Cardiorespiratory Fitness, Sleep Patterns and Chronic Obstructive Pulmonary Disease Risk Prediction: Development and Validation in UK Biobank Participants

Little is known about the effect of cardiorespiratory fitness (CRF) and sleep on chronic obstructive pulmonary disease (COPD) risk. We aimed to investigate whether adding either or both CRF and sleep patterns to conventional risk factors for COPD improve their predictive abilities. A total of 79,512 participants without baseline COPD from the UK Biobank were included. Healthy sleep scores, including sleep duration, insomnia, snoring, chronotype and daytime sleepiness were used to generate sleep patterns. The associations of CRF and sleep patterns with COPD events were investigated using Fine–Gray competing risk models, and predictive utility was determined by the C-index and continuous net reclassification index (NRI). Prediction models were further internally validated. During a median of 9.94 years of follow-up, 2,453 COPD events occurred. Both higher CRF and healthier sleep patterns were associated with reduced risk of COPD after adjusting for conventional risk factors. An initial model incorporating all conventional risk factors obtained satisfactory discrimination (C-index=0.832), and addition of CRF (C-index=0.836), sleep patterns (C-index=0.835), and both combined (C-index=0.838) improved the C-index and NRI (CRF +8.2%; sleep patterns +4.3%; combined +8.6%). The calibration was excellent, and the models still discriminated well after internal validation (C-indexes: initial, 0.818; CRF, 0.821; sleep patterns, 0.822; combined, 0.824). Both higher CRF levels and healthier sleep patterns were associated with reduced risks of COPD. The performance of the COPD prediction model developed in the study could be improved by adding CRF and sleep patterns, with an additive effect when both are added.