled trials with subjects with either type 2 diabetes or prediabetes. The odds ratio for all heart failure adverse events was 2.10 (95% CI 1.08 to 4.08). Four observational studies produced an odds ratio1.55 (95% CI 1.33 to 1.80). These authors also examined case reports, including 162 case subjects with 99 analyzable cases. Among these cases, the median time to onset of congestive heart failure was 24 weeks, although failure could occur early and did not appear to relate to dosage. Heart failure was not limited to the elderly; 26% of cases were in subjects less than 60 years of age.

Hospital admission for heart failure was elevated with thiazolidinediones compared with other treatments (pooled odds ratio 1.13 [95% CI 1.04 to 1.22], P=0.004; 4 studies, including 3 of observational design).115

In a Cochrane review of placebo-controlled trials of rosiglitazone,119 the authors identified data only from the ADOPT trial.147

In a review of oral hypoglycemic agents, Bolen and colleagues89 noted that the risk for congestive heart failure was higher with thiazolidinediones as either monotherapy or combination therapy than with metformin or sulfonylureas, with a range of 0.8% to 3.6% for thiazolidinediones and 0 to 2.6% for nonthiazolidinediones.

In a systematic review of thiazolidinediones use in diabetes and prediabetes,116 Lago, Singh, and Nesto noted an increased risk of congestive heart failure compared with controls (placebo-controlled and active-control trials): relative risk 1.72 (95% CI 1.21 to 2.42). For placebo-controlled trials only, the relative risk was 1.97 (95% CI 0.94 to 4.13). When examined separately, the relative risk for pioglitazone was 1.32 (95% CI 1.04 to 1.68); for rosiglitazone the relative risk was 2.18 (95% CI 1.44 to 3.32). The overall event rate for congestive heart failure with thiazolidinediones was 2.3% and with the comparison drugs 1.4%. The number needed to harm for congestive heart failure was 107 over the 29.7-month follow-up (number needed to harm ranged across studies from 35 to 491). Although the risk of heart failure was increased, the risk of cardiovascular death was not significant: relative risk 0.93 (95% CI 0.67 to 1.29); placebo-controlled trials only: relative risk 1.08 (95% CI 0.66 to 1.76); pioglitazone only: relative risk 1.01 (95% CI 0.51 to 2.09); rosiglitazone only: relative risk 0.91 (95% CI 0.63 to 1.3).

An analysis of 40 randomized control trials of pioglitazone use in 10,171 patients with type 2 diabetes by Mannucci et al205 showed no statistically significant increase in the risk of non-fatal heart failure (relative risk 1.32, 95% CI 0.88 to 1.98). When PROACTIVE177 was included in the analysis, an increased risk of non-fatal heart failure with the use of pioglitazone became statistically significant (results reported in graph only).

Monami and colleagues 206 found an increased risk of chronic heart failure for rosiglitazone compared with other treatments (M-H odds ratio 1.59, 95% CI 1.11 to 2.28) in an analysis of randomized control trials that reported at least 1 occurrence of chronic heart failure. The risk ratio for chronic heart failure in rosiglitazone-treated patients was lower in trials enrolling subjects with higher HbA1c. This correlation did not remain statistically significant after controlling for the duration of the trials. Correlations between duration of diabetes and higher triglycerides with a lower risk of chronic heart failure were statistically significant after adjusting for duration of trials.

Edema

Bolen and colleagues89 noted that the risk for edema was higher with thiazolidinediones than metformin or second generation sulfonylureas. Although few cases were considered serious, withdrawals secondary to edema were common. Both pioglitazone and rosiglitazone were associated with higher rates of edema than placebo; the between-group difference (in favor of placebo) was 0% to 3.4% for pioglitazone and 2.5% to 17% for rosiglitazone.

In a Cochrane review of pioglitazone,120 the authors pooled data on all available randomized controlled trials regardless of comparisons and noted a relative risk of edema of 2.86 (95% CI 1.14 to 3.18). Richter and colleagues did a similar review of rosiglitazone119 and noted an odds ratio for edema of 4.62 (95% CI 2.28 to 9.38).

Berlie and colleagues114 examined the risk of edema in a systematic review and the odds ratio for pioglitazone and rosiglitazone combined (from comparisons with various drugs) was 2.26 (95% CI 2.02 to 2.53, P<0.00001). These authors attempted to compare the rates with rosiglitazone and pioglitazone and found the rates higher with rosiglitazone (odds ratio 3.75 [95% CI 2.70 to 5.20]) compared with pioglitazone (odds ratio 2.42 [95% CI 1.90 to 3.08]).

Hypoglycemia

Hypoglycemia was fairly uncommon with both thiazolidinediones. The combination of insulin and a thiazolidinedione increased rates of hypoglycemia.89, 120, 212 Hypoglycemia rates with thiazolidinediones were lower than rates with sulfonylureas.89, 119, 120 Thiazolidinediones cause less hypoglycemia than second generation sulfonylureas, with risk differences ranging between 0.3 and 0.25 (overall risk difference 0.09, 95% CI 0.03 to 0.25). Rates with metformin were similar to those with thiazolidinediones (obtained from indirect comparisons).89

Pinelli and colleagues81 conducted a systematic review to compare the efficacy and safety of adding TZDs or exenatide to oral agents. They conducted a meta-analysis of 17 randomized control trials comparing a TZD with placebo or active control. Four of these studies reported results for hypoglycemic events. The risk of experiencing nonsevere hypoglycemia with TZD use compared with other treatments was not statistically significant (odds ratio 1.59, 95% CI 0.76 to 3.32). Severe hypoglycemia was rare in all of the trials.

Phung207 conducted a systematic review with meta-analysis to compare the addition of various noninsulin antidiabetic drugs to ongoing metformin. They found that there was no statistically significant increased risk of hypoglycemia with TZDs (RR 2.04, 95% CI 0.5 to 8.23).

Elevated serum aminotransferase levels

Bolen and colleagues89 found that rates of significant increases in serum aminotransferase levels (> 1.5 to 2 times normal) were low (<1%) and were similar to rates with metformin and second generation sulfonylureas. Other systematic reviews reached similar conclusions.106, 111, 213

Weight change

Thiazolidinediones caused similar weight gain compared with sulfonylureas either as mono- or combined therapy. Metformin consistently caused weight loss compared with thiazolidinediones and other oral agents.89 These authors identified 2 head-to-head randomized controlled trials and noted similar increases in weight with pioglitazone and rosiglitazone.

The review by Pinelli and colleagues81 assessed weight change with TZD use. Data from 5 studies with sufficient data on change in weight from baseline, showed that TZD use was associated with a statistically nonsignificant increase in weight (weighted mean difference 1.51 kg, 95% CI −0.12 to 3.15). When 3 trials comparing a TZD to an insulin secretagogue or muraglitazar were removed from the analysis, TZDs were associated with a statistically significant increase in body weight (weighted mean difference 2.19 kg, 95% CI 1.24 to 3.14).

The review by Phung and colleagues207 comparing the addition of various noninsulin antidiabetic drugs to ongoing metformin found that TZDs, sulfonylureas, and glinides were associated with weight gain (WMD 2.3 kg, 95% CI 1.7 to 2.9 for TZDs compared with placebo), whereas GLP-1 analogs, alpha-glucosidase inhibitors, and DPP-IV inhibitors were associated with weight loss or no weight change.

Fractures

One systematic review that analyzed data on the occurrence of fractures among patients using TZDs met our inclusion criteria.204 It reported a statistically significant increased risk of fracture among patients who were exposed to rosiglitazone and pioglitazone (odds ratio 1.45, 95% CI 1.18 to 1.79). The analysis was based on 10 randomized controlled trials involving 13,715 participants. The authors reported a significantly increased risk of fractures among women (odds ratio 2.23, 95% CI 1.65 to 3.10) but not among men (odds ratio 1.00, 95% CI 0.73 to 1.39) (data from 5 trials). The review concluded that long-term use of TZDs doubles the risk of fractures among women with type 2 diabetes, without a significant increase in risk among men.

Other reviews

In addition to the systematic reviews identified for the previous Drug Effectiveness Review Project TZDs report, 2 reviews were described (See Evidence Table 1 for 2008 TZD Report) in the previous Drug Effectiveness Review Project TZDs report that were not systematic and therefore did not fulfill inclusion criteria for the previous report.214, 215 We found 1 additional review216 that used the same dataset from the review by Nissen et al.

Nissen and Wolski215 examined the cardiovascular morbidity and mortality associated with rosiglitazone in a meta-analysis of 42 trials which included data from the Food and Drug Administration Web site, a clinical trials registry maintained by GlaxoSmithKline, and a search of the published literature. This paper was not determined to be a systematic review and therefore did not fulfill inclusion criteria. Evidence of a comprehensive literature search and data synthesis was not provided in the publication. Two large trials (DREAM and ADOPT) were the only included trials from the published literature. The authors noted an odds ratio for myocardial infarction of 1.43 (95% CI 1.03 to 1.98) and for death from cardiovascular causes of 1.64 (95% CI 0.98 to 2.74).

Diamond and colleagues216 reanalyzed the 42 trials included in the review by Nissen and Wolski to demonstrate variation in results from using various meta-analytic approaches.

Lincoff and colleagues214 examined the effect of pioglitazone on ischemic cardiovascular disease complications in diabetes using a database of individual patient data from Takeda Pharmaceuticals, the manufacturers of pioglitazone. The primary composite endpoint (death, nonfatal myocardial infarction, and nonfatal stroke) was decreased with pioglitazone as mono- or combination therapy with a variety of antidiabetic drugs (hazard ratio 0.82 [95% CI 0.72 to 0.94; P=0.005]). For placebo-controlled trials the hazard ratio was 0.09 (95% CI 0.01 to 0.84). The risk of serious heart failure was increased with pioglitazone (hazard ratio 1.41 [95% CI 1.14 to 1.76; P=0.002]).

One additional review by Padwal and colleagues217 examined various drugs in the prevention of diabetes and included several studies on troglitazone, but none on pioglitazone or rosiglitazone.

Direct evidence: Pioglitazone compared with rosiglitazone head-to-head trials

Eight head-to-head efficacy trials with adverse event data were identified.92, 93 90, 99–102 In one,92 719 patients with both type 2 diabetes and dyslipidemia were randomized to treatment with pioglitazone 30 mg daily for 12 weeks followed by 45 mg for an additional 12 weeks, or rosiglitazone 4 mg daily followed by 8 mg for the same intervals. There were no differences between the drugs in adverse events including weight change (2.0 ±0.2 kg for pioglitazone compared with 1.6 ±0.2 kg for rosiglitazone, P=0.164), liver function tests, creatine phosphokinase level, blood pressure and heart rate, hemoglobin and hematocrit, hypoglycemic episodes, edema, or congestive heart failure. Data on the incidence of specific adverse events were not reported. Total withdrawals (19.0% for pioglitazone compared with 21.9% for rosiglitazone) and withdrawals due to adverse events (2.7% for both drugs) were similar.

A second study included patients who were switched to pioglitazone or rosiglitazone from troglitazone.93 There was no information reported about adverse events in this study, with the exception of a similar weight gain in both groups (data not reported).

In a head-to-head trial in patients with type 2 diabetes and metabolic syndrome,90 there was no significant difference in the increase in body mass index after 12 months of treatment with pioglitazone 15 mg (1.2 kg/m2) or rosiglitazone 4 mg (1.5 kg/m2), with both groups receiving glimepiride. Of the 87 patients (96%) who completed the study, 6.7% of subjects in the pioglitazone group and 11.9% in the rosiglitazone group had mild to moderate adverse events (transient headache and flatulence), with none resulting in withdrawal. There were no significant differences between treatment groups in serum alanine (ALT) or aspartate (AST) aminotransferase at 12-month follow-up. In 1 subject in the pioglitazone group (N=45) ALT and AST increased to 1.5 times the upper limit of normal but returned to normal range after 15 days. With rosiglitazone (N=42) 2 subjects increased AST.

One of the head-to-head studies identified for the updated report (2008) presented both tolerability and adverse events data. Derosa and colleagues94–96, 142 noted among study completers (93% completion rate) that the rate of any side effect was 8.3% in the pioglitazone group and 10.4% in the rosiglitazone group (between-group P value >0.05), with both groups also taking metformin. These adverse events were transient headache and flatulence (metformin was new to some of the study subjects).98 In this trial, there were no significant differences between treatment groups in ALT or AST at 12-month follow-up. In 2 subjects in the pioglitazone group (N=48) ALT and AST increased to 1.5 times the upper limit of normal, but regressed to normal range after 15 days. With rosiglitazone (N=48) in 3 subjects AST and ALT increased to 2.0 times the upper limit of normal and also regressed. No other adverse events were reported in this study. Hematocrit decreased significantly in both treatment groups (P<0.05): Change with pioglitazone was −2.3 umol/L and with rosiglitazone was −2.4umol/L.

The detailed adverse event results for the other 4 can be found in Evidence Table 10.99–102 Briefly, all 4 of these were small studies (Numbers of 12, 35, 50, and 60) ranging from 12 to 20 weeks and were not designed to assess harms. Three of them reported slightly greater improvements in some lipid measures with pioglitazone than with rosiglitazone.99, 101, 102

Indirect evidence

For this report, we did not update the comparisons of pioglitazone or rosiglitazone compared with placebo. This information was included in the 2008 Drug Effectiveness Review Project drug class review on TZDs. We briefly summarize the findings of that report here.18

Overall withdrawals

Nine placebo-controlled trials of pioglitazone160, 163–165, 177, 180, 218–220 and 16 of rosiglitazone166, 167, 169–172, 176, 221–232 reported overall withdrawal rates. Treatment group withdrawal rates ranged from 7% to 33% in pioglitazone trials and 0 to 28% in rosiglitazone trials. Pooled risk differences showed trends for lower overall withdrawals in treatment groups than placebo groups for both pioglitazone (−1.0%; 95% CI −3.0% to 1.0%) and rosiglitazone (−5.0%; 95% CI −10.0 to 0.0). There was significant heterogeneity among rosiglitazone trials.

Withdrawals due to adverse events

The previous Drug Effectiveness Review Project TZDs report18 found that the proportion of patients who withdrew due to adverse events was similar for the 2 drugs: 4.7% in pioglitazone trials and 5.3% in rosiglitazone trials. Pooled risk differences showed no differences from placebo in either pioglitazone (0%; 95% CI −2 to 2) or rosiglitazone (−1%; 95% CI −3 to 0) trials. The proportion of withdrawals due to adverse events in the placebo groups differed between these groups of studies (4.4% in pioglitazone studies compared with 6.8% in rosiglitazone studies), so the pooled risk differences were not directly comparable.

Specific adverse events reported in placebo-controlled trials

The previous Drug Effectiveness Review Project TZDs18 indicated that the quality of reporting of adverse events in randomized controlled trials designed to measure efficacy was fair to poor. Most studies did not prespecify which events were evaluated and did not report details about ascertainment methods. In most cases, there was no difference from placebo in the number of patients reporting an adverse event. The most frequently reported adverse events were edema, hypoglycemia, and weight gain.

Edema

The incidence of edema reported in 16 placebo-controlled trials ranged from 0% to 27%. The incidence of edema was significantly greater with both pioglitazone and rosiglitazone than placebo. The pooled risk difference was significantly greater than placebo in pioglitazone trials (4%, 95% CI 2 to 7). Rosiglitazone was also associated with an increased risk of edema. The pooled risk difference in 7 placebo-controlled trials170, 176, 221, 227, 232–234 was 8% (95% CI 3 to 13). There was significant heterogeneity among the rosiglitazone trials, due to a higher incidence of edema in 2 of the trials (23% and 24%).176, 232 The incidence in the other 5 trials ranged from 3% to 8%, with differences from placebo ranging from 2% to 6%.

Hypoglycemia

The incidence of hypoglycemic episodes was reported in 11 placebo-controlled efficacy trials. The incidence ranged from 0 to 37.5% in 7 studies of pioglitazone and from 5.2% to 52.5% in 4 studies of rosiglitazone. The pooled risk difference between treatment and placebo was not significantly different for either drug, however.

The trials of rosiglitazone examined combination therapy with sulfonylureas176, 233, 235 or triple therapy with sulfonylurea and metformin.170 In pioglitazone trials, 3 used monotherapy,163, 236, 237 1 used combination therapy with sulfonylureas,164 and 3 used combination therapy with insulin.165, 218, 238 Pooled risk differences were not significantly different from placebo in pioglitazone trials using monotherapy (1%, 95% CI −1 to 2), combination therapy with sulfonylureas (1%, 95% CI −1 to 2), or insulin (7%, 95% CI −4 to 19). The highest rates of hypoglycemic events in pioglitazone studies were noted where pioglitazone was combined with insulin.165, 218

Weight gain

Twenty-six placebo-controlled trials provided information about weight gain in patients taking pioglitazone or rosiglitazone. A pooled estimate was not calculated for all of those studies to make indirect comparisons because of differences in the methods of measuring the outcome (for example, body mass index, change in weight, or patients gaining >5% of body weight) and limited reporting of results (for example, means were reported without a measure of dispersion). Trials with several doses found increased weight gain associated with higher doses.

Only 4 trials provided sufficient information to calculate a weighted mean difference. The pooled estimates for these trials were very similar for pioglitazone (3.69 kg, 95% CI 2.48 to 4.89)224, 231 and rosiglitazone (3.50 kg, 95% CI 2.25 to 4.75),220, 239 indicating that the drugs cause a similar amount of weight gain. This evidence is consistent with the findings of no difference between the drugs in weight gain reported in head-to-head trials.90, 92, 93

A 2004 meta-analysis112 found similar results in an analysis of 11 trials. Within 6 months of initiating therapy, the average weight gain was 2.7 kg (95% CI 1.8 to 3.7 kg), and drug grouping was not a predictor of heterogeneity (P>0.10).

The range of weight gain reported in active control trials found patients taking pioglitazone or rosiglitazone gained more weight than those taking a sulfonylurea or metformin.

Liver function abnormalities

The first thiazolidinedione approved for use in the United States, troglitazone, was withdrawn from the United States market in 2000 due to concerns about liver damage. Elevations in ALT (>3 times the upper limit of normal) were rare in efficacy trials of pioglitazone and rosiglitazone, with either no cases or reported incidences of less than 1%.

Risk of fracture

Based on data from ADOPT, in February 2007 GlaxoSmithKline issued a safety warning regarding increased risk of fractures associated with use of rosiglitazone. An analysis of these data240 found significantly more female patients who received rosiglitazone experienced fractures than did female patients who received either metformin or glyburide (9.3% compared with 5.1% and 3.5% respectively). The incidence in women was 2.74 per 100 patient-years with rosiglitazone, 1.54 per 100 patient-years with metformin, and 1.29 per 100 patient-years with glyburide. The majority of these fractures were in the upper arm (humerus), hand, or foot. The observed incidence of fractures for male patients in ADOPT was similar among the 3 treatment groups.

At GlaxoSmithKline’s request, an independent safety committee reviewed an interim analysis of fractures in another large ongoing, long-term, controlled rosiglitazone clinical trial, which compared rosiglitazone in combination with either metformin or sulfonylurea to combination therapy with metformin and sulfonylurea. The results of the preliminary analysis were reported to GSK as being consistent with the observations from ADOPT.

Heart failure and other cardiac adverse events

The product label states that rosiglitazone is not indicated in combination with insulin based on an increased incidence of cardiac failure and other cardiovascular adverse events observed in patients on insulin plus rosiglitazone compared with patients using insulin plus placebo.241 Patients who experienced heart failure were on average older, had a longer duration of diabetes, and were for the most part taking rosiglitazone 8 mg daily.

Two placebo-controlled trials of pioglitazone added to insulin reported incidences of congestive heart failure of 12.5%218 and 1%.238

The pioglitazone product label242 cites a 24-week postmarketing study comparing pioglitazone with glyburide in patients with New York Heart Association class II and III heart failure. Over the course of the study, overnight hospitalization for congestive heart failure was reported in 9.9% of patients on pioglitazone compared with 4.7% of patients on glyburide. This adverse event associated with pioglitazone was more marked in patients using insulin at baseline and in patients over 64 years of age. No difference in cardiovascular mortality between the treatment groups was observed.

In the PROACTIVE trial,177 rates of any report of congestive heart failure were increased with pioglitazone compared with placebo (P<0.0001), but rates of fatal heart failure were not different between groups (P=0.634)

Adverse events reported in active-control trials

Overall withdrawals, withdrawals due to adverse events, and specific adverse events reported in active-control trials are shown in Evidence Table 10.

Observational studies of adverse events
Direct evidence comparing pioglitazone with rosiglitazone: Harms
Overview

The previous Drug Effectiveness Review Project TZDs report18 included 12 observational studies that compared adverse events in patients taking pioglitazone with those in patients taking rosiglitazone. Five of these were designed to assess specific adverse events; in the others, adverse events were reported but were not the primary outcome. In this update, we did not include additional observational studies aiming to assess the risk of cardiovascular adverse events or fractures for people taking TZDs because it was felt that sufficient, and stronger, evidence from systematic reviews was already available. We did include 2 additional observational studies in this section of the report for the current update.243, 244 The main results of these studies related to this key question are summarized in Table 65 and Evidence Table 21.

Table 65. Observational studies comparing adverse events associated with thiazolidinediones to adverse events associated with active controls.

Table 65

Observational studies comparing adverse events associated with thiazolidinediones to adverse events associated with active controls.

Lower extremity and pulmonary edema

The prevalence of edema was the primary outcome in a retrospective chart review of 99 patients receiving thiazolidinediones in combination with insulin.245 The prevalence of edema was 12.7% for patients taking rosiglitazone 4 mg and 5.1% in those taking rosiglitazone 8 mg. Among patients taking pioglitazone, there was an increase in edema with increasing dose (1.3% with 15 mg and 6.3% with 30 mg). There was 1 case of pulmonary edema in a patient taking rosiglitazone.

In a retrospective chart review,246 pulmonary edema was noted in 2 patients (1.9%) taking pioglitazone and 3 taking rosiglitazone (3.1%). Four of these had existing congestive heart failure treated with diuretics. Another study247 reported edema in patients with documented heart failure. Fluid retention was seen with the use of both pioglitazone (15.6%) and rosiglitazone (14.3%) across all dosages. Two patients (11%) had physical signs of pulmonary edema, but the study does not report which drug the patients were taking.

Macular edema

The manufacturer of rosiglitazone issued a warning letter in December 2005 regarding post-marketing reports of new onset and worsening diabetic macular edema for patients receiving rosiglitazone.248 The incidence is not reported, but the warning letter states that reports were very rare. In the majority of these cases, the patients also reported concurrent peripheral edema. We identified no reports of macular edema in placebo-controlled trials or observational studies. Abnormal vision was reported in 2.3% of patients in 1 trial of rosiglitazone in combination with sulfonylureas,232 but this was lower than the rate in the placebo group (5.4%).

Heart failure

A retrospective cohort study used claims data to assess the risk of developing heart failure in patients taking pioglitazone (N=1347) or rosiglitazone (1882) for up to 40 months.249 Compared with a control group of patients who did not take thiazolidinediones, the hazard ratio for pioglitazone was 1.92 (95% CI 1.24 to 2.97), and for rosiglitazone 2.27 (95% CI 1.65 to 3.13). There was no significant difference in the risk of developing heart failure between these 2 drugs (P=0.091).

A retrospective database study designed to assess the prevalence of edema found no documentation of new-onset heart failure or exacerbations of existing heart failure in patients initiating thiazolidinediones therapy plus insulin.245 The study authors caution, however, that documentation of heart failure was poor and that the data may be unreliable.

Weight gain

Seven comparative observational studies reported weight gain in follow-up periods ranging from 8 weeks to 1 year (Table 64).244, 246, 250–255 There was no difference in the amount of weight gain in patients taking pioglitazone compared with rosiglitazone in any study.

Table 64. Range of weight gain reported in comparative observational studies.

Table 64

Range of weight gain reported in comparative observational studies.

Evidence comparing pioglitazone or rosiglitazone to active controls: Harms

Ten observational studies reported adverse events associated with thiazolidinediones compared with other active drugs (Table 65, Evidence Table 21).243, 256–264 The adverse events they examined included mortality, coronary heart disease events, heart failure, cancer or adenoma incidence, edema, weight gain and progression to insulin use. Because these studies did not report results separately for pioglitazone and rosiglitazone or they included only 1 of the thiazolidinediones, they do not provide information about the comparative safety of the thiazolidinediones. They do provide information about thiazolidinediones as a class compared with other antidiabetic agents.

In 2 studies, thiazolidinediones were not associated with increased mortality compared with other oral hypoglycemic agents.258, 261 In 1 study, pioglitazone was associated with reduced all-cause mortality compared with other oral antidiabetic medications.243 In older patients with heart failure thiazolidinediones, either alone or combined with metformin, were associated with a lower risk of death over a 15-month period compared with patients not treated with an insulin sensitizer.261

Two studies reported the incidence of coronary heart disease events (myocardial infarction or revascularization) with thiazolidinediones compared with metformin or sulfonylureas. A good-quality study using United States health insurance data found no increased risk of coronary heart disease events in patients initiating thiazolidinedione monotherapy compared with those initiating metformin plus sulfonylurea combination therapy.257 The other found similar risks with rosiglitazone compared with sulfonylureas, metformin, or insulin, either alone or in combination.262 Both studies also found no increased risk in the individual components of the composite outcome with thiazolidinedione use.

Hospital admission for congestive heart failure was the main outcome in a fair-quality cohort study that used data from a Kaiser Permanente diabetes registry.259 Relative to patients initiating therapy with sulfonylureas, patients initiating therapy with thiazolidinediones were no more likely to experience a hospitalization for heart failure after an average of 10.2 months of follow-up. A case-control study based on Oregon Medicaid claims data, in contrast, found a trend suggesting increased risk of hospitalization for heart failure associated with exposure to thiazolidinediones within the previous 60 days.265 Increased risk was also found with exposure to insulin and to the combination of insulin plus thiazolidinediones, but not for other oral antidiabetic agents.

A series of nested case-control studies found no difference in the incidence of breast, colon, or prostate cancer associated with exposure to thiazolidinediones compared with other oral diabetic medications or insulin.260 A case-control study found a slightly higher odds of having an adenoma on first colonoscopy for subjects with type 2 diabetes exposed to TZDs compared with those not exposed to TZDs.263

A study conducted in 500 primary care patients in Germany found fewer patients progressed to insulin therapy when taking pioglitazone than when taking a sulfonylurea.256 However, because this study did not control for confounders and did not clearly report its recruitment strategy and other methods, these results may have a high risk of bias.

The previous Drug Effectiveness Review Project TZDs report identified 43 additional uncontrolled studies of adverse events associated with individual thiazolidinediones.266–303 The results of these studies were consistent with evidence from randomized controlled trials and comparative observational studies. Conclusions that can be drawn from this body of evidence are limited because the studies do not provide information about comparative harms.

III. Fixed-dose Combination Products (FDCPs) or Dual Therapy

Summary of findings for Fixed Dose Combination Products or Dual Therapy: Harms

Harms in children
  • We did not find any evidence meeting inclusion/exclusion criteria for children.
Harms in adults
  • We found no head-to-head trials that compared harms between any 2 FDCPs (insufficient strength of evidence).
  • We found no studies that evaluated long-term harms beyond 15 months for any available FDCP (insufficient strength of evidence).
Avandamet® or dual therapy with metformin plus rosiglitazone
  • Similar rates of withdrawals due to adverse events with Avandamet®/dual therapy groups and monotherapy groups (3 trials ranging from 24 to 32 weeks, low strength of evidence).
  • Similar or slightly higher rates of hypoglycemia with Avandamet®/dual therapy groups compared with monotherapy groups (3 trials ranging from 24 to 32 weeks, low strength of evidence).
  • Similar rates of adverse cardiovascular events with Avandamet®/dual therapy and monotherapy, but duration of studies may not have been sufficient to reliably assess adverse cardiovascular events (3 trials ranging from 24 to 32 weeks, low strength of evidence).
  • Gastrointestinal adverse effects were the most frequently reported adverse events with Avandamet® and dual therapy with metformin plus rosiglitazone. Rates of gastrointestinal adverse effects with Avandamet® or dual therapy were high (28 to 47%), but were the same or slightly lower than those with metformin monotherapy (moderate strength of evidence).
  • Higher rates of edema with Avandamet® or dual therapy than with metformin monotherapy (moderate strength of evidence).
  • In 2 included trials of Avandamet®, subjects receiving Avandamet® reported virtually no change in weight from baseline (0.0 kg to 0.01 kg) compared with slight weight gain with rosiglitazone monotherapy, and slight weight gain (1.9 kg) or weight loss (−2.9 kg) with metformin monotherapy (moderate strength of evidence).
Avandaryl® or dual therapy with rosiglitazone and glimepiride
  • Few definitive conclusions about comparative harms for Avandaryl® or dual therapy with rosiglitazone and glimepiride can be drawn from direct evidence. The 2 included trials were a 28 week trial (N=874) comparing 2 dosages of Avandaryl® with glimepiride monotherapy and rosiglitazone monotherapy, and a 20 week trial (N=40) comparing concurrent use of rosiglitazone and glimepiride with rosiglitazone monotherapy.
  • Rates of hypoglycemia were greater with Avandaryl® or dual therapy than with monotherapy (moderate strength of evidence).
  • Weight gain was slightly greater with Avandaryl® or dual therapy than with monotherapy (moderate strength of evidence).
Actoplus Met® or dual therapy with pioglitazone and metformin
  • Evidence was limited to one large trial (N=600) comparing Actoplus Met® with component monotherapies and a 15 month trial comparing dual therapy with pioglitazone and metformin to monotherapy with either that reported very little harms information.
  • Overall incidence of adverse events, including serious adverse events, was similar across treatment groups (low strength of evidence).
  • Headache was reported more frequently with Actoplus Met® than with either component monotherapy (low strength of evidence).
  • Patients on Actoplus Met® gained less weight than patients on pioglitazone alone but gained more weight than patients on metformin alone (low strength of evidence).
  • Patients on Actoplus Met® experienced lower rates of edema than patients on pioglitazone alone but higher rates of edema than patients on metformin alone (low strength of evidence).
  • Diarrhea, hypoglycemia, and gastrointestinal events were reported most frequently in patients on metformin monotherapy and least frequently in patients on pioglitazone alone, with patients on Actoplus Met® reporting rates in between those for metformin and pioglitazone (low strength of evidence).
Janumet® or dual therapy with sitagliptin and metformin
  • No studies including Janumet® were found that met inclusion criteria. Evidence was limited to 1 trial (N=1,091, with outcomes reported at 24 and 54 weeks) including dual therapy with sitagliptin and metformin.31, 32
  • Gastrointestinal adverse effects were commonly reported (15% to −31% across all treatment arms) and were similar between sitagliptin 100 plus metformin 2000 and metformin 2000 monotherapy at 24 weeks (24.7 compared with 25.3%) and at 54 weeks (29 compared with 31%). Rates were slightly higher for sitagliptin 100 plus metformin 1000 compared with sitagliptin 100 monotherapy or with metformin 1000 monotherapy at 24 weeks (17.9 compared with 15.1 compared with 15.9%, respectively) and at 54 weeks (26 compared with 20 compared with 20%) (low strength of evidence).
  • Weight loss for subjects treated with sitagliptin plus metformin (−0.7 to −1.7 kg) was similar to that for subjects treated with metformin monotherapy (−1.0 to −1.5 kg) (low strength of evidence).
  • The combination of sitagliptin plus metformin resulted in slightly greater improvements in total cholesterol (at 24 weeks: −3.2 to −7.1; at 54 weeks: −6.6 to −8.8 mg/dL) compared with metformin or sitagliptin monotherapy (at 24 weeks: −1.5 to +2.7; at 54 weeks: −0.2 to +0.5 mg/dL) (low strength of evidence).

Detailed assessment for FDCPs and Dual Therapy: Harms

We identified studies that have been conducted specifically using fixed-dose combination tablets comprised of rosiglitazone/metformin (Avandamet®),183, 185, rosiglitazone/glimepiride (Avandaryl®),186 and pioglitazone/metformin (Actoplus Met®).139 Two of these were new since the 2007 Drug Effectiveness Review Project report on FDCPs.139, 183 We found no head-to-head studies comparing FDCPs.

We also included studies using dual therapy of rosiglitazone plus metformin,184 rosiglitazone plus glimepiride,187 pioglitazone plus metformin,188 and sitagliptin plus metformin. 31, 32 All of these were new for this report.

No studies were identified that used the fixed-dose combination tablets comprised of pioglitazone/glimepiride (Duetact®) 189 or sitagliptin/metformin (Janumet ®).190 The safety of Duetact® and Janumet ® have been established basedon trials using the co-administration of their separate components.

More detailed descriptions and summary tables for the studies in this section are provided in the corresponding section of Key Question 1 (Detailed assessment for FDCPs and Dual Therapy) related to efficacy. Details of included studies are found in Tables 37, 39, 41, and 43 and in Evidence Tables 5, 11. Throughout this section, meta-analyses were not performed due to an insufficient number of studies or heterogeneity of study populations, interventions, outcomes, and designs.

Avandamet® and dual therapy with metformin and rosiglitazone

Three randomized controlled trials including either Avandamet® or dual therapy with metformin and rosiglitazone met inclusion criteria. No comparative cohort studies, case-control studies or systematic reviews were identified reporting harms. Table 66 summarizes adverse events of Avandamet® (metformin + rosiglitazone) and rosiglitazone/metformin dual therapy in adults with type 2 diabetes.

Table 66. Adverse events of Avandamet® (metformin + rosiglitazone) and rosiglitazone/metformin dual therapy in adults with type 2 diabetes.

Table 66

Adverse events of Avandamet® (metformin + rosiglitazone) and rosiglitazone/metformin dual therapy in adults with type 2 diabetes.

Head-to-head trials

We found no head-to-head trials of Avandamet® or dual therapy with metformin and rosiglitazone comparing them with other FDCPs that met inclusion criteria.

Trials comparing Avandamet® or dual therapy with component monotherapy

Three fair-quality trials compared Avandamet® or dual therapy with metformin and rosiglitazone to monotherapy with metformin or rosiglitazone. Two trials compared Avandamet® with metformin monotherapy; 1 of them also compared Avandamet® with rosiglitazone monotherapy. The dual therapy trial compared concurrent use of metformin and rosiglitazone with metformin monotherapy. Study duration ranged from 24 to 32 weeks.

Mortality and withdrawals

One death occurred in the dual therapy arm of 1 trial; no other deaths during or shortly after treatment were reported. Rates of withdrawal due to adverse events ranged from 1% to 7.3% for Avandamet®/ dual therapy groups and from 2% to 9.6% in monotherapy groups. In the Avandamet®/ dual therapy groups, rates of withdrawal due to adverse events were consistently slightly numerically lower than or equal to those in the monotherapy arms.

Across trials, 1 rosiglitazone-treated patient and 2 dual therapy patients withdrew due to edema. One patient on metformin was withdrawn due to hypoglycemia. Gastrointestinal symptoms led 11 Avandamet® and 7 metformin-treated patients to withdraw from studies. One metformin and 2 dual therapy patients withdrew due to cardiovascular events; 1 dual therapy patient experienced abnormal liver function values and withdrew.

Only 1 trial reported the total rate of adverse events: 62% of patients on Avandamet® and 59% of those on metformin monotherapy. In the 2 Avandamet® trials, rates of serious adverse events were equivalent between the Avandamet® (3% to 4%) and monotherapy arms (3% to 4%). Other adverse events were mild to moderate in intensity.

Hypoglycemia

In both Avandamet® trials, subjects on Avandamet® reported slightly higher rates of hypoglycemia (7% to 12%) compared with metformin monotherapy (4% to 9%) or rosiglitazone monotherapy (8%). Most hypoglycemic symptoms were reported as mild or moderate and the majority required no intervention or minor dietary intervention. Finger stickglucose values indicated that confirmed hypoglycemia (glucose <50 mg/dL) was rare across arms.

Cardiovascular events

Adverse cardiovascular events were somewhat rare in these trials. In 2 of the 3, patients on Avandamet® or dual therapy reported higher rates of cardiovascular events (2% and 1.3%) compared with those on metformin monotherapy (0% and 0.8%). In the third trial, 1% of patients on Avandamet® reported such events, compared with 3% in each monotherapy arm. Across trials, cardiac ischemia occurred in 5 patients on Avandamet®, 5 patients on dual therapy, 2 patients on rosiglitazone, and 5 patients on metformin.

Gastrointestinal events

In all 3 studies, gastrointestinal events were the most commonly reported across treatment groups, ranging from 28% to 51%. Rates with Avandamet® or dual therapy were the same or slightly lower than those with metformin monotherapy. More patients on Avandamet® reported nausea/vomiting (16%) and dyspepsia (10%), compared with patients on metformin (13% and 8%, respectively) or rosiglitazone (8% and 9%). Metformin was associated with the highest incidence of diarrhea (21% compared with 14% with Avandamet® and 7% with rosiglitazone).

Edema

In 2 of the 3 trials, patients on Avandamet® or dual therapy reported higher rates of edema (2% and 4.7%) compared with those on metformin monotherapy (0% and 1.3%). In the third trial, 7% of patients on rosiglitazone monotherapy reported edema, compared with 6% on Avandamet® and 3% on metformin monotherapy.

Weight change

In the 2 Avandamet® trials, patients receiving Avandamet ® reported virtually no change in weight from baseline (0.0 kg to 0.01 kg). Rosiglitazone monotherapy was associated with a slight weight gain, and metformin monotherapy was associated with slight weight gain in 1 trial (1.9 kg) and weight loss (2.9 kg) in the other. In the dual therapy trial, patients on metformin lost a mean 1.78 kg and patients on dual therapy gained 1.79 kg from baseline.

Total cholesterol

Across all 3 trials, metformin monotherapy was consistently associated with a decrease in total cholesterol. Avandamet® had mixed results for total cholesterol; 1 trial reported a slight decrease while the other reported an increase.

Other adverse events

Headache was reported in 1 trial, with incidence being roughly equivalent across treatment arms. Of the patients in the dual therapy arm of that trial, 1.6% reported anemia, compared with no such reports by patients on metformin monotherapy.

Avandaryl® and dual therapy with rosiglitazone and glimepiride

Two randomized controlled trials including either Avandaryl® or dual therapy with rosiglitazone and glimepiride met inclusion criteria. No comparative cohort studies, case-control studies or systematic reviews were identified reporting long-term benefits.

Head-to-head trials

We found no head-to-head trials of Avandaryl® or dual therapy with rosiglitazone and glimepiride comparing them with other FDCPs that met inclusion criteria.

Trials comparing Avandaryl® or dual therapy with rosiglitazone and glimepiride

Two fair- or good-quality trials (2 articles) compared Avandaryl® or dual therapy with rosiglitazone and glimepiride to active treatment arms. Study durations were 28186 and 20 weeks.187

One good-quality trial (N=874) compared 2 dosages of Avandaryl® with glimepiride monotherapy and with rosiglitazone monotherapy.186 One fair-quality dual therapy trial (N=40) compared concurrent use of rosiglitazone and glimepiride with rosiglitazone monotherapy.187 Table 67 summarizes adverse effects of Avandaryl® (rosiglitazone + glimepiride) and rosiglitazone/glimepiride dual therapy in adults with type 2 diabetes.

Table 67. Adverse effects of Avandaryl® (rosiglitazone + glimepiride) and rosiglitazone/glimepiride dual therapy in adults with type 2 diabetes.

Table 67

Adverse effects of Avandaryl® (rosiglitazone + glimepiride) and rosiglitazone/glimepiride dual therapy in adults with type 2 diabetes.

Mortality and withdrawals

No deaths occurred in either trial. Rates of withdrawal in the Avandaryl® trial due to adverse events ranged from 1.4% (Avandaryl® 8 mg/4 mg) to 4.3% (rosiglitazone). In the 2 Avandaryl® arms, a total of 10 patients withdrew due to adverse events: 7 on the lower dose and 3 on the higher dose. Six patients on glimepiride and 10 patients on rosiglitazone withdrew due to adverse events. Roughly 50% of patients in each arm of the Avandaryl® trial reported at least 1 adverse event (excluding hypoglycemia), with the majority being mild to moderate. No serious adverse events were noted in the dual therapy trial, and the overall number of events was not reported.

Hypoglycemia

In the Avandaryl® (4 mg/4 mg) trial, over 19% of patients across arms reported a hypoglycemic episode. Of patients receiving Avandaryl®, 29% (4 mg/4 mg) and 22.5% (8 mg/4 mg) reported hypoglycemia. Similarly, 21.6% of patients on glimepiride reported such symptoms. Only 5.2% of those in the rosiglitazone monotherapy arm reported them. After finger stickglucose was tested, between 3.6% and 5.5% of patients on Avandaryl or glimepiride had confirmed hypoglycemia (blood glucose <50 mg/dL), compared with 0.4% of patients in the rosiglitazone arm.

In the small dual therapy trial, patients receiving the 2 drugs reported a total of 59 episodes of hypoglycemia; rosiglitazone monotherapy-treated patients reported only 4 episodes in total.

Cardiovascular events

Two patients in the Avandaryl® trial reported congestive heart failure: 1 in the glimepiride group and 1 in the higher-dose Avandaryl®. No cardiovascular events were reported in the dual therapy study.

Gastrointestinal events

None were reported in either trial.

Edema

Edema reports were fairly consistent across arms of the Avandaryl® trial and ranged from 2.3% (glimepiride group) to 3.2% (higher-dose Avandaryl). No episodes of edema were reported in the dual therapy study.

Weight change

Patients in all arms of the Avandaryl® trial gained weight. Glimepiride-treated patients gained a mean 1.1 kg; patients on rosiglitazone gained 1.0 kg. There appeared to be a dose-repose association between the 2 Avandaryl® arms: patients on lower-dose Avandaryl® gained 2.0 kg, and those on the higher dose gained 3.4 kg. 1 lower-dose Avandaryl® patient withdrew due to weight gain. In the dual therapy v rosiglitazone trial, all patients gained weight: 5.1 kg and 2.4 kg, respectively, with the difference between arms being statistically insignificant.

Total cholesterol

In the Avandaryl trial, mean total cholesterol increase was significant in the rosiglitazone and Avandaryl arms. The largest increase was in the rosiglitazone arm (21.8 mg/dL); more modest but still significant increases were found in the lower- and higher-dose Avandaryl® arms (8.7 mg/dL and 14.4 mg/dL, respectively). There was no significant difference in cholesterol levels between dual therapy and rosiglitazone.

Other adverse events

Headache and nasopharyngitis were reported in roughly 4% of patients in each arm of the Avandaryl® trial. 1.8% of patients in the higher-dose Avandaryl® arm and 2.2% in the rosiglitazone arm experienced anemia, compared with <1% in the glimepiride and lower-dose Avandaryl® arms.

Actoplus Met® and dual therapy with pioglitazone and metformin

We found one active-control trials of Actoplus Met® compared with monotherapies of its components that met inclusion criteria. This 24-week RCT (N=600) compared Actoplus Met® (30 mg/1,700 mg daily) with pioglitazone alone (30 mg daily) and metformin alone (1,700 mg daily).139 Overall incidences of adverse events were similar across treatment arms: 50.7%, 52.1% and 53.1% for the Actoplus Met®, pioglitazone monotherapy, and metformin monotherapy arms, respectively. Reports of severe adverse events were also similarly distributed among the arms: 1.0% for Actoplus Met®, 1.6% for pioglitazone monotherapy, and 1.4% for metformin monotherapy.

A 15 month trial (N=271) compared dual therapy with pioglitazone and metformin to monotherapy with each component (3 month run-in/titration phase, 12 month full-dose treatment and follow-up phase). Very little harms information was reported in this trial.188 Table 68 summarizes adverse effects of Avandaryl® (rosiglitazone + glimepiride) and rosiglitazone/glimepiride dual therapy in adults with type 2 diabetes.

Table 68. Adverse events of Actoplus Met® or pioglitazone plus metformin dual therapy in adults with type 2 diabetes.

Table 68

Adverse events of Actoplus Met® or pioglitazone plus metformin dual therapy in adults with type 2 diabetes.

Mortality and withdrawals

Fewer withdrawals from the FDCP study due to adverse events occurred in the Actoplus Met® and pioglitazone alone arms compared with the metformin alone arm (3.0%, 3.2%, and 4.8%, respectively). There were no deaths during the FDCP trial.

Hypoglycemia

In the FDCP trial, rates of hypoglycemia (defined as fasting plasma glucose <60 mg/dL) were low in all treatment arms: 1.0%, 0.5%, and 1.4% in the Actoplus Met®, pioglitazone alone, and metformin alone arms, respectively. In the dual therapy study, 3 patients in the dual therapy arm withdrew due to hypoglycemia (defined as above).

Cardiovascular events

There were no episodes of congestive heart failure during the FDCP trial. Three patients in the monotherapy arms (2 on pioglitazone and 1 on metformin) showed clinically significant worsening ECG results from baseline to end of follow-up. One pioglitazone patient was found to have coronary artery disease and myocardial infarction; the other was diagnosed with arterial branch block. The metformin patient was determined to have myocardial ischemia.

Gastrointestinal events

Diarrhea and gastrointestinal events were reported less frequently in patients taking Actoplus Met® (9.0% and 17.9%, respectively) compared with patients on metformin alone (15.3% and 25.8%, respectively) but more often than in patients on pioglitazone alone (2.6% and 105%,m respectively). In the dual therapy trial, 5 patients in the metformin arm withdrew due to gastrointestinal events.

Edema

Incidence of peripheral edema in the FDCP trial was highest for pioglitazone alone (4.2%), lowest for metformin alone (1.4%) and in between for patients on Actoplus Met® (3.0%).

Weight change

In the FDCP trial, patients on Actoplus Met® reported less weight gain (0.69 kg)from baseline than patients on pioglitazone alone (1.64 kg) but greater weight gain than patients on metformin alone (−1.28 kg). In the dual therapy study, 3 pioglitazone monotherapy patients withdrew from the study due to excessive weight gain. While weight change itself was not reported for this trial, dual therapy and metformin were associated with significant decreases in body mass index, compared to an increase in the pioglitazone group.

Total cholesterol

Neither the FDCP trial nor the dual therapy trial reported outcomes related to cholesterol.

Other adverse events

In the FDCP trial, headache was reported more frequently with than Actoplus Met ® (5.5%) than with pioglitazone (2.6%) or metformin (4.8%) monotherapy. In the same trial, bone fractures occurred in 1 metformin monotherapy patient (traffic accident) and 1 patient on Actoplus Met® (unspecified cause).

Janumet® and dual therapy with sitagliptin and metformin

No studies including Janumet® were found that met inclusion criteria. One randomized controlled trial including dual therapy with sitagliptin and metformin met inclusion criteria. This trial resulted in 3 publications; one reporting results after 24 weeks31 (N=1,091), one reporting results after 54 weeks,32 and the other after a total of 104 weeks33 No comparative cohort studies, case-control studies or systematic reviews were identified reporting long-term benefits.

Table 69 summarizes adverse events of metformin/sitagliptin dual therapy in adults with type 2 diabetes. Incidences of adverse events were generally similar between treatment arms over the 104-week study period. One patient in the higher-dose dual therapy group died of an electrical shock during the continuation phase, and 1 patient withdrew from the lower-dose metformin monotherapy arm due to esophageal carcinoma and died during the study period. Withdrawals due to adverse events were low (1.1% to 2.8% during the first 24 weeks and 2% to 4% during the entire study period) and were similar across treatment arms. There was slightly more variation in the incidence of severe adverse events between groups. At 24 weeks, fewer patients in the higher-dose dual therapy arm reported serious events (0.5%) than patients receiving sitagliptin monotherapy (5.0%), lower-dose metformin monotherapy (2.2%), higher-dose metformin monotherapy (1.1%) and lower-dose dual therapy (3.2%). After 104 weeks, sitagliptin monotherapy was associated with the highest rate of serious adverse events (7.3%), and lower-dose metformin monotherapy with the lowest rate (3.8%). Incidence of severe events in both dual therapy arms was 6% and 6.3% at the end of the study.(Table 69)

Table 69. Adverse events of metformin/sitagliptin dual therapy in adults with type 2 diabetes.

Table 69

Adverse events of metformin/sitagliptin dual therapy in adults with type 2 diabetes.

After 104 weeks, fewer sitagliptin monotherapy patients reported adverse events (60.3%), compared with the other treatment arms, including a high of 75.3% in the higher-dose sitagliptin-metformin combination arm. Seventy-one percent of patient in the lower-dose combination arm reported adverse effects.

Hypoglycemic events were rare across treatment groups at 24, 54 and 104 weeks and were of mild or moderate severity. At both points of measurement, more higher-dose dual therapy patients reported hypoglycemia (2.2%, 3%, and 4.9% at 24, 54 and 104 weeks, respectively). After 104 weeks, rates of hypoglycemia across the 3 monotherapy arms ranged from 1.1 to 2.2% with the lowest reported in the sitagliptin monotherapy arm

Gastrointestinal events were reported with similar frequency across treatment arms, with the higher-dose metformin monotherapy patients reporting the highest rates at both measurement points (25.3% at 24 weeks, 31% at 54 weeks, and 33% at 104 weeks). Patients in either dual therapy arm reported adverse gastrointestinal events more frequently than patients on sitagliptin monotherapy or lower-dose metformin. Nausea/vomiting and abdominal pain were reported most frequently in the higher-dose metformin monotherapy group, and diarrhea was reported most frequently by higher-dose dual therapy patients.

After 104 weeks, there was no change in weight from baseline for the sitagliptin monotherapy patients. Body weight decreased by small but statistically significant amounts in the other arms, ranging from 0.7 kg in the lower-dose dual therapy group to 1.7 kg in the higher-dose dual therapy arm.

The only arm in which total cholesterol changed significantly from baseline was higher-dose dual therapy; total cholesterol decreased by 3.0 mg/dL on average.

Key Question 3. Are there subgroups of patients based on demographics (age, racial groups, gender), comorbidities (drug-disease interactions, obesity), or other medications (drug-drug interactions) for which newer diabetes medications differ in efficacy/effectiveness or frequency of adverse events?

I. Newer Drugs for the Treatment of Diabetes Mellitus: Amylin Agonists, DPP-4 Inhibitors, and GLP-1 agonists

Summary of Findings for Newer Drugs

  • We found insufficient evidence to draw any firm conclusions about whether there are subgroups of patients based on demographics, comorbidities, or other medications for which newer diabetes medications differ in efficacy/effectiveness or frequency of adverse events.
  • The evidence that was found is generally hypothesis-generating, using post hoc pooled analyses or post hoc subgroup analyses in an exploratory manner.

Detailed Assessment for Newer Drugs

Pramlintide for type 1 diabetes

There was insufficient evidence to perform subgroup analyses based on age, sex, race, ethnicity, or baseline HbA1c in individual studies.

One randomized controlled trial conducted subgroup analyses that were not all prespecified, and 1 post hoc pooled-analyses was identified.21, 304 Results from these hypothesis-generating analyses should be used with caution. Further prospective research with larger sample sizes will need to be conducted to verify these findings.

Baseline body mass index

Pramlintide appeared to inhibit weight gain in patients with baseline body mass index ≤ 23 kg/m2 while producing mild weight loss for patients with body mass index > 23 kg/m2 (baseline to week 26).21 Data at 52-week follow-up were not reported.

Baseline HbA1c < 8%

Data from 3 studies that included patients with baseline HbA1c between 7% and 8.5% receiving pramlintide 30 mcg or 60 mcg were pooled and reported in a separate publication.304 Two of the 3 studies were identified and included in our review.19, 21 The third study was in abstract form and was excluded. The pooled publication reported results up to 26 weeks. In this subgroup, the pooled change in HbA1c was −0.3% and the change in weight was −1.6 kg (both placebo-corrected; both P<0.0009). There was no overall increased risk in hypoglycemia. The improvement in HbA1c in this pooled subgroup analysis was similar to the change in HbA1c noted for all subjects (across a range of HbA1c) in the original studies. Thus, it appears that patients with good but not optimal baseline HbA1c of 7% to 8.5% experienced similar degrees of HbA1c reduction as the populations included in the original trials, with no increased risk of hypoglycemia at 26 weeks.

Pramlintide for type 2 diabetes
Age, sex, total daily insulin dose, and prior use of oral hypoglycemic agents

None of the randomized controlled trials conducted subgroup analyses evaluating whether pramlintide had differential effects in these populations.

Race and ethnicity

A post hoc analysis305 of two 52-week trials25, 26 pooled subjects of various ethnic groups. Black and Hispanic patients tended to have higher baseline HbA1c (9.2% to 9.7%) than white patients (8.9% to 9.1%). Pramlintide produced larger reductions in HbA1c and weight from baseline in black patients (0.7%, 4.1 kg) than white patients (0.5%, 2.4 kg) and Hispanic patients (0.3%, 2.3 kg). Changes in total daily insulin requirement and baseline oral hyperglycemic use were not different among the different races and ethnicities.

Nausea and weight loss and effects of weight on HbA1c

Weight loss experienced with pramlintide 90 or 120 mcg appeared to be independent of nausea, as weight loss was similar in patients never experiencing nausea (90 or 150 mcg, −1.1 to −1.5 kg) and patients experiencing nausea at any time (90 or 150 mcg, −0.3 to −2.0 kg).25 In addition, improvements in HbA1c observed with pramlintide appeared to be independent of weight lost or gained during the trial (subjects who gained weight, change in HbA1c −0.29% to −0.53%; subjects who lost weight, change in HbA1c −0.22% to −0.58%).

A pooled analysis306 of overweight and obese patients also evaluated whether weight loss associated with pramlintide 120 mcg was influenced by nausea. Like the other, this post hoc subgroup analysis suggested that weight loss was independent of nausea (change in weight in group reporting “never nausea,” −1.3 kg; “nausea at any time,” −1.9 kg). None of the studies explored to see if there were any correlations between anorexia and weight loss.

Overweight and obese patients

A post hoc analysis306 pooled data from 2 randomized controlled trials comparing pramlintide 120 mcg with placebo when both were added to insulin. At 26-week follow-up overweight and obese (body mass index > 25 kg/m2) patients receiving pramlintide showed greater reductions in HbA1c and weight than similar patients receiving placebo. Approximately 2% of overweight and obese patients on pramlintide plus insulin achieved weight loss of ≥10% change from baseline compared with 0% in those on placebo plus insulin. Markedly obese patients (baseline body mass index 35–40 kg/m2 and >40 kg/m2) had the greatest weight loss (−2.4 kg and −3.2 kg, respectively).

Baseline HbA1c

When patients were stratified by baseline HbA1c,24 at 16 weeks patients with baseline HbA1c > 8.5% who received pramlintide plus insulin glargine showed larger improvements in HbA1c, fasting plasma glucose, and postprandial glucose than patients receiving placebo plus glargine (pramlintide change in HbA1c −1.19%, fasting plasma glucose −44.4 mg/dL, postprandial glucose −23 mg/dL, and weight −1.0 kg compared with placebo plus glargine HbA1c −0.69%, fasting plasma glucose −18.4 mg/dL, postprandial glucose +3.2 mg/dL, weight +1.1 kg). Among subjects with lower baseline HbA1c (≤ 8.5%), improvements in HbA1c (−0.36%) and weight ( 2.0 kg) were also larger in pramlintide-treated patients than those who took placebo plus glargine. Overall, reductions in HbA1c were greatest in those with baseline HbA1c >8.5%.

Another post hoc analysis307 pooled data from 2 trials at 26-week follow-up and examined patients with baseline HbA1c of 7.0% to 8.5%. Pramlintide plus insulin was better than placebo plus insulin for HbA1c (placebo-corrected change in HbA1c −0.43, P<0.0009) and weight (placebo-corrected change in weight −2.0 kg, P<0.0003).

Sitagliptin

There was insufficient evidence to perform subgroup analyses based on age, sex, race, ethnicity, baseline HbA1c, or other characteristics at the study level. Subgroup data not available in publications were supplemented by data provided by the manufacturer. The results from this section should be considered with caution until larger prospective trials evaluating these populations verify the findings.

Age, sex, race, body mass index, and prior use of oral hypoglycemic agents

Four published trials36, 43, 44, 46, 50, 51 reported no significant differences in changes in HbA1c based on subgroups defined by age, sex, race, and body mass index. Data on file from 3 additional trials 42, 48, 49, 308 also showed similar findings.

Data on file from 2 trials 45, 47, 308 showed a significant interaction between treatment effect and race for those on sitagliptin monotherapy and placebo. In one study,47 Hispanic patients experienced the largest decline in HbA1c (placebo-corrected difference in HbA1c from baseline: −1.04%, 95% CI −1.38 to −0.70) followed by White patients (placebo-corrected difference: in HbA1c from baseline: −0.69%, 95% CI −0.84 to 0.55), and Other patients (placebo-corrected difference in HbA1c from baseline: −0.44%, 95% CI, −0.82 to −0.07). In another study,45 the greatest HbA1c reduction was seen in Indian (placebo corrected change −1.4%, 95%CI −1.7 to −1.0) and Korean patients (placebo corrected change −1.4%, 95% CI −1.9 to −0.8).

Of 7 studies 30, 31, 44–47, 49, 308 that stratified groups by prior oral hypoglycemic agent use, only 1 trial 31, 308 showed a large numerical difference in treatment effect. Patients who were not taking an oral hypoglycemic agent prior to this trial experienced greater decline in HbA1c across all treatment arms compared with patients who were using oral agents before enrolling into the study. For instance, the change in HbA1c from baseline for “no prior oral agent use” for sitagliptin compared with placebo was −1.11% compared with −0.13% compared with −0.26% compared with +0.52% for those “treated with prior oral agents.” Between-group difference calculations were not conducted.

Baseline HbA1c

Subgroup information stratified by baseline HbA1c was found in 14 trials. Some data were available from published studies31, 34, 36, 38, 42–46, 48–51 and additional information from 6 trials (Scott 2007, Charbonnel 2006, Nauck, 2006, Scott 2008, Vilsboll 2010, and Aschner 2010) were obtained from data on file.38, 45, 51, 308

Four trials 43, 44, 47, 49, 51 found no significant differences in the change in baseline HbA1c among those in the following subgroups: <7.5%, <8%, 8–8.9%, >7.5%, ≥8.5%, and ≥9%. One trial42 showed significant interaction (P<0.001) in the change in HbA1c stratified by baseline HbA1c <8% and ≥9%. In patients with baseline HbA1c ≥9%, placebo-corrected reductions of −1.52% were observed for sitagliptin 100 mg/d compared with about −0.6% decrease in those with baseline HbA1c <8%. Data from Goldstein, et al.308 also showed consistent findings for sitagliptin 100 mg/d compared with placebo. For this study, interaction analyses were not conducted (change in HbA1c from baseline for those with HbA1c <8%: sitagliptin, −0.37% compared with placebo, +0.15% compared with those with HbA1c ≥9%: sitagliptin, −0.88% compared with placebo, +0.08%). In another trial, greater reduction was seen with sitagliptin 50mg twice daily (placebo corrected change −1.15%, 95% CI −2.27 to −0.03) and sitagliptin 100mg daily (placebo corrected change −1.18%, 95% CI −2.26 to −0.09 for patients with HbA1c >9% at baseline. Similarly, in patients with HbA1c ≥10% there HbA1c reduction than in the entire cohort with average HbA1c of 8.7%(−1.4% as compared to −1.0%).45 Compared to metformin, the greatest HbA1c reduction was see in patients with baseline HbA1c ≥8%, however there was no difference between sitagliptin and metformin (0.12%, 95% CI −0.07 to 0.3).38

Duration of diabetes

Two trials 43, 51 reported a potential interaction between median baseline duration of diabetes and HbA1c effects in patients randomized to sitagliptin 100 mg compared with placebo. One trial 43 reported patients with diabetes of ≤ 3 years’ duration had significantly greater reductions in HbA1c than patients who had diabetes for > 3 years (placebo-corrected mean change HbA1c for ≤ 3years −0.90%, 95% CI −1.21 to −0.60 compared with mean change HbA1c for > 3years −0.28%, 95% CI −0.59 to +0.20). Another trial 51 reported greater HbA1c reduction for patients having diabetes for >12 years as compared to <12 years (−0.64% compared to −0.5%).

Saxagliptin
Age, sex, race, body mass index, and prior use of oral hypoglycemic agents

One published study reported consistent HbA1c reduction across subgroups defined by age, sex, race, and body mass index.55

Baseline HbA1c

Three published studies reported results stratified by baseline HbA1c.54–56 All three studies reported statistically significant interaction for treatment effect with higher baseline HbA1c. Two studies did not report the HbA1c cut off that was statistically significant,55, 56 however one study reported the HbA1c reduction for saxagliptin groups for baseline HbA1c ≥9% ranged from −0.84 to −1.25% compared to +0.13% for placebo (P<0.05).54

Duration of diabetes

One published study reported HbA1c lowering effects to be consistent across a subgroup of patients defined by duration of diabetes, however the duration was not specified.55

Exenatide

Two publications examined subgroups based on demographic characteristics. A pooled analysis78 of 3 placebo-controlled trials reported that reductions in HbA1c were not related to age and that hypoglycemia was not more frequent in subjects ≥ 65 years of age. No primary study examined the efficacy or effectiveness of exenatide in subgroups defined by age or other characteristics.

Another study aiming to evaluate the efficacy and safety of sitagliptin as an add-on to metformin therapy (compared with adding placebo to metformin) in patients with moderately severe (HbA1c ≥ 8.0% and ≤ 11.0%) type 2 diabetes mellitus (T2DM) randomized 190 subjects for 30 weeks of treatment.50 Post hoc subgroup analyses were conducted for change in HbA1c for the following groups: differences by age (≤55 compared with >55), body mass index (≤30.1 compared with >30.1), sex, duration of diabetes, (≤6 years compared with >6), and previous metformin or metformin-based combination therapy. The study found that treatment effects were consistent across subgroups.

Liraglutide

We found no studies of liraglutide meeting inclusion criteria that examined differences in efficacy/effectiveness or adverse events for subgroups.

II. Thiazolidinediones (TZDs)

Summary of Findings for Thiazolidinediones (TZDs)

  • We found insufficient evidence to draw any firm conclusions about whether there are subgroups of patients based on most demographic characteristics, comorbidities, or other medications for which newer diabetes medications differ in efficacy/effectiveness or frequency of adverse events.
  • The evidence that was found is generally hypothesis-generating, using post hoc pooled analyses or post hoc subgroup analyses in an exploratory manner.
  • Some studies reported that the risk of fractures is increased with TZD use in women, but not in men.204, 309 On analysis of data from ADOPT found hazard ratios comparing rosiglitazone with metformin and glyburide were 1.81 (95% CI 1.17 to 2.80) and 2.13 (1.30 to 3.51), respectively.309 A systematic review and meta-analysis reported an increased risk among women (OR 2.23, 95% CI 1.65 to 3.10), but not in men (OR 1.00, 95% CI 0.73, 1.39).204

Detailed Assessment for Thiazolidinediones (TZDs)

Studies examining subgroups based on demographic characteristics or comorbidities are summarized in Table 70. Most studies were conducted in the United States or in Western Europe and examined white populations. Some studies included minority populations but did not present subgroup analyses on these populations.310 Thus, there are very limited data on the comparative effectiveness of pioglitazone and rosiglitazone among persons with various demographic characteristics and no conclusions can be drawn as to which drug is more efficacious or effective, or associated with fewer side effects in population subgroups.

Table 70. Studies examining subgroups based on demographic characteristics or comorbidities.

Table 70

Studies examining subgroups based on demographic characteristics or comorbidities.

Most of the studies identified in this review examined persons with type 2 diabetes without significant comorbidities such as coronary heart disease, heart failure, or renal insufficiency. Thus there is a paucity of data on the interaction of TZDs and micro- and macrovascular diseases that are highly prevalent among persons with diabetes, and no conclusions can be drawn on the comparative effectiveness of the 2 drugs under review among populations with significant comorbidities.

Subgroups based on demographic characteristics

Kreider and colleagues311 pooled the results of 8 randomized controlled trials examining monotherapy with rosiglitazone and examined subgroups of age less than and greater than 70 years. They found no differences between the 2 age groups for HbA1c and found rosiglitazone well tolerated in both age groups. The percentage of persons with at least 1 adverse event was comparable between the rosiglitazone and placebo groups, and between persons older and younger than 70 years. The incidence of anemia was higher in older patients taking rosiglitazone than in younger patients taking the drug and treatment patients had higher rates of anemia than patients in the placebo group. Weight gain was higher in the under-seventy group (2.14 kg) than the over-seventy group (1.66 kg) and the placebo groups (<70 years, −0.41 kg; >70 years, −1.34 kg).

Several studies examined racial or ethnic minorities. King compared Mexican Americans with non-Hispanic persons in a retrospective cohort study and found that HbA1c and weight changed to a similar degree in both populations. Jun and colleagues312 examined 100% Hispanics, and pioglitazone produced a decrease in HbA1c of 2.0% at 6 months. Twelve Chinese persons with nephropathy and type 2 diabetes were exposed to rosiglitazone over 15.5 months with improved HbA1c, a nonsignificant increase in weight, and no adverse events.313 Pioglitazone was as effective as glimepiride among 244 Mexican patients.314

Barnett and colleages235 examined the use of rosiglitazone in an Indian and Pakistani population in the United Kingdom and noted results and adverse events comparable to other placebo-controlled trials discussed above. Vongthavaravat et al.315 examined a mixed Asian and white population and their results were also consistent with findings in largely white populations in other studies of rosiglitazone.

In the updated Drug Effectiveness Review Project TZDs report (2008), several additional studies of rosiglitazone provided data on subgroups based on demographic data.143, 147, 170, 176 In a combination therapy, double-blind study (N=365) both groups received combination tablets of glyburide/metformin. The addition of rosiglitazone achieved greater reduction in HbA1c than the addition of placebo (between-group difference −1.0%, P<0.001). An improvement in HbA1c was demonstrated across age, sex, and racial subgroups.170

In a study of older adults with type 2 diabetes,176 HbA1c improved with rosiglitazone plus glipizide 10 mg twice a day compared with glipizide alone at 2-year follow-up (between-group change in HbA1c −0.79%, P<0.0001).

In a double-blind study (N=318) in subjects who had failed to achieve adequate control on metformin,143 metformin 1000 mg/glibenclamide 5 mg was compared with metformin 1500–2000 mg plus rosiglitazone 4 mg daily. Reduction in HbA1c was greater in the glibenclamide group at 24 weeks follow-up as noted above. This larger decrease in HbA1c occurred in the glibenclamide group across strata defined by sex, race, age, baseline HbA1c, or entry metformin dose.

In ADOPT,147 rosiglitazone was more effective than glyburide in all subgroups for the primary outcome of monotherapy failure: age ≤ 50 years, between 50 and 59 years, and ≥ 60 years; males and females; body mass index ≤ 30 kg/m2, between 30 and 35 kg/m2, and ≥35 kg/m2; baseline fasting plasma glucose ≤ 140 mg/dL and > 140 mg/dL; and waist circumference ≤ 99 cm, >99 – 110 cm, and > 110 cm.

An analysis using data from 1,840 women and 2,511 men randomly assigned in ADOPT to rosiglitazone, metformin, or glyburide examined time to first fracture, rates of occurrence, and sites of fractures.309 In men, fracture rates did not differ between treatment groups. In women, the study identified an increased risk of fractures with rosiglitazone. The cumulative incidence of fractures at 5 years was 15.1% (95% CI 11.2 to 19.1) with rosiglitazone, 7.3% (4.4 to 10.1) with metformin, and 7.7% (3.7 to 11.7) with glyburide. Thus, in women, the hazard ratios comparing rosiglitazone with metformin and glyburide were 1.81 (1.17 to 2.80) and 2.13 (1.30 to 3.51), respectively.

A systematic review and meta-analysis of 10 randomized controlled trials and 2 observational studies found similar results, concluding that long-term TZD use doubles the risk of fractures among women with type 2 diabetes, without significant increase in risk of fractures among men with type 2 diabetes.204 The risk of fractures overall in the 10 randomized controlled trials was increased with TZDs (odds ratio 1.45, 95% CI 1.18 to 1.79). Five randomized controlled trials showed an increased risk among women (odds ratio 2.23, 95% CI 1.65 to 3.10), but not in men (odds ratio 1.00, 95% CI 0.73, 1.39).

Comorbidities and other population characteristics

Patients with impaired renal function were examined in several studies. Agrawal and colleagues233 examined patients with renal impairment (creatinine clearance 30–80 mL/min) and found that rosiglitazone had similar effects on HbA1c in patients with and without renal impairment. In a retrospective chart review316 of patients on dialysis with end stage renal disease, rosiglitazone was associated with weight gain and a decrease in hematocrit at 3-month follow-up compared with pioglitazone. Data for pioglitazone, however, were not presented, limiting conclusions that can be drawn.

In a fair-quality study pooling 2 randomized controlled trials that compared rosiglitazone plus metformin combined therapy with metformin monotherapy, Jones and colleagues317 examined subgroups with body mass index < 25 kg/m2, 25–30 kg/m2, and >30 kg/m2. They noted greater improvement in HbA1c with rosiglitazone 4 or 8 mg daily plus metformin than with metformin monotherapy (P=0.025). Safety profiles were similar in all 3 subgroups. Weight gain was noted in the obese group (body mass index > 30 kg/m2) receiving metformin plus rosiglitazone (2.5 kg), while weight loss of 0.9 kg was found in obese patients on metformin alone. Weight change was not reported for the other body mass index subgroups.

Patients with diagnosed coronary artery disease were examined in 3 studies which were described above in Key Question 2, as these were the only studies that reported cardiovascular outcomes. Wang and colleagues168 examined 70 Chinese with coronary artery disease and type 2 diabetes and noted significant improvement in HbA1c with rosiglitazone with change in weight similar to the no-treatment control group. The primary and composite endpoint of coronary events (including death) was significantly decreased in the rosiglitazone group (P value reported as both <0.05 and <0.01). Wang and colleagues230 also examined Chinese persons with metabolic syndrome and found that fasting plasma glucose did not improve significantly in either the rosiglitazone or the placebo group (HbA1c was not presented).

In a poor-quality study, Choi and colleagues178 compared treatment with rosiglitazone plus conventional antidiabetic therapy among patients undergoing coronary catheterization to conventional treatment. At 6-month follow-up there were no significant differences in glycemic control or lipid concentrations between the 2 groups. The rate of restenosis and the stenosis diameter were less in the rosiglitazone group (between-group P=0.03).

Thirty-one postmenopausal women were examined in a poor-quality, placebo-controlled trial of rosiglitazone 4 mg daily.166 Results were similar to other placebo-controlled trials and no adverse events were reported.

No studies explicitly examined populations with a history of hypoglycemic episodes. Nor were studies identified that examined the effect of concomitant medications on the comparative effectiveness of pioglitazone and rosiglitazone. Most studies permitted the use of a variety of antihypertensive, cardiac, and cholesterol-lowering medications among participants. Subgroup or other stratified analyses were not performed to allow examination of drug-drug interactions with the thiazolidinediones.

In the updated Drug Effectiveness Review Project TZDs report, new data on the use of thiazolidinediones in persons with comorbidities was identified, particularly with cardiovascular disease. Since the publication of the large PROACTIVE study177 (discussed above) which compared pioglitazone with placebo, several additional subgroup analyses have been published, including of subjects with prior myocardial infarction318 or stroke.319 In the subgroup of patients with a previous myocardial infarction at baseline318 (N=2445) pioglitazone had a significant beneficial effect on fatal and nonfatal myocardial infarction (28% risk reduction, P=0.045) and acute coronary syndrome (37% risk reduction, P=0.035). There were no significant differences between groups for cardiovascular death or nonfatal myocardial infarction, or stroke, although event rates in the pioglitazone group were consistently lower than with placebo. Rates of heart failure requiring hospitalization or fatal heart failure were not significantly different between the pioglitazone and placebo groups, but heart failure occurred in a greater proportion of patients in the myocardial infarction subgroup (11.6%) than in subjects without prior myocardial infarction (7.0%, P<0.0001). The change in HbA1c was −0.8% (interquartile range −1.6% to −0.1%) in the pioglitazone group and −0.4% (interquartile range −1.1% to 0.3%) in the placebo group (between-group P<0.0001).

In another prespecified subgroup analysis of the PROACTIVE trial, pioglitazone was examined in subjects with (N=984) and without (N=4254) a prior stroke.319 In subjects with prior stroke, there was a trend (although not statistically significant) towards benefit with pioglitazone for the primary composite endpoint (all-cause death, nonfatal myocardial infarction, acute coronary syndrome, and cardiac interventions, stroke, amputation above the ankle, or revascularization) (hazard ratio 0.78, 95% CI 0.60 to 1.02). Also in the group with prior stroke, pioglitazone reduced fatal or nonfatal stroke (hazard ratio 0.53, 95% CI 0.34 to 0.85). In the subgroup without prior stroke, pioglitazone did not reduce the risk of first stroke.

Several other smaller recent trials also examined comorbidity subgroups with pioglitazone. In a small, open-label study in subjects with overt diabetic nephropathy (mean creatinine 2.6 mg/dL and 2.4 mg/dL in the pioglitazone and glipizide groups, respectively), HbA1c decreased more with pioglitazone (change −0.1 [standard deviation 1.2]) than with glipizide (change −0.4 [standard deviation 1.8]) (between-group P value 0.52).125 A small, placebo-controlled pioglitazone monotherapy study in persons newly diagnosed with type 2 diabetes and coronary heart disease found was no significant difference between groups in change in HbA1c.320

In a small randomized controlled trial (N=47) patients with impaired glucose tolerance or type 2 diabetes in addition to nonalcoholic steatohepatitis received either pioglitazone 45 mg daily or placebo, in addition to a weight loss intervention.180 Glycemic control improved with pioglitazone compared with placebo (P<0.001), with a decrease in weight and body mass index with pioglitazone compared with placebo (P=0.003 and 0.005, respectively). Liver aminotransferase levels normalized with pioglitazone, and plasma aspartate and alanine aminotransferase levels, along with hepatic fat content, all decreased with pioglitazone compared with placebo (P<0.05). Histologic changes in the liver also improved significantly with pioglitazone. In this fair-quality trial, patients were not stratified with respect to type 2 diabetes or impaired glucose tolerance status.

In another small study, patients with acute coronary syndrome received pioglitazone or no additional treatment starting 2 weeks after percutaneous, bare metal stent placement.181 Determined from quantitative angiography at 6 months, the late luminal loss was less in the pioglitazone group than in the control group (P=0.0008) and the restenosis rate was decreased (between-group P=0.0052). Major cardiac events (myocardial infarction or revascularization of the target lesion) were significantly decreased in the pioglitazone group at 6 months compared with the control group (7.7% compared with 60.7%, P<0.0001). No deaths occurred in either group.

Several studies in the updated report examined rosiglitazone with comorbidities. In a very small (N=16), poor-quality randomized controlled trial, subjects with coronary stent implantation were randomized to rosiglitazone 4–8 mg daily or placebo for 6 months. Rosiglitazone did not reduce in-stent restenosis. There were no differences in cardiac events between the groups.174 Lautamaki and colleagues noted a decrease in HbA1c compared with placebo in a study of combination therapy in patients with coronary artery disease (P<0.0001 compared with placebo).172