Background: We previously identified five clinical phenotypes of COPD patients with different clinical characteristics and outcomes, and used Classification and Regression Trees (CART) to develop an algorithm to allocate patients to these phenotypes based on the presence of cardiovascular comorbidities and/or diabetes, mMRC dyspnea grade (threshold: 2), FEV1, age and Body Mass Index (Burgel et al., Eur Respir J 2017; 50: 1701034). The resulting categories were validated using all-cause mortality.

Objective: To determine whether CAT score could replace mMRC dyspnea grade in this algorithm, and which CAT threshold should be used.

Methods: We identified patients with CAT score and mMRC dyspnea grade assessed at stable state within 1 month in a real-life cohort of COPD patients (Initiatives BPCO cohort, n=1760). Patients were categorized using the original mMRC-based algorithm and a derived algorithm where mMRC dyspnea grade was replaced by CAT score (with an initial tested threshold of 10 as per GOLD). Concordance between the algorithms was analysed.

Results: 428 patients with the required assessments were identified. Overall concordance between mMRC-based and CAT-based phenotyping algorithms was 69% with the CAT 10 threshold, with concordance rates <75% for 3 out of the 5 phenotypic categories. Therefore, a second CAT threshold of 20 was tested. Concordance increased to 78%, with >80% concordance for 4 categories and 65% for the 5th.

Conclusion: In study populations where mMRC is not available, CAT with a threshold of 20 can replace it in CART-developed decision trees to allocate patients to previously validated clinically relevant phenotypes.