Introduction
Despite advances in critical care medicine over the past decades, the mortality rate for acute respiratory distress syndrome (ARDS) remains high [
1‐
3]. Because dysregulated inflammation is the cardinal feature of ARDS [
2,
4], systemic corticosteroids have been considered a potentially beneficial therapy. However, previous randomized trials have failed to provide convincing evidence to prove the efficacy of corticosteroids in decreasing the mortality of ARDS [
5‐
8]. Only secondary outcomes, such as oxygenation improvement and reduction of the duration of mechanical ventilation, have shown consistent findings in favor of corticosteroid therapy.
Published meta-analyses about corticosteroid therapy for ARDS reported inconsistent conclusions [
9‐
13]. Different study selections and heterogeneity on mortality endpoints and etiologies of ARDS may account for the inconsistent study results in previous meta-analyses. Measuring the treatment effects at short-term or longer-term follow-up may influence study results, because therapeutic effects of corticosteroids develop early but some adverse effects, such as infection, develop late. Using short-term outcome as the study endpoint may underestimate the risk of corticosteroid therapy and overestimate the overall benefit. In addition, the mechanisms of lung injury and fibroproliferative response to injury vary in pulmonary and extrapulmonary ARDS [
14]. Therefore, the treatment response to corticosteroids in ARDS may be different in ARDS of different etiologies. However, the influence of the etiologies of ARDS on outcomes of corticosteroid therapy has not been evaluated in previous studies.
We conducted a systematic review and meta-analysis of corticosteroid therapy in ARDS with the aim of updating the best available evidence and exploring the source of observed heterogeneity.
Discussion
Our study found that the mortality outcomes of corticosteroid therapy in ARDS differed by duration of outcome measures. Corticosteroids had a possible but statistically insignificant effect on short-term mortality in RCTs but did not decrease longer-term mortality in either RCTs or cohort studies (Table
1). Within-study observation in studies reporting two mortality endpoints also suggested that the benefit of corticosteroid therapy decreased when follow-up was prolonged [
7,
8,
23]. This raises a concern that corticosteroid therapy in ARDS may bring initial benefits by suppressing the inflammatory process and reducing alveolocapillary permeability [
4,
33,
34], but the beneficial effects are soon counteracted by the delayed onset of adverse effects, such as immunosuppression and altered tissue repair [
35,
36]. We also found that the effect of corticosteroid therapy differed among different populations of ARDS patients. Corticosteroids may cause harm in certain ARDS subgroups, such as influenza-related ARDS. Taken together, current data do not support routine use of corticosteroids in ARDS. Given the heterogeneous nature of ARDS and the pleiotropic effects of corticosteroids, more clinical trials are needed to specify the favorable and unfavorable subgroups for corticosteroid therapy. For more comprehensive assessment of the effects of corticosteroid therapy, future studies should evaluate a mortality endpoint of adequate duration and the 60-day mortality used by the ARDSnet appears to be a reasonable study endpoint [
6].
Our study demonstrated the diverse treatment effects of corticosteroids among different etiologies of ARDS (Table
2). It is biologically plausible that different etiologies of ARDS have different responses to corticosteroid therapy because the pulmonary fibroproliferative response to injury may occur in an injury-specific rather than a stereotyped manner [
14]. The main damage targets differ in ARDS caused by different etiologies [
37,
38]. Therefore, it is not surprising that the efficacy of corticosteroid therapy differs among different etiologies. Additionally, our analysis showed that corticosteroids significantly increased mortality in influenza-related ARDS. The poorer outcome may be attributed to prolonged viral shedding and an increased risk of superinfection [
39,
40]. An expert review also advised against use of corticosteroids in the management of H1N1 influenza A infection [
41].
The dosage and timing of corticosteroid therapy in ARDS has changed over the last decades. Based on the equivalent doses of methylprednisolone, studies before 1990 usually used a high daily dose (30 mg/kg) and short period (≤2 days) regimen to prevent or treat ARDS. In the following two decades, most studies used a protocol of daily dose of ≤2 mg/kg with a gradual taper. Some investigators suggested different treatment dosages for early ARDS and persistent ARDS in which a duration of ARDS ≤3 days was considered as early ARDS and ≥5 days as persistent or unresolving ARDS [
4]. The treatment dose was suggested to be ≤1 mg/Kg for early ARDS [
7,
23], and 2 mg/Kg for persistent ARDS [
6,
8]. We summarized the treatment outcomes of corticosteroid therapy initiated at different stages of ARDS (Table
3). Our analysis found that patients with persistent ARDS seemed more likely to benefit from corticosteroid therapy. In addition, the subgroup analysis of the ARDSnet steroid study suggested against starting corticosteroid therapy >14 days after the onset of ARDS [
6]. These findings suggest a subgroup of ARDS might benefit most from corticosteroid therapy: persistent ARDS with the onset of ARDS <14 days. Persistent ARDS indicates a specific subgroup or phenotype of ARDS that is characterized by exaggerated or unresolving lung inflammation [
4]. In this subgroup, the benefit of corticosteroid therapy may outweigh the treatment risk. Dosage and administration schedules may also affect treatment outcomes of corticosteroids. However, it was difficult to evaluate the effects of these two factors in the study-level analysis. Cohort studies did not use standardized treatment protocols and individual-level data are needed to conduct such analyses. In RCTs, therapeutic trials usually used a low-dose regimen (1 to 2 mg/kg/day) and similar administration schedules.
Table 3
Subgroup analysis by timing of starting corticosteroid therapy
Preventive therapy
| | | | |
Randomized controlled trials | 3 | 154 | 1.24 (0.57 to 2.72) | 80% |
Early ARDS (≤3 days)
| | | | |
Randomized controlled trials | 3 | 367 | 0.86 (0.71 to 1.04) | 17% |
Cohort studies | 4 | 303 | 1.00 (0.24 to 4.20) | 70% |
Persistent ARDS (≥5 days)
| | | | |
Randomized controlled trials | 2 | 204 | 0.52 (0.11 to 2.52) | 79% |
Cohort studies | 3 | 105 | 0.73 (0.44 to 1.23) | 0% |
Published meta-analyses reached inconsistent conclusions on the role of corticosteroids for ARDS [
9‐
13]. We summarize the study conclusions and study selection strategy of previous meta-analyses in Table
4. Among the included studies in these meta-analyses, an extremely protective effect for corticosteroids was observed in the two studies published by Meduri [
7,
8], but the effects were neutral and modest in other trials. In addition, inclusion or exclusion of trials of using corticosteroids in severe pneumonia into analysis also had an impact on the results of meta-analysis. Confalonieri and Meduri reported remarkable mortality reduction by corticosteroid therapy in a severe pneumonia study [
42]. Previous meta-analyses reporting significant mortality reduction of corticosteroid therapy usually included this pneumonia study [
12,
13]. However, a large cohort study using a registry database reported that low-dose corticosteroids were associated with an increased mortality in pneumonia with septic shock [
43].
Table 4
Comparisons of published meta-analyses
Adhikari et al.(2004) [ 9] | Three RCTs | Early high-dose corticosteroids had no effect on early mortality. Corticosteroids given for late phase ARDS reduced hospital mortality. | Study interest not focused on corticosteroids; few studies and small sample size. |
Agarwal et al. (2007) [ 10] | Four RCTs and two cohort studies | Current evidence does not support a role for corticosteroids in the management of ARDS in either the early or late stages of the disease. | Excluding the RCTs of preventive use of corticosteroids; including high-dose corticosteroid study. |
| Nine RCTs (eight RCTs for mortality analysis) | A definitive role of corticosteroids in the treatment of ARDS in adults is not established. | Including the RCTs of preventive use of corticosteroids; excluding pneumonia studies; using Bayesian random effects models for data pooling. |
| Four RCTs (three ARDS studies and one pneumonia study) and five cohort studies | The use of low-dose corticosteroids was associated with improved mortality and morbidity outcomes without increased adverse reactions. | Including a RCT of pneumonia; excluding studies of high-dose and preventive use of corticosteroids. |
Lamontagne et al. (2010) [ 13] | Twelve RCTs (six ARDS studies and six pneumonia studies) | Corticosteroids administered within 14 days of disease onset may reduce all-cause mortality. | Including six studies of pneumonia. |
Systemic corticosteroid therapy may bring several unfavorable side effects [
36,
44], and one major concern in patients with ARDS is an increased risk of nosocomial infection secondary to immunosuppression. Because symptoms and signs of early infection may be masked by corticosteroids, previous RCTs performed intensive infection surveillance procedures during the study to reduce the risk of superinfection [
6,
7]. However, these intensive surveillance procedures are not always performed outside of clinical studies. Our analysis showed that the infection risks reported in RCTs and cohort studies were conflicting (Figure
3). It is not known whether restrictive patient selection and infection surveillance procedures in RCTs played a role in making such a difference. Similar to the concern for mortality outcomes, the infection risk of corticosteroid therapy should be evaluated in an adequate time frame because the immunosuppressive effect may develop late in the clinical course. However, most studies evaluated infectious complications in a short duration (Additional file
1: e-Table S6) and the infection risk of corticosteroid therapy might, therefore, be underestimated.
External validity should be noted for this meta-analysis. Included individual RCTs reported numerous exclusion criteria for patient enrollment. The results of this meta-analysis should not be generalized to patients with particular comorbidities. Most RCTs excluded patients with underlying diseases that might benefit from corticosteroids, such as inflammatory airway diseases or vasculitis. Were these patients enrolled, the study outcome might be more likely to favor the corticosteroid group. On the other hand, clinical trials also excluded patients with conditions that militate against the use of corticosteroids, such as active gastrointestinal bleeding, disseminated infections, extensive burns or immunocompromised status. Outside the scope of the generalizability of current data, the use of corticosteroids in ARDS should be individually evaluated. Underlying diseases are important considerations to justify the use of corticosteroids.
Strengths and limitations
The strengths of our study include a comprehensive search strategy to include all studies analyzed in previous meta-analyses but not be restricted to these studies and to evaluate short-term and longer-term outcomes of corticosteroid therapy. The diversity of treatment outcomes among different etiologies of ARDS was also evaluated. These analyses help explore the causes of inconsistency among previous meta-analyses and achieve a more concrete suggestion for the use of corticosteroids in ARDS. With inclusion of the data from cohort studies, the disparity between clinical trials and real-world practice was disclosed, and their consistent results or trends helped to increase the robustness of this meta-analysis. In addition, we performed a sensitivity analysis to test the influence of study pooling strategy, an analysis that previous meta-analyses did not perform. Our study also has limitations. The number of RCTs and sample size were relatively small. There are only two studies in some subgroup analyses and underpower is a concern. Sparse data are another concern for data pooling by the random-effects model. With respect to the evaluation of an etiology-specific response to corticosteroids, the classification of ARDS etiologies was limited because the mix of study populations was diverse among several studies. Finally, study quality might be a confounding factor that we were unable to control in the subgroup analysis when we try to explore the association between follow-up duration and mortality. Earlier studies tend to be of poor quality and their follow-up duration was also shorter.
Sheng-Yuan Ruan, MD; Graduate Institute of Epidemiology and Preventive Medicine, National Taiwan University, Taipei, Taiwan; and Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan; Hsien-Ho Lin, MD, ScD; Graduate Institute of Epidemiology and Preventive Medicine, National Taiwan University, Taipei, Taiwan; Chun-Ta Huang, MD, Ping-Hung Kuo, MD, Huey-Dong Wu, MD and Chong-Jen Yu, MD, PhD; Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
SYR and HHL designed the study. SYR, HHL and CTH conducted the analysis and interpretation of the data and drafted the manuscript. PHK, HDW and CJY contributed to the interpretation of the data and critical revision of the manuscript for important intellectual content. All authors read and approved the final manuscript. SYR and HHL are guarantors.