Identifying a hyperinflammatory subphenotype of ARDS associated with worse outcomes: may ferritin help?

Attributable mortality of acute respiratory distress syndrome (ARDS) is considerable.^1,2 Yet, no survival benefit has been shown in randomized controlled trials (RCTs) of pharmacologic strategies to treat ARDS. This is assumed to be a consequence of the heterogeneity of clinical and biological processes amongst patients meeting criteria for ARDS.³

In an attempt to address heterogeneity, recent efforts have led to elucidation of the role of respiratory microbiome⁴ and to identification both in patients at risk of ARDS⁵ and in patients with ARDS⁶ of reproducible subphenotypes, such as the “hypoinflammatory” and “hyperinflammatory” subphenotypes; the latter being associated with worse outcomes. Those subphenotypes were identified through post-hoc analyses of clinical and plasma biomarker data of patients enrolled in RCTs or in observational studies.^5,6 Implementation of subphenotypes in clinical practice has so far been limited due to lack of prospective validation as well as the need for rapid, real-time measurement of multiple biomarkers needed to stratify patients. Moreover, measurement of those biomarkers can only be undertaken in the research laboratory setting. Therefore, it would be desirable if, instead of multiple biomarkers, a single marker that is routinely available in the clinical setting could be used to stratify patients and, specifically, identify patients with ARDS at risk for worse outcomes.

In this issue of the Journal, Mehta et al evaluated whether ferritin, a routinely available marker, could identify patients with ARDS at risk for mortality.⁷ The authors took advantage of individual patient-level data from subjects previously enrolled in either HARP-2 (the derivation cohort; a RCT of simvastatin versus placebo) or ROSE (the validation cohort; a RCT of continuous cisatracurium with deep sedation for 48 hours versus usual care without neuromuscular blockade and with lighter sedation).^8,9 Although inclusion criteria were similar, the partial pressure of arterial oxygen to fraction of inspired oxygen ratio for enrolment was higher in HARP-2 (<300 mm Hg) than ROSE (<150 mm Hg).^8,9 Both RCTs had plasma collected in the first 48 hours after ARDS onset and prior to randomization. Ferritin levels were measured in each RCT using commercially available ELISA kits.^8,9 Using a logistic regression model with restricted cubic splines, the authors found that a log-fold increase in ferritin was associated with an odds ratio of 1.71 for 28-day mortality.⁷ They also determined that a threshold of ferritin >1380 ng/mL (present in 28% of HARP-2 patients and 24% of ROSE patients) was associated with higher mortality. Finally, the authors performed a mediation analysis demonstrating that the association between ferritin and mortality was mediated by interleukin (IL)-18 to a small but statistically significant effect, after adjustment for confounders, such as etiology of ARDS and APACHE II score. The rationale behind the focus on IL-18 was based on previous evidence indicating that ferritin promotes inflammasome activation, that IL-18 is a surrogate marker for inflammasome activity, and that IL-18 levels are elevated in patients with ARDS.^10,11

There are several strengths in the work by Mehta et al.⁷ Firstly, the authors used data and biospecimens from two RCTs.^8,9 Collecting biospecimens as routine practice in RCTs greatly facilitates investigative work into subphenotyping ARDS and understanding trajectories of subphenotypes. Secondly, the authors selected a single and widely available in the clinical practice marker (namely, ferritin) to identify patients with ARDS at risk for high mortality. Elevated ferritin levels have previously been linked to hyperinflammatory pathology and poor outcomes in hemophagocytic lymphohistiocytosis,¹² in sepsis with features of macrophage-like activation syndrome (where hyperferritinemia has been used to predict treatment response to anakinra),¹³ and in COVID-19.¹⁴ Although the role of hyperferritinemia in these disease states is not fully understood, previous links between elevated ferritin levels, inflammasome activation, and pyroptotic cell death provide a scientific premise for evaluating the causal relationship between ferritin, IL-18, and mortality. The finding of Mehta et al⁷ that IL-18 contributes as an intermediate pathway between ferritin and mortality may justify further investigation targeting the IL-18 pathway in hyperferritinemic patients with ARDS.

While perusing the elegant study by Metha et al,⁷ one may keep in mind two considerations. Firstly, detection of IL-18 in plasma may not necessarily be indicative of inflammasome activation and pyroptotic cell death, per se. IL-18 has a multitude of established roles including the induction of cell-mediated immunity during infection.¹⁵ Furthermore, it remains unclear whether targeting IL-18 would produce a clinically meaningful benefit given the partial (yet statistically significant) mediation estimated in this analysis.⁷ Secondly, there is currently a lack of understanding on how and whether ferritin levels change during the clinical course. Data regarding the trajectories of subphenotypes in critical illness is limited so far.¹⁶ One study extended latent class analysis in two RCTs of patients with ARDS on day 3, where two subphenotypes were again evident.¹⁷ That study found that >94% of patients stayed in their corresponding subphenotype class on day 3.¹⁷ However, whether the inflammatory burden and disease implications of the subphenotypes identified at baseline and day 3 are the same is unknown.¹⁸ Use of trajectory data will be of critical importance in future studies to gain deeper understanding of the temporal kinetics of a particular subphenotype, as well as whether disease states may be modified by treatment. Resolution of a detrimental subphenotype, if consistently linked to longer-term patient-centered outcomes, may be a more suitable endpoint to study than mortality.¹⁸

In conclusion, Mehta et al are complimented for following a pragmatic approach to the subphenotyping of ARDS with hyperinflammatory pathobiology using a single, readily available marker.⁷ As efforts to identify subphenotypes with clinical utility in ARDS continue, there is certain to be emergence of alternative biological mechanisms to classify them that may or may not overlap with previous discoveries. Nonetheless, this “reverse translation” approach, or post-hoc analysis of RCT biospecimens that harnesses the power of randomization, holds incredible potential for new subphenotype discoveries.^19,20 Finally, mediation analysis may serve as an important statistical approach to identify novel, mechanistically based biomarkers for future therapeutic targeting and/or monitoring of treatment response.