Polypharmacy and Risk of Major Cardiovascular Events Among Chinese Patients With Diabetes: A Retrospective Cohort Study Using Instrumental Variable Analysis
Author(s)
Nan Peng, Ph.D.1, Ermo Chen, Ph.D.2, Jinbo Zhang, BS3, Jiale Yang, BS3, Yuqing Fan, BS4, Dongning Yao, Ph.D.4, Pei Gao, Ph.D.2, Jing Wu, PhD1, Gordon Liu, PhD2, Beini Lyu, Ph.D.2.
1Tianjin University, Tianjin, China, 2Peking University, Beijing, China, 3China Pharmaceutical University, Nanjing, China, 4Nanjing Medical University, Nanjing, China.
1Tianjin University, Tianjin, China, 2Peking University, Beijing, China, 3China Pharmaceutical University, Nanjing, China, 4Nanjing Medical University, Nanjing, China.
OBJECTIVES: This study examined the nonlinear associations between polypharmacy and composite major adverse cardiovascular events (MACE) in Chinese patients with diabetes, using causal inference methods to address endogeneity.
METHODS: We used data from the Yinzhou Health Information System, a regional population-based electronic health record platform with 98% coverage in Yinzhou District, Ningbo, China, spanning from Jan 2015 to Oct 2021. Adults with at least two outpatient diabetes diagnoses (≥30 days apart) or one inpatient diagnosis were included, with follow-up from the date of diagnosis until death or their last recorded prescription. The primary outcome was MACE. To mitigate confounding, we applied an instrumental variable (IV) approach within a Cox proportional hazards framework, using practice-level prescribing preference as the instrument. Threshold regression was used to detect nonlinear exposure-response patterns. IV validity was confirmed via Hausman tests and Kleibergen-Paap statistics.
RESULTS: Among 73,766 eligible patients, 5,116 (6.94%) experienced MACE. Those with MACE were older (69.3 vs. 61.3 years), had higher unemployment rates (29.4% vs. 23.8%), and exhibited a greater proportion of unmarried individuals (12.9% vs. 9.4%) compared to patients without MACE. A nonlinear relationship was observed: when the average number of medications was below four, medication use was associated with reduced MACE risk (HR = 0.78, p < 0.01). However, when medication use exceeded four drugs, risk increased significantly (HR = 1.75, p < 0.01). Instrument validity and strength were confirmed (Hausman p < 0.01; Kleibergen-Paap F = 18.4).
CONCLUSIONS: Using robust causal methods and longitudinal data, our study suggests an optimal threshold for medication use among patients with diabetes. Both undertreatment and excessive polypharmacy were associated with elevated risks of major cardiovascular events.
METHODS: We used data from the Yinzhou Health Information System, a regional population-based electronic health record platform with 98% coverage in Yinzhou District, Ningbo, China, spanning from Jan 2015 to Oct 2021. Adults with at least two outpatient diabetes diagnoses (≥30 days apart) or one inpatient diagnosis were included, with follow-up from the date of diagnosis until death or their last recorded prescription. The primary outcome was MACE. To mitigate confounding, we applied an instrumental variable (IV) approach within a Cox proportional hazards framework, using practice-level prescribing preference as the instrument. Threshold regression was used to detect nonlinear exposure-response patterns. IV validity was confirmed via Hausman tests and Kleibergen-Paap statistics.
RESULTS: Among 73,766 eligible patients, 5,116 (6.94%) experienced MACE. Those with MACE were older (69.3 vs. 61.3 years), had higher unemployment rates (29.4% vs. 23.8%), and exhibited a greater proportion of unmarried individuals (12.9% vs. 9.4%) compared to patients without MACE. A nonlinear relationship was observed: when the average number of medications was below four, medication use was associated with reduced MACE risk (HR = 0.78, p < 0.01). However, when medication use exceeded four drugs, risk increased significantly (HR = 1.75, p < 0.01). Instrument validity and strength were confirmed (Hausman p < 0.01; Kleibergen-Paap F = 18.4).
CONCLUSIONS: Using robust causal methods and longitudinal data, our study suggests an optimal threshold for medication use among patients with diabetes. Both undertreatment and excessive polypharmacy were associated with elevated risks of major cardiovascular events.
Conference/Value in Health Info
2025-09, ISPOR Real-World Evidence Summit 2025, Tokyo, Japan
Value in Health Regional, Volume 49S (September 2025)
Code
RWD110
Topic Subcategory
Health & Insurance Records Systems
Disease
SDC: Diabetes/Endocrine/Metabolic Disorders (including obesity)