Cardiovascular safety of phentermine and topiramate in a United States claims database

Abstract

BACKGROUND: Increased heart rate was seen in clinical trials for Qsymia (Q, phentermine [P] and topiramate [T]), an oral medication for weight management, but the effect on major adverse CV events (MACE) is unknown. OBJECTIVES: This study compared MACE incidence among current users of P, T, and P + T (including unbranded and Q) with former users of any of these medications. METHODS: A retrospective cohort study was conducted in MarketScan Commercial Claims and Medicare Supplemental data. Entry criteria were age ≥18 years and enrollment duration ≥6 months before the index date. Current use periods began at the first prescription dispensing date after a ≥180‐day clean period (initial entry) or after a gap of >60 days (subsequent use). Unexposed periods began after 60 days without exposure to any study medication. MACE comprised hospitalization for acute myocardial infarction (AMI) or stroke and in‐hospital CV‐related death as determined via ICD‐9‐CM diagnoses. MACE incidence rates (IR) and IR ratios (IRR) with 95% confidence intervals (CIs) were estimated for P, T, P+T, and Q current vs unexposed time in former users. Cohort trimming and stratification by propensity score were used to adjust for confounding. RESULTS: Current users of P + T (n = 19 184 people) contributed 3245 person‐years (py) (including 2587 py with Q); current P (n = 124,334) contributed 24 107 py; and current T (n = 316 388) contributed 64 607 py. Former users (n = 386 136) contributed 310 665 py. MACE and its components were generally increased in patients who were older, male, and who had prior CV events. As expected, AMI and stroke IR were higher than IR of in‐hospital CV death. Crude MACE IR/1000 py (95% CI) were 0.92 (0.19‐2.70) for P + T; 0.39 (0.01‐2.15), Q; 0.91 (0.57‐1.38), P; 3.37 (2.94‐3.85), T; and 2.00 (1.85‐2.17), unexposed. Adjusted IRR (95% CI) vs unexposed were 0.57 (0.19‐1.78) for P + T; 0.24 (0.03, 1.70), Q; 0.56 (0.34‐0.91), P; and 1.58 (1.33‐1.87), T. Results were quantitatively similar when limiting the comparison of current to former users within each specific medication cohort (eg, current vs former P + T use), although the point estimates were closer to the null. CONCLUSIONS: Unexpectedly, MACE risk was lower with current P and higher with current T vs unexposed. MACE risk in current P + T (including Q) users was lower than that among former users; however, the small number of events produced considerable statistical uncertainty, and results are consistent with a range of effects from a strong negative association to a modest positive association.