• ARDS

Adult patients with acute respiratory distress syndrome (ARDS) are nowadays considered to be a heterogeneous population. Profound differences in systemic inflammation, lung morphology and physiological response have been observed. The use of unsupervised statistical techniques using multiple variables have been used to identify homogeneous subgroups.1 Such subgroups derived from multidimensional data using unbiased algorithms are also called subphenotypes. They can be used to improve trial design and, finally, patient care through prognostic and predictive enrichment.2 In the adult population, two subphenotypes based on biomarkers of systemic inflammation referred to as ‘hypoinflammatory’ and ‘hyperinflammatory’ ARDS have been identified consistently in several thousands of patients.3 4 Patients with the hyperinflammatory subphenotype have a higher mortality and require longer duration of mechanical ventilation. Heterogeneity of treatment effect (HTE) has been shown for simvastatin, positive and expiratory pressure (PEEP) strategy, fluid strategy and possibly steroid exposure, suggesting the potential for predictive enrichment. More recently, paediatric ARDS populations have also been subphenotyped using clinical and biological data.5 The concordance between adult-derived inflammatory subphenotypes and heterogeneity in the inflammatory response in paediatric patients was yet unknown.

In their Thorax paper, Nadir Yehya and colleagues identified a hypoinflammatory and a hyperinflammatory subphenotype in two independent cohorts of a total of 626 paediatric patients with ARDS by using latent class analysis (LCA), an unsupervised multidimensional statistical technique that aims to generate homogeneous subpopulations, on plasma inflammatory biomarker levels.6 Patients …