U.S. flag

An official website of the United States government

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Bennett WL, Wilson LM, Bolen S, et al. Oral Diabetes Medications for Adults With Type 2 Diabetes: An Update [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2011 Mar. (Comparative Effectiveness Reviews, No. 27.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Oral Diabetes Medications for Adults With Type 2 Diabetes: An Update

Oral Diabetes Medications for Adults With Type 2 Diabetes: An Update [Internet].

Show details


This systematic review addresses the comparative effectiveness and safety of diabetes medications used most frequently in the United States as monotherapy and in combination therapy with each other and with insulin preparations. This review updates and adds to a previous comparative effectiveness review (CER)21 published in 2007 comparing the effectiveness and safety of oral diabetes medications, mainly as monotherapy.

Prior to beginning this update, we conducted an extensive preliminary literature review and assessed evidence gaps identified in the 2007 review. We built upon the prior systematic review by focusing on the head-to-head comparisons of medications that should be of greatest relevance to clinicians and their patients (Table 2). We broadened the scope by including two newer medication classes, namely the Glucagon-like peptide-1 (GLP-1) receptor agonists and the Dipeptidyl peptidase-4 (DPP-4) inhibitors and two-drug combinations of medications. We identified 166 articles, which included 75 trials and 19 observational studies that have been published since we completed our 2007 review. We included 19 articles with newer medication class comparisons, 77 articles that contained either metformin or a thiazolidinedione in combination with another medication, and 8 articles with comparisons that included insulin preparations in combination with oral medications (Table 2). Our comprehensive review of the newer medications classes in comparison to other medications and comparisons of combination therapies is an important contribution to the literature because it is the first to address this many comparisons for a wide range of outcomes in patients with type 2 diabetes mellitus.

We defined our key questions similarly to the 2007 review, focusing on intermediate outcomes (Key Question 1), long-term clinical outcomes (Key Question 2), adverse events (Key Question 3) and subpopulations (Key Question 4). As expected, intermediate clinical outcomes such as hemoglobin A1c (HbA1c) levels were studied more frequently in randomized controlled trials (RCTs) than long-term clinical outcomes of diabetes, with 121 RCTs included in Key Question 1 about glycemic control and other intermediate outcomes, 66 articles that applied to Key Question 2 on long-term clinical outcomes, 107 articles for Key Question 3 on adverse events, and 28 articles that contained information for Key Question 4, addressing medication effectiveness and safety in subpopulations.

Key Findings and Implications

Overall, we were unable to definitively support one drug or combination of drugs over another for mortality, macrovascular and microvascular complications of diabetes. Compared with other medications, metformin alone and in combination, had the highest benefit to risk ratio for intermediate outcomes, with similar efficacy for HbA1c reduction as other drugs, but less weight gain and less risk of hypoglycemia.

Intermediate Outcomes

Hemoglobin A1c (HbA1c). Most diabetes medications (metformin, thiazolidinediones, sulfonylureas, and repaglinide) reduced HbA1c to a similar degree by about 1 absolute percentage point when compared with baseline values. Metformin reduced HbA1c more than the DPP-4 inhibitors as monotherapy by about 0.4 absolute percentage points. Combination therapies with metformin (such as metformin plus thiazolidinediones, metformin plus sulfonylureas, and metformin plus DPP-4 inhibitors) generally were more effective at reducing HbA1c compared with metformin monotherapy by about 1 absolute percentage point. These results were consistent with the 2007 systematic review,21 except that we did not have any data on the DPP-4 inhibitors at that time because they were not yet Food and Drug Administration (FDA)-approved. Although we included comparisons with the GLP-1 agonists, evidence for these comparisons was graded as insufficient or low, limiting our ability to draw firm conclusions. Although we could not draw firm conclusions on the comparative effectiveness of 2-drug combinations due to few head to head studies, we did find that most combination therapies showed similar reductions in HbA1c.

Two other recent systematic reviews compared HbA1c with add-on treatments to metformin.235,236 One review identified 16 placebo-controlled trials and 11 comparisons with active comparators of metformin combination therapy and concluded that sulfonylureas were superior to thiazolidinediones in reducing HbA1c in combination with metformin.235 In our pooled analysis of direct comparisons, we did not detect a significant difference in these combinations, which was confirmed in a recent network meta analysis.236 Our review adds to these recently published reviews by including add-on therapies to thiazolidinediones, including more articles and additional meta-analyses.

Weight. Diabetes medications varied in their effects on body weight. Notably, weight gain was small to moderate, even in the longer duration RCTs such as U.K. Prospective Diabetes Study (UKPDS)8 and A Diabetes Outcome Progression Trial (ADOPT)38 (less than 5 kg). However, even small amounts of weight gain (5 percent to 10 percent of body weight) may be associated with increased insulin resistance.237

Metformin consistently had a more favorable effect on weight when compared with other diabetes medications such as thiazolidinediones, sulfonylureas, and DPP-4 inhibitors. As monotherapy, metformin was associated with between-group differences of −2.6 kg when compared with thiazolidinediones, −2.7 kg when compared with sulfonylureas and −1.4 kg when compared with DPP-4 inhibitors. Our results on weight related to comparisons among thiazolidinediones, metformin, and sulfonylureas were consistent with the 2007 review, which showed weight gain for thiazolidinediones and sulfonylureas when compared with placebo, and weight neutrality when metformin was compared with placebo.21 The findings on the GLP-1 agonists and their associated weight loss were similar with another systematic review.238

We also found high strength of evidence for some combination therapies. For example, metformin plus sulfonylurea had a slightly more favorable effect on weight than either metformin plus a thiazolidinedione or a thiazolidinedione plus a sulfonylurea. Drug effects on weight may impact the choice of drug added for second line combination therapy in a patient not well controlled on a single agent. One explanation for metformin’s favorable effect on weight is that it was due to the removal of pretrial medications that increased weight in the run-in period. This suggests that a beneficial effect on weight is seen in direct comparisons between medications only when the other medication has a clearly adverse effect on weight. The mechanism of weight loss for the GLP-1 agonists is not yet well understood, but animal studies suggest a centrally mediated anorectic effect of GLP-1.239,240

Lipids. Effects on lipid levels varied across medication type, but most effects were small to moderate. For instance, pooled analyses showed between-group differences of around 5 to 10 mg/dL in low-density lipoproteins (LDL), 10 to 30 mg/dL in triglycerides (TG), and 3 to 5 mg/dL in high-density lipoproteins (HDL).

In general, we found that metformin had favorable effects on all the lipid classes; it decreased LDL and TG, and modestly increased HDL. Metformin decreased LDL relative to sulfonylureas, rosiglitazone and pioglitazone, and decreased TG relative to sulfonylureas and rosiglitazone. However, pioglitazone decreased TG more than metformin. Compared with metformin alone, the combination of rosiglitazone and metformin increased LDL and HDL, but also increased TG. The addition of pioglitazone to metformin also increased HDL but decreased TG over the combination of metformin and a sulfonylurea. The addition of DPP-4 inhibitors to metformin did not have an effect on HDL relative to metformin monotherapy. Our updated review contributes to the literature by including DPP-4 inhibitors and GLP-1 agonists for lipid outcomes. However, we found insufficient or low strength of evidence for most of these comparisons because of the limited number of studies. Similar to our 2007 review,21 we noted that one medication or class may have favorable effects on one lipid outcome and unfavorable effects on another lipid outcome. For instance, rosiglitazone increased LDL more than pioglitazone, and increased HDL less than pioglitazone, but both favorably decreased TG. Varying effects on lipid fractions such as these may account for differences in cardiovascular risk between medications. Decisions regarding medications that may adversely affect lipids are important because of the importance of cardiovascular disease risk reduction in patients with diabetes.241

Long-Term Clinical Outcomes

Despite the inclusion of two additional large RCTs16,38 and 39 other studies since the 2007 systematic review, we found, overall, low or insufficient strength of evidence to support conclusions about the comparative effectiveness of diabetes medications, either in monotherapy and combination therapy, on all-cause mortality, or macrovascular and microvascular long-term diabetes complications. Compared with the 2007 review, we have additional trials for each drug-drug comparison specifically for metformin versus a thiazolidinedione, metformin versus a sulfonylurea, and comparisons with meglitinides.

Using the trials identified in the 2007 review, Selvin et al. conducted meta-analyses of each drug versus any other drug comparators.242 Treatment with metformin was associated with a decreased risk of ischemic heart disease (pooled OR 0.74; 95 percent CI 0.62 to 0.89) compared with any other oral diabetes agent or placebo, although the results for all-cause mortality and cardiovascular morbidity were not significant. Rosiglitazone was the only diabetes agent associated with an increased risk of cardiovascular morbidity or mortality when compared to any other comparator or placebo, but this result was not statistically significant and had a wide confidence interval.242

In September 2010, the FDA placed restrictions on the use of rosiglitazone, through a Risk Evaluation and Mitigation Strategy, which in part, will require clinicians to attest to and document that the drug’s benefits outweigh the cardiovascular risks. This decision was made after a federal medical advisory panel concluded that rosiglitazone was associated with myocardial ischemia, but voted to keep it on the market.243 Their conclusion was based on recent observational data18,210 and meta-analyses by Nissen and Wolski,15,19 as well as increased understanding of the pharmacology of rosiglitazone.244 Although the FDA acknowledged the limitations of the study designs, there was little evidence to clearly disprove the concerns.245 Other analyses including the original 2007 review21,242,246,247 have not shown an elevated risk of myocardial ischemia, but had very imprecise point estimates.

A notable addition to this update was the Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes (RECORD) trial, which reported that the combined arms of rosiglitazone plus metformin and rosiglitazone plus sulfonylurea were noninferior to metformin plus sulfonylurea for the primary endpoint of hospitalization or death from cardiovascular disease.16 However, these findings were inconclusive for myocardial infarction, for which there was a nonsignificant slightly increased risk in the two arms that included rosiglitazone (combined with metformin or sulfonylurea).16 As the FDA acknowledged, the RECORD trial was open label with a noninferiority design which may have limited its ability to ascertain the cardiovascular effects of rosiglitazone.245

Our updated review informs the debate around rosiglitazone by providing a comprehensive comparative risk and benefit assessment in relation to all other hypoglycemic agents on a wide range of outcomes, not only cardiovascular ischemic risk. We followed a prespecified protocol and engaged a research team that was not invested in either side of the rosiglitazone debate. Other than the risk of heart failure associated with the thiazolidinediones, we found no conclusive evidence of excess ischemic cardiovascular risk associated with rosiglitazone, consistent with the original review. However, the methods for this review differed from those by Nissen and Wolski.15,19 We included studies that occurred only in people with type 2 diabetes and had active comparators, while Nissen et al. included studies in people with other chronic diseases and placebo-controlled trials. In light of the potential ischemic risk of rosiglitazone and the multiple other available medications to treat diabetes, clinicians will need to determine when the benefits of rosiglitazone outweigh the potential risk for individual patients, in keeping with the FDA’s recommendations.

In addition to comparisons with the controversial drug, rosiglitazone, we included other drugs and comparisons of high clinical interest for long-term clinical outcomes. Several large, well-done cohort studies concluded that the risk of all-cause mortality and cardiovascular disease morbidity and mortality was decreased for metformin compared with sulfonylureas, either alone or in combination with other medications, consistent with the analysis by Selvin et al.242 However, the large A Diabetes Outcome Progression Trial (ADOPT),38 which followed participants for a median of 4 years, did not identify any difference in risk between the sulfonylurea and metformin arms. The cohort studies are subject to confounding by indication, as sicker patients may be more likely to take sulfonylureas, and use them in combination. However, trials like ADOPT38 often exclude patients with comorbidities who are at highest risk for long-term complications.

Unfortunately, no studies reporting the outcome of retinopathy met our inclusion criteria. In the 2007 review, six studies reported this outcome, which were all excluded from this updated review because participants were either taking additional background medications or because there was no comparison of interest (e.g., gliclazide versus glibenclamide). In the 2007 review, three studies reported on the outcome of retinopathy. The most notable study was the U.K. PDS, which reported no difference in progression to retinopathy between a sulfonylurea and metformin at 12 years of followup.8 Unfortunately, we found no additional studies examining this clinically important outcome.

Also, few studies reported on the outcomes of nephropathy or neuropathy. We found pioglitazone had greater reductions in the albumin-to-creatinine ratio as compared with metformin, with unclear implications for long-term effects on diabetic nephropathy or chronic kidney disease progression. We were unable to make conclusions about neuropathy because of small sample sizes and inconsistent definitions of the outcome. Because few studies have considered neuropathy and its profound implications for patient quality of life, this will be an important area for future research.

The paucity of robust evidence on cardiovascular outcomes and other important clinical outcomes for diabetes medications may reflect the emphasis of most studies on glycemic control, a surrogate marker, for drug approval. Future research and longer studies will be needed to address this evidence gap.

Adverse Events

We focused our review on the comparative safety of diabetes medications as monotherapy and in combination therapy, and refer readers to our 2007 review for additional details about specific adverse effects reported in placebo-controlled trials. In this update, we confirmed the elevated risk of hypoglycemia associated with sulfonylureas, either alone or in combination, compared with the other hypoglycemic agents. For example, we showed a more than four-fold higher risk of hypoglycemia associated with sulfonylureas compared with metformin alone, and an almost 6-fold higher risk of hypoglycemia for metformin plus a sulfonylurea compared with metformin plus a thiazolidinedione. We also demonstrated that the newer drug class, DPP-4 inhibitors had a lower risk of hypoglycemia than sulfonylureas, and a risk comparable to that of metformin.

We confirmed a doubling of the risk of heart failure with the thiazolidinedione class of medications, particularly compared with sulfonylureas, which was also reported in two recent meta-analyses.248,249 In fact, both the thiazolidinediones, rosiglitazone and pioglitazone, are contraindicated in patients with serious or severe heart failure (Stage 3 or Stage 4) according to the FDA boxed warnings on the thiazolidinediones.250,251 The excess deaths and hospitalizations associated with heart failure with the thiazolidinediones in RECORD16 indicates that heart failure induced by thiazolidinediones is clinically important.

We included four new safety outcomes in addition to the others we addressed in the 2007 review: macular edema, cholecystitis, pancreatitis, and fractures, because of safety concerns that emerged after the review. The 2007 review reported an increased risk of cholecystitis with pioglitazone in an unpublished pooled analysis from the FDA.21 However, in this updated review we found no additional evidence on this outcome for the comparisons of interest. Several case reports and case series have reported spontaneous macular edema associated with the thiazolidinedione class.252–253 However, clinical trials are underpowered to detect rare adverse events and hence we did not detect any significant difference in the rates of macular edema, as we only identified one trial reporting on this outcome. A recently published prospective cohort study in the Kaiser Permanente database of over 17,000 users of the thiazolidinediones reported an increased odds of macular edema with the thiazolidinediones (OR 2.6; 95 percent CI 2.4 to 3.0) compared to nonusers, significant even after adjustment for age and glycemic control. Notably, this cohort study also reported an increased risk of macular edema with insulin and meglitinides.254

Patients with diabetes may have an increased baseline risk of acute pancreatitis.255 The current drug labels for exenatide and sitagliptin have been strengthened with information from spontaneous post-marketing reports of severe pancreatitis including hemorrhagic pancreatitis occurring after exenatide therapy.256 The clinical trials with the GLP-1 agonists may have been underpowered to detect these rare occurrences of pancreatitis. However, a recent claims database study failed to show any significant relationship between the GLP-1 agonists, DPP-4 inhibitors and pancreatitis.257

Our results for lactic acidosis support the results from the 2007 review, as well as the Cochrane systematic review on this topic258 showing no increased risk of lactic acidosis among metformin users. The Cochrane review reported similar rates between metformin users (5.1 cases per 100,000 patient-years) and those on other oral hypoglycemic agents or placebo (5.8 cases per 100,000 patient-years). Further, there was no statistically significant difference in the net change of lactate levels from baseline in metformin users compared to those on other oral hypoglycemic agents or placebo suggesting no increased risk of lactic acidosis with metformin compared to other oral hypoglycemic agents or placebo.258

As with the 2007 systematic review, we evaluated cancer as an outcome. We included four trials with inconclusive results. One retrospective cohort study not included for this outcome, because of uneven use of insulin, evaluated cancer mortality among the sulfonylurea cohort compared to the metformin cohort using the administrative data from Saskatchewan Health, Canada.259 The mortality from cancer was higher in the sulfonylurea cohort (9.7 per 1000 person-years) than the metformin cohort (6.3 per 1000 person-years), with a hazard ratio for cancer mortality of 1.3 (95 percent CI 1.1 to 1.6), adjusted for age, sex, insulin use, and comorbidities. This study was limited by the use of administrative data and high risk for residual confounding. Although we did not identify additional evidence about diabetes medications and cancer risk, several recent studies have highlighted that this is an area of active research.260,261 In particular, a large German cohort study published in 2009 showed a positive association between cancer incidence and insulin for all insulin types. Another study suggested a relationship between cancer risk and treatment with insulin glargine compared with human insulin,260 while another study did not observe the association.262 A recent study extracted cancer diagnosis information from ADOPT and RECORD, with nearly 39,000 person-years of drug exposure, and showed no advantage of metformin over rosiglitazone and sulfonylureas in terms of cancer rates.263

We found high strength of evidence for comparative safety in terms of fracture risk. The RECORD study reported significantly increased risk of upper and lower limb fractures in women randomized to rosiglitazone combination therapy arms compared with metformin plus sulfonylureas. A prior systematic review that included ten studies evaluating the long-term effect of thiazolidinediones on fracture risk showed a significant increase in fracture risk, most apparent in women.264 Fractures reported with the thiazolidinediones have been mainly those of the upper and lower limb and not hip fractures. Several recent observational studies have also reported an increased risk of fractures with the thiazolidinediones among men as well, but the risk appears to be higher among women and those of advanced age.265

We confirmed the results of our 2007 review showing more frequent gastrointestinal adverse events for metformin compared with thiazolidinediones and sulfonylureas.21 We also reported higher gastrointestinal side effects with metformin compared with the newer DPP-4 inhibitors, but graded the strength of evidence as low because of inconsistency of effects.

Two-Drug Combinations, Including Addition of Insulin Preparations

In this update, we included comparisons of two-drug combinations of medications that contained either metformin or a thiazolidinedione in combination with another medication, two-drug combinations compared to metformin alone, and combinations of a medication with either basal or premixed insulin preparations compared with non-insulin two-drug combinations (Table 2). Overall, we found that most combinations of two drugs when compared to monotherapy had additive effects, both in terms of improved glycemic control, but also risk for adverse events and weight gain, confirming the 2007 review and other reviews.21

Comparative benefit of a two-drug combination over another was less clear, and several combinations had evidence of similar effects on glycemic control. Our conclusion is similar to a recent network meta-analysis of the effect of non-insulin medications added to metformin.236 One combination comparison favored metformin plus GLP-1 analogs over metformin plus DPP-4 inhibitors, showing a 0.6 absolute percentage point greater reduction in HbA1c. The clinical meaning of this small between-group difference is unclear. Despite little to no difference in HbA1c among the combination therapies, we found that some combinations clearly had increased risk for adverse events and weight gain. Thiazolidinediones in combination with either metformin or sulfonylureas increased weight gain compared with metformin plus sulfonylurea. In contrast, metformin plus a GLP-1 agonist decreased weight compared with several other two-drug combinations, but we found low strength of evidence because of the paucity of studies using the same comparators (see below).

Although this review does not provide a comprehensive review of the addition of insulin preparations to oral medications, we did include several clinically relevant comparisons. We were unable to draw any firm conclusions about the use of premixed insulin preparations compared with basal insulin, in combination with oral agents, with regard to glycemic control or long term clinical outcomes. There was a modestly lower risk of hypoglycemia when metformin was combined with a basal insulin rather than a premixed insulin preparation, confirming a recent CER on premixed insulin analogues, also commissioned by AHRQ.266 In addition, two recent systematic reviews compared NPH insulin with longer-acting synthetic insulins, glargine or detemir. Most studies had combined insulin with oral medications. They reported no difference in glycemic control between the two insulin products, and also found slightly lower hypoglycemia with the longer-acting insulins.267,268

Newer Diabetes Classes of Medications: DPP-4 Inhibitors and GLP-1 Agonists

Eight articles contained comparisons with the new GLP-1 receptor agonists, exenatide or liraglutide, and 12 articles contained comparisons with the DPP-4 inhibitors, sitagliptin or saxagliptin, either as monotherapy or combination therapy. The American Diabetes Association Consensus/European Association for the Study of Diabetes consensus statement has suggested the use of a GLP-1 receptor agonist as an add-on treatment to metformin,22 a comparison of interest we included for this updated review, but did not have explicit recommendations for the DPP-4 inhibitor class. The American Association of Clinical Endocrinologists/American College of Endocrinology’s consensus algorithm recommends consideration of a DPP-4 inhibitor either as initial monotherapy or second line therapy, and a GLP-1 agonist as initial combination therapy with metformin when the HbA1c is greater than or equal to 7.6 percent.23

We found that the DPP-4 inhibitors improved HbA1c to a lesser extent than metformin as monotherapy, but that when added to metformin there was improved HbA1c without additional hypoglycemia risk. These findings are consistent with a Cochrane systematic review269 and another recent systematic review.270

The majority of comparisons with the GLP-1 agonists for the intermediate outcomes (KQ1) were graded with low strength of evidence because of few studies within each comparison, and evidence was insufficient for the long-term outcomes and most safety outcomes. Despite this limitation, the GLP-1 agonists combined with metformin showed similar HbA1c reduction, when compared to metformin plus basal insulin or metformin plus a thiazolidinedione. In addition, the GLP-1 agonists showed decreases in weight compared with sulfonylureas alone, as well as in combination with metformin compared with other standard combination therapies. The largest recent systematic review of the GLP-1agonists identified 21 RCTs (six unpublished) and showed a reduction in HbA1c by one absolute percentage point in comparison with placebo, with weight loss, as well as a low risk of hypoglycemia.238 Since exenatide’s release, the FDA published alerts about postmarketing case reports of pancreatitis271 and acute renal failure and insufficiency.272 In the studies we included, the event rates for these complications were too low to draw any conclusions.


Several important limitations to our updated systematic review deserve mention. Because this was an update of a comprehensive review published in 2007, we focused our update a priori on studies with active control comparators, which are most relevant for clinical practice. Placebo-controlled trials had been included in the original 2007 review. However, the majority of placebo-controlled trials are short-term and lacking long-term outcomes. However, the exclusion of placebo-controlled trials has implications for the review, including missed rare adverse events, such as macular edema and acute pancreatitis. To conclude from an active-control study that one medication is more effective than another requires prior knowledge that the active-control drug has been studied previously and is known to be more effective than placebo. Because this was an update of the 2007 review that had included placebo controlled trials, for most drug comparisons this was probably true.273 However, this assumption may be less valid for the newer medications of saxagliptin, sitagliptin, nateglinide, exenatide and liraglutide, where evidence from other systematic reviews, such as Cochrane Reviews, will be also be helpful in making conclusions, and further studies will be needed.

In addition, our inclusion criteria required that all studies fit into one or more of the prespecified comparisons of interest (Table 2), which identified specific drug-drug or two-drug comparisons. For example, studies that included any number of “background medications” were excluded. Our rationale was to avoid contamination by use of background medications with unclear interactions with the intervention medications. This was especially important because of our goal of evaluating two-drug combinations. Applying the inclusion criteria, which required prespecified comparisons of interest, had several implications. This criteria required the exclusion of several large trials,8,9,12,72–74,274–277 some of which compared HbA1c lowering strategies, not individual medications, as well as some smaller trials and observational studies. Of note, the PROspective pioglitAzone Clinical Trial In macroVascular Events (PROactive) study was included in the 2007 systematic review but excluded from this updated review.274 Another unintended consequence of requiring these prespecified comparisons of interest was that some of the recent studies of exenatide278–281 as add-on therapy to oral medications did not fit our inclusion criteria.

Another implication of the requirement of specific medication comparisons was the exclusion of several case-control studies that did not report outcomes of interest by drug comparison. Although we applied very broad search terms and did not exclude studies by study design, we only identified seven case control studies and six of these were subsequently excluded from the review because they did not report their results to fit with the prespecified drug comparisons of interest for this review. For example, five studies282–286 compared a drug of interest with any other unspecified drug for an adverse event outcome, and this was not a comparison of interest.

We selected key questions focused on intermediate and long-term clinical outcomes through an extensive topic refinement process at the beginning of this process, which involved input from stakeholders on the Technical Expert Panel. Diabetes care is a rapidly growing and very extensive field, and we note the omission of key outcomes. For example, we did not collect information about several patient-reported outcomes, such as medication adherence and barriers to adherence, or health-related quality of life. These outcomes are important because they may mediate the efficacy of treatment outcome, and also have significant value to patients and clinicians. Future reviews with methodologies designed to capture many different study designs, including qualitative studies, and use of a wide range of measures, are most needed to address these outcomes. Although we assessed the mean difference in HbA1c between intervention groups in Key Question 1, we did not include the durability of HbA1c changes over time as an outcome, which may best be addressed using long-term well-designed observational studies.

Limitations within the included studies have presented challenges to how we reported their outcomes and our ability to combine them in meta-analyses. For example, several studies failed to report the significance of between-group differences and the measures of dispersion, thereby hindering efforts to estimate effect size across trials. Some trials underdosed comparison medications, limiting our ability to draw conclusions about efficacy. In addition, because of our interest in the comparative effectiveness of drugs, we focused primarily on the relative differences between drugs in our forest plots. In the forest plots, however, we also included footnotes with information about the range of absolute differences from baseline to followup in the comparison arms for readers who wish to estimate the magnitude of effect in absolute terms. Finally, many included trials were industry-sponsored, raising the possibility of publication bias and other forms of bias, such as selective reporting of outcomes. While publication bias generally was not found, these analyses have limited power due to small numbers of studies for many comparisons.

Future Research

In this updated systematic review, we synthesized current literature about the comparative effectiveness and safety of diabetes medications when used as monotherapy and in two-drug combinations. We identified some deficiencies in the published literature that need to be addressed by future research to meet the decision making needs of patients, providers, and policy makers. We organized these deficiencies and recommendations using the PICOTS format for specifying research questions: patient populations, interventions, comparators, outcome measures of interest, timing, and settings.


Studies often employed narrow inclusion criteria, enrolling patients at lowest risk for complications, and commonly used run-in periods to avoid enrolling patients with adverse effects or poor adherence, which may limit applicability. We identified the following research gaps related to target patient populations:

  1. The literature is deficient in studies enrolling people with varying levels of underlying cardiovascular and renal disease risk.
  2. Results reported in subgroups of the population were rare, especially the elderly and people with multiple comorbid conditions, such as underlying chronic kidney disease.

Interventions and Comparators

We identified the following gaps in the literature, where future studies could address additional medication comparisons to support clinicians in decision making.

  1. The published literature is deficient in studies of the comparative effectiveness of two-drug combinations, focused either on their effectiveness or the safety and thus, interaction between two medications.
  2. The comparative effectiveness literature is sparse on monotherapy and combination therapy comparisons of meglitinides, DPP-4 inhibitors, and GLP-1 agonists, with other first line diabetes medications.
  3. Few studies used comparisons with a basal or premixed insulin added to metformin or thiazolidinediones.

Outcomes of Interest

Overall, few studies contained sufficient data on event rates for major clinically important adverse events and long-term complications of diabetes.

  1. We identified few published studies on long-term clinical outcomes such as cardiovascular disease, stroke, nephropathy, and neuropathy.
  2. Few studies used standard measures for diabetic nephropathy and kidney function, such as estimated glomerular filtration rate, or clinical outcomes like time to dialysis, as outcomes in the comparison of these medications.
  3. We identified few observational studies that examined macular edema, cancer and fractures for thiazolidinediones, insulin, and other medications.


We identified several key deficiencies in study timing and duration of followup.

  1. The literature is relatively deficient in studies of the short-term benefits, if any, of the addition of insulin to oral agents, and the long-term effects on mortality and cardiovascular disease, from the addition of insulin to a regimen relative to the addition of another oral agent.
  2. Few studies on harms lasted greater than 2 years, a shorter duration of exposure than typically seen in clinical practice, where these drugs may be prescribed for decades. Some adverse effects, like congestive heart failure, may take years to develop, and others like fractures, may be due to cumulative exposure. The FDA approval process focuses on short-term harms, providing less incentive for pharmaceutical companies to engage in longer term trials.


Study settings are relevant to understanding the applicability of the findings to the general U.S. population of patients with diabetes.

  • Few trials reported the study setting or source for participant recruitment, such as an outpatient clinical or subspecialty clinical setting, which is relevant because the majority of patients with diabetes are cared for by primary care physicians.

We also identified methodological problems and made recommendations to consider for future research:

  1. We recommend studies consistently report between-group comparisons of changes from baseline, as well as measures of dispersion such as standard errors, to improve interpretation of the significance of their findings.
  2. We recommend improved adverse event and long-term outcome reporting, with predefined outcomes and definitions, and a description of methods for ascertainment.
  3. We recommend trials report the steps taken to ensure randomization and allocation concealment.
  4. We recommend that observational studies of the comparative effectiveness and safety of diabetes medications report details of the treatment type, dose, timing and duration of use of the medication, when available.
  5. We recommend that studies consistently report the number of deaths in each study arm, even if there were none.
  6. We recommend that studies allowing use of “background” medications report which medications were allowed and stratify results by the combination therapy, which includes the background medication(s) plus the study drug(s).
  7. We recommend conducting a network meta-analysis to assess indirect comparisons, which were not addressed in this report.


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (9.0M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...