Reproductive
Evaluating Measurement Error in the Relationship Between Uterine Fibroids and Atherosclerotic Cardiovascular Disease Julia DiTosto* Julia DiTosto Sunni Mumford Jennifer Lewey Jarcy Zee Anuja Dokras Kyle Busse Snigdha Alur-Gupta Stefanie Hinkle Enrique Schisterman Ellen Caniglia
Preliminary claims data suggest that uterine fibroids, a common gynecologic condition, increase the risk of atherosclerotic cardiovascular disease (ASCVD). However, risk estimates may be affected by misclassification.
Females aged 18-50 with uterine fibroids (N=450,177), age-matched to 5 females with gynecologic claims in the same month, were identified in Optum Clinformatics DataMart (2000-2022). Pooled logistic regression models estimated Crude and Adjusted 1, 3, 5, and 10-year RR and 95% CI for fibroids on ASCVD, weighted for baseline confounding, incident fibroid diagnoses in controls, and Optum disenrollment. Probabilistic bias analyses with 1,000,000 simulations addressed measurement error using validation study parameters for misclassification of exposure (fibroids) and a confounder (obesity), assuming non-differential misclassification. Sensitivity and specificity were modeled using beta distributions. Multiple bias modeling sequentially adjusted for biases in reverse order of occurrence and accounted for random error. The distribution median was reported as the point estimate, with the 2.5th and 97.5th percentiles defining 95% simulation intervals (SI). Final estimates were obtained by dividing bias-adjusted RRs by the RR from measured confounding (calculated by dividing the Adjusted by the Crude RR).
The Adjusted RRs before probabilistic bias analyses ranged from 2.46 (1-year) to 1.81 (10-years). Adjusting for exposure misclassification shifted results away from the null (1-year RR: 3.27, 95% SI 3.00, 3.69), while obesity misclassification shifted results toward the null (1-year RR: 1.98, 95% SI 1.85, 2.23). The multiple bias-adjusted RR attenuated the Adjusted RR (1-year RR: 2.17, 95% CI 2.04, 2.42). Analysis results were consistent across timeframes for the 3, 5, and 10-year RRs.
Probabilistic bias analyses for measurement error attenuated results but confirmed the robust relationship between uterine fibroids and ASCVD.
