FormalPara Take-home message

In this double-blind, randomized, placebo-controlled clinical trial of 584 participants hospitalized with severe community-acquired pneumonia, prolonged methylprednisolone treatment did not significantly reduce 60-day all-cause mortality or improve secondary outcomes during initial hospitalization or up to 1 year of follow-up. The risk for complications was similar to the control group.


Pneumonia is the leading cause of community-acquired infection requiring intensive care unit (ICU) admission and a common precipitant of septic shock and acute respiratory distress syndrome (ARDS) [1]. Hospital mortality is higher for patients who are older, bacteremic [2], have more comorbidities [3], meet criteria for healthcare-associated pneumonia (HCAP), require mechanical ventilation (MV) or vasopressor support, or are transferred to the ICU from a medical ward [4]. Most hospital deaths occur after eradication of bacteria from tracheal secretions and the bloodstream [5, 6], implying that adequate antibiotic treatment alone may be insufficient in further improving outcomes. Importantly, patients surviving hospitalization remain at risk for long-term morbidity [7], re-hospitalizations [4], and increased post-discharge mortality at 1 year (21–40%) [4] and up to 5 years [8]. Evidence points to the host’s inability to fully down-regulate systemic inflammation and restore tissue homeostasis as the dominant pathophysiologic processes contributing to acute and chronic adverse outcomes in community-acquired pneumonia (CAP) [9, 10].

Glucocorticoids were investigated in multiple randomized trials, with a signal for benefit in patients with severe pneumonia [11, 12]; however, a large confirmatory study was lacking. The Department of Veterans Affairs (VA) Cooperative Study #574 evaluated the efficacy of prolonged methylprednisolone treatment on short- and long-term morbidity and mortality in patients admitted to the ICU with severe CAP. We hypothesized that a 20-day low-dose methylprednisolone treatment would reduce 60-day mortality and improve clinical outcomes. The rationale for a 20-day treatment was to support the resolution phase of the disease [13], incorporate adequate glucocorticoid tapering [14], and to reduce post-hospitalization low-grade systemic inflammation.


Trial design and oversight

A double-blind, randomized, placebo-controlled trial was conducted at 42 VA Medical Centers from January 1, 2012 to August 31, 2016. Eligible patients were randomly assigned in a 1:1 ratio to either methylprednisolone or placebo. The trial protocol and the statistical analysis plan are provided in the Supplement Appendix.

The trial was approved by the VA Central Institutional Review Board and conducted in accordance with Good Clinical Practice Guidelines. An independent Data Monitoring Committee monitored patient safety, study conduct, and data. The authors vouch for the accuracy and completeness of the data and statistical analyses and for the fidelity of the trial to the protocol.


Adult patients presenting with a clinical diagnosis of severe CAP/HCAP were enrolled within 72–96 h (additional 24 h in patients not yet meeting severity criteria) of hospital presentation. Inclusion criteria required the presence of one major or three minor modified American Thoracic Society/Infectious Diseases Society of America (ATS/IDSA) criteria for severe pneumonia [15] as well as admission to intensive or intermediate care. Eligibility criteria are detailed in the Trial Protocol (Supplement Appendix).

Treatment and other trial procedures

Written informed consent was obtained from each participant or their legally authorized representative if they were unable to provide consent. Participants were randomly assigned in a 1:1 ratio to receive methylprednisolone or placebo using random permuted blocks of sizes 2 and 4, stratified by study site and need for MV at enrollment.

Methylprednisolone or placebo was given in a double-blind fashion. On the day of randomization (day 0), an intravenous loading dose of 40 mg was given, followed by maintenance infusion. The full 20-day treatment course included 40 mg/day on days 1–7, 20 mg/day on days 8–14, 12 mg/day on days 15–17 and 4 mg/day on days 18–20. Study drug was given by continuous infusion during ICU stay and changed to twice per day, via intravenous or enteral administration, after ICU discharge. Participants in both groups received standardized care following consensus recommendations [15, 16].

Participants were assessed daily up to day 8 during the initial ICU stay, at hospital discharge, and on days 28, 60, and 180. The final 1-year follow-up for mortality and re-hospitalizations was performed through review of records. We attempted to assess all participants regardless of treatment continuation. Monitoring for serious adverse events (SAEs) continued until the final follow-up contact. Safety monitoring and reporting procedures are detailed in the Trial Protocol.


The primary outcome was all-cause mortality at 60 days. Secondary outcomes included: (1) During hospitalization: post-randomization development of vasopressor-dependent shock or ARDS; number of multiple organ dysfunction syndrome (MODS)-free days to day 8; MV-free days up to days 8 and 28; duration of ICU and hospital stay; potential complications associated with methylprednisolone treatment; and hospital mortality; (2) Post-discharge: cardiovascular complications within 180 days of randomization; quality of life and functional status at days 28, 60, and 180; number and causes of re-hospitalization at VA hospitals within 1 year; SAEs and complications; and all-cause mortality at days 180 and 365. Exploratory outcome included duration of MV. MODS was assessed using the Sequential Organ Failure Assessment score [17]. Health-related quality of life was measured by the Veterans RAND 12 Item Health Survey [18, 19]. Functional status was measured by the Activities of Daily Living Scale and the Instrumental Activities of Daily Living Scale [20, 21]. Outcome definitions are detailed in the protocol.

Statistical analysis

We estimated that 1406 participants randomized 1:1 to the two treatment groups would provide 85% power to detect a 7% absolute reduction in 60-day mortality (21% in the methylprednisolone group vs. 28% in the placebo group). The original plan was to randomize 1420 participants (accounting for 1% attrition in primary outcome) over 5 years (January 2012–December 2016) and conduct two interim analyses at approximately 50% and 75% of the target number of participants to allow early discontinuation for efficacy (based on two-sided boundaries[22]) or futility (based on conditional power). Because of low recruitment, an ad hoc interim futility analysis was conducted on April 8, 2015 based on data as of February 6, 2015. At that time, 431 participants were randomized and the primary outcome was available for 372 participants. A one-sided non-binding futility boundary was calculated [22]. Conditional power was calculated for a range of differences in 60-day mortality (0–10%) and for two different target numbers of patients with 60-day mortality (the original target 1406 and the projected sample size 800 by December 2016). Based on the information, the DMC supported continued recruitment until the end of the planned recruitment period (December 2016). Study enrollment was stopped on April 30, 2016 due to persistent low recruitment; the final number of randomizations was 586. No additional interim analysis was done. Study follow-up ended in August 2016, which allowed collection of primary outcomes for all randomized participants. We report data for 584 participants, because two participants were improperly consented, and their data cannot be used for analyses.

Primary analyses were performed on the intention-to-treat sample (n = 584). Sixty-day all-cause mortality was compared by Chi-square test. The difference in percentages of 60-day mortality and the 95% confidence interval (CI) were calculated. Generalized linear mixed effect models were used to adjust for site (as a random effect) and baseline patient characteristics (as fixed effects), including MV status at randomization, age, Acute Physiology and Chronic Health Evaluation (APACHE) III score, bacteremia, use of anti-inflammatory medications, and use of macrolide antibiotics at baseline. Pneumonia Severity Index [23] class and Simplified Acute Physiology Score (SAPS) III score [24] were not included in the model to reduce collinearity of the covariates. Sensitivity analyses were performed to assess robustness of results and included using the per-protocol sample and different imputation methods for 21 participants missing primary outcomes (all due to early study withdrawal). The results from imputations were similar and not shown. Kaplan–Meier estimate of survival probability at day 60 was also calculated. Pre-specified subgroup analyses included MV status at randomization, APACHE III score quartiles, and CAP versus HCAP; post hoc subgroup analyses included severity of CAP, adequacy of initial antibiotic treatment, ARDS at baseline, and time of study treatment initiation (within 48 h vs. > 48 h of hospital presentation). Logistic regression was used to examine subgroups by treatment interactions.

Secondary outcomes were compared using Chi-square test or Fisher’s exact test for categorical outcomes, two-sample t tests or Wilcoxon rank-sum tests for continuous outcomes, and log rank tests and Kaplan–Meier curves for time to death and duration of MV up to day 28. Survival up to 180 days was compared by the restricted mean survival time (RMST) [25].

All p values are two-sided. The p values for secondary and exploratory outcomes were adjusted for multiplicity by the Bonferroni method, separately for in-hospital outcomes and post-discharge outcomes. The widths of the confidence intervals for the treatment differences in secondary and exploratory outcomes were not adjusted for multiplicity. SAS 9.4 (SAS Institute, Cary, NC, USA) was used for analysis. Unless specified otherwise, results are reported as methylprednisolone vs. placebo.



Of the 3936 patients who were assessed for eligibility, 584 were randomized; 70% were randomized within 48 h of hospital presentation and 94% within 72 h (median time to randomization, 37 h). Two hundred and ninety-seven participants were assigned to the methylprednisolone group and 287 to the placebo group (Fig. 1); 193 (33%) were receiving MV at the time of randomization. A total of 382 (65%) participants started study treatment within 48 h of hospital presentation and 513 (88%) within 72 h (median time from hospital presentation to study treatment initiation, 40 h). The study flow diagram is shown in Fig. 1, which also provides information on study drug withdrawal and reasons.

Fig. 1
figure 1

Enrollment, randomization, and follow-up. ¥Participants who were consented improperly are not included in this diagram. §The reasons for failing eligibility criteria were “select all that apply,” so one patient may have more than one reason for exclusion. Five patients who did not meet eligibility criteria (three did not meet inclusion criteria and two met exclusion criteria) were randomized. *Reasons for study drug withdrawal were check all that apply. ¶Active gastrointestinal bleeding requiring transfusion of at least 5 units of PRBC’s. ¶¶Such as exacerbation of COPD or asthma, and vasculitis

The two treatment groups were balanced in demographics and baseline patient characteristics (Table 1). The mean age was 68.8 years, 96% were male, and 83% were White. Patients had an average of four major comorbidities (Table S1). Thirty-four percent of participants met HCAP criteria, 69% had multi-lobar involvement on chest radiograph, 15% had bacteremia, 11% had ARDS at enrollment, and 13% had vasopressor-dependent shock at enrollment. Pathogens potentially responsible for the pneumonia were identified in 250 (43%) of the 577 participants with specimens from the respiratory tract, pleural fluid, blood or urine. The most common pathogens isolated were Staphylococcus aureus (10%), Streptococcus pneumoniae (9%), Pseudomonas aeruginosa (3%), and Escherichia coli (3%). Initial antibiotic treatment was deemed adequate in 96% of the participants based on ATS/IDSA guideline recommendations (Fig. 2).

Table 1 Baseline characteristics
Fig. 2
figure 2figure 2

Kaplan–Meier estimate of survival. Kaplan–Meier estimates of survival are shown in the overall population (A), in patients who were receiving mechanical ventilation at randomization (Patients on MV; B), and in those not receiving mechanical ventilation at randomization (Patients not on MV; C). The inset in each panel shows the same data on an enlarged y axis and up to day 60

Primary outcome

There was no significant difference in 60-day all-cause mortality (16% vs. 18%; unadjusted absolute risk difference − 2%, 95% CI − 8 to 5%; unadjusted odds ratio (OR) 0.89, 95% CI 0.58–1.38; p = 0.61) (Table 2). The result was similar when adjusted for site and MV status at randomization (adjusted OR 0.90; 95% CI 0.57–1.40; p = 0.63) and when also adjusted for baseline patient characteristics (adjusted OR, 0.87; 95% CI 0.53–1.42; p = 0.58). Kaplan–Meier estimate of 60-day mortality was 16% (95% CI 12–21%) in the methylprednisolone group and 18% (95% CI 14–23%) in the placebo group. No significant variation was found in the treatment effect across study sites. Results were similar in the per-protocol sample (Table S2 and Table S3). There was no significant between-group difference in the subgroup analyses (Table S4 and Fig. 3).

Table 2 Primary and secondary outcomes
Fig. 3
figure 3

60-day all-cause mortality according to subgroup. The odds ratios and 95% confidence intervals are based on logistic regression with treatment as the single covariate. The widths of the confidence intervals have not been adjusted for multiplicity and therefore cannot be used to infer treatment effects

Secondary outcomes

In-hospital morbidity and mortality

There were no significant differences between the treatment groups in development of vasopressor-dependent shock, development of ARDS, MV-free days up to days 8 or 28, duration of ICU stay (median 3 vs. 4 days; p = 1.00), duration of hospital stay (median 7 vs. 8 days; p = 1.00), or hospital mortality (12% vs. 10%; p = 1.00) (Table 2). Among the 25 (12 vs. 13) participants who developed new shock or ARDS, 5 (1 vs. 4) stopped study medication to receive open label glucocorticoid treatment. Among participants who required MV at randomization, there was a 3-day reduction in median duration of MV (median 4 vs. 7 days; hazard ratio (HR) 1.44; 95% CI 1.04–1.99; p = 0.21 after Bonferroni correction).

Post-discharge morbidity and mortality

There were no significant between-group differences in cardiovascular complications, quality of life, functional status, or re-hospitalizations (Table 2). The most common reasons for re-hospitalization were pneumonia (20%), congestive heart failure (18%), and chronic obstructive pulmonary disease (COPD) (17%).

The two treatment groups had similar 180-day mortality (21% vs. 24%; OR 0.86; 95% CI 0.58–1.29; p = 1.00) and RMST up to day 180 (151 days vs. 149 days; difference 2.5 days; 95% CI − 7.7 to 12.6 days; p = 1.00) (Table 2). Kaplan–Meier estimate of mortality by 180 days was 20% (95%CI, 16% to 26%) in the methylprednisolone group and 23% (95% CI 19–29%) in the placebo group. The two groups also had similar 1-year mortality (30% vs. 33%; OR 0.88; 95% CI 0.61–1.27; p = 1.00) and time to death (HR 0.90; 95% CI 0.66–1.22; p = 1.00) (Table 2 and Fig. 2A). Results of secondary outcomes were similar in the per-protocol sample (Table S2) and within the MV (Table S5 and Fig. 2B) and non-MV strata (Table S6 and Fig. 2C). Within each stratum, the two treatment groups had similar baseline characteristics (Table S7 and Table S8).

Cause of death

No apparent between-group differences in immediate or underlying cause of death were observed for all deaths, deaths up to day 60, deaths during initial hospitalization, or deaths after discharge from initial hospitalization (Tables S9 and S10).

Adverse events

During the 180 days after randomization, 365 SAEs occurred in 167 (56.2%) participants in the methylprednisolone group, and 342 SAEs occurred in 162 (56.4%) participants in the placebo group (Table S11). There were no significant differences between treatment arms in SAEs (Table S11) or complications (Table S12) during 180 days after randomization or in in-hospital or post-discharge complications (data not shown).


The ESCAPe trial showed that, in participants admitted to the ICU with severe CAP or HCAP, a 20-day treatment with low-dose methylprednisolone did not significantly reduce all-cause 60-day mortality, the primary outcome. We observed a 3-day reduction in median duration of MV in participants who required MV at randomization, although the certainty of this finding may be low given the small sample size in this subgroup, the imprecision of the estimated difference, and lack of multiplicity correction. No other significant differences were found in morbidity or mortality outcomes or complications during 1 year of follow-up.

To our knowledge, this is the largest trial investigating the efficacy of adjunct glucocorticoids on patients with severe pneumonia requiring ICU admission and the first randomized controlled trial (RCT) designed to evaluate both short- and long-term outcomes. We review our findings in the context of recent literature. In the last 15 years, 11 published RCTs investigated prolonged glucocorticoid treatment in patients hospitalized with bacterial CAP (n = 1808) [11, 26]; six of the largest RCTs (n = 1506) were part of an individual patient data meta-analysis [27].

We did not find a significant reduction in 60-day mortality or mortality up to 1 year, which is contrary to the observed reduction in 30-day mortality in severe CAP meta-analyses [11, 26]. The timing for glucocorticoid administration in this study may have missed the optimal window for intervention. Our study allowed for randomization up to 72–96 h after hospital admission. While 65% of study participants initiated study treatment within 48 h of hospital presentation and 88% within 72 h, the inherent delay in the initiation of anti-inflammatory therapy occurred during the initial peaks of inflammatory mediators in response to invasive microbial pathogens [28] and may have attenuated potential benefits [29]. Second, the methylprednisolone dose of 40 mg/day may be inadequate to achieve the level of glucocorticoid receptor saturation necessary for optimal anti-inflammatory response; a higher dose was found effective in ARDS (most attributed to pneumonia) [30]. Third, compared to the prior largest RCT on severe CAP [31], our patient population was sicker, as evidenced by oxygenation indices, need for MV, and a greater burden of comorbidities associated with glucocorticoid resistance such as chronic pulmonary and cardiovascular diseases [32]. Fourth, the observed mortality in the control group was substantially lower than what was used for the power calculation. Fifth, the broad range of severity across our study cohort likely represented different pathophysiologic processes of which corticosteroids possibly have a heterogeneous effect.

For secondary and exploratory outcomes, the 1-day reduction in median hospitalization duration (95% CI − 2.3 to 0.3 days) was similar to that reported in meta-analysis [27] and mainly driven by a 2.6-day reduction in the MV stratum (95% CI − 6.2 to 1.1 days). Contrary to prior investigations, we did not observe significant reduction in progression to shock or ARDS [26], increased risk for re-hospitalization [27], or lower myocardial infarction incidence [33].

The longer duration of methylprednisolone treatment in our trial was not associated with an increased risk of SAEs or complications within 180 days after randomization. These findings are consistent with those of updated meta-analyses of ICU patients with pneumonia [26], septic shock [34], and ARDS [30], underscoring the safety of prolonged glucocorticoid treatment in this population.

Response to glucocorticoid treatment may be affected by the severity of dysregulated systemic inflammation [31, 35, 36]. In a RCT in patients with severe CAP and C-reactive protein (CRP) levels > 150 mg/L, methylprednisolone was found to reduce treatment failure [31]. In a retrospective cohort study in patients with severe CAP admitted to ICU and receiving glucocorticoid treatment, the subgroup with CRP levels > 150 mg/L had faster recovery of hypoxemia and increased ICU- and hospital-free days [35]. These findings suggest that biologic markers may help identify patients most likely to benefit from glucocorticoid treatment. The blood samples collected in ESCAPe will allow examination of the relationship between clinical outcomes and markers of systemic inflammation over time, which may provide the groundwork for development of personalized glucocorticoid treatment strategies [30].

Evidence of glucocorticoid benefits in severe pneumonia due to coronavirus disease 2019 (COVID-19) [37, 38] and ARDS [39] has generated greater interest in this field of research. The safety of prolonged methylprednisolone treatment has been confirmed [11, 26, 27]. However, notable treatment heterogeneity in published protocols [30], such as the specific glucocorticoid, timing of initiation, dosage, duration, mode of administration, and tapering strategy, underscore the need for a more uniform approach. Further studies are required to clarify how these treatment components impact clinical outcomes and host responses. During the pandemic, variability in response to glucocorticoid treatment was observed, leading clinicians to adjust dosage and duration based on markers of inflammation and oxygenation. This has called attention to an underappreciated aspect of glucocorticoid treatment, the great interindividual variability in (i) achieved blood drug levels [40] and (ii) intracellular glucocorticoids receptor sensitivity [41], areas in need of research [30].

This trial has several limitations. First, enrollment was stopped before reaching the target sample size 1420 because of low recruitment. The main contributing factor to low recruitment was that the proportion of the patient population meeting study eligibility criteria was lower than anticipated (26% versus anticipated 70%), even though the consent rate for eligible patients was higher than anticipated (57% versus anticipated 30%). Another contributing factor was 2 years of relatively low influenza activity during the recruitment period. Second, the certainty of our overall study findings may be limited given that the sample size was lower than target and the analyses may be underpowered. Third, delayed initiation of anti-inflammatory therapy may have attenuated the differences between the treatment groups [29]. Fourth, the VA population is predominantly older, male, and with multiple comorbidities compared to the general population [27]; therefore, the trial’s results may not be generalizable to non-Veterans. Fifth, the high proportion of patients excluded due to physician opinion of not being a viable candidate might indicate a potential risk of selection bias. Sixth, this study excluded patients with recent or concurrent use of glucocorticoids; thus, it cannot determine if patients with severe CAP who require a short course of glucocorticoids for co-morbid diseases (such as COPD) would benefit from prolonged glucocorticoid treatment.


In patients with severe CAP, prolonged low-dose methylprednisolone treatment did not significantly reduce 60-day mortality. The risk for complications was similar to the control group.

Data sharing statement

See Supplement Appendix.