Sex-specific machine learning improves prediction of incident and prevalent COPD.

BACKGROUND: Computed tomography imaging with machine learning can predict incident and prevalent chronic obstructive pulmonary disease (COPD), however, it is unknown if sex-specific models improve performance. RESEARCH QUESTION: Do sex-specific machine learning models using computed tomography (CT) derived disease features improve prediction of incident and prevalent chronic obstructive pulmonary disease (COPD) and identify sex-specific predictors? STUDY DESIGN AND METHODS: Canadian Cohort Obstructive Lung Disease (CanCOLD) study participants underwent baseline CT imaging, and spirometry at baseline/follow-up. Models predicted incident and prevalent COPD using demographics and CT features (lung density/texture/shape, airway shape) for the combined-sex, male-only, and female-only datasets, and externally tested in Subpopulations and Intermediate Outcome Measures in COPD (SPIROMICS). Performance was evaluated using area under the receiver operating characteristic curve (AUC). RESULTS: 1283 CanCOLD and 1840 SPIROMICS participants were included. For incident COPD, the female-only model outperformed the male-only and combined-sex models in internal (AUC=0.86 vs. 0.76 and 0.78; p<0.05) and external tests (AUC=0.83 vs. 0.71 and 0.77; p<0.05). For prevalent COPD, the female-only model again outperformed the male-only and combined-sex model in internal (AUC=0.84 vs. 0.78 and 0.78; p<0.01) and external tests (AUC=0.84 vs. 0.70 and 0.73; p<0.05). Female-only models selected parenchymal texture and lung-shape features not selected in combined-sex models, whereas male-only models selected airway-based features. INTERPRETATION: Sex-specific models outperform combined-sex models for identifying those with and at risk of COPD, particularly in females, by capturing different disease-relevant features. These findings highlight that combined-sex models can obscure sex-specific biology and adopting sex-specific prediction strategies may improve early detection, risk stratification, and allow for treatment targeting.