OBJECTIVE

We evaluated the effect of optimizing metformin dosing on glycemia and body weight in type 2 diabetes.

RESEARCH DESIGN AND METHODS

This was a prespecified analysis of 6,823 participants in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) taking metformin as the sole glucose-lowering drug who completed a 4- to 14-week (mean ± SD 7.9 ± 2.4) run-in in which metformin was adjusted to 2,000 mg/day or a maximally tolerated lower dose. Participants had type 2 diabetes for <10 years and an HbA1c ≥6.8% (51 mmol/mol) while taking ≥500 mg of metformin/day. Participants also received diet and exercise counseling. The primary outcome was the change in HbA1c during run-in.

RESULTS

Adjusted for duration of run-in, the mean ± SD change in HbA1c was −0.65 ± 0.02% (−7.1 ± 0.2 mmol/mol) when the dose was increased by ≥1,000 mg/day, −0.48 ± 0.02% (−5.2 ± 0.2 mmol/mol) when the dose was unchanged, and −0.23 ± 0.07% (−2.5 ± 0.8 mmol/mol) when the dose was decreased (n = 2,169, 3,548, and 192, respectively). Higher HbA1c at entry predicted greater reduction in HbA1c (P < 0.001) in univariate and multivariate analyses. Weight loss adjusted for duration of run-in averaged 0.91 ± 0.05 kg in participants who increased metformin by ≥1,000 mg/day (n = 1,894).

CONCLUSIONS

Optimizing metformin to 2,000 mg/day or a maximally tolerated lower dose combined with emphasis on medication adherence and lifestyle can improve glycemia in type 2 diabetes and HbA1c values ≥6.8% (51 mmol/mol). These findings may help guide efforts to optimize metformin therapy among persons with type 2 diabetes and suboptimal glycemic control.

Metformin is widely recommended as first-line therapy for management of hyperglycemia in patients with type 2 diabetes (1). Metformin is inexpensive, rarely associated with hypoglycemia when used alone, has beneficial effects on body weight and lipids, and appears to reduce the risk of cardiovascular events (2). Most individuals tolerate metformin well, although gastrointestinal side effects may require a switch to long-acting formulations, dose reduction, or discontinuation. While vitamin B12 deficiency can complicate therapy (3), lactic acidosis appears to be extremely rare when the drug is used appropriately (4).

Metformin is approved up to a total daily dose of 2,550 mg (or 2,000 mg/day for the extended-release [XR] form), although 2,000 mg/day is often considered the optimal dose, because higher doses may be associated with increased gastrointestinal side effects with marginal glycemic benefit (5). When administered as monotherapy to participants not receiving any other antidiabetic drug, metformin can reduce HbA1c by up to 2.0% (6,7) depending on baseline HbA1c levels and dose. However, the incremental effect of metformin when the dose is increased in persons taking less than a maximum dose is not clear. This is a relevant clinical issue, particularly for patients taking <2,000 mg/day who may have HbA1c values within a designated target range but still above normal and, hence, have suboptimal glycemic control. Metformin, when used alone, rarely causes hypoglycemia and is generally safe even in patients who do not have diabetes (8,9). Therefore, in most cases, there is little downside to increasing the dose to improve diabetes control. However, to improve our understanding of this issue, it would help to better define the extent to which average glycemia can be improved by optimizing metformin dosing.

The ongoing Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) offers insight in this respect. GRADE is designed to determine the relative effectiveness of four commonly used glucose-lowering medications when added to metformin. A run-in phase was conducted prior to randomization. To enter run-in, participants had to be taking metformin as the sole glucose-lowering medication and to have HbA1c levels ≥6.8% (51 mmol/mol). Although a large portion of participants entered run-in on a prescribed dosage of 2,000 mg/day, the actual dosage varied. During run-in, the dose was either maintained at 2,000 mg/day or adjusted, as tolerated, toward a goal of 2,000 mg/day. The dose adjustments included attempts to increase the dose to 2,000 mg/day in participants on less than that amount and to decrease the dose to 2,000 mg/day in participants taking more than that amount. In this study, we report the impact of optimizing metformin dose during the run-in phase of the GRADE study on glycemic control and body weight and evaluate predictors of glycemic response.

Study Design

GRADE is a multicenter randomized trial designed to assess the comparative effectiveness of glargine insulin, glimepiride, sitagliptin, and liraglutide when added to metformin. The GRADE protocol was described in detail in 2013 (10). A total of 11,230 volunteers underwent screening, including measurement of HbA1c; 7,764 participants returned for an initial run-in visit. Of those who had an initial run-in visit and a screening HbA1c, 6,850 (88.2%) completed the run-in. Recruitment for GRADE began in May 2013 and concluded in July 2017.

Eligibility

To be eligible for run-in, participants had to have had type 2 diabetes for ≤10 years and HbA1c ≥6.8% (51 mmol/mol) while taking ≥500 mg of metformin daily for at least 8 weeks prior to starting the run-in. To be eligible for randomization, participants had to complete a run-in period of 6–14 weeks if they were taking metformin at a dose other than 2,000 mg/day prior to screening or at least 4 weeks if taking 2,000 mg/day at the time of screening. For participants taking metformin 2,000 mg daily at screening, the daily dose was maintained at that level unless it had to be reduced due to intolerance. For participants taking a dose <2,000 mg/day or >2,000 mg/day, the dose was adjusted to 2,000 mg/day as tolerated. Participants were advised to take metformin with meals. Some participants who could not tolerate 2,000 mg/day of the immediate-release (IR) formulation were switched to the XR formulation to facilitate titration to 2,000 mg/day.

A total of 914 participants entered but did not complete run-in. A total of 300 reasons could be ascertained in 252 participants; 63 no longer met eligibility criteria, 83 were not judged an acceptable candidate, and 154 declined further participation (16 cited lack of time, 24 cited conflicting responsibilities, 24 cited concern about being assigned to an injectable medication, and 23 cited side effects from metformin).

Study Outcomes

The primary outcome was the change in HbA1c between the screening and final run-in visits. HbA1c was measured in blood samples obtained at the screening visit or within 30 days of that visit. Screening HbA1c measurements were performed at local clinical laboratories. HbA1c at final run-in was determined on samples sent to the study core laboratory at the University of Minnesota. Secondary outcome was change in weight, measured twice in light clothing, with the average used. Body weight was determined at the screening visit and again at either the final run-in or randomization visit. Adherence was determined based on participants’ self-reported responses to a questionnaire administered at the final run-in visit.

Data Analysis and Statistics

This study was performed as a prespecified analysis based on a proposal written before any analysis was done and reviewed and approved by the GRADE Study Publications and Presentations Subcommittee. This report was restricted to the run-in phase (mean ± SD duration of 7.9 ± 2.4 weeks).

Data are expressed as means and SDs for quantitative variables and counts and column percentages for qualitative variables. Comparisons between males and females were performed using the χ2 test of independence for qualitative variables and the Student t test with unequal variances for quantitative variables. The means and CIs in Fig. 1A and B are least-squares means and their 95% CIs from least squares regression models, which included duration of run-in as a covariate to adjust for nonuniform run-in periods. Supplementary Figure 1 plots change in HbA1c against screening HbA1c, and regression lines were estimated over the range from 6.8% (51 mmol/mol) to 8.0% (64 mmol/mol) by fitting an ANCOVA model for change in HbA1c as a function of screening HbA1c, an indicator for metformin change <1,000 mg/day, and an interaction between screening HbA1c and the indicator. The P value for the interaction term in this model was used to test equality of slopes. Dropping the interaction term in this model and obtaining the P value for the indicator for metformin <1,000 mg/day provided a test for a difference in intercepts. The trend lines in Supplementary Fig. 2 are from locally weighted scatterplot smoothers. Data were analyzed using R version 3.5.1.

Figure 1

Dose-dependent effects of metformin on glycemia and body weight. Changes with 95% CIs in HbA1c (A) and body weight (B) by magnitude of metformin dose change during run-in. Numerical mean values are listed within the bars. Actual mean ± SD dose changes for reduced, unchanged, and <500, 500–999, and ≥1,000 mg/day, respectively, were −602 ± 245, 0 ± 0, 295 ± 22, 503 ± 28, and 1,084 ± 183 in A and −608 ± 250, 0 ± 0, 296 ± 20, 504 ± 29, and 1,080 ± 180 in B. C: Bars representing binary variable based on whether HbA1c (A1c) had improved by 0.3% (3.3 mmol/mol) or better (−0.3% is the median change); P < 0.001 by χ2 test. D: Proportion of participants with an HbA1c <7% (53 mmol/mol) before and after the run-in. The data in A and B are based on a regression model with HbA1c change or weight change as the response and the category of metformin change as a predictor, adjusted for duration of run-in. The estimates and error bars are from the least squares means for metformin change and their 95% CIs. The P values are calculated from contrasts between the least squares means and are adjusted for multiple comparisons using a Dunnett adjustment. HbA1c values in percentage units can be converted to millimoles per mole using the NGSP HbA1c converter at https://www.ngsp.org/convert1.asp.

Figure 1

Dose-dependent effects of metformin on glycemia and body weight. Changes with 95% CIs in HbA1c (A) and body weight (B) by magnitude of metformin dose change during run-in. Numerical mean values are listed within the bars. Actual mean ± SD dose changes for reduced, unchanged, and <500, 500–999, and ≥1,000 mg/day, respectively, were −602 ± 245, 0 ± 0, 295 ± 22, 503 ± 28, and 1,084 ± 183 in A and −608 ± 250, 0 ± 0, 296 ± 20, 504 ± 29, and 1,080 ± 180 in B. C: Bars representing binary variable based on whether HbA1c (A1c) had improved by 0.3% (3.3 mmol/mol) or better (−0.3% is the median change); P < 0.001 by χ2 test. D: Proportion of participants with an HbA1c <7% (53 mmol/mol) before and after the run-in. The data in A and B are based on a regression model with HbA1c change or weight change as the response and the category of metformin change as a predictor, adjusted for duration of run-in. The estimates and error bars are from the least squares means for metformin change and their 95% CIs. The P values are calculated from contrasts between the least squares means and are adjusted for multiple comparisons using a Dunnett adjustment. HbA1c values in percentage units can be converted to millimoles per mole using the NGSP HbA1c converter at https://www.ngsp.org/convert1.asp.

Close modal

Participant Characteristics (Baseline and Final Run-in)

A total of 6,823 participants completed the run-in and had reported metformin doses and HbA1c measurements at screening and at a final run-in visit. This included 5,039 out of 5,047 randomized GRADE participants, as 8 of the 5,047 were missing HbA1c values at screening and could not be included in the analyses. This also included 1,784 who completed run-in but were not randomized because of ineligibility due to a final screening A1C outside the range of 6.8–8.5% (51–69 mmol/mol) inclusive (83% of the 1,784) or because they decided not to take a second drug, not to accept injections, or not to comply with other study requirements. Baseline characteristics determined at the initial run-in visit as well as changes in HbA1c, weight, metformin dosing, and metformin adherence during run-in are reported by sex in Table 1. Weight change could be calculated on 6,127 out of the 6,823 participants with complete HbA1c and metformin information. The predominance of male participants reflects participants recruited at Veterans Health Administration sites at which the male patient population is substantially higher. There was no sex difference in initial (screening) HbA1c.

Table 1

Characteristics by sex of GRADE participants at initial run-in and findings at the final run-in visit

AllFemaleMaleP value
Characteristics at initial run-in visit     
N 6,823 2,519 4,304  
 Age (years)* 56.1 ± 10.1 54.5 ± 9.7 57.1 ± 10.2 <0.001 
 Duration of run-in (weeks) 7.9 ± 2.4 8.1 ± 2.5 7.8 ± 2.3 <0.001 
 Race    <0.001 
  Caucasian 4,319 (63.3) 1,350 (53.6) 2,969 (69.0)  
  Native American 193 (2.8) 112 (4.4) 81 (1.9)  
  Asian 240 (3.5) 61 (2.4) 179 (4.2)  
  Hawaiian/Pacific Islander 47 (0.7) 9 (0.4) 38 (0.9)  
  African American 1,478 (21.7) 747 (29.7) 731 (17.0)  
  Other/multiple 450 (6.6) 200 (7.9) 250 (5.8)  
  Unknown/not reported 96 (1.4) 40 (1.6) 56 (1.3)  
 Hispanic 1,302 (19.2) 601 (24.1) 701 (16.4) <0.001 
 Weight (kg) 100.3 ± 22.5 93.4 ± 21.7 104.4 ± 21.9 <0.001 
 Height (cm) 170.3 ± 10.0 161.6 ± 7.1 175.5 ± 7.5 <0.001 
 BMI (kg/m234.5 ± 6.8 35.6 ± 7.4 33.9 ± 6.3 <0.001 
 Creatinine (mg/dL) 0.82 ± 0.20 0.69 ± 0.16 0.90 ± 0.18 <0.001 
 Lipids (mg/dL)*     
  Total cholesterol 163.8 ± 37.8 172.7 ± 37.1 158.7 ± 37.2 <0.001 
  HDL 43.3 ± 12.1 48.2 ± 13.0 40.5 ± 10.6 <0.001 
  LDL 90.7 ± 31.5 96.4 ± 32.0 87.3 ± 30.7 <0.001 
  Triglycerides 125.0 (88.0, 184.0) 122.0 (86.0, 175.0) 128.0 (89.0, 190.0) <0.001 
 HbA1c (%) at screening visit 8.1 ± 1.2 8.1 ± 1.2 8.1 ± 1.2 0.509 
 Metformin dose (mg/day) at screening visit    <0.001 
  <1,000 516 (7.6) 204 (8.1) 312 (7.2)  
  1,000–1,499 2,115 (31.0) 861 (34.2) 1,254 (29.1)  
  1,500–1,999 672 (9.8) 229 (9.1) 443 (10.3)  
  2,000 3,424 (50.2) 1,201 (47.7) 2,223 (51.6)  
  ≥2,000 96 (1.4) 24 (1.0) 72 (1.7)  
Findings at final run-in visit     
 HbA1c (%)     
  Final run-in visit 7.6 ± 0.9 7.6 ± 0.9 7.6 ± 0.9 0.523 
  Change −0.52 ± 0.94 −0.51 ± 0.91 −0.52 ± 0.95 0.822 
 Weight change (kg) during run-in −0.74 ± 2.10 −0.70 ± 1.95 −0.77 ± 2.19 0.185 
 Metformin dose (mg/day)     
  Final run-in visit    <0.001 
   <1,000 2 (0.0) 0 (0.0%) 2 (0.0)  
   1,000–1,499 284 (4.2) 133 (5.3%) 151 (3.5)  
   1,500–1,999 348 (5.1) 188 (7.5%) 160 (3.7)  
   2,000 6,189 (90.7) 2,198 (87.3%) 3,991 (92.7)  
  Change    <0.001 
   Reduced 192 (2.8) 70 (2.8) 122 (2.8)  
   Unchanged 3,548 (52.0) 1,250 (49.6) 2,298 (53.4)  
   <500 193 (2.8) 53 (2.1) 140 (3.3)  
   500–999 721 (10.6) 321 (12.7) 400 (9.3)  
   ≥1,000 2,169 (31.8) 825 (32.8) 1,344 (31.2)  
 Metformin type at final run-in    <0.001 
  IR 4,665 (68.4) 1,596 (63.4) 3,069 (71.3)  
  XR 2,158 (31.6) 923 (36.6) 1,235 (28.7)  
 Percent of pills missed last week    <0.001 
  0 4,852 (71.1) 1,725 (68.5) 3,127 (72.7)  
  0–20 1,848 (27.1) 749 (29.7) 1,099 (25.5)  
  >20 120 (1.8) 44 (1.7) 76 (1.8)  
AllFemaleMaleP value
Characteristics at initial run-in visit     
N 6,823 2,519 4,304  
 Age (years)* 56.1 ± 10.1 54.5 ± 9.7 57.1 ± 10.2 <0.001 
 Duration of run-in (weeks) 7.9 ± 2.4 8.1 ± 2.5 7.8 ± 2.3 <0.001 
 Race    <0.001 
  Caucasian 4,319 (63.3) 1,350 (53.6) 2,969 (69.0)  
  Native American 193 (2.8) 112 (4.4) 81 (1.9)  
  Asian 240 (3.5) 61 (2.4) 179 (4.2)  
  Hawaiian/Pacific Islander 47 (0.7) 9 (0.4) 38 (0.9)  
  African American 1,478 (21.7) 747 (29.7) 731 (17.0)  
  Other/multiple 450 (6.6) 200 (7.9) 250 (5.8)  
  Unknown/not reported 96 (1.4) 40 (1.6) 56 (1.3)  
 Hispanic 1,302 (19.2) 601 (24.1) 701 (16.4) <0.001 
 Weight (kg) 100.3 ± 22.5 93.4 ± 21.7 104.4 ± 21.9 <0.001 
 Height (cm) 170.3 ± 10.0 161.6 ± 7.1 175.5 ± 7.5 <0.001 
 BMI (kg/m234.5 ± 6.8 35.6 ± 7.4 33.9 ± 6.3 <0.001 
 Creatinine (mg/dL) 0.82 ± 0.20 0.69 ± 0.16 0.90 ± 0.18 <0.001 
 Lipids (mg/dL)*     
  Total cholesterol 163.8 ± 37.8 172.7 ± 37.1 158.7 ± 37.2 <0.001 
  HDL 43.3 ± 12.1 48.2 ± 13.0 40.5 ± 10.6 <0.001 
  LDL 90.7 ± 31.5 96.4 ± 32.0 87.3 ± 30.7 <0.001 
  Triglycerides 125.0 (88.0, 184.0) 122.0 (86.0, 175.0) 128.0 (89.0, 190.0) <0.001 
 HbA1c (%) at screening visit 8.1 ± 1.2 8.1 ± 1.2 8.1 ± 1.2 0.509 
 Metformin dose (mg/day) at screening visit    <0.001 
  <1,000 516 (7.6) 204 (8.1) 312 (7.2)  
  1,000–1,499 2,115 (31.0) 861 (34.2) 1,254 (29.1)  
  1,500–1,999 672 (9.8) 229 (9.1) 443 (10.3)  
  2,000 3,424 (50.2) 1,201 (47.7) 2,223 (51.6)  
  ≥2,000 96 (1.4) 24 (1.0) 72 (1.7)  
Findings at final run-in visit     
 HbA1c (%)     
  Final run-in visit 7.6 ± 0.9 7.6 ± 0.9 7.6 ± 0.9 0.523 
  Change −0.52 ± 0.94 −0.51 ± 0.91 −0.52 ± 0.95 0.822 
 Weight change (kg) during run-in −0.74 ± 2.10 −0.70 ± 1.95 −0.77 ± 2.19 0.185 
 Metformin dose (mg/day)     
  Final run-in visit    <0.001 
   <1,000 2 (0.0) 0 (0.0%) 2 (0.0)  
   1,000–1,499 284 (4.2) 133 (5.3%) 151 (3.5)  
   1,500–1,999 348 (5.1) 188 (7.5%) 160 (3.7)  
   2,000 6,189 (90.7) 2,198 (87.3%) 3,991 (92.7)  
  Change    <0.001 
   Reduced 192 (2.8) 70 (2.8) 122 (2.8)  
   Unchanged 3,548 (52.0) 1,250 (49.6) 2,298 (53.4)  
   <500 193 (2.8) 53 (2.1) 140 (3.3)  
   500–999 721 (10.6) 321 (12.7) 400 (9.3)  
   ≥1,000 2,169 (31.8) 825 (32.8) 1,344 (31.2)  
 Metformin type at final run-in    <0.001 
  IR 4,665 (68.4) 1,596 (63.4) 3,069 (71.3)  
  XR 2,158 (31.6) 923 (36.6) 1,235 (28.7)  
 Percent of pills missed last week    <0.001 
  0 4,852 (71.1) 1,725 (68.5) 3,127 (72.7)  
  0–20 1,848 (27.1) 749 (29.7) 1,099 (25.5)  
  >20 120 (1.8) 44 (1.7) 76 (1.8)  

Continuous variables are shown as the mean ± SD with a P value comparing males to females based on a two-sample t test with unequal variances. As an exception, triglycerides are shown as median (interquartile range) and analyzed by nonparametric rank test due to skew in the data. For categorical values, the table shows the number of participants and percentages for each category and a χ2P value comparing males to females. HbA1c values in percentage units can be converted to millimoles per mole using the NGSP HbA1c converter at https://www.ngsp.org/convert1.asp.

*

Age is indicated as reported at screening. Lipid measurements were carried out only at randomization, so HDL, LDL, and triglyceride values exclude 1,784 of the 6,823 participants who completed run-in but were not randomized.

Metformin Formulation and Adherence

Adherence data (Table 1) showed that 77.1% of all participants reported never missing a medication dose in the past week, while 1.8% of the participants reported missing ≥20% of the doses. Male participants reported slightly better adherence than females. At the initial run-in, 73.5% of the participants were on immediate-acting metformin IR, and 26.5% were on long-acting metformin XR. During run-in, 8.0% of participants initially on metformin IR and 2.9% of participants on XR switched metformin type. At completion of run-in, the proportions on IR and XR were 68.4% and 31.6%, respectively. There were no reports of severe hypoglycemia (requiring assistance from a third party) or gastrointestinal effects judged as severe adverse events during run-in.

Effect of Changes in Metformin Dosing on Glycemic Control and Body Weight

HbA1c levels, adjusted for duration of run-in (Fig. 1) or unadjusted (Table 2), decreased dependent on the magnitude of change in metformin dose. As shown in Fig. 1A, this included a decline in HbA1c even in participants whose dose of metformin did not change (−0.48 ± 0.02% [−5.2 ± 0.2 mmol/mol]). Therefore, in assessing the effect of metformin dose changes on glycemic control (Fig. 1A), the group that did not change metformin dose was used as an internal control in the sense that tests of contrasts were done comparing the metformin dose change groups with the internal control group (adjusting for duration of run-in). The decrease in HbA1c was significantly greater than this internal control group only for participants who increased the metformin dose by ≥1,000 mg/day (−0.65 ± 0.02% [−7.1 ± 0.2 mmol/mol]; P < 0.001). Of additional note, participants who reduced their metformin dose had a significantly smaller drop in HbA1c than the internal control group that did not reduce the metformin dose (−0.23 ± 0.07% [−2.5 ± 0.8 mmol/mol]; P = 0.002). Similar to changes in HbA1c, there was a decrease in weight in participants whose dose of metformin did not change (−0.66 ± 0.04 kg) (Fig. 1B). We again used the group that did not change metformin dose as an internal control (adjusting for duration of run-in). Weight decreased significantly compared with the internal control group only in participants who increased their metformin dose by ≥1,000 mg/day (−0.91 ± 0.05 kg; P < 0.001), while weight loss was nonsignificantly lower in participants whose metformin dose was reduced (−0.33 ± 0.16 kg; P = 0.146).

Table 2

Participant data by magnitude of metformin dose change

Metformin dose change 
ReducedUnchanged<500500–999≥1,000P value
N 192 3,548 193 721 2,169  
Duration of run-in (weeks) 7.2 ± 2.4 7.0 ± 2.3 8.5 ± 1.9 8.7 ± 1.9 9.1 ± 2.0 <0.001 
HbA1c (%)       
 Screening visit 7.8 ± 0.9 8.1 ± 1.2 8.1 ± 1.3 7.9 ± 1.1 8.2 ± 1.3 <0.001 
 Final run-in visit 7.6 ± 0.9 7.6 ± 0.9 7.7 ± 0.9 7.4 ± 0.9 7.5 ± 1.0 <0.001 
 Change −0.19 ± 0.67 −0.42 ± 0.90 −0.45 ± 1.05 −0.50 ± 0.87 −0.72 ± 0.99 <0.001 
Metformin dosing (mg/day)       
 Screening visit 2,219.3 ± 366.2 1,948.3 ± 217.1 1,668.4 ± 170.8 1,257.4 ± 326.5 904.9 ± 192.9 <0.001 
 Final run-in visit 1,617.2 ± 419.5 1,948.3 ± 217.1 1,963.7 ± 187.4 1,760.7 ± 325.9 1,988.7 ± 74.3 <0.001 
Metformin type at final run-in      <0.001 
 IR 116 (60.4) 2,616 (73.7) 148 (76.7) 409 (56.7) 1,376 (63.4)  
 XR 76 (39.6) 932 (26.3) 45 (23.3) 312 (43.3) 793 (36.6)  
Adherence to metformin       
 Percentage of pills missed last week      <0.001 
 0 144 (75.0) 2,610 (73.6) 148 (76.7) 487 (67.5) 1,463 (67.5)  
 0–20 43 (22.4) 891 (25.1) 43 (22.3) 209 (29.0) 662 (30.5)  
 >20 5 (2.6) 44 (1.2) 2 (1.0) 25 (3.5) 44 (2.0)  
Metformin dose change 
ReducedUnchanged<500500–999≥1,000P value
N 192 3,548 193 721 2,169  
Duration of run-in (weeks) 7.2 ± 2.4 7.0 ± 2.3 8.5 ± 1.9 8.7 ± 1.9 9.1 ± 2.0 <0.001 
HbA1c (%)       
 Screening visit 7.8 ± 0.9 8.1 ± 1.2 8.1 ± 1.3 7.9 ± 1.1 8.2 ± 1.3 <0.001 
 Final run-in visit 7.6 ± 0.9 7.6 ± 0.9 7.7 ± 0.9 7.4 ± 0.9 7.5 ± 1.0 <0.001 
 Change −0.19 ± 0.67 −0.42 ± 0.90 −0.45 ± 1.05 −0.50 ± 0.87 −0.72 ± 0.99 <0.001 
Metformin dosing (mg/day)       
 Screening visit 2,219.3 ± 366.2 1,948.3 ± 217.1 1,668.4 ± 170.8 1,257.4 ± 326.5 904.9 ± 192.9 <0.001 
 Final run-in visit 1,617.2 ± 419.5 1,948.3 ± 217.1 1,963.7 ± 187.4 1,760.7 ± 325.9 1,988.7 ± 74.3 <0.001 
Metformin type at final run-in      <0.001 
 IR 116 (60.4) 2,616 (73.7) 148 (76.7) 409 (56.7) 1,376 (63.4)  
 XR 76 (39.6) 932 (26.3) 45 (23.3) 312 (43.3) 793 (36.6)  
Adherence to metformin       
 Percentage of pills missed last week      <0.001 
 0 144 (75.0) 2,610 (73.6) 148 (76.7) 487 (67.5) 1,463 (67.5)  
 0–20 43 (22.4) 891 (25.1) 43 (22.3) 209 (29.0) 662 (30.5)  
 >20 5 (2.6) 44 (1.2) 2 (1.0) 25 (3.5) 44 (2.0)  

Continuous variables are shown as the mean ± SD with a P value based on an overall F test from an ANOVA. For categorical variables, the table shows the number of participants and percentages for each category and a χ2P value comparing the metformin dose-change categories. HbA1c values in percentage units can be converted to millimoles per mole using the NGSP HbA1c converter at https://www.ngsp.org/convert1.asp.

Among the participants in the no-change control group, 199 entered run-in taking <2,000 mg metformin/day. Therefore, we further analyzed the changes in HbA1c and weight, excluding those participants (Supplementary Fig. 3). The results were similar (compare Fig. 1 and Supplementary Fig. 3) with no changes in statistical significance.

Because 914 participants entered but did not complete run-in, the data in Fig. 1A and B were reanalyzed in a sensitivity analysis incorporating all participants entering run-in by estimating inverse probability weights based on the propensity to finish run-in and repeating the analysis, applying these weights to the analytic sample of 6,823 participants who did complete run-in. This analysis revealed similar findings with no change in significance (data not shown). A further sensitivity analysis was performed on the data in Fig. 1A and B, stratified by whether participants received metformin IR or XR. Trends for the associations of metformin dose change with change in HbA1c and change in weight were similar for those who received both IR and XR metformin. However, one exception was that those who received metformin XR and increased their dose <500 mg/day experienced a considerably smaller decrease in both HbA1c and weight, although estimates for this subgroup were imprecise due to a small (n = 45 for HbA1c and n = 40 for weight) subgroup size (data not shown).

Because of the large variability in change in HbA1c, a binary variable was created based on whether HbA1c had improved by 0.3% (3.3 mmol/mol) or better (−0.3% was the median change). There was a significant difference in the percent of participants achieving an HbA1c decrease of at least 0.3% among metformin dose-change groups (P < 0.001 by χ2) (Fig. 1C). There were also metformin dose-dependent differences in the percentage of participants reaching an HbA1c <7.0% (53 mmol/mol) (Fig. 1D).

Parameters Predicting HbA1c Change With Incremental Metformin

Regression analyses were carried out for several parameters expected to affect the change in HbA1c following increments in metformin dosage (Table 3 and Supplementary Fig. 2). The dominant factor in multivariate analyses was the HbA1c level prior to dose change. These relationships are depicted using nonlinear correlation for participants whose metformin dose was increased by ≥1,000 mg/day or changed by <1,000 mg/day (Supplementary Fig. 1). An inflection point is apparent just above an HbA1c value of 8.0% (64 mmol/mol), suggesting a steeper relationship above that point. Linear regression was carried out in order to look at the change in HbA1c as related to screening HbA1c values between 6.8% and 8.0% (51 and 64 mmol/mol) in Supplementary Fig. 1. These values are in the range of often-cited target levels (1113). Although the slopes did not differ, the differences in elevation of the curves indicated greater HbA1c-lowering effect for dose increments ≥1,000 mg/day.

Table 3

Factors associated with change in HbA1cby univariate and multivariate analyses

PredictorHbA1c (%) change estimateP value (univariate)HbA1c (%) change estimateP value (multivariate)
HbA1c (%) at screening −0.4981 <0.001 −0.5003 <0.001 
Age 0.0108 <0.001 −0.0052 <0.001 
Race     
 American Indian/Alaska Native −0.854 <0.001 −0.4019 <0.001 
 Asian 0.0352 0.566 −0.0583 0.230 
 Hawaiian/Pacific Islander −0.0241 0.859 0.0641 0.533 
 Black or African American −0.1324 <0.001 −0.0915 <0.001 
 Other/multiple −0.2919 <0.001 −0.0915 0.012 
 Unknown/not reported −0.2769 0.004 −0.1966 0.010 
Male −0.0052 0.824 0.0289 0.184 
Weight at screening 0.0026 <0.001 0.0022 <0.001 
Hispanic −0.2811 <0.001 −0.0816 0.001 
Creatinine 0.2728 <0.001 −0.3141 <0.001 
Diabetes duration 0.0405 <0.001 0.0444 <0.001 
PredictorHbA1c (%) change estimateP value (univariate)HbA1c (%) change estimateP value (multivariate)
HbA1c (%) at screening −0.4981 <0.001 −0.5003 <0.001 
Age 0.0108 <0.001 −0.0052 <0.001 
Race     
 American Indian/Alaska Native −0.854 <0.001 −0.4019 <0.001 
 Asian 0.0352 0.566 −0.0583 0.230 
 Hawaiian/Pacific Islander −0.0241 0.859 0.0641 0.533 
 Black or African American −0.1324 <0.001 −0.0915 <0.001 
 Other/multiple −0.2919 <0.001 −0.0915 0.012 
 Unknown/not reported −0.2769 0.004 −0.1966 0.010 
Male −0.0052 0.824 0.0289 0.184 
Weight at screening 0.0026 <0.001 0.0022 <0.001 
Hispanic −0.2811 <0.001 −0.0816 0.001 
Creatinine 0.2728 <0.001 −0.3141 <0.001 
Diabetes duration 0.0405 <0.001 0.0444 <0.001 

The coefficient estimates and P values are from least squares regression models. HbA1c values in percentage units can be converted to millimoles per mole using the NGSP HbA1c converter at https://www.ngsp.org/convert1.asp.

In addition to initial HbA1c level, other factors associated with change in HbA1c were examined in a multivariable model (Table 3). Factors significantly associated with a decrease in HbA1c included older age; African American, Native American, or other/multiple race (compared with white race); Hispanic ethnicity; and higher creatinine. Factors significantly associated with an increase in HbA1c included higher baseline weight and longer duration of diabetes.

We report the effect of adjusting metformin dosage in 6,823 participants with an initial HbA1c level of ≥6.8% (51 mmol/mol) over a mean run-in period of 7.9 ± 2.4 weeks. Adjusting dosage to 2,000 mg/day or to a maximally tolerated lower dose was associated with a mean decrease in HbA1c of 0.52 ± 0.94% (5.7 ± 10.3 mmol/mol). Greater increases in dosing were associated with greater reductions in HbA1c. However, participants with no change in metformin dosing, and even those with a dose reduction, also exhibited decreases in HbA1c levels, suggesting that improvement in adherence to the medication and/or lifestyle behavior also contributed.

Several reports have compared the effects of different doses of metformin on HbA1c levels (5,7,1419). For total daily doses of 1,000 mg versus 2,000 mg (15,16,19) or 1,500 mg versus 3,000 mg (17), improvements in HbA1c were in the range of 0.2–0.3% (2.2–3.3 mmol/mol). However, another study (5) reported a difference of ∼0.8% (8.7 mmol/mol) for 1,000 versus 2,000 mg daily after subtracting out the change in HbA1c in placebo-treated participants. In that study, the placebo group had a 1.2% (13.1 mmol/mol) increase in HbA1c over 14 weeks, while baseline HbA1c values for all groups were relatively high at ∼10.0% (86 mmol/mol).

Our current study differs from the above reports in an important way. Those studies compared the effectiveness of metformin in participants not previously taking the drug who were randomly assigned to different doses. In this study, we adjusted metformin dosing in persons already taking the drug, typical of what would be done in clinical practice. In some of the previous studies (14,18), comparative doses of metformin were examined in persons also taking another glucose-lowering drug. We are not aware of other studies that compared the effects of differential adjustments in metformin dosage in participants already taking the drug and not using another glucose-lowering drug.

Three smaller studies compared increments in metformin dosing from 1,000 to 2,000 mg/day to continuation of metformin 1,000 mg/day with additions of sitagliptin (20), vildagliptin (21), or rosiglitazone (22). Respectively, for participants whose metformin dose was increased, these studies showed changes in HbA1c of −0.80% (−8.7 mmol/mol) from a baseline mean of 8.7% (72 mmol/mol), −0.37% (−4.0 mmol/mol) from a baseline mean of 7.3% (56 mmol/mol), and −0.71% (−7.8 mmol/mol) from a baseline mean of 8.0% (64 mmol/mol). These reports did not examine differential increments in metformin dosing and, importantly, did not provide a control group (metformin unchanged) to estimate the effect of lifestyle and adherence recommendations or of trial entry. Moreover, the relationships of increments in HbA1c to pretreatment HbA1c or other factors affecting metformin responsiveness were not assessed, beyond one report showing a greater effect of incrementing metformin in participants whose baseline HbA1c was >8% (64 mmol/mol) compared with <8% (change of −0.46% or −5.0 mmol/mol vs. −0.31% or −3.4 mmol/mol). Taken together, the data from these studies are roughly in agreement with our findings.

We acknowledge that we adjusted metformin dosing in the context of a clinical trial in which the drug was provided free of cost and with emphasis on medication adherence, diet, and exercise. These factors may in part explain why participants with no change in metformin dose had a decrease in HbA1c and weight. Moreover, there is evidence that glycemia improves upon entry into a clinical trial per se (23). Despite this, adherence and lifestyle modification alone cannot explain all of the improvement in HbA1c with optimization of dosage, because HbA1c fell significantly more in participants whose dose was increased by 1,000 mg/day, when compared with the group for which dose was unchanged (Fig. 1).

We observed that in participants whose metformin dose was reduced, either because they were taking >2,000 mg at the time of initial run-in or because of intolerance, HbA1c improved less than in participants whose dose was unchanged (Fig. 1). The reduction in metformin dose was from an average of 2,219 to 1,617 mg/day, while the average metformin dose of those who were unchanged was 1,948 mg/day. This differs from an older report in which there was no difference in HbA1c levels, with 2,500 mg daily compared with 2,000 mg/day (5). However, that study compared different doses in different individuals, whereas we made dose adjustments within the same participants. This discrepancy cannot be attributed to the emphasis on medication adherence and lifestyle change in our study, as that should not have differed among participants.

Although multiple factors beyond metformin dose change per se were predictive of the effect of metformin on HbA1c levels, by far the strongest was the HbA1c level prior to metformin dose adjustment (Table 3 and Supplementary Fig. 2). Baseline HbA1c levels contributed independently both among participants whose dose was increased by ≥1,000 mg/day and for participants whose dose was not increased or increased by lesser amounts. In multivariate analyses, older age, African American, Native American, or other/multiple race (compared with white), Hispanic ethnicity, and higher creatinine levels were less strongly associated with a decrease in HbA1c (Table 3), while higher weight and longer diabetes duration were associated with a mild increase. Consistent with our findings, an examination of electronic health records of 19,672 patients showed that African American heritage was associated with a greater reduction in HbA1c than European American following prescription of metformin (24). However, an association with race/ethnicity was not reported in other studies (25,26).

Adherence to metformin therapy during run-in appeared to be strong in that 71.1% of participants reported no missed pills in the past week, and 27.1% reported missing only 0–20% (Table 1). Further supporting tolerance to metformin, 90.7% of participants were at 2,000 mg/day (Table 1), and the mean dose at final screening was 1,932 mg/day. Importantly, intolerance was expected to be low, because all participants were required to be taking metformin in order to be eligible for screening. This may explain why nonadherence was lower in this study than in other reports (27).

The American College of Physicians (ACP) recently suggested a target HbA1c value of 8.0% (64 mmol/mol) as appropriate for “most patients with type 2 diabetes” (12). Although many consider this ACP target of 8.0% as too high for many individuals (13), we point out that the curvilinear relationships in Supplementary Fig. 1 show that the impact of metformin dose adjustment was greater for those with an initial HbA1c level of ≥8.0% (64 mmol/mol). Therefore, increasing metformin may be particularly effective for those with an HbA1c level above the ACP target.

Our findings have broad clinical implications. Figure 1 shows that there is potential for metformin dose adjustment to improve glycemic control even in individuals with lower HbA1c values or values already within guidelines (1113). Moreover, Table 2 shows that participants whose dose was increased by ≥1,000 mg/day entered screening taking a mean of 905 mg metformin daily with a mean HbA1c of 8.2% (66 mmol/mol). These observations, in common, suggest that there may be benefit from metformin dose adjustment in many patients with type 2 diabetes. In support of this concept, there were no reports of severe hypoglycemia during the GRADE run-in. Moreover, although not specifically tabulated, participant-reported hypoglycemia of any degree was very unusual, as expected based on other reports indicating that metformin causes little hypoglycemia, when used as the sole glucose-lowering drug (2830). Further, we point out that metformin is often used in the absence of diabetes for prevention of the disease (8) or for therapy of polycystic ovarian disease in persons without diabetes (31).

In contrast, beyond hypoglycemia, there are other well-known adverse effects of metformin (4), and benefit must be considered relative to risk. We found that the fall in HbA1c was very modest for individuals entering run-in with HbA1c values in the lower range (Supplementary Fig. 1).

There are some limitations to this study. The run-in period was variable and as low as 4 weeks for participants entering the study already taking 2,000 mg/day, in whom the metformin dose was not changed. Thus, changes in HbA1c in those with short duration of follow-up may have underestimated the effect that might have been seen if a usual interval HbA1c had been obtained (i.e., ≥2 months). However, this effect may have been mitigated in that all participants had to be taking metformin for at least 8 weeks at the time of screening. As noted, we adjusted metformin dosing in the context of a clinical trial and, as per study protocol, this was done along with recommendations for medication adherence and changes in behavior, diet, and exercise. The initial HbA1c values were determined by local laboratories, while the HbA1c values at final run-in were all determined by our study core laboratory. However, the methodology for HbA1c is now well standardized in relation to average glycemia and would not be expected to differ by baseline HbA1c or metformin dose (32,33). Adherence to therapy was based on participant self-report rather than pill counts. Moreover, adherence to therapy should not have been affected by affordability, because metformin was provided free of cost to all participants.

In conclusion, adjusting metformin dosing to 2,000 mg/day or to a maximally tolerated lower dose combined with promoting lifestyle changes and medication adherence improved glycemic control by an average of 0.52% (5.7 mmol/mol) in patients who had an average HbA1c of ≥6.8% (51 mmol/mol) and reported taking an average of 1,543 mg/day at baseline. The improvement was greater in those with a higher initial HbA1c. These findings serve as a guide that could help to improve management in persons on submaximal metformin therapy.

Clinical trial reg. no. NCT01794143, clinicaltrials.gov

Funding. The GRADE Study is supported by the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health (U34-DK-088043 and U01-DK-098246). The American Diabetes Association supported the initial planning meeting for the U34 proposal. The National Heart, Lung, and Blood Institute and the Centers for Disease Control and Prevention are also providing funding support. Educational materials have been provided by the National Diabetes Education Program. Material support in the form of donated medications and supplies has been provided by BD Biosciences, Bristol-Myers Squibb, Merck, Novo Nordisk, Roche Diagnostics, and Sanofi.

Duality of Interest. W.I.S. reports support from the Veterans Health Administration. L.S.P. has served on scientific advisory boards for Boehringer Ingelheim and Janssen and has or had research support from Merck, Amylin, Eli Lilly, Novo Nordisk, Sanofi, PhaseBio, Roche, Abbvie, Vascular Pharmaceuticals, Janssen, Glaxo SmithKline, and the Cystic Fibrosis Foundation outside the submitted work. L.S.P. is also a cofounder, officer, board member, and stockholder of a company, Diasyst, Inc., that is developing software aimed to help improve diabetes management. L.S.P. is supported in part by the Veterans Health Administration. D.J.W. reports that grants from the National Institute of Diabetes and Digestive and Kidney Diseases funded the trial during the conduct of the study and that other support from Novo Nordisk was also received outside the submitted work. S.P.F. reports grants from the National Institute of Diabetes and Digestive and Kidney Diseases during the conduct of the study. V.R.A. reports grants from Calibra Bioceuticals, Janssen Pharmaceuticals, and Theracos, Inc.; personal fees from BD Biosciences and Zafgen; grants and personal fees from AstraZeneca/Bristol-Meyers Squibb, Novo Nordisk, and Sanofi; and other support from Adocia and Merck outside the submitted work. M.H.T. reports grants from the University of Michigan (Ann Arbor, MI) during the conduct of the study and other support from Eli Lilly and Company outside the submitted work. No other potential conflicts of interest relevant to this article were reported.

This work is not intended to reflect the official opinion of the Veterans Health Administration or the U.S. government.

Author Contributions. W.I.S. contributed to the design, acquisition and interpretation of data, supervision and management of the research, and drafting and critical review of the manuscript. L.S.P. contributed to the design, interpretation of data, and critical revision of the manuscript. D.J.W. contributed to the design, interpretation of data, and critical revision of the manuscript. S.P.F. contributed to the design, acquisition and interpretation of data, supervision and management of the research, and critical review of the manuscript. A.W.C. contributed to the acquisition of data, interpretation of data and results, supervision and management of research, and critical revision of the manuscript. M.T. contributed to the acquisition of data, supervision and management of research, and critical revision of the manuscript. M.P. contributed to the acquisition of data and drafting of the manuscript. J.C. contributed to the acquisition of data, supervision and management of research, and critical revision of the manuscript. P.A.H. contributed to the design, acquisition, analysis, and interpretation of data, supervision and management of research, and drafting and critical review of the manuscript. A.C. contributed to the acquisition and interpretation of data, supervision and management of research, and critical revision of the manuscript. V.R.A. contributed to the acquisition and interpretation of data, supervision and management of research, and critical review of the manuscript. M.H.T. contributed to the acquisition and interpretation of data, supervision and management of research, and drafting and critical revision of the manuscript. J.K. contributed to the design, acquisition and interpretation of data, and critical revision of the manuscript. N.R. contributed to the design, interpretation of data, supervision and management of research, and critical revision of the manuscript. N.M.B. and N.Y. contributed to the statistical analysis and interpretation of data and results, drafting of the statistical sections, tables, and figures, and critical revision of the manuscript. All authors affirm that authorship is merited based on the International Committee of Medical Journal Editors authorship criteria. W.I.S. and N.Y. are the guarantors of this work and, as such, had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented in abstract form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22–28 June 2018.

1.
American Diabetes Association
.
8. Pharmacologic approaches to glycemic treatment: Standards of Medical Care in Diabetes—2018
.
Diabetes Care
2018
;
41
(
Suppl. 1
):
S73
S85
2.
Holman
RR
,
Paul
SK
,
Bethel
MA
,
Matthews
DR
,
Neil
HA
.
10-year follow-up of intensive glucose control in type 2 diabetes
.
N Engl J Med
2008
;
359
:
1577
1589
3.
Aroda
VR
,
Edelstein
SL
,
Goldberg
RB
, et al.;
Diabetes Prevention Program Research Group
.
Long-term metformin use and vitamin B12 deficiency in the diabetes prevention program outcomes study
.
J Clin Endocrinol Metab
2016
;
101
:
1754
1761
4.
Bailey
CJ
.
Metformin: historical overview
.
Diabetologia
2017
;
60
:
1566
1576
5.
Garber
AJ
,
Duncan
TG
,
Goodman
AM
,
Mills
DJ
,
Rohlf
JL
.
Efficacy of metformin in type II diabetes: results of a double-blind, placebo-controlled, dose-response trial
.
Am J Med
1997
;
103
:
491
497
6.
Nathan
DM
,
Buse
JB
,
Davidson
MB
, et al.;
American Diabetes Association
;
European Association for Study of Diabetes
.
Medical management of hyperglycemia in type 2 diabetes: a consensus algorithm for the initiation and adjustment of therapy: a consensus statement of the American Diabetes Association and the European Association for the Study of Diabetes
.
Diabetes Care
2009
;
32
:
193
203
7.
Hirst
JA
,
Farmer
AJ
,
Ali
R
,
Roberts
NW
,
Stevens
RJ
.
Quantifying the effect of metformin treatment and dose on glycemic control
.
Diabetes Care
2012
;
35
:
446
454
8.
Hostalek
U
,
Gwilt
M
,
Hildemann
S
.
Therapeutic use of metformin in prediabetes and diabetes prevention
.
Drugs
2015
;
75
:
1071
1094
9.
Radosh
L
.
Drug treatments for polycystic ovary syndrome
.
Am Fam Physician
2009
;
79
:
671
676
10.
Nathan
DM
,
Buse
JB
,
Kahn
SE
, et al.;
GRADE Study Research Group
.
Rationale and design of the Glycemia Reduction Approaches in Diabetes: a Comparative Effectiveness Study (GRADE)
.
Diabetes Care
2013
;
36
:
2254
2261
11.
American Diabetes Association
.
6. Glycemic targets: Standards of Medical Care in Diabetes—2018
.
Diabetes Care
2018
;
41
(
Suppl. 1
):
S55
S64
12.
Tung
EL
,
Davis
AM
,
Laiteerapong
N
.
Glycemic control in nonpregnant adults with type 2 diabetes
.
JAMA
2018
;
319
:
2430
2431
13.
Abbasi
J
.
For patients with type 2 diabetes, what’s the best target hemoglobin A1C
?
JAMA
2018
;
319
:
2367
2369
14.
Bosi
E
,
Dotta
F
,
Jia
Y
,
Goodman
M
.
Vildagliptin plus metformin combination therapy provides superior glycaemic control to individual monotherapy in treatment-naive patients with type 2 diabetes mellitus
.
Diabetes Obes Metab
2009
;
11
:
506
515
15.
Fujioka
K
,
Brazg
RL
,
Raz
I
, et al
.
Efficacy, dose-response relationship and safety of once-daily extended-release metformin (Glucophage XR) in type 2 diabetic patients with inadequate glycaemic control despite prior treatment with diet and exercise: results from two double-blind, placebo-controlled studies
.
Diabetes Obes Metab
2005
;
7
:
28
39
16.
Goldstein
BJ
,
Feinglos
MN
,
Lunceford
JK
,
Johnson
J
,
Williams-Herman
DE
;
Sitagliptin 036 Study Group
.
Effect of initial combination therapy with sitagliptin, a dipeptidyl peptidase-4 inhibitor, and metformin on glycemic control in patients with type 2 diabetes
.
Diabetes Care
2007
;
30
:
1979
1987
17.
Grant
PJ
.
The effects of high- and medium-dose metformin therapy on cardiovascular risk factors in patients with type II diabetes
.
Diabetes Care
1996
;
19
:
64
66
18.
Lewin
A
,
Lipetz
R
,
Wu
J
,
Schwartz
S
.
Comparison of extended-release metformin in combination with a sulfonylurea (glyburide) to sulfonylurea monotherapy in adult patients with type 2 diabetes: a multicenter, double-blind, randomized, controlled, phase III study
.
Clin Ther
2007
;
29
:
844
855
19.
Williams-Herman
D
,
Johnson
J
,
Teng
R
, et al
.
Efficacy and safety of sitagliptin and metformin as initial combination therapy and as monotherapy over 2 years in patients with type 2 diabetes
.
Diabetes Obes Metab
2010
;
12
:
442
451
20.
Frias
JP
,
Zimmer
Z
,
Lam
RLH
, et al
.
Double-blind, randomized clinical trial assessing the efficacy and safety of early initiation of sitagliptin during metformin uptitration in the treatment of patients with type 2 diabetes: the CompoSIT-M study
.
Diabetes Obes Metab
2019
;
21
:
1128
1135
21.
Filozof
C
,
Schwartz
S
,
Foley
JE
.
Effect of vildagliptin as add-on therapy to a low-dose metformin
.
World J Diabetes
2010
;
1
:
19
26
22.
Weissman
P
,
Goldstein
BJ
,
Rosenstock
J
, et al
.
Effects of rosiglitazone added to submaximal doses of metformin compared with dose escalation of metformin in type 2 diabetes: the EMPIRE Study
.
Curr Med Res Opin
2005
;
21
:
2029
2035
23.
Gale
EA
,
Beattie
SD
,
Hu
J
,
Koivisto
V
,
Tan
MH
.
Recruitment to a clinical trial improves glycemic control in patients with diabetes
.
Diabetes Care
2007
;
30
:
2989
2992
24.
Williams
LK
,
Padhukasahasram
B
,
Ahmedani
BK
, et al
.
Differing effects of metformin on glycemic control by race-ethnicity
.
J Clin Endocrinol Metab
2014
;
99
:
3160
3168
25.
Diabetes Prevention Program Research Group
.
Long-term effects of metformin on diabetes prevention: identification of subgroups that benefited most in the Diabetes Prevention Program and Diabetes Prevention Program Outcomes Study
.
Diabetes Care
2019
;
42
:
601
608
26.
Goonesekera
SD
,
Yang
MH
,
Hall
SA
,
Fang
SC
,
Piccolo
RS
,
McKinlay
JB
.
Racial ethnic differences in type 2 diabetes treatment patterns and glycaemic control in the Boston Area Community Health Survey
.
BMJ Open
2015
;
5
:
e007375
27.
McGovern
A
,
Tippu
Z
,
Hinton
W
,
Munro
N
,
Whyte
M
,
de Lusignan
S
.
Comparison of medication adherence and persistence in type 2 diabetes: a systematic review and meta-analysis
.
Diabetes Obes Metab
2018
;
20
:
1040
1043
28.
UK Prospective Diabetes Study (UKPDS) Group
.
Effect of intensive blood-glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34)
[published correction appears in Lancet 1998;352:1558].
Lancet
1998
;
352
:
854
865
29.
Holstein
A
,
Egberts
EH
.
Risk of hypoglycaemia with oral antidiabetic agents in patients with type 2 diabetes
.
Exp Clin Endocrinol Diabetes
2003
;
111
:
405
414
30.
Lalau
JD
,
Race
JM
,
Brinquin
L
.
Lactic acidosis in metformin therapy. Relationship between plasma metformin concentration and renal function
.
Diabetes Care
1998
;
21
:
1366
1367
31.
Yang
PK
,
Hsu
CY
,
Chen
MJ
, et al
.
The efficacy of 24-month metformin for improving menses, hormones, and metabolic profiles in polycystic ovary syndrome
.
J Clin Endocrinol Metab
2018
;
103
:
890
899
32.
Hanas
R
,
John
G
;
International HBA1c Consensus Committee
.
2010 consensus statement on the worldwide standardization of the hemoglobin A1C measurement
.
Diabetes Care
2010
;
33
:
1903
1904
33.
Hanas
R
,
John
WG
;
International HbA1c Consensus Committee
.
2013 update on the worldwide standardization of the hemoglobin A(1c) measurement
.
Clin Chem Lab Med
2013
;
51
:
1041
1042
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at https://www.diabetesjournals.org/content/license.

Supplementary data