Study cohort and design

Details of the TECOS ( ClinicalTrials.gov : NCT00790205 ) design have been previously described. In brief, individuals ≥50 years of age with relatively well-controlled T2D (glycated hemoglobin [HbA _1c ] 6.5-8.0%) and prior ASCVD (history of coronary artery disease, ischemic cerebrovascular disease, or atherosclerotic peripheral arterial disease) were eligible for the trial. Participants were randomized 1:1 to sitagliptin versus placebo, and open-label use and titration of glucose-lowering agents (other than dipeptidyl peptidase-4 inhibitors or GLP-1 receptor analogues) was encouraged to aim for an individually appropriate glycemic target in all participants. The primary outcome was a four-point composite of CV death, myocardial infarction (MI), stroke, or hospitalization for unstable angina, each component of which was confirmed by blinded central adjudication. All patients provided written informed consent, and the study protocol was approved by the ethics committees for each of the 673 participating trial sites in 38 countries.

This post hoc analysis included all patients in the TECOS intention-to-treat cohort who had baseline BMI data available and used the same primary outcome. Secondary outcomes included the components of the primary outcome, time to all-cause death, and hospitalization for heart failure (hHF), mirroring the parent clinical trial.

Statistical analyses

Baseline characteristics are displayed by World Health Organization BMI categories: underweight/normal weight (BMI <25 kg/m ), overweight (BMI 25-29.9 kg/m ), obese class I (BMI 30-34.9 kg/m ), obese class II (BMI 35-39.9 kg/m ), and obese class III (BMI ≥40 kg/m ). Asian-specific BMI categories were applied to Asian individuals (as self-reported on the case report form; 22.1% of the study population): underweight/normal weight (BMI <23 kg/m ), overweight (BMI 23-27.4 kg/m ), obese class I (BMI 27.5-32.4 kg/m ), obese class II (BMI 32.5-37.4 kg/m ), and obese class III (BMI ≥37.5 kg/m ). The underweight and normal weight categories were combined because of the relatively small number of underweight individuals in the overall cohort (n = 69, 0.5%). Categorical variables are presented as number of participants (and percentage) and continuous variables are presented as mean (and standard deviation) or median (with 25th and 75th percentiles). Characteristics were compared across BMI categories using the chi-square test for categorical variables and the Student t test or Kruskal-Wallis test for continuous variables.

To examine whether obesity at baseline (defined as BMI ≥30 kg/m in the general population or ≥ 27.5 kg/m in Asians) was associated with glycemic control during the trial, the relationship between obesity and HbA _1c collected at different time points was modeled using repeated measures in a mixed model that included randomized study treatment and region.

Kaplan-Meier estimated cumulative incidence of the primary composite CV outcome was plotted according to baseline BMI category with differences among the categories assessed using the log-rank test. The unadjusted and adjusted relative associations between baseline BMI categories and the primary or secondary outcomes were evaluated using Cox proportional hazards models stratified by region, with a priori specified adjustment including sex, age, race, duration of diabetes, history of coronary disease, history of peripheral artery disease, history of chronic obstructive pulmonary disease, history of heart failure, history of atrial fibrillation/flutter, baseline estimated glomerular filtration rate, baseline HbA _1c , baseline urinary albumin:creatinine ratio, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, smoking status, baseline medications (e.g., aspirin, beta-blocker, statins, and ACEi/ARB), and randomized study group. The proportional hazards assumption was evaluated graphically using standardized score process. Hazard ratios (HRs) and 95% confidence intervals (CIs) are reported using combined underweight/normal weight group as the reference category.

To evaluate the possible modifying effect of sitagliptin on the relationship between baseline BMI categories and the primary composite outcome, we tested an interaction term of baseline BMI category and randomized assignment to sitagliptin versus placebo in the Cox proportional hazards model. Baseline BMI was also considered in its continuous form by evaluating the shape of the relationship between continuous baseline BMI and the primary composite outcome using restricted cubic spline transformation.

All analyses were based on the intention-to-treat population, with 2-sided statistical testing and P < .05 considered statistically significant. Statistical analyses were performed using SAS software (version 9.4; Cary, NC). SAS PROC MI was used to impute missing baseline data, and these imputed datasets were used for all modeling procedures.

The TECOS trial was funded by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.