Corticosteroids for adult patients hospitalised with non-viral community-acquired pneumonia: a systematic review and meta-analysis

Abstract

Purpose:

International clinical practice guidelines addressing corticosteroid treatment for patients hospitalised with non-viral community-acquired pneumonia (CAP) are inconsistent.

Methods:

We conducted a systematic review of randomized controlled trials (RCTs) evaluating the use of corticosteroids in hospitalised adult patients with suspected or probable CAP. We performed random effects pairwise, Bayesian, and dose–response meta-analyses using the restricted maximum likelihood (REML) heterogeneity estimator. We assessed certainty of evidence using GRADE methodology.

Results:

We identified 30 eligible RCTs, including a total of 7519 patients. The prednisone-equivalent doses ranged between 29 mg/day and 100 mg/day. Corticosteroids probably reduced short-term (28–30 days) mortality (RR 0.82 [95% CI 0.74–0.91]; moderate certainty) while the reduction in longer term (60–90 day) mortality is less certain (RR 0.89 [95% CI 0.76–1.03]; low certainty). Corticosteroids reduced the need for invasive mechanical ventilation (IMV) (RR 0.63 [95% CI 0.48–0.82]; high certainty) and may reduce duration of ICU stay (MD 1.53 days fewer [95% CI 0.31–2.75 days fewer]; low certainty), and hospital stay (MD 2.30 days fewer [95% CI 0.81–3.81 days fewer]; low certainty). Corticosteroids probably increased hyperglycaemia requiring intervention (RR 1.32 [95% CI 1.12–1.56]; moderate certainty) but probably have no effect on secondary infections (RR 0.97 [95% CI 0.85–1.11]; moderate certainty).

Conclusion:

Corticosteroids probably reduced short-term mortality and reduce the need for invasive mechanical ventilation in hospitalised patients with CAP.

PROSPERO registration number:

CRD42024521536.

Background

Community acquired pneumonia (CAP) remains a common cause of hospitalisations worldwide [1, 2], an important cause of sepsis and acute respiratory distress syndrome (ARDS), intubation, invasive mechanical ventilation, and death worldwide [3, 4].

As the immunomodulatory effects of corticosteroids could reduce the risk of death from pneumonia, adjunctive use of corticosteroids for CAP has been examined in several randomized clinical trials (RCTs) [5]. However, there is a wide variability in adoption to clinical practice [6], as the evidence base remains relatively inconsistent. A part of this clinical variability may be explained by heterogeneity in the study populations within RCTs examining corticosteroids in CAP spanning seven decades including a wide spectrum of fatality (mortality ranges from 10 to 50%), and patients with septic shock and acute respiratory distress syndrome (ARDS).

Three recent studies illustrate this persistent uncertainty [7–9], with potential benefits observed in the CAPE-COD trial and the pneumonia subgroup of the APROCCHSS trial, not observed in the non-pandemic arm of the REMAP-CAP trial. In this context, we worked with an international group of investigators to generate an updated high-level systematic review and meta-analysis summarizing the current treatment effects and adverse events of corticosteroids in hospitalized patients with CAP.

Methods

We registered the protocol for this review on the international prospective register PROSPERO (CRD42024521536) on March 13, 2024. We have prepared this manuscript following the PRISMA guidelines for systematic reviews [10]. The overall objective of this systematic review was to estimate the effect of corticosteroid therapy compared with control (either placebo or usual care) on mortality in hospitalised adult (defined as ≥ 16 years of age) patients with suspected or confirmed CAP.

Search strategy and trial eligibility criteria

We identified trials through systematic searching of Cochrane Acute Respiratory Infections Group’s Specialised Register, MEDLINE, Embase, LILACS, references of identified trials, using the terms as reported in the Cochrane review of corticosteroids in pneumonia and a recent published systematic review [11, 12]. We excluded studies primarily assessing COVID-19, Influenza, and other viral pneumonias. We last updated our search on April 1, 2024. We also searched clinicaltrials.gov, EudraCT, and the WHO ISRCTN registry for ongoing and/or unpublished trials. We did not restrict searches by language, trial status (ongoing or completed), publication status, or date. We performed screening independently and in duplicate in two stages. First, we reviewed all titles and abstracts, and any citation deemed potentially relevant by either reviewer was advanced to full text review. At full text review, we resolved disagreements using discussion and 3rd party adjudication, when required.

We sought additional relevant trial data through contact with research networks, and by full text screening of cited references from relevant published systematic reviews or randomized trials evaluating corticosteroids in CAP [5, 12, 13]. Table S1 presents the search strategy. As patients enrolled into RCTs examining the role of corticosteroids in sepsis and ARDS syndrome may also have pneumonia, we included those RCTs identified through recent systematic reviews by our group [13–15], but only if mortality data from the CAP subgroup was reported separately.

Data collection

We collected trial data from published manuscripts. For the REMAP-CAP corticosteroid trial, which was not published at the time of data collection, we asked the investigators to provide trial level data using pre-specified data collection forms. When possible, we extracted pneumonia subgroup data from all studies included in recently published systematic review of sepsis and ARDS RCTs performed by our group [14]. If not reported in the individual RCTs, we approached the primary trial author(s) to provide these data for the outcomes of interest. Data extraction was conducted in duplicate by two independent reviewers (TP and EP).

Outcomes

We analysed short-term mortality, considering the longest reported up to 60 days after randomization. Additional outcomes included longer-term mortality (longest reported after 60 days), progression to invasive mechanical ventilation (IMV) by 28 days, progression to extracorporeal membrane oxygenation (ECMO), duration of IMV, ventilator free days (VFD) at 28 days, ICU length of stay and hospital length of stay.

For safety outcomes, we captured upper gastrointestinal bleeding, nosocomial or secondary infections, hyperglycaemia requiring interventions, delirium/psychiatric disorders and all types of serious adverse effects. We captured these safety outcomes as defined by individual study authors.

Risk of bias

Two reviewers, independently and in duplicate (TP and EP), assessed the risk of bias for individual randomized controlled trials (RCTs) at an outcome level using an updated version of the Cochrane tool (RoB 2.0) [12]. We categorised risk of bias into four levels: (i) low risk of bias, (ii) probably low risk of bias, (iii) probably high risk of bias, and (iv) high risk of bias. We made these four assessments across several domains: bias from the randomization process, deviations from the intended interventions, incomplete outcome data, outcome measurement, and selection of reported results.

Trials assessed as either or probably low risk of bias across all domains were considered to have an overall low risk of bias. We resolved discrepancies through discussion, and if needed, by involving a third reviewer for adjudication (BR).

Statistical analyses

For all outcomes, we conducted a random-effects meta-analysis using the restricted maximum likelihood (REML) estimator for heterogeneity. We reported the effects of interventions as relative risk (RR) for dichotomous outcomes and mean differences (MD) for continuous outcomes, each with 95% confidence intervals (CIs). To aid the interpretation of dichotomous outcomes, we calculated absolute risk differences (RD) per 1000 patients, along with 95% CIs. We determined baseline risk using the median risk from the standard care arms of included trials. For cluster randomized trials, we adjusted the effective sample size using the design effect as outlined in the Cochrane guidelines [16].

We performed a dose–response meta-analysis for short-term mortality using one-stage approach, random-effects and the restricted maximum likelihood (REML) heterogeneity estimator and methods proposed by Greenland et al. [17, 18]. Dose–response meta-analysis estimates the association between doses of an exposure and the relative risk or mean difference of an outcome. We used the mean daily dose of corticosteroids used in each trial with the following corticosteroid conversions: 1 mg of dexamethasone = 26.7 mg of hydrocortisone = 5.3 mg of methylprednisolone/prednisolone = 6.7 mg of prednisone [19, 20]. For analyses with five or more studies, we assessed for non-linearity by using restricted cubic splines with knots at 10%, 50%, and 90% percentiles and a Wald-type test [21]. Restricted cubic splines accommodate non-linear relationships by splitting the independent variable (i.e., dose) at “knots” and fitting separate curves between knots. For analyses in which we observed statistically significant non-linear associations, we planned to report the findings from the non-linear model (however, none were observed).

To help inform imprecision assessments, we performed trial sequential analysis (TSA), using the random effects model for short- and longer-term mortality. For TSA, we used a statistical significance level of 5%, a power of 80% and a relative risk reduction (RRR) of 15% to represent a clinically important difference. We performed TSA analyses using Trial Sequential Analysis version 0.9.5.10 beta (Copenhagen Trial Unit, Centre for Clinical Intervention Research, Rigshospitalet, Copenhagen, Denmark, www.ctu.dk/tsa)[22].

Subgroup and sensitivity analysis

For short-term mortality, we performed subgroup analysis using meta-regression based on age (continuous variable), sex, severity of CAP (severe CAP was defined using the Official Clinical Practice Guideline of the American Thoracic Society and Infectious Diseases Society of America [23] or need for ICU admission), dose (continuous variable of prednisone equivalent dose), and type of corticosteroid. We hypothesized that younger patients, those with more severe disease, and those treated with higher doses would respond more favourably to corticosteroids. For sex and molecule, we hypothesized no difference between groups.

We performed subgroup analysis based on primary patient group captured by the trial (sepsis, pneumonia, and ARDS). We also performed regression analyses based on the year of publication, to address the changes in standard care over the past two decades, since interventions such as low tidal volume ventilation has emerged. We performed subgroup analysis using fixed-effects meta-regression (‘across-trial’ approach). For any subgroup that was found to demonstrate statistically significant effect modification, we planned to assess subgroup credibility using the ICEMAN tool [24].

As sensitivity analysis, for short- and longer-term mortality, we report Bayesian inverse-variance weighted meta-analyses using the following posterior probabilities: OR < 1; OR > 1; OR < 0.9; OR > 1.11; OR < 0.8; OR > 1.25. Table S2 presents more details on the methods used to perform the Bayesian analysis. Lastly, for short-term mortality, we performed a post hoc analysis comparing earlier vs later administration of corticosteroids. We performed two analyses, once defining early as within 24 h and the later within 48 h.

Certainty of the evidence

For all outcomes, we evaluated the certainty of evidence using GRADE methodology. We rated certainty as high, moderate, low, or very low, considering factors such as risk of bias, inconsistency, indirectness, imprecision, and publication bias. In assessing imprecision, we adopted a minimally contextualized framework, focusing solely on whether confidence intervals encompass a minimally important effect, without considering the range of plausible effects indicated by the confidence intervals [25]. We used the following threshold for imprecision rating: short- and longer-term mortality (null), need for invasive mechanical ventilation (1%), serious adverse events (2%), and ventilatory free days/ duration of hospitalisation (1 day). We contextualize findings using the GRADE narrative summary guidance, reflecting the certainty of evidence and the effect’s size (for example, corticosteroids decrease mortality [high certainty], corticosteroids likely decrease mortality [moderate certainty], corticosteroids may decrease mortality [low certainty], and the impact of corticosteroids on mortality is highly uncertain [very low certainty]) [26].

Results

Our updated search identified 834 unique citations including one relevant post-hoc RCT subgroup analysis and one new relevant study (REMAP-CAP trial) based on a conference presentation. We included 18 RCTs from our previously published systematic review examining the same question [12]. From previous reviews evaluating corticosteroids in patients with sepsis or ARDS, we identified 59 potentially eligible studies, including four studies that explicitly reported CAP subgroup data in the primary publication. Also, 17 RCTs in sepsis and ARDS reported on patients with pneumonia but did not specify whether this was community-acquired and after contacting study authors, we included four of them for which we were able to acquire CAP specific outcome data. From these sources, we have included ten additional studies. In total, we included 30 RCTs (n = 7519 patients) for this systematic review evaluating the role of corticosteroids in patients hospitalised with CAP [7–9, 27–52]. Figure 1 presents more detail on the inclusion and exclusion process.

Fig. 1.

PRISMA diagram presenting the inclusion and exlcusion criteria

Trial characteristics

We included trials published between 1956 and 2024, representing seven decades of research, with 79% of the data from studies published in the last decade (since 2014) (Figure S3). Most trial participants were male (ranging from 31.6 to 96%) and middle-aged (mean ages varying from 36.4 to 76.1 years). At the time of randomization, 51% of patients were admitted into ICU, and 30% were mechanically ventilated. Twenty-two trials (4730 patients) had severe CAP while eight trials primarily enrolled hospitalised patients with non-severe CAP (2789 patients). Only 18% of patients met ARDS criteria at enrolment.

Regarding the type of corticosteroids, three trials examined dexamethasone (852 patients), 13 studied hydrocortisone (3005 patients), 8 methylprednisolone (1129 patients), 5 prednisone/prednisolone (1971 patients), and one trial investigated a combination of hydrocortisone with fludrocortisone (562 patients). Daily prednisone equivalent dose ranged between 29 and 100 mg. Table 1 presents the trial characteristics.

Table 1.

Characteristics of key clinical trials on corticosteroid use severe and non-severe patients

Study	Number randomized	Male (%)	Age	ICU (%)	Mechanical ventilation	Severity	CRP (mg/L)	ARDS	Intervention
Angus et al. 2024 (REMAP-CAP)	455	61.3	62	100	40.2	Severe [all ICU]	226	0	Hydrocortisone
Annane et al. 2006	177	31.6	61	100	NR	Severe [all ICU]	NR	100	Hydrocortisone
Blum et al. 2015 (STEP)	785	62	74	0	0	Non-severe [PSI I–III > 50%]	161.5	0	Prednisone
Confalonieri et al.2005	46	69.5	63.5	100	73.9	Severe [majority ICU patients]	420	0	Hydrocortisone
Dequin et al. 2023 (CAPE COD)	795	69.4	67	NR	22.2	Severe [PSI scores IV–V > 50%]	250	0	Hydrocortisone
El-Ghamrawy et al. 2006	34	61.8	61.8	100	NR	Severe [majority ICU patients]	NR	0	Hydrocortisone
Fernández Serrano et al. 2011	45	66.7	63	0	0	Severe [fine scores IV–V > 50%]	NR	0	Methylprednisolone
Gang et al. 2016	58	NR	NR	100	NR	Severe [majority ICU patients]	NR	NR	Methylprednisolone
Gordon et al. 2016 (VANISH)	104	61.54	65	100	76.92	Severe [all ICU]	NR	NR	Hydrocortisone
Heming et al. 2024 (APROCCHSS)	562	70	65	100	96	Severe [all ICU]	NR	41.67	Hydrocortisone/Fludrocortisone
Keh et al. 2016 (HYPRESS)	54	75.9	70	100	7.4	Severe [all ICU]	220.8	22.20%	Hydrocortisone
Lloyd et al. 2019 (IMPROVe-GAP)	816	57	76.1	10.5	NR	Non-severe [50% of patients with CORB scores < 2; all GIM patients]	88.2	0	Prednisone
Marik et al. 1993	30	NR	36.44	100	NR	Severe [mean Apache II score 13, all ICU patients]	NR	NR	Hydrocortisone
McHardy and Schonell et al. 1972	126	48.4	60.3	0	NR	Non-severe [defined by trials, most patients classified as mild-moderate]	NR	NR	Prednisolone
Meduri et al. 1998	24	37.5	48.3	100	100	Severe [all ICU]	NR	100	Methylprednisolone
Meduri et al. 2007	91	51.6	51	100	100	Severe [all ICU]	254.3	100	Methylprednisolone
Meduri et al. 2022 (ESCAPe)	584	96	68.8	100	33	Severe [PSI scores IV–V > 50%]	NR	11%	Methylprednisolone
Meijvis et al. 2011 (Ovidius)	304	56.5	63.6	0	0	Non-severe [PSI I–III > 50%]	217	0	Dexamethasone
Mikami et al. 2007	31	74.2	72	0	0	Non-severe [PSI I–III > 50%]	19.7	0	Prednisolone
Nafae et al. 2013	80	56.2	49	NR	0	Severe [based on baseline vitals indicating mean CORB score > 2]	92.3	0	Hydrocortisone
Rezk et al. 2013	27	85.2	45.3	100	100	Severe [all ICU]	232.67	100	Methylprednisolone
Sabry et al. 2011	80	72.5	62.2	100	75	Severe [majority ICU patients]	568.5	0	Hydrocortisone
Snijders et al. 2010	213	58.2	63.5	10.3	NR	Non-severe [PSI I–III > 50%]	235.9	0	Prednisolone
Steinberg et al. 2006 (LaSRS)	180	49.1	49.1	100	100	Severe [all ICU]	NR	100	Methylprednisolone
Tongyoo et al. 2016	197	51.2	64.4	100	100	Severe [all ICU]	NR	100	Hydrocortisone
Torres et al. 2015	120	61.4	65.3	75	2.5	Severe [PSI scores IV–V > 50%]	258.7	0	Methylprednisolone
Venkatesh et al. 2018 (ADRENAL)	840	61	62	100	NR	Severe [all ICU]	NR	NR	Hydrocortisone
Villar et al. 2020 (DEXA-ARDS)	147	66.7	54	100	100	Severe [all ICU]	NR	100	Dexamethasone
Wagner et al. 1956	113	67.3	NR	NR	NR	Non-severe [as defined by authors]	NR	0	Hydrocortisone
Wittermans et al. 2021	401	67.4	67.5	0	0	Non-severe [PSI I–III > 50%]	204.5	0	Dexamethasone

Risk of bias

For short-term mortality, we judged 11 trials to be at a high risk of bias. We judged eight trials to be at risk of bias due to issues arising from the randomization process, nine at risk due to deviations from the intended interventions, three at risk of bias due to missing data, and four at risk of bias due to selective reporting. Figure 2 presents the risk of bias assessments for short-term mortality.

Fig. 2.

Risk of bias assessment for short-term mortality

All-cause mortality

Our search found 29 trials (7494 patients) reporting short-term mortality, including 1185 events. Based on pooled analysis, corticosteroids probably reduced short-term mortality (RR 0.82 [95% CI 0.74–0.91]) (moderate certainty) and may reduce long-term mortality compared to usual care (RR 0.89 [95% CI 0.76–1.03]) (low certainty). Trial sequential analysis showed that we did not meet the optimal information size and this helped to inform imprecision assessments, ultimately lowering the certainty of evidence (Figure S4). Figure 3 and Fig. 4 illustrate the forest plot for short- and long-term mortality, respectively. Table 2 presents the summary of findings.

Fig. 3.

Short-term mortality forest plot, with relative risk (RR) and 95% confidence inervals (CI)

Fig. 4.

Long-term mortality forest plot, with relative risk (RR) and 95% confidence inervals (CI)

Table 2.

Summary of findings on corticosteroid use in community acquired pneumonia patients across outcomes

Outcomes	Number of participants (studies)	Certainty of the evidence (GRADE)	Relative effect (95% CI)	Anticipated absolute effects*
				Risk with control	Risk difference with steroids
Short-term mortality (30–60 days)	7494 (29 RCTs)	⊕⊕⊕◯ Moderatea	RR 0.82 (0.74–0.91)	175 per 1000	31 fewer per 1000 (45 fewer to 16 fewer)
Long-term mortality (90–180 days)	4363 (10 RCTs)	⊕⊕◯◯ Lowa,b	RR 0.89 (0.76–1.03)	278 per 1000	31 fewer per 1000 (67 fewer to 8 more)
Progression to ECMO	602 (2 RCTs)	⊕⊕◯◯ Lowc	RR 0.90 (0.26–3.13)	21 per 1000	2 fewer per 1000 (16 fewer to 45 more)
Progression to IMV	2774 (10 RCTs)	⊕⊕⊕⊕ High	RR 0.63 (0.48–0.82)	100 per 1000	37 fewer per 1000 (52 fewer to 18 fewer)
Duration of ICU stay	1825 (11 RCTs)	⊕⊕◯◯ Lowa,b	–	–	MD 1.53 days fewer (2.75 fewer to 0.31 fewer)
Duration of hospitalization	3546 (14 RCTs)	⊕⊕◯◯ Lowa,b	–	–	MD 2.3 days fewer (3.81 fewer to 0.8 fewer)
Duration of IMV	1777 (9 RCTs)	⊕⊕⊕◯ Moderatea	–	–	MD 3.15 days fewer (4.16 fewer to 1.61 fewer)
Ventilator-free days	2980 (7 RCTs)	⊕⊕◯◯ Lowa,b	–	–	MD 2.03 days more (0.05 fewer to 4.1 more)
Serious adverse events	2544 (6 RCTs)	⊕⊕◯◯ Lowb,d	RR 0.75 (0.57–0.99)	395 per 1000	99 fewer per 1000 (170 fewer to 4 fewer)
Hyperglycaemia	4513 (16 RCTs)	⊕⊕⊕◯ Moderatea	RR 1.32 (1.12–1.56)	261 per 1000	84 more per 1000 (31 more to 146 more)
Gastrointestinal bleeding	4304 (14 RCTs)	⊕⊕⊕◯ Moderateb	RR 0.86 (0.58–1.29)	26 per 1000	4 fewer per 1000 (11 fewer to 7 more)
Secondary infections	4485 (15 RCTs)	⊕⊕⊕◯ Moderateb	RR 0.97 (0.85–1.11)	150 per 1000	4 fewer per 1000 (22 fewer to 16 more)
Neuropsychiatric effects	2850 (7 RCTs)	⊕⊕⊕◯ Moderateb	RR 1.10 (0.89–1.36)	77 per 1000	8 more per 1000 (8 fewer to 28 more)

CI confidence interval, MD mean difference, RR risk ratio Explanations

GRADE Working Group grades of evidence

High certainty: we are very confident that the true effect lies close to that of the estimate of the effect

Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different

Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect

Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect

^*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI)

^aSignificant heterogeneity in the effect size

^bCrosses the MCID

^cCrosses the MCID in both directions

^dHeterogenous definitions of serious adverse events

Other clinical outcomes

Corticosteroids reduced the need for IMV (RR 0.63 [95% CI 0.48–0.82]) (high certainty) and probably reduced the duration of IMV (MD 3.15 days fewer [95% CI 1.61–4.16 days fewer]) (moderate certainty) compared to patients not receiving corticosteroids. Therapy with corticosteroids may reduce the duration of ICU stay (MD 1.53 days fewer [95% CI 0.31–2.75 days fewer] and duration of hospitalisation (MD 2.30 days fewer [95% CI 0.81–3.81 days fewer]) (both low certainty).

Corticosteroids may increase ventilatory-free days (MD 2.03 days more [95% CI 0.05 fewer to 4.10 more]) (low certainty) while the effect on the need for ECMO (RR 0.90 (95% CI 0.26–3.13) is uncertain (very low certainty). Figures S5–S10 present the forest plots.

Adverse events

Only 6 of 30 trials reported serious adverse events (2544 participants; 841 events). Corticosteroids may reduce the risk of serious adverse events (RR 0.75 [95% CI 0.57–0.99]) (low certainty) but probably increase the risk of hyperglycaemia (RR 1.32 [95% CI 1.12–1.56]) (moderate certainty) as compared to patients that did not receive corticosteroids. The impact of corticosteroids on gastrointestinal bleeding, secondary infections and neuropsychiatric effects remains uncertain (all low to very low certainty). Table 2 and Tables S11–15 presents the summary of findings.

Subgroups and sensitivity analyses

We did not observe evidence of effect modification based on type of corticosteroid, age or sex of the patients (S16–18). We found a non-linear dose response relationship to corticosteroid regimens and short-term mortality benefit, with the most effective prednisone-equivalent dose of 60 mg/day (RR 0.78 [95% CI 0.68–0.89]). In Figures S19–20, we reported the results of these analyses.

We did not observe evidence of effect modification based on the primary trial population ARDS, pneumonia, sepsis (Figures S21–S22). We also observed no difference in response to corticosteroids when accounting for year of publication in our regression analysis or based on high or low risk of bias (Figures S23–S24). The Bayesian meta-analysis showed posterior probability of 86.8% for OR < 0.9 for short-term mortality with corticosteroids and 78.7% for OR < 0.9 for long-term mortality (Figures S25–S26). There was no evidence that early administration affected mortality as compared to later administration (Tables S27–S28).

Discussion

In this updated systematic review and meta-analysis of evidence examining the treatment effects and safety of corticosteroid treatment in hospitalised adult patients with CAP participating in RCTs, we showed new data from the REMAP-CAP Trial, the pre-planned subgroup analysis of the APROCCHSS study, and RCTs evaluating sepsis or ARDS reporting CAP-specific subgroup data. These RCTs spanned seven decades and differed in the type and dose of corticosteroid evaluated, although 79% of our data are from studies published in the last decade (since 2014). Our analyses included the largest sample to date (7519 patients from 30 RCTs) of any published meta-analysis, thereby gaining precision and certainty of evidence examining pooled treatment effects. We report that in hospitalised adult patients with CAP, corticosteroids probably reduce short-term mortality, reduce the need for invasive mechanical ventilation, and likely reduce the duration of ICU stay. The impact of corticosteroid use on ventilator-free days, the need for ECMO, on the duration of ICU, and hospital stay were uncertain. We found a non-linear dose response relationship, with the estimated most effective prednisone-equivalent dose of 60 mg/day for short term mortality benefit. Aside from the well reported risk of hyperglycaemia with corticosteroids in patients with CAP receiving corticosteroid therapy, limited adverse event data precludes meaningful inferences on clinically important adverse events such as GI bleeding, secondary infections and neuropsychiatric effects.

Among recent RCTs, the CAPE-COD trial stopped early for benefit. It is recognised that RCTs that stop early for benefit may are often associated with greater effect sizes compared with untruncated trials [53, 54]. Of note, mortality rate seen in the control group in the CAPE-COD trial was lower than expected. Similarly, subgroup analysis of moderately sized RCTs could also result in overestimation of treatment effects [55], which may be a consideration when interpreting the benefits with CAP observed in the exploratory subgroup analysis of the APROCCHSS phase-3 trial. In contrast, the benefits of corticosteroids for CAP were not observed in the REMAP-CAP and ESCAPe studies. The ESCAPe trial was underpowered to detect a mortality difference and was stopped early due to slow enrolment during the COVID-19 pandemic. It has been suggested that discrepancies of these relatively large RCTs may be due to variability in corticosteroid type or dose. However, we did not observe a difference in treatment effect across type of corticosteroid used in this analysis [12].

Our study has several strengths. Although there have been many recent systematic reviews on this topic, we provide an extensive update, including the recently published REMAP-CAP data [12, 56–58]. We have taken a comprehensive approach in including previously reported studies evaluating sepsis or ARDS reporting CAP-specific subgroup data [12, 13, 15]. As a result, we have how included a total of 30 RCTs, and 7519 patients, this as compared to previously published largest meta-analysis which includes about half of the total patients and RCTs (15 RCTs, n = 3367). In contrast to previous systematic reviews that have reported findings with overlapping populations, (e.g. CAP with and without septic shock, CAP with and without ARDS, more and less severe CAP), our comprehensive approach to subgroup analysis has allowed us to formulate more specific and nuanced conclusions regarding groups that have previously been hypothesized to benefit from corticosteroid exposure. The use of GRADE and ICEMAN tool have allowed for a careful assessment of certainty of evidence, and credibility of subgroup findings and have helped to contextualize findings for readers.

We also acknowledge limitations of our study. We acknowledge the heterogeneity in terms of RCTs spanning seven decades where usual care has changed and outcomes from CAP have improved. There were differences in the eligibility criteria between RCTs and in the dose and steroids drug tested. However, our subgroup analyses aimed to address these differences. We also acknowledge that despite contacting all eligible trial investigators, not all were able to provide data for analysis resulting in a small number of sepsis and ARDS trials providing CAP subgroup data. There was marked variation in defining ‘severe’ CAP across the included studies. Evidence for several important outcomes was based on low or very low certainty precluding definitive conclusions. Our results are application primarily to all-cause CAP. Thus, it remains unclear whether the effect of corticosteroids is consistent across different microbiological causes of pneumonia, and including viral pneumonia which we excluded, and whether our results can be extrapolated to ARDS. We also did not study modifiable factors related to ICU admission including country, health care systems (and resources available), and staffing. As this information was not available in the trials included in this study, we were unable to study whether the effects of corticosteroids differed between patients treated with low versus high tidal volume ventilation. Similarly, variation in staffing and expertise between centres, number of studies reporting the effect of corticosteroids on progression to ECMO was small and therefore conclusions on this were limited.

Importance of our findings and next steps

Existing clinical practice guidelines are inconsistent in their recommendations for corticosteroid use in patients with CAP [5, 6]. The current review highlights further research is needed to delineate the effect of corticosteroids in patients with less severe CAP, patients with viral pneumonia and in patients with all-cause ARDS. Further research is also needed to better delineate if certain patient groups with CAP are more likely to benefit from corticosteroids than others, based on specific severity criteria, biomarkers, microbiologic subgroup, or other considerations.

Conclusions

In this meta-analysis of clinical trials of adult patients hospitalized for CAP, administration of corticosteroids, compared with usual care or placebo, probably reduce short-term mortality. Limited adverse event data preclude meaningful inferences on clinically important adverse events such as GI bleeding, secondary infections and neuropsychiatric effects, due to administration of corticosteroids.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1007/s00134-025-07912-2

Data availability

Data request can be made to the corresponding author.