To Örebro University

Background: In 2017, Burgel and colleagues developed an algorithm to identify clinical phenotypes that predict mortality in COPD. Our study aimed to 1) investigate whether the phenotypes can predict acute exacerbations of COPD (AECOPD) and 2) validate their ability to predict mortality.

Methods: The Tools Identifying Exacerbations (TIE) cohort study recruited participants with spirometry-verified COPD from primary and secondary care in three Swedish regions. Participants were allocated to phenotypes 1-5 using the previously developed algorithm containing comorbidities (heart failure, coronary artery disease, hypertension, and/or diabetes), dyspnoea, age, forced expiratory volume in 1 s (FEV1), and body mass index (BMI). Data on AECOPDs and deaths during the three-year follow-up were collected from medical records and analysed with Cox proportional hazards regressions. Harrel's C-index was used to assess the models' discriminative ability.

Results: Among the 566 participants, 59 % were female, and the mean +/- SD FEV1 was 57 +/- 18 % of predicted. The hazard ratios (HRs) [95 % CI] for time to AECOPD were 3.04 [1.93-4.79], 2.38 [1.54-3.66], and 3.52 [1.73-7.15] in phenotypes 1, 2, and 4 compared with 5 (C-index = 0.61). When AECOPD history was used to predict future AECOPD the C-index was 0.65. The HRs [95 % CI] for mortality were 8.24 [1.93-35.3], 6.26 [1.40-28.0], and 16.7 [3.25-86.3] in phenotypes 1, 3, and 4 compared to 5 (C-index = 0.68). For AECOPD history, the C-index was 0.55.

Conclusion: Clinical COPD phenotypes based on comorbidities, dyspnoea, age, FEV1, and BMI predict AECOPDs but do not perform better than AECOPD history. However, they perform better than AECOPD history in predicting mortality.