Prediction of COPD risk accounting for time-varying smoking exposures.


Chang JT(1), Meza R(1), Levy DT(2), Arenberg D(3), Jeon J(1).
Author information:
(1)Department of Epidemiology, School of Public Health, University of Michigan, Ann Arbor, Michigan, United States of America.
(2)Department of Oncology, Georgetown Lombardi Comprehensive Cancer Center, Washington D.C., DC, United States of America.
(3)Division of Pulmonary and Critical Medicine, Department of Internal Medicine, University of Michigan Medical School, Ann Arbor, Michigan, United States of America.


RATIONALE: Chronic Obstructive Pulmonary Disease (COPD) is the fourth leading cause of death in the United States. Studies have primarily assessed the relationship between smoking on COPD risk focusing on summary measures, like smoking status. OBJECTIVE: Develop a COPD risk prediction model incorporating individual time-varying smoking exposures. METHODS: The Nurses' Health Study (N = 86,711) and the Health Professionals Follow-up Study (N = 39,817) data was used to develop a COPD risk prediction model. Data was randomly split in 50-50 samples for model building and validation. Cox regression with time-varying covariates was used to assess the association between smoking duration, intensity and year-since-quit and self-reported COPD diagnosis incidence. We evaluated the model calibration as well as discriminatory accuracy via the Area Under the receiver operating characteristic Curve (AUC). We computed 6-year risk of COPD incidence given various individual smoking scenarios. RESULTS: Smoking duration, year-since-quit (if former smokers), sex, and interaction of sex and smoking duration are significantly associated with the incidence of diagnosed COPD. The model that incorporated time-varying smoking variables yielded higher AUCs compared to models using only pack-years. The AUCs for the model were 0.80 (95% CI: 0.74-0.86) and 0.73 (95% CI: 0.70-0.77) for males and females, respectively. CONCLUSIONS: Utilizing detailed smoking pattern information, the model predicts COPD risk with better accuracy than models based on only smoking summary measures. It might serve as a tool for early detection programs by identifying individuals at high-risk for COPD.