PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Shekelle P, Munjas B, Romanova M, et al. Self-Monitoring of Blood Glucose in Patients with Type 2 Diabetes Mellitus: Meta Analysis of Effectiveness [Internet]. Washington (DC): Department of Veterans Affairs (US); 2007 Sep.


Literature Flow

In total, we examined 52 titles. Seventeen articles were identified from prior systematic reviews. The electronic update search identified 23 articles. An additional 11 articles were identified through reference mining. One was identified by a content expert.

Of the titles identified through our electronic literature search, 14 were rejected as not relevant to the project. This left 38 from all sources. Ten articles were excluded at abstract review. In January 2006 we received the Balk and colleagues draft report of a review of SMBG in type 2 diabetes. 7

We performed an update search in July of 2007 that resulted in two additional articles, one of which was excluded since it did not test SMBG. In total we reviewed 30 articles.

We compared trials identified for our review with those identified in the three recent systematic reviews (Table 1). Our review included 10 trials, compared to six in the review by Welschen and colleagues, 6 seven in the review by Balk and colleagues, 7 and 13 in the review by Jansen. 8 Jansen included studies of self-monitoring of urine glucose that we did not, and we included studies rejected by Balk and/or by Welschen for a variety of reasons, detailed in Table 1.

Table 1. Comparison of RCTs we included in our review with those included in three recent systematic reviews.

Table 1

Comparison of RCTs we included in our review with those included in three recent systematic reviews.

Initial screening of the articles resulted in 14 RCTs that measured the effect of SMBG compared to a group not receiving SMBG and monitored A1c levels with at least three months of follow-up. Four were excluded; one because the trial presented duplicate data, the other three because the trials compared a control group of SMBG to an intervention group of SMBG plus other components. (Figure 1) We identified five observational studies that assessed the effectiveness of SMBG in diabetic Veterans.

Figure 1. Self-monitoring of Blood Glucose Literature Flow.

Figure 1

Self-monitoring of Blood Glucose Literature Flow.

Description of the Efficacy Evidence

The 10 RCTs ranged in size from 29 to 988 subjects. All patients had type 2 diabetes, the mean duration of which was three to 13 years. All trials but one included only patients treated without insulin, the one exception being a trial from Bangladesh that included patients on oral hypoglycemic agents or insulin and not specifying how many of each type. The average age of patients was between 50 and 66. Almost all trials included counseling/education with SMBG in the intervention group, but other components of the intervention were varied (Table 2). All trials measured A1c as an outcome; five trials assessed this at six months, three trials assessed this at three months, and five trials assessed this at one year or more. The quality of trials varied; most trials scored positively on less then half of the criteria on the Delphi list. 10 Details of each trial are presented in the Evidence Table (Appendix C). We now present a brief synopsis of each trial.

Table 2. Components of each arm of the 10 RCTs.

Table 2

Components of each arm of the 10 RCTs.

Wing RR et al. (1986) 19 This study assessed the usefulness of SMBG in improvement of dietary compliance for obese patients with type 2 diabetes. Authors from the University of Pittsburgh School of Medicine enrolled 50 patients (mean weight 98 kg, 78% women) with adult-onset diabetes who were treated with oral hypoglycemic agents and insulin. All patients received weekly behavioral weight control counseling for the first three months, then monthly for six months, and twice more until week 62. Monetary stimulation was used. The intervention group was asked to monitor blood glucose on average 5.4 times a week to provide feedback of dietary modifications. Patients’ compliance with the diet and SMBG were monitored. Medication adjustments were made in similar mode in both groups according to study protocol. Five patients were excluded from the analyses with reasons explained by the authors. By week 12 there was a slight improvement of glycemic control in both groups (A1c values: SMBG group: 10.19% to 9.68%, control group: 10.86% to 10%). No statistically significant difference between groups was observed in glycosylated hemoglobin measurements by the one-year follow up. Both groups lost a significant amount of weight (6.1 kg). Patients with a high level of compliance to either SMBG or weight loss lost twice the amount of weight compared to poorly compliant patients. A large and not significantly different number of patients in both groups had their medications or insulin dose decreased during the study.

Fontbonne A et al. (1989) 20 This is a French study of 208 non-insulin treated patients with long-term and poorly controlled diabetes (mean duration=13 years, mean A1c=8.3%). Patients were randomized into three groups (SMBG, urinary glucose monitoring and control), and seen by their respective physicians every two months with either A1c results or SMBG measurements. At each visit medication and/or dietary modification was allowed. Forty-four patients were lost to follow-up. A1c values at the end of six months were not significantly different between the three groups. However, the degree of compliance to SMBG appeared to relate to outcome; the more blood strips used, the larger the decrease in A1c values.

Rutten G et al. (1990) 23 This study from the Netherlands looked at the feasibility and effect of a diabetes type 2 protocol with SMBG in general practice. One hundred forty-nine patients (66 in intervention and 83 in control groups) from eight practices were studied over 12 months. Ten patients were excluded from each group with reasons explained for all. Some patients in the treatment group tested their fasting glucose and reported it on a monthly basis. In case of elevated readings, they were referred to a study doctor, where the protocol was followed and medications were possibly changed. The protocol included weight reduction counseling and medication changes with up to two oral hypoglycemic agents used. Other patients from the treatment group did not check their blood glucose, but had it measured during quarter-annual visits with a doctor. At the end of the study, the treatment group decreased A1c values from 9.7% to 9.2%, whereas the control group increased from 8.9% to 9.4%.

Muchmore DB et al. (1994) 24 This study tested the hypothesis that combined use of SMBG and dietary carbohydrate counting is beneficial in managing type 2 diabetes. The Scripps Clinic enrolled 29 overweight patients (BMI=34, 61% women) with diet or treated with oral hypoglycemics diabetes. Six were excluded for reasons not described. Patients participated in a 28-week behavioral weight loss program, with emphasis on glycemic response to carbohydrate intake and exercise. Medication adjustment was not included in the study protocol, but it was done for a similar amount of patients in both groups by their own physicians. Although A1c improved more in the intervention group than in the control group (a decrease of 1.54% vs. 0.84% absolute), the difference was not statistically significant. Quality-of-life measures were similar in both groups. Weight loss was equivalent (~6 kg) in both groups by week 44.

Jaber LA et al. (1996) 25 This study was performed in a university-affiliated internal medicine outpatient clinic. It enrolled 45 obese African-American patients with non-insulin dependent diabetes mellitus (NIDDM). All patients were treated with sulfonylurea agents. The mean age of the patients was 62, 70% were women, the mean BMI was 33, and mean duration of diabetes was six years. During four months of follow-up, the intervention group received diabetic education, medication counseling, instructions on dietary regulation, exercise, and SMBG, as well as evaluation and adjustment of their hypoglycemic regimen by the pharmacist. Patients were instructed to monitor blood glucose eight times a week. The control group was followed by their physicians. Six patients withdrew or dropped out from the study with reasons explained by the authors. The intervention group patients had on average 2.2 changes in drug therapy, with an increase in oral hypoglycemics dose on most visits. The final A1c value in the intervention group decreased from 11.5% to 9.2%, and in the control group it decreased from 12.2% to 12.1%. This difference between groups was statistically significant. Quality-of-life analyses revealed no significant differences in any of the domains tested between or within groups. No significant changes were noted within or between groups in blood pressure, body weight, serum lipid measurements and renal function parameters.

Kibriya MG et al. (1999) 27 Physicians of the Bangladesh Institute of Research and Rehabilitation in Diabetes Endocrine and Metabolic disorders (BIRDEM) recruited 64 type 2 diabetic patients of “higher-middle class to rich socio-economic class and having completed secondary school certificate level education” to participate in this randomized study and followed them for 18 months. Their aim was to evaluate the cost effectiveness of SMBG in the management of type 2 diabetes in developing countries. All patients received education on diet and how to adjust insulin or oral antidiabetic medications. Patients in the SMBG group were advised to check their blood glucose two to three times a day every two weeks and adjust their medications accordingly if fasting values were >6.0 mmol/L. They were also asked to visit the physician at three month intervals for blood glucose and A1c measurements. Patients in the control group visited the physician at one month intervals and had their antidiabetic regimen modified if needed, based on fasting blood glucose measurements at the monthly visits and A1c values at each three month visit. Cost analysis was performed using conveyance cost, patient wage loss, costs of test strips, glucometer, laboratory tests, and manpower. For the control group, results demonstrated a decrease of 0.43% in A1c after 18 months which was statistically significant. Statistical comparisons between groups were not reported. The SMBG group demonstrated a 1.37% drop in A1c after 18 months which was also significant. Cost analysis revealed comparable results for both groups ($134.55 for the control group vs. $134.75 for the SMBG group). Conclusions were that SMBG with proper diabetes education is a cost effective strategy in the management of type 2 diabetes.

Schwedes U et al. (2002) 29 This study was a randomized multicenter trial that recruited subjects in Germany and Austria and followed them for six months. A total of 250 patients were randomized within blocks of eight to one of two groups: one group used SMBG, kept a blood glucose/eating diary, and received standardized counseling; the control group received only nonstandardized counseling on diet and lifestyle. Two hundred and twenty three patients were included in the final analysis. Patients in the SMBG group were instructed to measure blood glucose six times a day (pre-and postprandially) on two days per week, and to document eating habits and state of well-being. Patients were seen every four weeks. Results showed a statistically significant difference in A1c reduction between the two groups. The control group had a 0.54% reduction of A1c compared to a 1.0% reduction in the SMBG group.

Guerci B et al. (2003) 30 The Auto-Surveillance Intervention Active (ASIA) study out of France followed 689 patients for six months. Patients were randomized to receive either a conventional laboratory work-up based solely on A1c measurements every 12 weeks (control group) or conventional laboratory work-up and SMBG at a frequency of at least six times a week (intervention group). Both groups received counseling on diet and exercise from their general practitioners during five visits throughout the course of the study. At the three month visit, practitioners could modify treatments of their patients based on their A1c value measured at that time. All but three of the patients were on at least one oral antidiabetic drug. Among those, the most widely prescribed drugs were sulfonylureas and biguanides. Results demonstrated a 0.28% absolute greater drop in A1c in the SMBG group as compared to the control group at the end of the study, which was statistically significant. This difference was most pronounced at three months, with a steady state reached in the last three months of the study. The authors concluded that SMBG was associated with better quality of metabolic control than usual recommendations alone in patients with type 2 diabetes. They noted that since no specific instruction for adjusting behavior to the results of SMBG was given to the patients.

Davidson MB et al.(2005) 32 This randomized controlled trial followed 88 patients for six months. Patients in the treatment group were instructed to measure pre- and post-prandial blood glucose levels six days a week. Patients in both groups received dietitian counseling five times during the study. A nurse, who was blinded to whether the patient was in the treatment group or not, followed a detailed algorithm to make her therapeutic decisions. Her goals were to lower fasting glucose concentrations to <130mg/dL by stepwise increases in metformin or a sulfonylurea agent every two weeks, and to achieve an A1c value <7.5%. If the goal A1c was not achieved, a thiazolidinedione was added. Results demonstrated a significant drop in A1c levels of both groups, but no statistically significant difference between the two groups.

Farmer A et al.(2007)33 This trial randomized 453 patients seen at 48 general practices in London into three groups (usual care, SMBG, SMBG plus training to use results for self care). Patients were included if they had type 2 diabetes, were at least 25 years old at diagnosis, were managed with diet or oral hypoglycemics, had an A1c level greater than or equal to 6.2% at the initial visit, and were able to independently perform daily living activities. Patients in the control group (n=152) received standardized usual care and were seen for A1c measurements once every three months for 12 months. Patients in the less intensive SMBG group (n=150) were instructed to monitor three times a day on two days of every week. In addition to being instructed about self-monitoring, patients in the more intensive SMBG group (n=151) were trained to use the results for self care. Patients in the more intensive group were not instructed to measure blood glucose a set number of times per week, but rather to “explore the effect of different activities... on their blood glucose level.” From diaries kept by the patients, those in the more intensive group tested their blood sugar on average six times per week early in the study, but by the end of the 12 months this had decreased to an average of five times per week. Patients in the less intensive group tested their blood sugar on average about five times per week throughout the trial. At three, six, nine, and 12 months no statistically significant differences in A1c levels were found between the three groups.

Key Question #1: Is regular self-monitoring of blood glucose effective in achieving target A1c levels for patients with type 2 diabetes?

Studies of Efficacy

Achieving Target A1c Levels

We identified a single trial that assessed the effect of SMBG (plus counseling and education) in 149 patients in general practice in the Netherlands at meeting target A1c levels. 23 In this study the target A1c level was 8.0%. Prior to the intervention 45 and 41 patients in the in control and intervention arms, respectively, had an A1c values greater than 8%. After the intervention, one patient (2%) and two patients (5%) in the control and intervention arms had A1c values of less than 8% (p=0.6). Thus there is little evidence to draw a conclusion about the efficacy of SMBG at achieving target A1c levels. We judged the strength of this evidence as very low. [GRADE: Very Low = Any estimate of effect is very uncertain.]

Improving Glycemic Control

While not directly answering this question but certainly relevant to it, is the effect of SMBG on the mean A1c level. All 10 trials reported this outcome. We grouped trials based on the duration of the intervention. The individual and pooled results are shown in Figure 2.

Figure 2. Analysis of Mean Difference between Control and SMBG Group at Follow-up.

Figure 2

Analysis of Mean Difference between Control and SMBG Group at Follow-up. *Subtotal does not include “Farmer et al., 2007 [#1039] Less intensively self monitoring” arm.

We identified three trials that reported A1c outcomes at three months. 19,25,33 The three trials reported variable results. 19,25 We did not pool the results of these three trials because their results were too heterogeneous, with an I2 statistic of 67%.

We identified five trials that reported outcomes at about six months. 20,29,30,32,33 Only one trial reported a statistically significant improvement in A1c30, although a second trial also yielded a statistically significant result after adjusting for baseline difference29. The random effects pooled estimate of effect of these five trials was a change in mean A1c value of −0.21% (95% CI: −0.38%, −0.04%). The I2 statistic for heterogeneity was 0.

We identified five trials that reported outcomes at about one year or longer. 19,23,24,27,33 No study reported a statistically significant difference between groups in the mean A1c value, although two studies reported statistically significant benefits after adjusting for baseline differences in A1c values. 23,27 The random effects pooled estimate of the effect of these five trials was a change in A1c value of −0.15%(95% CI: −0.36%, 0.06%). The I2 statistic for heterogeneity was 0.

We performed several additional analyses. First we compared studies scores, four or more Delphi items positively (which we called “High quality”) with those scoring less than four items positively (“Low quality”). The pooled results showed no statistically significant differences between high and low quality studies.

We repeated our primary analysis using as the outcome the difference in A1c levels between groups adjusted for baseline A1c levels (whether or not to do such adjusting in the results of an RCT is controversial). When analyzed this way there was much greater heterogeneity between studies, with I2 statistics of 49% and 75% for studies with six month and 12 month outcomes respectively. However, despite this our primary pooled results were remarkably similar: a modest and statistically significant effect on A1c at six months of 0.19 (compared to a pooled result of 0.21 in the main analysis); and a nonsignificant effect at 12 months. In the difference of difference analysis, high quality studies reported lower estimates of effect than low quality studies, an observation seen in other conditions.34 This re-analysis supported our primary analysis (Figure 3).

Figure 3. Analysis of Difference of Differences between Control and SMBG Group at Follow-up.

Figure 3

Analysis of Difference of Differences between Control and SMBG Group at Follow-up. *Subtotal does not include “Farmer et al., 2007 [#1039] Less intensively self monitoring” arm.

Meta-regression on baseline values of A1c level showed differential effectiveness (p value for difference = 0.05), with higher baseline values of A1c being associated with lesser efficacy of SMBG. Each 1% increase in A1c was associated with a 0.3% decrease in efficacy of SMBG. Thus, indirect evidence suggests that SMBG results in a smaller percent change in A1c for patients with higher baseline values of A1c.

We attempted to identify other components of the intervention or characteristics of the patients associated with greater effectiveness. The trials did not have sufficient similarity in intervention components to permit a meta-regression analysis (See Figure 2). Almost all studies included SMBG and counseling/education, making an assessment of the effect of one without the other impossible, and other intervention components were too sparsely distributed to support meta-regression. Meta-regression using the quality assessment (as either a continuous variable or dichotomous at a threshold value of four) also did not demonstrative differences between results. An analysis of the frequency of SMBG testing is discussed in Key Question #4.

The funnel plot for publication bias is shown in Figure 4. Neither Begg’s test nor Eggar’s test yielded evidence of unexplained heterogeneity.

Figure 4. Publication Bias for the 10 RCTs.

Figure 4

Publication Bias for the 10 RCTs.

Therefore, we found that adding SMBG along with education, counseling, (and some times other components) results in a statistically significant decrease in A1c level of an absolute 0.21% at six months. Results at three months and one year are more variable, although there is a suggestion that this benefit may continue out to at least one year (absolute reducation = −0.15%, 95% CI: −0.36, 0.06). Indirect evidence suggests SMBG maybe less efficacious in subjects with higher baseline A1c values.

We judged the strength of evidence for this outcome as moderate, because individual trials did not in general report significant results and interventions were heterogeneous. [GRADE: Moderate= Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.]

Studies of Effectiveness in Veterans

We identified six publications assessing the value of SMBG specifically in Veterans. One publication was an RCT comparing SMBG with urine glucose monitoring, 22 and did not include a comparison with Veterans who did no monitoring. The other five studies were observational in design, mostly retrospective chart reviews that sought to compare a variety of outcomes between Veteran patients receiving supplies for SMBG with those not receiving such supplies. 35–39 Details of all studies are in Table 3.

Table 3. Studies of Self-monitoring of Blood Glucose in Veterans.

Table 3

Studies of Self-monitoring of Blood Glucose in Veterans.

All studies reported that there was no difference in A1c levels between groups. Although most studies attempted to try and control for baseline differences between patients, the observational study design cannot inform what the A1c values of Veteran patients currently using SMBG would be if they did not receive SMBG supplies. It is possible, for example, that Veterans are selected by their clinicians for receipt of SMBG because they are more difficult to control.

Therefore, the results of the effectiveness studies do not negate the efficacy evidence from RCTs that the addition of SMBG and education can result in a decrease in A1c levels of about 0.3% absolute at six months and up to one year. However, these studies do raise the question of whether Veteran patients are receiving the full possible benefits of SMBG. For example, the RCTs reporting benefit also all included counseling and education. If this is necessary for SMBG to have an effect, one explanation of the difference in the results between the RCTs and the observational studies in veterans is that there is inadequate counseling and education of the Veterans.

We judge the strength of evidence for this outcome as very low because these are observational studies with serious limitations in study quality. [GRADE: Very Low = Any estimate of effect is very uncertain.]

Key Question #2: Is regular self-monitoring of blood glucose effective in maintaining target A1c levels for patients with type 2 diabetes?

We did not identify any trials that directly assessed this question. Therefore, we draw no conclusion. [GRADE: Very Low = Any estimate of effect is very uncertain.]

Key Question 3: Does regular self-monitoring of blood glucose reduce the frequency of hypoglycemia in patients with type 2 diabetes?

We identified four trials that reported hypoglycemia as an outcome. More details of these trials were presented previously. A brief synopsis of each study follows, with respect to the hypoglycemia outcomes.

Jaber LA et al. (1996) 25 There were 17 reported hypoglycemic reactions in the intervention group and two in the control group. All were rated as mild to moderate, and successfully self-treated. The authors report: “High rate of reported hypoglycemia is partly inherent to the study design and execution. Intervention group patients were repeatedly instructed to and trained on recognition and documentation of hypoglycemia. They were also questioned about the occurrence of these reactions at every clinic visit. Subjects in the control group were not given any specific instructions regarding hypoglycemia and data on its occurrence were collected at the end of the study where capture of this information may have been hindered by the duration of the elapsed time”.

Kibriya MG et al. (1999) 27 In this trial the patients in the SMBG group were instructed to perform testing every two weeks, and adjust the dose of anti-diabetic medication accordingly. Control group patients were seen by their doctors on a monthly basis, and had their anti-diabetic treatment modified if needed. During 18 months of follow-up, ten patients in the SMBG intervention group had a total of 17 episodes of hypoglycemia, and five patients from the control group had seven similar episodes. Two patients from the SMBG group needed hospitalization for hyperglycemia compared to none in the control group.

Guerci B et al. (2003) 30 During this trial 78 patients re ported at least one episode of hypoglycemia, either symptomatic or asymptomatic: 53(10%) patients in the SMBG group and 25(5%) patients in the control group. These proportions were significantly different due to asymptomatic hypoglycemia alone (P=0.001). There was no serious episode of hypoglycemia reported.

Farmer A et al. (2007)33 This trial classified hypoglycemia as grade 2 (mild symptoms requiring minor intervention), grade 3 (moderate symptoms requiring immediate third party intervention), and grade 4 (unconscious). Fourteen patients in the control group had at least one grade 2 hypoglycemia episode, compared to 33 patients in the less intensive intervention group and 43 patients in the more intensive intervention group and 43 patients in the more intensive intervention group (p<0.001). One patient in the control group had a grade 3 hypoglycemic episode.

Thus, the limited evidence available indicates that SMBG increases the frequency of recognized hypoglycemia. This is due to an increase in asymptomatic low blood sugar readings, and also an increase in mild-to-moderate symptomatic episodes. There is scant evidence about the effect of SMBG on more clinically significant hypoglycemia. We judge the strength of evidence for SMBG increasing asymptomatic and mildly symptomatic hypoglycemia as moderate. [Moderate = Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.]

Key Question 4: Is there evidence that different frequencies of testing result in differences in improvements in A1c?

We did not identify any study that explicitly tested the effect of different frequency of SMBG on outcomes. This could have been accomplished either as an RCT (randomizing patients to differing frequencies and comparing outcomes) or as an analysis of outcomes within a cohort of patients using SMBG. We therefore were compelled to use an indirect method to examine this question. The indirect method compares the outcomes of studies that vary in the frequency of reported use of SMBG. Indirect methods have only a limited ability to control for other study level differences.

We used meta-regression to assess the effect of the reported frequency of SMBG use in the RCTs (measured as times/week) on differences in A1c level compared to control. No association was found (p=0.99). Therefore we draw no conclusion about the effect of frequency of SMBG monitoring on A1c values, and judge the strength of the evidence to be very low. [GRADE: Very Low = Any estimate of effect is very uncertain.]