Effect of Long-Term Metformin and Lifestyle in the Diabetes Prevention Program and Its Outcome Study on Coronary Artery Calcium

Abstract

Background: Despite the reduced incidence of coronary heart disease with intensive risk factor management, people with diabetes mellitus and prediabetes remain at increased coronary heart disease risk. Diabetes prevention interventions may be needed to reduce coronary heart disease risk. This approach was examined in the DPP (Diabetes Prevention Program) and the DPPOS (Diabetes Prevention Program Outcome Study), a long-term intervention study in 3234 subjects with prediabetes (mean±SD age, 64±10 years) that showed reduced diabetes risk with lifestyle and metformin compared with placebo over 3.2 years.

Methods: The DPPOS offered periodic group lifestyle sessions to all participants and continued metformin in the originally randomized metformin group. Subclinical atherosclerosis was assessed in 2029 participants with coronary artery calcium (CAC) measurements after an average of 14 years of follow-up. The CAC scores were analyzed continuously as CAC severity and categorically as CAC presence (CAC score >0) and reported separately in men and women.

Results: There were no CAC differences between lifestyle and placebo intervention groups in either sex. CAC severity and presence were significantly lower among men in the metformin versus the placebo group (age-adjusted mean CAC severity, 39.5 versus 66.9 Agatston units, P=0.04; CAC presence, 75% versus 84%, P=0.02), but no metformin effect was seen in women. In multivariate analysis, the metformin effect in men was not influenced by demographic, anthropometric, or metabolic factors; by the development of diabetes mellitus; or by use/nonuse of statin therapy.

Conclusions: Metformin may protect against coronary atherosclerosis in prediabetes and early diabetes mellitus among men.

Clinical trial registration: URL: http://www.clinicaltrials.gov. Unique identifier: NCT00038727.

Keywords: clinical trial [publication type]; coronary artery disease; prediabetic state; therapeutics.

Conflict of interest statement

Disclosures: Kieren Mather received a research grant from Novo IIT and has also medication or supply donations with aggregate value >$10K for other research studies from Novo, Sanofi, Merck, and Abbott. The other authors declare no conflict of interest.

© 2017 American Heart Association, Inc.

Figures

Fig 1. Consort diagram

* CAC exclusion criteria are not mutually exclusive.

Fig 2. Distribution of CAC Scores by Treatment Group and Sex

Each bar depicts the percent prevalence of CAC severity by color-coded severity category. There were no differences between CAC severity categories among treatment groups overall (p=0.69) or among men (p=0.08) or women (p=0.50)

Fig 3. Spearman correlations of CAC severity and covariates among all participants and by treatment group and sex

A. Baseline covariates B. Mean covariates values during follow-up Each circle represents the correlation coefficient of the bivariate analysis between CAC severity and the covariate, with the diameter of the circle proportionate to the correlation shown as a superscript to the circle. The color shading in the circle depicts the p value (black for p 0.05). Smooth circles indicates a positive correlation while scalloped circles depict a negative correlation.

Fig 3. Spearman correlations of CAC severity and covariates among all participants and by treatment group and sex

A. Baseline covariates B. Mean covariates values during follow-up Each circle represents the correlation coefficient of the bivariate analysis between CAC severity and the covariate, with the diameter of the circle proportionate to the correlation shown as a superscript to the circle. The color shading in the circle depicts the p value (black for p 0.05). Smooth circles indicates a positive correlation while scalloped circles depict a negative correlation.

Fig 4. Adjusted treatment effects of metformin vs. placebo (MET vs PLAC) and metformin vs. lifestyle (MET vs. ILS) on CAC severity stratified by sex

The treatment effects are stratified by sex along the columns, displayed along the rows and expressed as geometric mean ratio of metformin compared to placebo (MET vs. PLAC) and metformin compared to lifestyle (MET vs. ILS). The geometric mean ratio is calculated by taking the anti-log of the treatment effect coefficient from obtained the Tobit regression model with ln(CAC+1) as the outcome. Adjustments to the treatment effects are sequentially modeled as defined below. Baseline risk factors include family history of premature CVD, BMI, HbA1c, ln(ACR), non-HDL-C, HDLc, SBP, eGFR by CKD-epi, ln(CRP), ever smoke status. Risk factors during follow-up include mean levels for BMI, HbA1c, ln (ACR), non-HDL-C, BMI, ln(CRP), SBP, ACR, eGFR and years of statin. MODEL 0 Unadjusted treatment effect MODEL 1 Adjusted for age only MODEL 2 Adjusted for demographics: Model 1 + demographics (age, race/ethnicity) MODEL 3a Adjusted for baseline RFs: Model 2 + baseline risk factors (Family history of premature CVD, BMI, HbA1c, ACR, non-HDL-C, HDL-C, SBP, eGFR by CKD-epi, CRP, ever smoke status) MODEL 3b Adjusted for demographics and diabetes status: Model 2 + diabetes status at the time of the scan MODEL 4a Adjusted for demographics, risk factors at baseline and during follow-up : Model 3a + risk factors during follow-up MODEL 4b Adjusted for demographics, risk factors at baseline and diabetes status: Model 3a + diabetes status MODEL 5 Fully adjusted: Model 4A + diabetes status and duration