Abstract 10050: Chronic Utilization of Medications with Metabolic Effects and Risk for Incident Diabetes in Postmenopausal Women

Abstract

BACKGROUND Studies suggest thiazide diuretics (TD), statins (ST), β blockers (BB), and antidepressants (AD) can impair glucose homeostasis and increase risk for diabetes. Data are needed on the prevalence of simultaneous drug use and the relation to risk for incident diabetes, since additive effects may increase diabetes development. Therefore, we investigated the association between multiple medication use and incidence of diabetes in the Women’s Health Initiative Clinical Trials and Observational Study.

METHODS We categorized self-reported baseline use of medications within the four drug classes, TD, ST, BB, and AD, according to those taking 1, 2, or 3-4 drugs from these classes at baseline and compared them to those taking 0 drugs. We identified incident diabetes annually during an average of 7.6 years of followup. Diabetes was self-reported as ever having received a physician diagnosis of and/or treatment for diabetes when not pregnant. Women with diabetes at baseline are excluded. RESULTS A total of 141,370 women without diabetes at baseline and with data regarding use of medications were included. There were 12,048 cases (8.5%) of incident diabetes identified over more than 1 million person-years of followup. A total 106,647 women took 0, 27,208 took 1, 6,639 took 2 and 876 took 3-4 drugs from the four classes. With increasing number of drug classes used, baseline age, BMI, waist, glucose and SBP were significantly higher, as were history of HTN, CVD, and hypercholesterolemia, all p<0.0001. Overall incidence of diabetes increased with number of drug classes used, log rank p<0.0001 (Fig A), and adjusted risk for diabetes increased with number of drug classes used. (Fig B) CONCLUSION This analysis in postmenopausal women demonstrates that risk for diabetes is higher when taking two or more TD, ST, BB or AD than when taking a single drug. Confounding by indication may play a role; however, this is less likely in our adjusted model.