Association Between Phentermine and Risk of Major Adverse Cardiovascular Events: A Suite of Self-Controlled and Comparative Cohort Studies to Conquer Confounding

Background: Obesity is a highly prevalent disease with few pharmacotherapy options. Phentermine, an amphetamine-like stimulant, is approved for short-term weight management and is the most common obesity drug in the US. Previous population-based studies on adverse cardiovascular effects of phentermine had a limited sample size and a high risk of bias.

Objectives: To evaluate the association between short-term phentermine use and major adverse cardiovascular events (MACE).

Methods: We conducted a case-crossover study in MarketScan® Research Databases (2005-2019) and identified MACE cases as acute myocardial infarction, stroke (ischemic or hemorrhagic), or in-hospital cardiovascular death. We used validated algorithms and appropriate crosswalks to maximize the accuracy of outcome measurement. We required insurance benefits for at least six months before the event to measure drug exposure and time-varying covariates. Our primary analysis assigned the risk, wash-out, and control windows in two-month increments before the event date. We considered a study window "exposed" when at least one day of supply for phentermine was present. Sensitivity analyses varied window length and used prescription fill as the exposure. Considering the seasonal and temporal trends in phentermine use, we conducted case-time-control and case-case-time-control analyses to improve effect estimation. We also conducted a post hoc retrospective cohort study (2013-2019) with a new-user, active-comparator design to contrast the risk of MACE between phentermine and other obesity drugs (liraglutide, lorcaserin, or bupropion/ naltrexone). We followed patients for four months in an intention-to-treat framework and adjusted for potential confounders via the high dimensional propensity score procedure.

Results: We identified 325,927 MACE, of whom 405 cases had exposure to phentermine in any window before MACE (262 acute myocardial infarctions, 133 strokes, 10 in-hospital cardiovascular deaths). The adjusted conditional odds ratio was 0.75 (0.57, 0.98) in the case-crossover analysis, 0.68 (0.50, 0.91) in the case-time-control analysis, and 0.97 (0.55, 1.71) in case-case-time-control analysis. Sensitivity analyses confirmed robustness in primary findings, and the adjusted risk ratio in the post hoc analysis was 0.89 (0.54-1.46).

Conclusions: Case-crossover and case-time-control designs showed a protective effect likely due to residual confounding. The case-case-time-control technique likely reduced confounding but resulted in less precision. Our cohort study was subject to a different confounding structure, yet it could not refute findings in the primary analysis. In conclusion, short-term exposure to phentermine was not associated with a higher risk of MACE.