Glucocorticoid treatment of acute respiratory distress syndrome (ARDS) remains contentious, and the available evidence remains contradictory [1–4]. In this context, the current analysis by Meduri et al. [5] is welcome. These authors conducted a two-part analysis—(1) individual patient data meta-analysis (IPDMA) from trials with methylprednisolone and (2) an updated trial-level meta-analysis including additional randomised controlled trials (RCTs) with hydrocortisone in early ARDS—and have reported that steroids accelerated the resolution of ARDS, leading to reduced ventilatory assistance, hospital mortality and health care utilisation [5]. However, these conclusions appear to contradict those of the ARDS Network LaSRS study [4], which contributed 56 % of the patients to the IPDMA. Furthermore, in addition to reporting no benefit from the routine use of methylprednisolone in patients with ARDS, the LaSRS investigators found that the use of methylprednisolone was associated with an increased risk of neuromuscular complications and that initiation of methylprednisolone treatment more than 2 weeks after the onset of ARDS led to an increase in the risk of death despite improved early cardiopulmonary physiology [4]. This discrepancy demands some consideration.

Meta-analysis offers some advantages over a single high-quality RCT, as the greater number of patients enrolled in the former, as well as the range of differing populations, circumstances and settings, facilitates generalisability. However, studies in critical care settings are particularly challenging due to the heterogeneity of both the cohort and the treatments, which can lead to misleading conclusions [6]. While this is minimised in IPDMA [7], which is considered the gold standard for meta-analysis [8], and therefore is the focus of our attention here, both the quality of the individual studies included in the IPDMA and of the analysis itself need to be considered.

Well-documented and published guidelines (PRISMA) for the conduct, reporting and transparency of meta-analysis [9]—and specifically for IPDMA [10]—have been developed. Despite Meduri et al.’s detailed description of the statistical methods used in their IPDMA [5], the information provided is insufficient to conclude that these guidelines were all followed. In addition, there was moderate heterogeneity in study outcomes (reported as an I 2 statistic, with likely wide 95 % confidence intervals), which the authors attribute to the LaSRS study. These methodological concerns suggest that the reader should be cautious in drawing conclusions.

There are important differences in trial design between the studies contributing data to the IPDMA of Meduri et al. [5]. The LaSRS study enrolled 180 patients (1:1 randomisation), with 60-day mortality as its primary outcome [4]; the remaining studies used a 2:1 randomisation (n = 24 [1], 27 [2] and 91 [3]), with reduction in the lung injury score at day 7 [3] or day 14 [1, 2] as the primary outcome (Table 1) and only mortality at day 28 reported. Despite early improvements in cardiopulmonary physiology and an increased number of ventilator-free days, intensive care unit-free days and shock-free days during the first 28 days of treatment, patients treated with methylprednisolone in the LaSRS study did not show improved outcomes at day 60 and day 180 and had greater neuromuscular weakness and an increase in mortality if the treatment had been started after 14 days. While Meduri et al. [5] cogently argue that the rapid cessation of methylprednisolone resulted in an exacerbation of lung inflammation, contributing to these adverse effects, these data also emphasise the importance of examining longer-term outcomes [11]. Mortality is both unambiguous and unarguably important, but it does depend upon when it is measured [12], and current data are limited to treatment day 28 or hospital mortality. Long-term functional disability is an equally important legacy in ARDS survivors [13] and is an increasing focus for both researchers and clinicians.

Table 1 Details of primary outcome, total number, tidal volume and plateau pressure for studies included in the individual patient data meta-analysis
Full size table

There are also differences in routine care (co-interventions) in the studies included in the IPDMA that may contribute to heterogeneous outcomes. Lung protective mechanical ventilation has generally become the standard of care for ARDS patients. It is of note that barring the LaSRS study [4], the other three studies [1–3] used tidal volumes that would not be considered lung protective (Table 1). Other factors, such as use of neuromuscular blocking agents [14] and fluid balance, can also affect the outcome of patients with ARDS. Taken together, these factors raise the question of standardisation, where possible, and the potential bias of co-interventions during a clinical trial. For example, while magnesium was found to improve outcome from myocardial infarction in LIMIT-2 (n = 2316) [15], this was not confirmed in the ISIS-4 mega-trial (n = 58,050) [16]. An important difference between these latter two studies was the much greater use of aspirin and attempted revascularisation in ISIS-4 [16]. It is unknown whether protective ventilation mitigates the beneficial effects of steroids in ARDS, but clinicians should consider the possible bias introduced by unbalanced co-interventions when interpreting data from both RCTs and meta-analyses.

The potential adverse effects of therapeutic steroids go beyond neuromuscular weakness, immunosuppression, superadded infection and higher blood glucose levels [17]. The mineralocorticoid effect of steroids contributes to fluid and sodium retention [18, 19], with both a positive fluid and sodium balance associated with adverse outcomes in patients with lung injury [20–22]. Prospective data examining this potential confounder should be considered in future clinical trials.

On the principle of primum non nocere (first, do no harm), we feel that there is currently insufficient evidence to advocate the routine use of steroids in patients with ARDS as potential short-term improvements appear to be mitigated by later adverse effects. If steroids are used, however, abrupt cessation should be avoided.