Identification of three classes of acute respiratory distress syndrome using latent class analysis
Objectives: Acute Respiratory Distress Syndrome (ARDS) is a highly heterogeneous syndrome which exhibits significant differences in the underlying causes, responses to treatment and prognosis. It is mandatory to make subtypes of ARDS to guideline clinical treatment and trial design. The study aimed to identify subtypes of ARDS using latent class analysis. Design: The study was a secondary analysis of the EDEN study which was a randomized, controlled, multicenter trial conducted from January 2, 2008 to April 12, 2011. The primary study endpoint was death through 90-day follow up. Latent class analysis was performed incorporating variables on day 1 after enrollment. The number of classes was chosen by maximum log-likelihood, Akaike Information Criterion and the number of patients in each class. Setting: secondary analysis of a multicenter randomized controlled trial. Patients: patients within 48 hours of developing ARDS requiring mechanical ventilation. Interventions: none. Measurements and main results: A total of 1000 patients were enrolled in the study, including 233 non-survivors and 767 survivors. The latent class analysis identified three classes of ARDS. Class 1 (hemodynamically stable type) had significantly higher survival rate (p=0.015) and cumulative incidence of unassisted breathing (p=0.016) than class 2 (non-SIRS type) and 3 (SIRS type) through 90 days follow up. There was significant interaction between cumulative fluid balance and the class (p=0.03). While more fluid balance was beneficial for class 2, it was harmful for class 1 and 3. Conclusions: The study identified three classes of ARDS, which showed different clinical presentations, responses to fluid therapy and prognosis. The classification system used simple clinical variables and could help to design ARDS trials in the future. Trial registration: NCT00609180. Registered February 6, 2008.