Abstract
Background
The number of immunocompromised patients with community-acquired pneumonia (CAP) admitted to intensive care unit (ICU) is rising, accompanied by high mortality. Although glucocorticoid (GC) therapy is recommended for severe CAP by Society of Critical Care Medicine (SCCM) in 2024, its efficacy in immunocompromised patients remains uncertain. This study aimed to evaluate the impact of GC therapy on mortality in immunocompromised patients with CAP in the ICU.
Methods
A retrospective cohort study was conducted using the MIMIC-IV 3.1 database. Primary outcomes was twenty-eight days (28-d) mortality. Secondary outcomes included six months (6-m) and twelve months (12-m) mortality, in-hospital mortality, ICU mortality, hospital and ICU length of stay, and 28-d ventilator-free days. Propensity score matching (PSM) was applied to minimize selection bias. Cox proportional hazards models were used to assess the association between GC therapy and mortality, also including analyses of timing, duration, and average daily dose of GC administration. Sensitivity analysis was performed using E-values.
Results
A total of 1770 patients were included, with 593 receiving antibiotics plus GC therapy. After PSM, 521 patients were included in both the GC and non-GC treatment groups. GC therapy was associated with increased 28-d mortality (GC vs non-GC, after PSM: 29% vs 23%, HR = 1.373, p < 0.05). Secondary outcome, GC therapy was associated with increased 6-m and 12-m mortality (GC vs non-GC, after PSM: 6-m, 47% vs 43%; 12-m, 56% vs 51%, HR = 1.215; p < 0.05). Six different Cox models confirmed GC therapy was associated with increased 28-d mortality (HR = 1.337–1.374, p < 0.05). Initiating GC therapy > 48 hours after ICU admission, treatment duration > 7 days, and low-dose GC (≤0.5 mg/kg/d) were also linked to higher 28-d mortality (p < 0.05). E-value analysis indicated the results were relatively robust.
Conclusion
Among immunocompromised CAP patients in the ICU, GC therapy may increase 28-d，6-m and 12-m mortality, an association potentially attributable to its late initiation, prolonged duration, and insufficient dosage. Further research is needed to define the optimal GC treatment regimen for this population.
Supplementary information
The online version contains supplementary material available at 10.1186/s12879-026-12682-3.
Keywords: Glucocorticoid, Immunocompromised, Community-acquired pneumonia, Mortality
Background
Community-acquired pneumonia (CAP) is an infection of the lung parenchyma acquired outside of hospital settings and represents one of the leading causes of global disease burden [1, 2]. When the inflammatory response in CAP becomes dysregulated, the disease may progress to severe CAP (SCAP) [3]. SCAP used to describe ICU-admitted patients with CAP as they might require organ support [4]. Despite early initiation of antibiotic therapy, SCAP remains associated with high mortality, with approximately 50% of ICU-admitted patients dying within one year [5].
SCAP is characterized by a dysregulated inflammatory response, which may lead to critical illness-related corticosteroid insufficiency (CIRCI) [6]. CIRCI results from dysfunction of the hypothalamic-pituitary-adrenal axis, altered cortisol metabolism, and glucocorticoid (GC) resistance in the context of severe inflammation [7]. GC supplementation can exert anti-inflammatory effects by activating and enhancing the innate immune system, promoting the resolution of inflammation, and facilitating tissue repair [8]. However, the use of GC therapy in CAP patients has been controversial.
From 2007 to 2019, guidelines for the management of CAP issued by American Thoracic Society (ATS)/Infectious Diseases Society of America (IDSA) shifted from a controversial to a nonrecommendation regarding GC use [9, 10]. With the advancement of randomized controlled trials (RCTs), evidence supporting the benefits of GC therapy in SCAP patients has increased. In the latest CAP guidelines independently released by ATS in 2025 [11], GC are recommended for adult SCAP of noninfluenza etiology. In parallel, Society of Critical Care Medicine (SCCM) has shown earlier inclination toward GC use—its 2017 guidelines conditionally recommended GC for hospitalized CAP patients [12], and the 2024 update strongly recommends GC therapy for bacterial SCAP patients [13]. In any case, the most recent guidelines from both ATS and SCCM support the efficacy of GC in treating SCAP. As for the treatment of COVID19 pneumonia, GC use is also recommended in the National Institutes of Health (NIH) management guidelines [14]. Specifically, current guidelines suggest that GC therapy may be beneficial for hospitalized COVID-19 patients receiving any form of oxygen therapy, patients with bacterial SCAP, and those with SCAP complicated by acute respiratory distress syndrome (ARDS) or sepsis [6, 13]. However, these guidelines and previous RCTs have excluded immunocompromised patients.
With advancements in healthcare, the survival of immunosuppressed individuals—such as transplant recipients and cancer patients—has improved, resulting in a growing number of immunocompromised patients in ICU [15]. The ICU and in-hospital mortality for CAP in immunocompromised patients have reached 32.4% and 44.1%, respectively [16]. To date, it remains unclear whether immunocompromised patients with SCAP can benefit from GC therapy.
This study aims to evaluate the impact of GC therapy on mortality among immunocompromised patients with CAP in the ICU, in order to provide further evidence for the use of GC in this specific population.
Methods
Study design and data source
This retrospective clinical cohort study utilized data from the Medical Information Mart for Intensive Care IV (MIMIC-IV), a publicly available critical care database [17]. Version 3.1 of the database includes records from 364,627 patients, encompassing 546,028 hospital admissions and 94,458 ICU admissions at the Beth Israel Deaconess Medical Center, affiliated with Harvard Medical School, between 2008 and 2022. Extract clinical data of immunocompromised CAP patients in the ICU from MIMIC-IV 3.1 database.
Study population
Patients were required to meet the 2007 ATS/IDSA diagnostic criteria for CAP [9] and be classified as immunocompromised according to the 2020 CAP consensus guidelines [18]. The inclusion criteria were adapted from the study by Reyes et al. [19]. Patients were eligible if they were aged > 18 years, admitted to the ICU within 48 hours of hospital admission, had a documented immunocompromising condition (Table S1), had pneumonia listed among the top ten diagnoses (Table S2), and received pneumonia-related antibiotic therapy for at least 96 hours initiated within 48 hours of admission. Exclusion criteria included patients hospitalized outside the ICU for more than 48 hours before ICU admission; patients who did not meet the antibiotic treatment criteria; immunocompetent individuals, HIV patients, patients with pneumocystis jirovecii pneumonia (PJP) (considering that there are many prospective studies on HIV and PJP). Pneumonia-related antibiotics included β-lactams (including carbapenems) and fluoroquinolones, receipt of at least one of these classes was required for inclusion. Pathogens were identified based on pneumonia diagnosis and sputum culture results obtained within 48 hours of ICU admission. Patients who received intravenous GC therapy within one week of admission were assigned to the GC treatment group.
Data collection
Collected data included patient demographics, Charlson Comorbidity Index, and indicators of disease severity within 48 hours after ICU admission. These indicators comprised the Simplified Acute Physiology Score II (SAPS II), Acute Physiology Score III (APS III), Sequential Organ Failure Assessment (SOFA) score, use of invasive mechanical ventilation, and the occurrence of septic shock and ARDS. All scores used were the maximum values recorded within 48 hours. Outcome variables included ICU, in-hospital, twenty-eight days (28-d), six months (6-m), and twelve months (12-m) mortality, as well as overall survival time. Laboratory parameters within the first 48 hours of ICU admission, results of the first two sputum cultures, and details regarding the use of GC (including timing of initiation, duration of therapy, and daily dose), as well as pneumonia-related antibiotic use were extracted. Data queries were conducted using Structured Query Language (SQL) and processed via Navicat Premium V16 software (License No. 65485839).
Study outcomes
The primary outcomes for the GC and non-GC groups was 28-d mortality. Secondary outcomes included 6-m and 12-m mortality, ICU mortality, in-hospital mortality, ICU and hospital length of stay, and 28-d ventilator-free days.
Statistical analysis
The Kolmogorov-Smirnov test was used to assess the normality of continuous variables, which were reported as mean ± standard deviation (SD) or median with interquartile range (IQR). Group comparisons were performed using the independent samples t-test for normally distributed variables and the Mann-Whitney U test for non-normally distributed variables. Categorical variables were summarized as counts (percentages) and compared using the chi-square test or Fisher’s exact test. Missing data were handled using multiple imputation via random forest, with the number of imputations set at 5.
To balance baseline characteristics between the GC and non-GC groups, propensity score matching (PSM) was performed using logistic regression to calculate propensity scores. A 1:1 nearest-neighbor matching algorithm was applied with a caliper width of 0.05. Standardized mean difference (SMD) and p-value were used to evaluate the balance between groups, with SMD > 0.1 indicating imbalance.
A multivariable Cox proportional hazards model was used to assess the effect of GC therapy on mortality before and after PSM. An extended Cox model was constructed based on matched variables to adjust for multiple covariates, including sex, age, Charlson Comorbidity Index, SOFA score, SAPS II, APS III, presence of ARDS and septic shock, use of invasive mechanical ventilation, type of immunosuppression, and type of pathogen. These covariates were selected based on their clinical relevance to GC treatment decision-making. The Cox regression model also evaluated the effects of GC initiation time, duration of use, and average daily dose on mortality.
Subgroup analyses after PSM were performed using Cox proportional hazards models to evaluate the effect of GC therapy on mortality across different strata, including sex, age, SOFA score, type of immunocompromising disease, pathogen category, presence of ARDS or septic shock, and use of invasive mechanical ventilation.
Sensitivity analyses were performed using the E-value to assess the robustness of results against unmeasured confounding [20, 21]. To evaluate potential bias, an additional multivariable Cox regression excluding patients with a prior history of steroid use was conducted. p value < 0.05 was considered statistically significant for all analyses. All analyses were done using R software (version 4.4.2).
Results
Patient characteristics
A total of 1,770 immunocompromised patients with CAP were included, comprising 593 patients in the GC treatment group and 1,177 in the non-GC group (Fig. 1). The GC types included dexamethasone, hydrocortisone, and methylprednisolone. Among them, 472 patients received GC within 48 hours of ICU admission, while 121 initiated GC therapy between 48 hours and 7 days. The average GC treatment duration was 6 days. Specifically, 204 patients received GC for less than 48 hours, 252 for 48 hours to 7 days, and 137 for more than 7 days. After converting all dosages to methylprednisolone equivalents, the average daily dose was 53 mg.
Fig. 1.
Patients’ enrollment flowchart
Regarding immunocompromised status, 721 patients had primary immunodeficiency, 640 had cancer and/or were undergoing chemotherapy, 169 were transplant recipients, 163 were on long-term GC therapy, and 77 were receiving other immunosuppressive medications. Among the overall cohort, 37% (652/1,770) received invasive ventilation, 3% (61/1,770) developed ARDS, and 22% (391/1,770) experienced septic shock.
Microbiological confirmation was obtained in 682 cases. Among them, viral pneumonia was diagnosed in 11% (74/682), including 12 SARS-CoV-2 cases. Bacterial pneumonia (including atypical bacteria) accounted for 78% (531/682), and fungal or aspergillus pneumonia for 11% (77/682). The five most frequently identified bacterial pathogens were Staphylococcus aureus (21%, 144/682), Pseudomonas aeruginosa (10%, 71/682), Klebsiella pneumoniae (5%, 34/682), Streptococcus pneumoniae (4%, 29/682), and Escherichia coli (3%, 19/682). A total of 93% (1,647/1,770) of patients received β-lactam antibiotics.
No significant difference in ARDS incidence was observed between the GC and non-GC groups (p > 0.05). However, significant differences existed in age, sex, Charlson Comorbidity Index, illness severity scores (SOFA, SAPS II, APS III), septic shock incidence, type of immunosuppression, and pathogen distribution (p < 0.05). To minimize selection bias, PSM was performed based on demographics, Charlson Comorbidity Index, severity of illness within 48 hours after ICU admission (SAPS II, APS III, SOFA score, mechanical ventilation rate, incidence of septic shock and ARDS), immunocompromised disease, pathogen, and laboratory parameters. After PSM, all SMD were < 0.1, indicating adequate balance between GC and non-GC groups (Table 1, Fig. 2), with 521 patients in each group.
Table 1.
Baseline characteristics of the GC and non-GC groups before and after PSM
| characteristic | Before PSM | After PSM | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| All patients (n = 1770) | Non-GC (n = 1177) | GC (n = 593) |
P value | SMD | All patients (n = 1042) | Non-GC (n = 521) |
GC (n = 521) |
P value | SMD | |
| Age, year | 67 (15) | 68 (15) | 65 (14) | < 0.001 | 0.172 | 66 (14) | 67 (15) | 66 (14) | 0.484 | 0.043 |
| Gender, male, % | 1001(57) | 685 (58) | 316 (53) | 0.049 | 0.099 | 561 (54) | 283 (54) | 278 (53) | 0.756 | 0.019 |
| Charlson comorbidity index | 6.61 (2.97) | 6.73 (3.01) | 6.36 (2.86) | 0.010 | 0.128 | 6.56 (2.94) | 6.56 (3.05) | 6.55 (2.82) | 0.958 | 0.003 |
| SOFA | 5.42 (3.59) | 5.17 (3.37) | 5.90 (3.96) | < 0.001 | 0.199 | 5.71 (3.69) | 5.72 (3.65) | 5.71 (3.74) | 0.973 | 0.002 |
| SAPS II | 40.55 (13.72) | 39.76 (12.96) | 42.13 (14.98) | 0.001 | 0.169 | 41.87 (14.18) | 41.98 (14.05) | 41.77 (14.32) | 0.814 | 0.015 |
| APS III | 51.42 (19.99) | 50.36 (18.62) | 53.52 (22.33) | 0.003 | 0.154 | 52.54 (21.03) | 52.47 (20.94) | 52.60 (21.14) | 0.924 | 0.006 |
| ARDS, % | 61 (3) | 42 (4) | 19 (3) | 0.692 | 0.020 | 30 (3) | 13 (3) | 17 (3) | 0.459 | 0.046 |
| Septic shock, % | 391 (22) | 234 (20) | 157 (26) | 0.002 | 0.157 | 252 (24) | 127 (24) | 125 (24) | 0.885 | 0.009 |
| Haemoglobin, g/dL | 10.26 (2.17) | 10.11 (2.12) | 10.55 (2.25) | < 0.001 | 0.203 | 10.36 (2.17) | 10.34 (2.21) | 10.37 (2.14) | 0.834 | 0.013 |
| Platelets, K/uL | 235.55 (146.39) | 242.31 (146.96) | 222.13 (144.43) | 0.006 | 0.139 | 225.39 (144.13) | 225.15 (141.23) | 225.64 (147.11) | 0.956 | 0.003 |
| Leukocyte, K/uL | 13.05 (11.61) | 13.05 (11.26) | 13.05 (12.29) | 0.994 | < 0.001 | 12.87 (11.59) | 12.88 (11.72) | 12.85 (11.46) | 0.970 | 0.002 |
| Neutrophils, % | 77.66 (18.87) | 78.67 (17.29) | 75.67 (21.54) | 0.003 | 0.154 | 76.09 (20.75) | 75.63 (20.66) | 76.54 (20.84) | 0.479 | 0.044 |
| Lymphocyte, % | 11.93 (13.05) | 11.78 (12.69) | 12.23 (13.74) | 0.504 | 0.034 | 12.30 (13.86) | 12.59 (14.18) | 12.00 (13.55) | 0.494 | 0.042 |
| Blood urea nitrogen, mg/dL | 31.61 (23.56) | 32.11 (23.60) | 30.62 (23.48) | 0.208 | 0.063 | 30.85 (23.51) | 30.80 (23.32) | 30.90 (23.72) | 0.944 | 0.004 |
| Serum creatinine, mg/dL | 1.68 (1.60) | 1.76 (1.68) | 1.51 (1.41) | 0.001 | 0.159 | 1.54 (1.43) | 1.55 (1.40) | 1.53 (1.47) | 0.881 | 0.009 |
| Glucose, mg/dL | 147.81 (69.63) | 148.96 (72.44) | 145.52 (63.69) | 0.306 | 0.050 | 145.30 (65.89) | 144.44 (67.63) | 146.16 (64.15) | 0.674 | 0.026 |
| Invasive ventilation, % | 681 (38) | 417 (35) | 264 (45) | < 0.001 | 0.186 | 436(42) | 213 (41) | 223.00 (43) | 0.530 | 0.039 |
| Immunocompromised category | < 0.001 | 0.559 | 0.853 | 0.072 | ||||||
| Primary immunodeficiency, % | 721 (41) | 554 (47) | 167 (28) | 336 (32) | 173 (33) | 163 (31) | ||||
| Malignancy and chemotherap, % | 640 (36) | 432 (37) | 208 (35) | 402 (39) | 202 (39) | 200 (39) | ||||
| Transplantation, % | 169 (10) | 91 (8) | 78 (13) | 124 (12) | 60 (12) | 64 (12) | ||||
| History of glucocorticoid use, % | 163 (9) | 55 (5) | 108 (18) | 118 (11) | 54 (10) | 64 (12) | ||||
| History of immunosuppressant use, % | 77 (4) | 45 (4) | 32 (5) | 62 (6) | 32 (6) | 30 (6) | ||||
| Pathogen | < 0.001 | 0.237 | 0.978 | 0.055 | ||||||
| Unknow, % | 1088 (61) | 747(63) | 341 (57) | 627 (60) | 318 (61) | 309 (59) | ||||
| Staphylococcus aureus, % | 144 (8) | 93 (8) | 51 (9) | 83 (8) | 40 (8) | 43 (8) | ||||
| Pseudomonas aeruginosa, % | 71 (4) | 41 (3) | 30 (5) | 47 (5) | 21 (4) | 26 (5) | ||||
| Other bacteria, % | 316 (18) | 219 (19) | 97 (16) | 178 (17) | 89 (17) | 89 (17) | ||||
| Virus, % | 74 (4) | 34 (3) | 40 (7) | 54 (5) | 27 (5) | 27 (5) | ||||
| Fungi & Aspergillus, % | 77 (4) | 43 (4) | 34 (6) | 53 (5) | 26 (5) | 27 (5) | ||||
APS, acute physiology score; SAPS, simplified acute physiology; SOFA, sequential organ failure assessment; ARDS, acute respiratory distress syndrome; GC, glucocorticoid; PSM, propensity score matching; SMD, Standardized mean difference
Fig. 2.
Differences in baseline variables before and after PSM. APS, acute physiology score; SAPS, simplified acute physiology; SOFA, sequential organ failure assessment; ARDS, acute respiratory distress syndrome; PSM, propensity score matching; SMD, standardized mean difference
Outcomes
With regard to primary outcomes (Table 2), before PSM, the GC group exhibited significantly higher 28-d mortality than the non-GC group (28% vs 20%, HR = 1.418, p = 0.001, Fig. S1). After PSM, GC therapy remained significantly associated with increased mortality at 28-d (29% vs 23%, HR = 1.373, p = 0.011; 6-m: 47% vs 43%, HR = 1.215, p = 0.037; 12-m: 56% vs 51%, HR = 1.215, p = 0.023; Fig. 3, Fig. S2).
Table 2.
Effect of GC on mortality of patients before and after PSM
| Outcome | Before PSM | After PSM | ||||||
|---|---|---|---|---|---|---|---|---|
| Non-GC (n = 1177) |
GC (n = 593) |
Adjusted HR (95% CI)a |
P value | Non-GC (n = 521) |
GC (n = 521) |
Adjusted HR (95% CI)a |
P value | |
| Primary outcome | ||||||||
| 28-d mortality, % | 238 (20) | 166 (28) | 1.418(1.114, 1.757) | 0.001 | 118 (23) | 150 (29) | 1.373 (1.076, 1.752) | 0.011 |
| Secondary outcome | ||||||||
| 6-m mortality, % | 486 (41) | 275 (46) | 1.222(1.042, 1.432) | 0.013 | 223 (43) | 246 (47) | 1.215 (1.012, 1.458) | 0.037 |
| 12-m mortality, % | 561 (48) | 325 (55) | 1.271(1.098, 1.472) | 0.001 | 264 (51) | 292 (56) | 1.215 (1.027, 1.437) | 0.023 |
| Hospital mortality, % | 516 (44) | 309 (52) | 1.126(0.968, 1.309) | 0.124 | 240 (46) | 276 (53) | 1.095 (0.919,1.306) | 0.309 |
| ICU mortality, % | 285 (24) | 198 (33) | 1.107(0.904, 1.348) | 0.313 | 139 (27) | 174 (33) | 1.032 (0.816, 1.304) | 0.795 |
| Length of hospital, mean(sd), day | 13.63 (12.25) | 15.80 (14.44) | NAb | 0.002 | 14.08 (11.97) | 15.31 (14.21) | NAb | 0.130 |
| Length of ICU, mean(sd), day | 5.39 (6.34) | 6.99 (7.59) | NAb | < 0.001 | 5.74 (6.67) | 6.63 (6.98) | NAb | 0.035 |
| 28-d ventilator-free days | 26.03 (4.23) | 25.26 (5.19) | NAb | 0.002 | 25.65 (4.42) | 25.43 (4.90) | NAb | 0.432 |
a. Multivariate Cox regression model analysis of all cohort data. b. HR value is not applicable, expressed in “NA”. GC, glucocorticoid; PSM, propensity score matching
Fig. 3.
Forest diagram of the effect of GC treatment on 28-d, 6-m and 12-m mortality
Regarding secondary outcomes (Table 2), before PSM, the GC group exhibited significantly higher 6-m and 12-m mortality than the non-GC group (6-m: 46% vs 41%, HR = 1.222, p = 0.013; 12-m: 55% vs 48%, HR = 1.271, p = 0.001, Fig. S1). There were no significant differences in in-hospital and ICU mortality between the two groups (p > 0.05, Fig. S1). However, patients in the GC group had longer hospital length of stay (GC vs non-GC: 15.80 ± 14.44 vs 13.63 ± 12.25 days, p = 0.002). Although statistically significant differences were observed in ICU length of stay and 28-d ventilator-free days, the differences in days were relatively small (ICU stay: 6.99 ± 7.59 vs 5.39 ± 6.34 days, p < 0.001; 28-d ventilator-free days: 26.03 ± 4.23 vs 25.26 ± 5.19 days, p = 0.002). After PSM, GC therapy remained significantly associated with increased mortality at 6-m and 12-m (6-m: 47% vs 43%, HR = 1.215, p = 0.037; 12-m: 56% vs 51%, HR = 1.215, p = 0.023; Fig. 3, Fig. S2). There was no difference in ICU and in-hospital mortality (Fig. S2, Fig. S3), length of hospital stays, and 28-d ventilator-free days. Although there was a statistical difference in length of ICU stay, the difference was modest (6.63±6.98 vs. 5.74 ± 6.67 days, p = 0.035).
In the extended multivariable Cox proportional hazards models (Table 3), six models were constructed. Model 1 included univariate analysis for GC treatment; Model 2 added baseline characteristics; Model 3 further included illness severity scores; Model 4 added complications; Model 5 incorporated pathogen type; and Model 6 included immunosuppression classifications. GC treatment was significantly associated with increased 28-d mortality in all models (HR 1.337–1.374, p < 0.05, Fig. S4). In the extended Cox regression analysis, more than 50% of patients had unknown pathogens, which may have affected outcomes. Excluding patients with unknown pathogens after PSM, GC treatment was not associated with increased 28-d mortality (HR = 0.935, p = 0.637, Table S3).
Table 3.
Effects of GC use, duration and starting time on 28-d mortality after PSM
| Variables | HR | 95% CI | P value |
|---|---|---|---|
| Model 1 | 1.337 | 1.050–1.701 | 0.018 |
| Model 2 | 1.374 | 1.079–1.749 | 0.010 |
| Model 3 | 1.367 | 1.073–1.740 | 0.011 |
| Model 4 | 1.364 | 1.071–1.738 | 0.012 |
| Model 5 | 1.366 | 1.072–1.740 | 0.012 |
| Model 6 | 1.370 | 1.074–1.746 | 0.011 |
| Start time of GC | |||
| ≤48 h (n = 407) | 1.178 | 0.905–1.532 | 0.223 |
| > 48 h (n = 114) | 1.952 | 1.384–2.752 | < 0.001 |
| Duration of GC | |||
| 0–48 h (n = 187) | 1.153 | 0.820–1.621 | 0.412 |
| 48 h-7d (n = 217) | 1.419 | 1.049–1.921 | 0.023 |
| > 7d (n = 117) | 1.469 | 1.023–2.110 | 0.037 |
| Average daily-dose of GC | |||
| ≤0.5 mg/kg (n = 343) | 1.421 | 1.088–1.856 | 0.010 |
| 0.5–1 mg/kg (n = 86) | 1.141 | 0.724–1.798 | 0.570 |
| > 1 mg/kg (n = 92) | 1.246 | 0.830–0.870 | 0.289 |
Model1: use GC; Model2: Model1+gender+age+charlson comorbidity index; Model3: Model2+SOFA+APSIII; Model4: Model3+ARDS+Septic-shock+Invasive ventilation; Model5: Model4+pathogen; Model6: Model5+immunocompromised category GC, glucocorticoid; APS, acute physiology score; SAPS, simplified acute physiology; SOFA, sequential organ failure assessment; ARDS, acute respiratory distress syndrome; PSM, propensity score matching
Initiating GC therapy more than 48 hours after ICU admission was significantly associated with increased 28-d mortality (HR = 1.952, p < 0.001). GC use for more than 7 days was similarly associated with higher 28-d mortality (HR = 1.469, p = 0.037). The average daily dose of GC was calculated based on patient’s body weight. After PSM, body weight was comparable between the two groups (non-GC vs GC: 76 [64, 90] vs 74 [62, 89] kg, p = 0.128). Patients were categorized into low, moderate, and high dose groups based on average daily GC dosage. A daily dose of < 0.5 mg/kg/day (low dose) was associated with increased 28-d mortality (HR = 1.421, p = 0.01).
Subgroup analysis
Subgroup analysis showed that GC use increased 28-d mortality among patients aged < 65 years, male patients, those with SOFA scores ≤6, those with primary immunodeficiency, those not receiving invasive ventilation, and those without ARDS or septic shock (HR = 1.82, 1.56, 1.66, 1.84, 1.47, 1.34, and 1.46, respectively; p < 0.05; Fig. 4A). Among patients with unknown pathogens, GC use was associated with increased 28-d mortality (HR = 1.58, p = 0.003, Fig. 4), while no such association was observed in infections with identified pathogens such as Staphylococcus aureus, P. aeruginosa, viruses, fungi, or aspergillus.
Fig. 4.
Forest diagram of the effect of GC treatment on 28-d mortality in different subgroups after PSM. SOFA, sequential organ failure assessment; ARDS, acute respiratory distress syndrome
Sensitivity analysis
For sensitivity analysis, a multivariable Cox proportional hazards model was constructed using known risk factors for mortality in immunocompromised patients with CAP in the ICU to evaluate the 28-d mortality (Table 4). E-value analysis indicated that the observed associations between GC and 28-d mortality would require HRs greater than 2.082, to be explained by unmeasured confounding factors. These results suggest that the impact of unmeasured confounders on the association between GC therapy and mortality is relatively small compared to known risk factors.
Table 4.
Multivariate cox regression analysis of risk factors affecting 28-d mortality after PSM
| Variables | HR | 95% CI | P value |
|---|---|---|---|
| Age, year | 1.001 | 0.990–1.012 | 0.861 |
| Gender, male | 0.967 | 0.755–1.239 | 0.790 |
| Charlson comorbidity index | 1.175 | 1.119–1.233 | < 0.001 |
| SOFA | 0.992 | 0.941–1.046 | 0.777 |
| SAPS II | 1.002 | 0.987–1.018 | 0.794 |
| APS III | 1.016 | 1.006–1.025 | 0.001 |
| ARDS | 1.408 | 0.743–2.671 | 0.294 |
| Septic shock | 1.358 | 1.017–1.814 | 0.038 |
| Leukocyte | 1.002 | 0.991–1.012 | 0.786 |
| Neutrophils | 0.999 | 0.990–1.007 | 0.738 |
| Lymphocyte | 0.990 | 0.975–1.004 | 0.158 |
| Invasive ventilation | 1.032 | 0.784–1.359 | 0.823 |
| Immunocompromise | 0.981 | 0.877–1.096 | 0.728 |
| Pathogen | 0.939 | 0.864–1.020 | 0.137 |
| glucocorticoid | 1.373 | 1.077–1.750 | 0.011 |
APS, acute physiology score; SAPS, simplified acute physiology; SOFA, sequential organ failure assessment; ARDS, acute respiratory distress syndrome; PSM, propensity score matching
Further analysis
Patients receiving long-term GC therapy may continue treatment after admission. To reduce bias, patients with a history of GC use were excluded from further analysis. After PSM, 118 had long-term GC use prior to admission. Multivariable Cox regression analysis was performed using the same covariates as in Table 2 yielded consistent results, GC therapy was associated with increased 28-d mortality (After PSM: HR = 1.436, p < 0.05, Table S4).
GC were recommended for use in patients with COVID-19 pneumonia. After PSM, a total of 10 patients with COVID-19 pneumonia were included (5 in each group). All patients in the non-GC group survived, with no cases of ARDS or septic shock. In contrast, three patients in the GC group died within one year; all had concurrent ARDS and septic shock, and their Charlson Comorbidity Index were higher than those of the survivors. After excluding the 10 patients with COVID-19 pneumonia, a multivariate Cox regression analysis was performed using the same covariates as in Table 2. The results showed that GC therapy remained a potential risk factor for 28-d mortality (HR = 1.346, p = 0.017, Table S4).
Discussion
This study is the first to retrospectively evaluate the impact of GC therapy on mortality in immunocompromised patients with CAP admitted to the ICU. The results indicated that GC use was associated with increased 28-d mortality of immunocompromised patients with CAP in ICU. Given the proven efficacy of GC therapy in COVID-19 pneumonia patients requiring oxygen support and its strong recommendation in the NIH guidelines [14, 22], further analysis was conducted after excluding COVID-19 pneumonia patients to minimize bias. The results still indicated an increased 28-day mortality. After excluding patients with a prior history of GC treatment, the results were consistent, suggesting the relative robustness of the association between GC use and increased 28-day mortality. The increased mortality may be related to delayed initiation, prolonged duration, or insufficient dosing of GC therapy. Regarding secondary outcomes, GC use was also associated with increased 6- and 12-m mortality, while its effects on ICU and in-hospital mortality, ICU and hospital length of stay, and 28-d ventilator-free days were not statistically significant. Subgroup analyses suggested that the increased mortality may be related to GC use in patients with milder disease severity or unknown pathogens. This is consistent with current recommendations for immunocompetent patients with CAP, in whom GC therapy is only advised for severe cases, particularly those complicated by ARDS or septic shock, and should consider the type of causative pathogen [13]. When the pathogen is unknown, GC administration may impair pathogen clearance, potentially leading to disease progression and clinical deterioration [23].
Previous studies, along with our results, suggest that GC therapy may not be suitable for critically ill immunocompromised patients with severe infections. A study evaluating adverse outcomes of GC use in immunocompromised patients with sepsis reported that high-dose GC therapy nearly doubled both hospital and 28-d mortality [24]. Similarly, an observational study of non-HIV immunocompromised patients with PJP found that those receiving GC had twice the mortality rate of those who did not [25]. A recent prospective RCT on post-transplant PJP also showed no benefit from GC treatment [26]. Furthermore, in cancer patients with COVID-19, GC use was associated with a higher incidence of adverse events [27]. Collectively, these findings indicate that GC therapy may result in worse outcomes in immunocompromised populations.
Regarding the timing, duration, and dosage of GC therapy, current guidelines and previous studies recommend early initiation within 48 hours, at a dose of 1 mg/kg/day, with treatment duration not exceeding one week [13, 28, 29]. In our study, initiation of GC therapy more than 48 hours after ICU admission and treatment duration longer than 7 days were significantly associated with increased 28-d mortality. In addition, a daily dose below 0.5 mg/kg/day was also linked to higher mortality risk. A meta-analysis of four randomized controlled trials found that patients with ARDS who received early methylprednisolone treatment (within 72 hours) demonstrated more rapid clinical improvement compared to those who received delayed GC therapy (after 7 days) [30]. In RCTs involving COVID-19 patients, a prolonged 23-day tapering regimen showed no benefit over a 10-day regimen and was associated with more adverse events [31]. Therefore, a 1-week course of GC therapy may be a safer option. In terms of dosing, patients in the early inflammatory stage may experience adrenal insufficiency, making low-dose GC ineffective in correcting CIRCI. Thus, in immunocompromised patients, early initiation, short duration, and non-low-dose GC regimens may help avoid increased mortality risk. However, whether such regimens confer clinical benefit requires further investigation.
This study has several limitations. First, as a single-center retrospective study with a relatively limited sample size, residual confounding may exist despite adjustments using PSM. Second, due to the retrospective nature of the study, GC dosing and duration were not standardized as RCT, and inconsistencies in GC use among patients were inevitable. To address this, we conducted subgroup analyses to explore the effects of GC initiation time, duration, and dosage on outcomes in immunocompromised patients. Third, data on GC-related adverse events—such as secondary infections, hyperglycemia, gastrointestinal bleeding, and changes in immune markers—were not available. Moreover, a subgroup analysis of COVID-19 patients was conducted after PSM, the results showed that the baseline characteristics between the GC and non-GC groups were still inconsistent among these patients. This indicated that even though PSM and the multivariate model had adjusted for some variables, residual confounding still existed—a major limitation inherent to retrospective studies. Therefore, further prospective research is needed to identify patient subgroups among immunocompromised CAP cases who may benefit from GC therapy.
Conclusion
In immunocompromised patients with CAP admitted to the ICU, GC therapy was associated with increased 28-d, 6-m, and 12-m mortality, potentially due to unidentified pathogens, delayed initiation, prolonged duration, and insufficient dosing of GC. The progression of pneumonia results from the complex interaction between pathogen virulence, antimicrobial therapy, and host immune defenses. Future treatment strategies for severe infections in immunocompromised patients should focus on achieving an appropriate balance between mitigating immune dysregulation and preserving effective pathogen clearance. Despite the inherent limitations of retrospective studies, this study provides valuable insights into GC treatment strategies for immunocompromised ICU patients with CAP. Further research is needed to identify patient subgroups who may benefit from GC therapy, with the ultimate goal of improving outcomes in this vulnerable population.
Electronic supplementary material
Below is the link to the electronic supplementary material.
Acknowledgements
Not applicable.
Abbreviations
- ARDS
Acute respiratory distress syndrome
- APS
Acute physiology score
- ATS
American Thoracic Society
- CAP
Community-acquired pneumonia
- CIRCI
Critical illness-related corticosteroid insufficiency
- GC
Glucocorticoid
- ICU
Intensive care unit
- IDSA
Infectious Diseases Society of America
- NIH
National Institutes of Health
- PJP
Pneumocystis jirovecii pneumonia
- PSM
Propensity score matching
- RCTs
Randomized controlled trials
- SCAP
Severe community-acquired pneumonia
- SAPS
Simplified acute physiology score
- SOFA
Sequential organ failure assessment
- SMD
Standardized mean difference
- SQL
Structured Query Language
- SCCM
Society of Critical Care Medicine
- 28-d
Twenty-eight days
- 6-m
Six months
- 12-m
Twelve months
Author contributions
Xue Tian wrote the main manuscript. Dongpo Wei designed this study. Shiyu Meng collected data from MIMIC-IV database. Changxing Chen and Shanshan Jin analyzed data. Ruilan Wang revised manuscript. All authors reviewed the manuscript.
Funding
This study was supported by the National Science Foundation of China (No.82072210), the Noncommunicable Chronic Diseases-National Science and Technology Major Project of China (2023ZD0506502), the National Key Clinical Specialist Construction Project of China (No.Z155080000004), the National Key Research and Development Program of China (No.2024YFC3044400), Clinical Research Plan of SHDC of China (SHDC2020CR5010-003), the Key Supporting Discipline of Shanghai Healthcare System of China (No.2023ZDFC0102), the Science and Technology of Shanghai Committee of China (21MC1930400, 23Y31900100) and Clinical Research Innovation Plan of Shanghai General Hospital (CCTR-2025C10, CCTR-2025C11).
Data availability
Data is provided within the manuscript or supplementary information files.
Declarations
Ethics approval and consent to participate
This study did not require IRB approval as used open source and ammonized data.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Xue Tian, Dongpo Wei and Shiyu Meng contributed equally to this work.
Contributor Information
Changxing Chen, Email: map_templar@126.com.
Shanshan Jin, Email: jinshanshan123@yeah.net.
Ruilan Wang, Email: wangyusun@hotmail.com.
References
- 1.Aliberti S, Dela Cruz CS, Amati F, Sotgiu G, Restrepo MI. Community-acquired pneumonia. Lancet. 2021 Sep 4;398(10303):906–19. [DOI] [PubMed] [Google Scholar]
- 2.Naghavi M, Ong KL, Aali A, Ababneh HS, Abate YH, Abbafati C, et al. Global burden of 288 causes of death and life expectancy decomposition in 204 countries and territories and 811 subnational locations, 1990-2021. A systematic analysis for the Global burden of disease study 2021. The Lancet. 2024 May;403(10440):2100–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.File TM, Ramirez JA. Community-acquired pneumonia. N Engl J Med. 2023 Aug 17;389(7):632–41. [DOI] [PubMed] [Google Scholar]
- 4.Martin-Loeches I, Torres A, Nagavci B, Aliberti S, Antonelli M, Bassetti M, et al. ERS/ESICM/ESCMID/ALAT guidelines for the management of severe community-acquired pneumonia. Intensive Care Med. 2023 Jun;49(6):615–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Fan G, Zhou Y, Zhou F, Yu Z, Gu X, Zhang X, et al. The mortality and years of life lost for community-acquired pneumonia before and during COVID-19 pandemic in China. The Lancet Reg Health - West Pac. 2024 Jan;42:100968. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Pirracchio R, Venkatesh B, Legrand M. Low-dose Corticosteroids for critically ill Adults with severe pulmonary infections: a review. JAMA. 2024 Jul 23;332(4):318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Annane D, Pastores SM, Arlt W, Balk RA, Beishuizen A, Briegel J, et al. Critical illness-related corticosteroid insufficiency (CIRCI): a narrative review from a Multispecialty Task Force of the Society of Critical Care Medicine (SCCM) and the European Society of Intensive Care Medicine (ESICM). Intensive Care Med. 2017 Dec;43(12):1781–92. [DOI] [PubMed] [Google Scholar]
- 8.Meduri GU, Annane D, Confalonieri M, Chrousos GP, Rochwerg B, Busby A, et al. Pharmacological principles guiding prolonged glucocorticoid treatment in ARDS. Intensive Care Med. 2020 Dec;46(12):2284–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Mandell LA, Wunderink RG, Anzueto A, Bartlett JG, Campbell GD, Dean NC, et al. Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of Community-acquired pneumonia in Adults. Clin Infect Dis. 2007 Mar;1(44(Suppl 2): S27–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Metlay JP, Waterer GW, Long AC, Anzueto A, Brozek J, Crothers K, et al. Diagnosis and treatment of adults with community-acquired pneumonia. An official clinical practice guideline of the American Thoracic Society and Infectious Diseases Society of America. Am J Respir Crit Care Med. 2019 Oct 1;200(7):e45–67. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Jones BE, Ramirez JA, Oren E, Soni NJ, Sullivan LR, Restrepo MI, et al. Diagnosis and management of Community-acquired pneumonia. An official American Thoracic Society clinical Practice Guideline. Am J Respir Crit Care Med. 2025 Jul 18. [DOI] [PubMed]
- 12.Pastores SM, Annane D, Rochwerg B. Corticosteroid Guideline Task Force of SCCM and ESICM. Guidelines for the diagnosis and management of critical illness-related corticosteroid insufficiency (CIRCI) in critically ill patients (part II): society of Critical Care Medicine (SCCM) and European Society of Intensive Care Medicine (ESICM) 2017. Intensive Care Med. 2018 Apr;44(4):474–77. [DOI] [PubMed]
- 13.Chaudhuri D, Nei AM, Rochwerg B, Balk RA, Asehnoune K, Cadena R, et al. Focused update: guidelines on use of corticosteroids in sepsis, acute respiratory distress syndrome, and community-acquired pneumonia. Crit Care Med. 2024. 2024 May;52(5):e219–33. [DOI] [PubMed]
- 14.Gulick RM, Pau AK, Daar E, Evans L, Gandhi RT, Tebas P, et al. National Institutes of Health COVID-19 treatment guidelines panel: perspectives and lessons learned. Ann Intern Med. 2024 Nov;177(11):1547–57. [DOI] [PMC free article] [PubMed]
- 15.Chean D, Windsor C, Lafarge A, Dupont T, Nakaa S, Whiting L, et al. Severe Community-acquired pneumonia in immunocompromised patients. Semin Respir Crit Care Med. 2024 Apr;45(2):255–65. [DOI] [PubMed] [Google Scholar]
- 16.for the Efraim investigators and the Nine-I study group, Azoulay E, Pickkers P, Soares M, Perner A, Rello J, et al. Acute hypoxemic respiratory failure in immunocompromised patients: the Efraim multinational prospective cohort study. Intensive Care Med. 2017 Dec;43(12):1808–19. [DOI] [PubMed] [Google Scholar]
- 17.Johnson AEW, Bulgarelli L, Shen L, Gayles A, Shammout A, Horng S, et al. MIMIC-IV, a freely accessible electronic health record dataset. Sci Data. 2023 Jan 3;10(1):1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ramirez JA, Musher DM, Evans SE, Dela Cruz C, Crothers KA, Hage CA, et al. Treatment of Community-acquired pneumonia in immunocompromised Adults. Chest. 2020 Nov;158(5):1896–911. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Reyes LF, Garcia E, Ibáñez-Prada ED, Serrano-Mayorga CC, Fuentes YV, Rodríguez A, et al. Impact of macrolide treatment on long-term mortality in patients admitted to the ICU due to CAP: a targeted maximum likelihood estimation and survival analysis. Crit Care. 2023 May 31;27(1):212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Haneuse S, VanderWeele TJ, Arterburn D. Using the E-Value to assess the potential effect of unmeasured confounding in observational studies. JAMA. 2019 Feb 12;321(6):602. [DOI] [PubMed] [Google Scholar]
- 21.VanderWeele TJ, Ding P. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med. 2017 Aug 15;167(4):268–74. [DOI] [PubMed] [Google Scholar]
- 22.RECOVERY Collaborative Group, Horby P, Lim WS, Emberson JR, Mafham M, Bell JL, et al. Dexamethasone in hospitalized patients with covid-19. N Engl J Med. 2021 Feb 25;384(8):693–704. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Taves MD, Ashwell JD. Glucocorticoids in T cell development, differentiation and function. Nat Rev Immunol. 2021 Apr;21(4):233–43. [DOI] [PubMed] [Google Scholar]
- 24.Boekhoud L, Schaap HMEA, Huizinga RL, Olgers TJ, Ter Maaten JC, Postma DF, et al. Predictive performance of NEWS and qSOFA in immunocompromised sepsis patients at the emergency department. Infection. 2024 Oct;52(5):1863–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Kamel T, Janssen-Langenstein R, Quelven Q, Chelly J, Valette X, Le MP, et al. Pneumocystis pneumonia in intensive care: clinical spectrum, prophylaxis patterns, antibiotic treatment delay impact, and role of corticosteroids. A French multicentre prospective cohort study. Intensive Care Med. 2024 Aug;50(8):1228–39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Hosseini-Moghaddam SM, Kothari S, Humar A, Albasata H, Yetmar ZA, Razonable RR, et al. Adjunctive glucocorticoid therapy for Pneumocystis jirovecii pneumonia in solid organ transplant recipients: a multicenter cohort, 2015-2020. Am J Transplant. 2024 Apr;24(4):653–68. [DOI] [PubMed] [Google Scholar]
- 27.Zhang L, Zhu F, Xie L, Wang C, Wang J, Chen R, et al. Clinical characteristics of COVID-19-infected cancer patients: a retrospective case study in three hospitals within Wuhan, China. Ann Oncol. 2020 Jul;31(7):894–901. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Confalonieri M, Urbino R, Potena A, Piattella M, Parigi P, Puccio G, et al. Hydrocortisone infusion for severe community-acquired pneumonia: a preliminary randomized study. Am J Respir Crit Care Med. 2005 Feb 1;171(3):242–48. [DOI] [PubMed] [Google Scholar]
- 29.Torres A, Sibila O, Ferrer M, Polverino E, Menendez R, Mensa J, et al. Effect of corticosteroids on treatment failure among hospitalized patients with severe community-acquired pneumonia and high inflammatory response: a randomized clinical trial. JAMA. 2015 Feb 17;313(7):677. [DOI] [PubMed] [Google Scholar]
- 30.Meduri GU, Bridges L, Shih MC, Marik PE, Siemieniuk RAC, Kocak M. Prolonged glucocorticoid treatment is associated with improved ARDS outcomes: analysis of individual patients’ data from four randomized trials and trial-level meta-analysis of the updated literature. Intensive Care Med. 2016 May;42(5):829–40. [DOI] [PubMed] [Google Scholar]
- 31.Salton F, Confalonieri P, Centanni S, Mondoni M, Petrosillo N, Bonfanti P, et al. Prolonged higher dose methylprednisolone versus conventional dexamethasone in COVID-19 pneumonia: a randomised controlled trial (MEDEAS). Eur Respir J. 2023 Apr;61(4):2201514. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data is provided within the manuscript or supplementary information files.