Methylprednisolone for heart surgery in pediatric patients: a meta-analysis of randomized trials

Abstract

Perioperative corticosteroids, particularly methylprednisolone, have been used for decades in pediatric cardiac surgery to attenuate the systemic inflammatory response associated with cardiopulmonary bypass. This systematic review and meta-analysis evaluated the efficacy and safety of perioperative methylprednisolone in pediatric patients undergoing cardiac surgery with cardiopulmonary bypass. This systematic review and meta-analysis followed PRISMA 2020 guidelines (PROSPERO CRD420251231338). We searched PubMed/MEDLINE, Embase, Scopus, Web of Science, and CENTRAL from inception to November 2025 for randomized controlled trials comparing perioperative intravenous methylprednisolone with placebo or standard care in patients aged < 18 years undergoing cardiac surgery with cardiopulmonary bypass. Risk of bias was assessed using the Cochrane Risk of Bias 2 tool. The primary outcome was all-cause in-hospital or 30-day mortality. Secondary outcomes included mechanical ventilation duration, cardiac intensive care unit length of stay, postoperative infections, and hyperglycemia. Random-effects models with Paule-Mandel estimator and Hartung-Knapp confidence intervals were used to pool risk ratios and mean differences. We included eight randomized controlled trials (1,735 pediatric patients; 867 on methylprednisolone, 868 controls). Overall, methylprednisolone was not associated with reduced mortality compared with control (RR 0.66, 95% CI 0.35–1.25; I² = 0%; 6 trials, 1,586 patients). However, methylprednisolone reduced the duration of mechanical ventilation (MD − 0.27 days, 95% CI − 0.46 to − 0.09; I²=0%; 5 trials) but significantly increased the risk of hyperglycemia (RR 2.28, 95% CI 1.38–3.78; I²=45.0%; 6 trials). No differences were observed in infection rates (RR 1.06, 95% CI 0.68–1.66; I²=0%; 5 trials) or ICU length of stay (MD − 0.13 days, 95% CI − 0.51 to 0.25; I²=0%; 5 trials). Perioperative methylprednisolone was not associated with reduced overall mortality in pediatric cardiac surgery. Among secondary outcomes, methylprednisolone was associated with a modest reduction in mechanical ventilation duration but with a significantly increased risk of postoperative hyperglycemia. These findings indicate that any potential benefit in ventilatory outcomes should be weighed against the higher metabolic risk. Further research is required to determine whether any patient subgroups may derive net clinical benefit from its use.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1038/s41598-026-57085-1.

Introduction

Congenital heart disease (CHD) is the most common congenital anomaly worldwide. It remains a leading cause of morbidity and mortality during infancy, despite major advances in surgical techniques and perioperative care¹. A substantial proportion of affected infants require early surgical correction using cardiopulmonary bypass (CPB), a procedure that continues to be associated with significant postoperative complications, prolonged intensive care unit (ICU) stay, and persistent mortality risk².

The systemic inflammatory response induced by cardiopulmonary bypass is a primary driver of postoperative organ dysfunction in pediatric cardiac surgery³. In neonates and young infants, CPB triggers a pronounced inflammatory cascade characterized by complement activation, pro-inflammatory cytokine release, endothelial injury, and ischemia–reperfusion injury, culminating in capillary leak and myocardial dysfunction⁴. These effects are amplified by developmental factors unique to early life, including immune immaturity, limited physiological reserve, and a disproportionately high extracorporeal circuit surface-to-blood volume ratio⁴.

To attenuate CPB-associated inflammation, perioperative corticosteroids, most commonly methylprednisolone, have been administered in pediatric cardiac surgery for several decades. Methylprednisolone exerts potent anti-inflammatory effects by suppressing cytokine signaling and modulating leukocyte activation⁵. Early experimental and clinical studies demonstrated consistent reductions in inflammatory biomarkers following corticosteroid administration, supporting widespread adoption of this practice⁶. However, whether these biological effects translate into clinically meaningful improvements in patient outcomes remains an ongoing debate.

Evidence from early randomized controlled trials was limited by small sample sizes, heterogeneous populations, and insufficient power to detect differences in relevant clinical outcomes⁷, such as mortality⁸, , duration of mechanical ventilation⁹, or ICU length of stay¹⁰. Consequently, corticosteroid use has varied considerably across institutions, with contemporary surveys confirming marked heterogeneity in practice regarding agent selection, dosing, timing, and patient selection, highlighting the absence of consensus or standardized recommendations¹¹.

High-quality contemporary evidence has challenged the routine use of perioperative methylprednisolone. The Steroids to Reduce Systemic Inflammation after Infant Heart Surgery (STRESS) trial—the largest randomized controlled trial to date (n = 1,200 infants)—demonstrated no reduction in mortality or significant postoperative morbidity with methylprednisolone compared with placebo, while identifying a significantly increased incidence of postoperative hyperglycemia requiring insulin therapy¹². These findings have raised concerns regarding potential metabolic harm without demonstrable clinical benefit.

Subsequent systematic reviews and meta-analyses incorporating contemporary randomized trials have not demonstrated a consistent survival benefit of prophylactic corticosteroids in pediatric cardiac surgery, while confirming an increased risk of metabolic adverse events^13–15. Although modest reductions in mechanical ventilation duration and ICU length of stay have been reported in some analyses, these effects were small, heterogeneous, and of uncertain clinical significance.

Despite recent high-quality trials and meta-analyses questioning its efficacy, perioperative methylprednisolone remains widely used in pediatric cardiac surgery, with substantial practice heterogeneity persisting across institutions. Given the publication of large contemporary trials and the ongoing clinical uncertainty regarding patient selection and risk-benefit balance, an updated comprehensive synthesis of randomized evidence is warranted. Although several systematic reviews have addressed perioperative corticosteroids in pediatric cardiac surgery, the existing evidence base has four important gaps: (i) prior syntheses have pooled heterogeneous corticosteroid classes (methylprednisolone, dexamethasone, hydrocortisone), introducing drug-class confounding; (ii) most have not incorporated the most recent post-STRESS evidence, including subpopulation re-analyses published in 2024–2025; (iii) statistical methods used in earlier work (typically DerSimonian–Laird with fixed continuity corrections) are known to under-cover in small-k meta-analyses with sparse events; and (iv) few prior syntheses have provided a complete GRADE Summary of Findings table with anticipated absolute effects. To address these gaps, we conducted a systematic review and meta-analysis restricted to randomized trials of methylprednisolone, applied a conservative inferential framework (inverse-variance with the Paule–Mandel estimator, Hartung–Knapp–Sidik–Jonkman confidence intervals, prediction intervals, and proportional treatment-arm continuity correction), incorporated the most recent evidence up to 30 November 2025, and produced a complete GRADE Summary of Findings. The primary outcome was all-cause mortality; secondary outcomes were duration of mechanical ventilation, cardiac ICU length of stay, postoperative infections, and hyperglycemia.

Methods

Study design and reporting standards

This study was conducted as a systematic review and meta-analysis of randomized controlled trials evaluating the efficacy and safety of methylprednisolone in pediatric patients undergoing cardiac surgery with cardiopulmonary bypass. The methodological approach conformed to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) statement and followed the standards of the Cochrane Handbook for Systematic Reviews of Interventions (version 6.5). The study protocol was registered in PROSPERO before initiating the systematic search (CRD420251231338).

Eligibility criteria

Eligible studies were randomized controlled trials (double-blind or single-blind) involving patients aged < 18 years who underwent corrective or palliative congenital heart surgery with cardiopulmonary bypass. Studies enrolling mixed age groups were included if age-stratified outcomes were reported or could be extracted. The intervention was perioperative intravenous methylprednisolone at any dose, administered at any time during the perioperative period (preoperative, intraoperative, or postoperative). The comparator was placebo or standard care without corticosteroids. Only randomized controlled trials published in peer-reviewed journals were eligible. Observational studies, case series, and conference abstracts were excluded.

Information sources and search strategy

A comprehensive electronic search was performed in PubMed/MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Central Register of Controlled Trials (CENTRAL) from inception to 30 November 2025. The search strategy combined Medical Subject Headings (MeSH) terms and keywords, including methylprednisolone, cardiopulmonary bypass, pediatric cardiac surgery, congenital heart disease, corticosteroids, and randomized controlled trial. The complete search strategy is provided in Supplementary Table S1.

No restrictions on language or publication year were applied. Reference lists of eligible studies, previous systematic reviews, and relevant narrative reviews were manually screened to identify additional trials. Authors of primary studies were contacted when necessary to obtain missing data or clarify reported outcomes.

Outcomes

The primary outcome was all-cause mortality, defined as in-hospital death or death within 30 days of surgery, whichever was reported. When both were available, in-hospital mortality was used as the outcome. Secondary outcomes included mechanical ventilation duration (days), cardiac intensive care unit length of stay (days), postoperative infections (any infection requiring antimicrobial therapy), and hyperglycemia (blood glucose > 150 mg/dL or requiring insulin therapy). When studies used different thresholds for hyperglycemia or infection definitions, we accepted study-specific criteria. Duration of mechanical ventilation and intensive care unit length of stay were analyzed as continuous outcomes. Mortality, infections, and hyperglycemia were analyzed as dichotomous outcomes. An exploratory subgroup analysis of mortality by age stratum (neonate vs. infant vs. older child) was performed. As individual patient-level data were not available — and as the largest contributing trial (STRESS) included approximately 30% neonates within an otherwise infant population — this analysis is reported in the Supplementary Materials only, and is interpreted as hypothesis-generating rather than confirmatory.

Study selection

Two reviewers independently screened titles and abstracts of all retrieved records to identify potentially eligible studies. Full-text articles of potentially eligible studies were obtained and assessed for inclusion. Disagreements were resolved through discussion or, if needed, by consultation with a third reviewer. The selection process was documented using a PRISMA flow diagram (Fig. 1).

Fig. 1.

PRISMA 2020 flow diagram of study selection. Flow diagram illustrating the process of study identification, screening, eligibility assessment, and inclusion in this systematic review and meta-analysis. A total of 693 records were identified through database searches (PubMed, Scopus, Web of Science, and Embase). After removal of 269 duplicate records, 424 unique records were screened by title and abstract. Of these, 398 were excluded for not meeting eligibility criteria. Twenty-six full-text articles were assessed for eligibility, and 18 were excluded for reasons including non-randomized design, non-pediatric population, absence of cardiopulmonary bypass, combined interventions, inappropriate comparator, duplicate publication, protocol-only reports, or conference abstracts lacking sufficient data. Ultimately, eight randomized controlled trials were included in the qualitative and quantitative synthesis.

Data extraction and management

Two reviewers independently extracted data using a standardized form. Extracted data included study characteristics (design, setting, sample size), participant characteristics (age, surgical complexity), intervention details (methylprednisolone dose, timing, comparator), and outcomes as defined above. For dichotomous outcomes, we extracted the number of events and the total number of participants in each group. For continuous outcomes, we extracted means and standard deviations. When continuous outcomes were reported as medians with interquartile ranges, we converted these to means and standard deviations using established methods. When multiple publications reported the same trial, we used data from the most complete or recent report. Missing data were requested from the study authors when necessary. When unavailable, missing standard deviations were imputed from reported confidence intervals, p-values, or standard errors following Cochrane methods.

Assessment and risk of bias

Risk of bias in individual studies was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool. Two reviewers independently evaluated each trial across five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each domain was rated as low risk, some concerns, or high risk of bias. Disagreements were resolved by consensus.

Statistical analysis

Meta-analyses were conducted using R version 4.3.2 (R Foundation for Statistical Computing) with the meta package. Dichotomous outcomes were pooled as risk ratios (RR) with 95% confidence intervals; continuous outcomes were pooled as mean differences (MD). Random-effects models were fitted to all outcomes to accommodate the expected clinical and methodological heterogeneity across trials. Pooled effect estimates were obtained using the generic inverse-variance (IV) method, as implemented in the R package meta. Between-study variance (tau-squared) was estimated with the Paule–Mandel estimator, and confidence intervals around the summary effect were derived using the Hartung–Knapp–Sidik–Jonkman (HKSJ) adjustment, which provides more conservative interval coverage when the number of studies is small. The IV–HKSJ combination was selected a priori because it has demonstrated superior coverage properties in meta-analyses that include few trials and rare events.

For studies with zero events, a treatment-arm continuity correction (TACC) was applied, as implemented in the meta package in R. This approach applies a proportional continuity correction rather than a fixed value and reduces bias in the presence of sparse data. Studies with zero events in both arms were retained in the analysis, allowing them to contribute to variance estimation under the random-effects model.

Statistical heterogeneity was quantified using the I² and τ² statistics. We interpreted I² values > 50% as indicating substantial heterogeneity. Ninety-five-percent prediction intervals (PI) were calculated to estimate the range of true effects expected in future similar settings, following the approach proposed by Higgins, Thompson and Spiegelhalter (2009): PI = mu-hat ± t(k − 2) × √(tau-squared-hat + SE²(mu-hat)), where mu-hat is the pooled effect, tau-squared-hat is the Paule–Mandel estimate of between-study variance, SE(mu-hat) is the standard error of the pooled effect under the Hartung–Knapp–Sidik–Jonkman adjustment, k is the number of studies contributing to the pooled estimate, and t(k − 2) is the corresponding quantile of the Student t distribution. Prediction intervals were computed only for outcomes with at least three contributing studies. Subgroup analyses were performed using random-effects models, and subgroup differences were evaluated using the test for interaction. Publication bias was assessed using funnel plots and Egger’s test when ≥ 10 studies were available for a given outcome. All statistical tests were two-sided with α = 0.05.

Sensitivity analyses were prespecified for any outcome in which a single study contributed > 40% of the pooled weight. For such outcomes, leave-one-out meta-analyses, alternative between-study variance estimators (DerSimonian–Laird and REML), and sample-size weighting were performed to assess the robustness of the pooled estimate.

Studies evaluating methylprednisolone at any dose were eligible. Among the included trials, the standard regimen of 30 mg/kg was consistently used. Two included trials had three arms. In the study by El Derh (2022), which compared low-dose methylprednisolone (10 mg/kg) with standard-dose methylprednisolone (30 mg/kg) and placebo, only the 30 mg/kg arm and the placebo arm were included because the review was restricted to the standard-dose regimen. In the trial by Keski-Nisula (2015), which evaluated two 30 mg/kg methylprednisolone regimens administered at different time points versus placebo, the two intervention arms were combined into a single group prior to analysis to avoid double-counting of participants, in accordance with Cochrane recommendations for multi-arm trials.

A prespecified era-based sensitivity analysis was performed restricted to trials published in 2000 or later, in order to evaluate the robustness of the mortality estimate to the inclusion of historical trials in which surgical and perioperative practice differs substantially from modern care.

Certainty of evidence (GRADE)

The certainty of evidence for each outcome was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach. This framework was used to assess overall confidence in the pooled effect estimates from the meta-analysis.

Because all included studies were randomized controlled trials, the initial certainty of evidence was high. The certainty was subsequently downgraded when concerns were identified across the following domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias, in accordance with the GRADE methodological guidance.

Risk of bias judgments were informed by the Cochrane Risk of Bias 2 (RoB 2) tool applied to each included trial. Downgrading for risk of bias was considered when the pooled estimate was predominantly driven by studies with high risk of bias or important methodological limitations. Inconsistency was evaluated by examining variability in effect estimates across studies and by considering the magnitude of statistical heterogeneity using the I² statistic. Indirectness was assessed by examining differences between the included studies and the review question with respect to population, intervention, comparator, or outcomes.

Imprecision was evaluated by examining the width of the 95% confidence intervals and whether the intervals crossed the line of no effect, indicating uncertainty regarding the magnitude or direction of the treatment effect. When confidence intervals included both clinically meaningful benefits and harms, the certainty of the evidence was downgraded.

Assessment of publication bias was planned using funnel plots and Egger’s test when at least 10 studies were available for a given outcome; however, for most outcomes, the number of included trials was insufficient to permit a reliable statistical assessment.

The overall certainty of evidence for each outcome was categorized as high, moderate, low, or very low, depending on the number and severity of downgrading decisions. The results of this assessment were summarized in a Summary of Findings table, which presents the pooled relative effect estimates, the anticipated absolute effects, the number of participants and studies contributing to each outcome, and the corresponding certainty of evidence ratings.

Results

Selection of studies

The electronic database search identified 693 records across bibliographic databases and trial registers. Of these, 152 records were retrieved from PubMed, 293 from Scopus, 72 from Web of Science, and 176 from Embase. After removal of 269 duplicate records, 424 unique citations remained and were screened by title and abstract.

During the initial screening phase, 398 records were excluded because they were clearly irrelevant to the study objectives, did not involve pediatric cardiac surgery, did not assess methylprednisolone, or did not meet the predefined eligibility criteria. Consequently, 26 reports were sought for full-text retrieval; all were successfully obtained and evaluated for eligibility.

Following full-text evaluation, 18 reports were excluded for the following reasons: four studies were not randomized controlled trials; three did not involve a pediatric population; one study did not include cardiopulmonary bypass; four studies evaluated methylprednisolone as part of a combined intervention, precluding isolation of its independent effect; two studies used an inappropriate comparator; one report represented a duplicate publication; one study was available only as a protocol without outcome data; and two reports were conference abstracts lacking sufficient quantitative information.

Ultimately, eight randomized controlled trials met al.l inclusion criteria and were included in the qualitative and quantitative synthesis^6,12,15–20. The study selection process is summarized in the PRISMA 2020 flow diagram (Fig. 1).

Characteristics of included studies

A total of eight randomized controlled trials published between 1980 and 2022 were included in the final qualitative and quantitative synthesis. These studies enrolled pediatric patients undergoing corrective or palliative congenital heart surgery with cardiopulmonary bypass (CPB). They evaluated the perioperative administration of intravenous methylprednisolone compared with placebo or standard care (Table 1).

Table 1.

Characteristics of included studies.

Author (Year)	Country	N (total; exp/ctrl)	Age (mean ± SD)	Dose / Timing	CPB Time	Comparator	Follow-up	Outcome (Mortality)
Toledo-Pereyra et al. (1980)	USA	Total: 95 patients. Experimental (exp): 47 patients. Control (ctrl): 48 patients	0–19 years	Dose: 30 mg/kg IV methylprednisolone (Solu-Medrol®). Timing: 4 doses administered: 1 h preoperatively, 5 min before cardiopulmonary bypass (CPB), immediately after CPB, and then every 6 h for 24 h.	Treated: 64.73 min (SD: 32.70). Control: 79.40 min (SD: 60.00). (P > 0.10)	Placebo	Mean duration until discharge was: Treated: 15.52 days. Control: 16.68 days.	Improvement in survival. Significant difference P < 0.05 in mortality: 2 deaths in Methylprednisolone group vs. 8 deaths in Control group.
Keski-Nisula et al. (2013)	Finland	Total: 40 neonates. Experimental (MP Group): 20 neonates. Control (Placebo Group): 20 neonates	9.9 ± 7.0 days (MP Group). 11.0 ± 7.2 days (Placebo Group). (P = 0.643)	Dose: 30 mg/kg IV methylprednisolone (MP). Timing: Single bolus dose after anesthesia induction.	168 ± 43 (MP Group). 198 ± 85 min (Placebo Group) (P = 0.234).	Saline solution (Placebo).	30 postoperative days	Mortality (30 days): 0 deaths (MP) vs. 3 deaths (Placebo). Not significant (P = 0.231)
Keski-Nisula et al. (2015)	Finland	Total: 45 children. Groups: 15 in MP Induction (exp 1), 15 in MP Prime (exp 2), 15 in Placebo (ctrl).	5.9 ± 3.8 months (Induction), 5.1 ± 2.3 months (Prime), 4.9 ± 3.3 months (Placebo). Age range: 1 to 18 months	Dose: 30 mg/kg methylprednisolone (MP). Intervention Groups: (1) MP Induction: IV after anesthesia induction. (2) MP Prime: In theCardiopulmonary Bypass(CPB) circuit solution.	Mean CPB time: 67 ± 31 min (Induction), 62 ± 23 min (Prime), 79 ± 27 min (Placebo). (P = 0.260).	Placebo (Saline solution)	30-day mortality was evaluated	Mortality (30 days): 0 deaths in all three groups.
Dalili et al. (2015)	Iran	Total: 100 children. Group A (Corticosteroid): 50 patients. Group B (Non-Users): 50 patients	39.8 ± 24.7 months (Group A). 38.2 ± 19.8 months (Group B). (P = 0.917)	Single dose of 30 mg/kg methylprednisolone IV. Timing: Injected at the moment of entry to the ICU post-surgery.	106.2 ± 77.8 min (Group A). 111.12 ± 31.8 min (Group B). (P = 0.505)	No drug administration	Until hospital discharge	Not reported
Graham et al. (2019)	USA	176 neonates included in the analysis. Experimental (Methylprednisolone): 81 neonates. Control (Placebo): 95 neonates.	Mean age at surgery: 8.6 days	Single dose of 30 mg/kg methylprednisolone IV. Timing: Administered after anaesthesia induction	185 min (Methylprednisolone) vs. 176 min (Placebo).	Placebo	Primary outcome until hospital discharge. Secondary outcomes up to 30 days post-surgery or 90 days after surgery	Mortality (Individual): 5 (6%) vs. 6 (6%). Not significant (P = 0.970).
Keski-Nisula et al. (2020)	Finland	29 infants. Experimental: 15 infants. Control: 14 infants.	6.0 months (IQR: 4.8–8.3) in MP vs. 6.7 months (IQR: 5.6–8.1) in Control. Range: 2 to 12 months	Single dose of 30 mg/kg methylprednisolone IV. Timing: Administered after anaesthesia induction	55 min (IQR: 46–103) in MP vs. 66 min (IQR: 59–81) in Control. Not significant (P = 0.20).	Saline solution	Measured at 4 time points up to 6 h postoperatively. Clinical outcome follow-up was until First Postoperative Day and ICU stay.	Mortality in PICU: 1 death (MP) vs. 0 deaths (Control). Not significant (P = 0.330).
Hill et al. (2022)	USA	1200 infants. Experimental (Methylprednisolone): 599 infants. Control (Placebo): 601 infants.	126.0 days (IQR: 14.0-191.0) in MP vs. 124.0 days (IQR: 14.0-182.0) in Placebo. Age range: Less than 1 year.	Dose: 30 mg/kg methylprednisolone IV. Timing: Administered in the pump-priming fluid of CPB.	122.0 min (IQR: 88.0-161.0) in MP vs. 121.0 min (IQR: 90.0-160.0) in Placebo.	Placebo	Until hospital discharge or 30 days after surgery	2.8% (MP) vs. 2.0% (Placebo). (not significant)
El Derh et al. (2022)	Egypt	75 patients. Group A (Low Dose): 25 patients. Group B (High Dose): 25 patients. Group C (Control): 25 patients.	5.2 ± 1.8 years (High Dose), 4.9 ± 2.1 years (Low Dose), 5.6 ± 2.0 years (Control). (actual population: 1–9 years).	Low Dose: 10 mg/kg MP. High Dose: 30 mg/kg MP. Timing: Both doses administered after anaesthesia induction.	48.6 ± 12.1 min (High Dose) vs. 48.7 ± 12.1 min (Low Dose) vs. 45.9 ± 10.9 min (Control). Not significant (P = 0.636).	Low Dose (10 mg/kg MP) and Placebo (Saline).	Measurement of laboratory outcomes up to 24 h postoperatively. ICU stay.	Not reported

The trials were conducted across diverse geographical settings, including North America, Europe, the Middle East, and Africa, with the United States and Finland being the most frequently represented countries. Sample sizes varied considerably, ranging from small single-center trials with fewer than 30 participants to the large, multicenter STRESS trial, which included 1,200 infants. Overall, the pooled population comprised neonates, infants, and children up to 19 years of age, with most contemporary studies focusing predominantly on neonates and infants younger than one year (Table 1).

Across studies, methylprednisolone was administered intravenously, predominantly at a dose of 30 mg/kg. Most trials employed a single-dose regimen, whereas earlier studies and selected contemporary trials used multiple-dose strategies or evaluated different administration timings. Timing of administration varied substantially and included delivery after anesthesia induction, incorporation into the CPB priming solution, administration immediately after CPB, or repeated dosing during the early postoperative period. Comparators consisted primarily of a saline placebo. One multi-arm study compared low-dose (10 mg/kg) and standard-dose (30 mg/kg) methylprednisolone versus placebo; however, only the 30 mg/kg arm was included in the quantitative synthesis (Table 1).

Cardiopulmonary bypass durations were broadly comparable between intervention and control groups within individual studies, with mean or median CPB times generally ranging from approximately 45 to 200 min, and no statistically significant intergroup differences were reported in most trials. Follow-up periods extended from immediate postoperative assessments and intensive care unit (ICU) stay to 30-day mortality or hospital discharge, depending on the outcome of interest and study design (Table 1).

Regarding clinical outcomes, mortality was variably reported across trials. Early studies suggested a potential survival benefit with methylprednisolone, whereas later, larger trials, including the STRESS trial, showed no significant difference in mortality between treatment and control groups. Low cardiac output syndrome and the need for postoperative inotropic support were evaluated in several studies, with mixed findings and no consistent reduction associated with methylprednisolone administration. Measures of postoperative recovery, such as duration of mechanical ventilation, ICU length of stay, and total hospital stay, were reported heterogeneously and did not show consistent between-group differences across trials (Table 2).

Table 2.

Characteristics of outcomes assessed.

Study	Mechanical ventilation	ICU stay	Hospital stay	IL-6	IL-8	Inotropes	Adverse events
Toledo-Pereyra et al. (1980)	NR	NR	Mean hospital stay was 16.68 days (Control) vs. 15.52 days (Treated). (P > 0.10).	NR	NR	NR	Yes (infections, stress ulcer), no differences (p > 0.10)
Keski-Nisula et al. (2013)	5.6 ± 4.2 days (MP) vs. 5.7 ± 4.6 days (Placebo). (P = 0.998)	9.3 ± 5.2 days (MP) vs. 8.2 ± 4.9 days (Placebo). (P = 0.515)	NR	IL-6 concentrations were significantly lower in the MP group	IL-8 concentrations were significantly lower in the MP group	Inotropic Score: 25.6 ± 13.7 vs. 27.4 ± 12.5; (P = 0.681)	MP group had significantly higher glucose levels and required significantly more insulin (12/20 vs. 2/17; P = 0.003)
Keski-Nisula et al. (2015)	1 ± 0.2 days (Induction), 1.3 ± 1.0 days (Prime), 1.3 ± 0.7 days (Placebo). (P = 0.220) no significant difference	2.4 ± 1.8 days (Induction), 2.7 ± 1.7 days (Prime), 2.5 ± 1.6 days (Placebo). (P = 0.70) no significant difference	NR	No significant differences in IL-6 concentrations (p = 0.62 to 0.29)	IL-8 concentrations were significantly lower in the MP Induction group compared to Placebo and Prime	No significant differences (p = 0.96) at arrival to ICU. (p = 0.17) ta 12:00 h	Significantly higher glucose levels in MP groups (Induction and Prime) compared to Placebo 6 h after CPB (P = 0.003). More insulin therapy was required in MP groups
Dalili et al. (2015)	18.2 ± 11.04 h (Group A) vs. 25.5 ± 31.3 h (Group B). Not significant (P = 0.688).	3.02 ± 1.9 days (Group A) vs. 3.22 ± 2.1 days (Group B). (P = 0.565)	NR	NR	NR	Low Cardiac Output (LCOS) requiring inotropes. 1 (2%) in Group A vs. 3 (6%) in Group B. (P = 0.307)	Hyperglycemia (Glucose > 150 mg/dL): Significantly more frequent in Group A (Corticosteroid) (62%) than Group B (42%) (P = 0.045). Bacteraemia: Significantly less frequent in Group A (0%) than Group B (12%) (P = 0.012).
Graham et al. (2019)	4 days (IQR: 3–7) vs. 5 days (IQR: 3–8). (P = 0.44) No significant difference	11 days (IQR: 7–18) vs. 11 days (IQR: 7–20). (P = 0.870)	20 days (IQR: 13–35) vs. 22 days (IQR: 12–35). (P = 0.980)	NR	NR	Maximum Vasoactive Inotropic Score (VIS): 18.2 ± 6.9 vs. 20.4 ± 6.9. Significantly lower in Methylprednisolone group (P = 0.030).	Infection: 4 (5%) vs. 8 (8%). Not significant (P = 0.360). Hyperglycaemia and insulin use: Highest glucose and insulin requirement were similar in both groups (unlike previous studies).
Keski-Nisula et al. (2020)	Median 2 days (IQR: 2–4) in MP vs. 2 days (IQR: 2–5) in Control. Not significant (P = 0.960).	Median 2 days (IQR: 2–4) in MP vs. 2 days (IQR: 2–5) in Control. Not significant (P = 0.960)	NR	Significantly lower in MP group at all post-CPB time points (P < 0.001).	Significantly lower in MP group at all post-CPB time points (P < 0.001).	Similar median in both groups at ICU arrival (P = 0.79) and on First Postoperative Day (P = 0.910).	Hyperglycaemia: Significantly higher in MP group at all post-CPB time points (P = 0.006). Insulin was required in 3 MP patients vs. 0 in Control (P = 0.080).
Hill et al. (2022)	Prolonged Ventilation (> 7 days): 6.8% (MP) vs. 8.5% (Placebo). Not significant	NR	Median postop LOS: MP 10 (6–20) days vs. Placebo 11 (6–23) days; adjusted discharge rate ratio 1.11 (0.99–1.25)	NR	NR	Low Cardiac Output Syndrome (LCOS): 5.2% (MP) vs. 6.2% (Placebo). Not significant	Hyperglycaemia with insulin use: Significantly higher in MP group (19.0%) vs. Placebo (6.7%) (P < 0.001). Infection: Not significant (5.2% vs. 4.0%).
El Derh et al. (2022)	NR	ICU stay: 10.3 ± 0.9 h (HD); 10.6 ± 0.8 h (LD); 10.4 ± 0.9 h (Control); p = 0.468 (no significant difference)	NR	Significant decrease in High Dose compared to Low Dose and Control (P < 0.001). Low Dose was also significantly lower than Control (P < 0.001).	NR	Significant decrease in High Dose and Low Dose compared to Control. High Dose was the lowest (VIS: 4.7 ± 1.5 vs. 9.2 ± 1.5 in Control) (P < 0.001).	Hyperglycaemia (blood glucose): ↑ Significant and hierarchical: High Dose > Low Dose > Control (P < 0.001). Heart Block: Less frequent in High Dose (0%) vs. Low Dose (8.0%) vs. Control (16.0%) (P = 0.155). Wound Infection: Not significant (20.0% in High Dose vs. 8.0% in Control).

Several studies assessed inflammatory biomarkers, particularly interleukin-6 (IL-6) and interleukin-8 (IL-8), and consistently demonstrated lower postoperative cytokine concentrations in patients receiving methylprednisolone. However, these biochemical effects did not uniformly translate into improved clinical outcomes. Adverse events were systematically reported in most contemporary trials, with postoperative hyperglycemia and increased insulin requirements occurring more frequently in methylprednisolone-treated patients. In contrast, rates of postoperative infection were generally similar between groups, and no consistent increase in infectious complications was observed. Other adverse outcomes, including acute kidney injury, thrombosis, and conduction abnormalities, were infrequently reported and showed no definitive association with corticosteroid exposure (Table 2).

Risk of bias assessment

The risk of bias of the included randomized controlled trials was assessed using the Cochrane Risk of Bias 2 (RoB 2) tool for the outcome of mortality. Overall, the methodological quality of the included studies was considered acceptable, although variability across domains and between individual trials was observed.

Regarding the randomization process (Domain 1), all included trials were judged to be at low risk of bias. Random sequence generation and allocation concealment were adequately described or inferred from the study design, particularly in the more recent trials, which employed centralized or computer-generated randomization procedures. No baseline imbalances suggestive of compromised randomization were identified (Fig. 2).

Fig. 2.

Risk of bias assessment of included randomized controlled trials using the Cochrane Risk of Bias 2 (RoB 2) tool. Summary of the risk-of-bias assessment across five methodological domains for each included randomized controlled trial evaluating perioperative methylprednisolone in pediatric cardiac surgery. The domains include: randomization process (D1), deviations from intended interventions (D2), missing outcome data (D3), measurement of the outcome (D4), and selection of the reported result (D5). Most studies were judged to be at low risk of bias across domains, while some concerns and one high-risk assessment were identified primarily in the domain related to deviations from intended interventions.

For deviations from intended interventions (Domain 2), most studies were assessed as having low risk of bias, reflecting appropriate blinding of participants, clinical staff, and outcome assessors, as well as adherence to assigned interventions. However, one large contemporary trial exhibited a high risk of bias in this domain due to deviations in post-randomization management and treatment adjustments, which may have influenced the mortality effect estimate.

Assessment of missing outcome data (Domain 3) indicated low risk of bias across all included studies. Mortality data were complete or nearly complete, with minimal loss to follow-up and no evidence that missing data were related to either the intervention or the outcome.

Measurement of the outcome (Domain 4) was consistently judged to be at low risk of bias. Mortality is an objective outcome, and its ascertainment was unlikely to be influenced by knowledge of the intervention assignment. Outcome definitions were clear and consistent within individual trials.

Regarding the selection of the reported result (Domain 5), most studies were rated low risk of bias because outcomes were reported according to prespecified protocols or clearly stated methods. Nevertheless, one study raised concerns about limited information on outcome pre-specification and analytical choices, which reduced certainty about selective reporting.

When considering the overall risk of bias, three studies were classified as having low risk of bias, two as having some concerns, and one as having high risk of bias, primarily due to deviations from the intended interventions. Despite these limitations, the overall body of evidence on mortality was deemed moderate to high in methodological quality.

Effects of methylprednisolone on primary outcomes

Figure 3 presents the pooled analysis of all-cause mortality comparing perioperative methylprednisolone with control or placebo in pediatric patients undergoing cardiac surgery with cardiopulmonary bypass. Six randomized controlled trials contributed data on this outcome, totalling 1,586 participants. The random-effects meta-analysis demonstrated no statistically significant difference in mortality associated with methylprednisolone administration (RR 0.66, 95% CI 0.35–1.25). No heterogeneity was detected (I² = 0.0%). The prediction interval was wide and crossed the no-effect line (0.26–1.23). The pooled estimate was driven mainly by the largest and most contemporary trial, which contributed the most significant statistical weight. Smaller studies showed imprecise estimates with wide confidence intervals, reflecting limited event numbers. Overall, these findings indicate that perioperative methylprednisolone is not associated with a statistically significant reduction in mortality; the 95% confidence interval is compatible both with no effect and with a clinically meaningful reduction, and the wide prediction interval (0.32–1.35) underscores the residual uncertainty.

Fig. 3.

Forest plot of the effect of methylprednisolone on all-cause mortality. Forest plot comparing perioperative methylprednisolone with placebo or standard care for all-cause mortality in pediatric patients undergoing cardiac surgery with cardiopulmonary bypass. Effect estimates are expressed as risk ratios (RR) with 95% confidence intervals using a random-effects model with the Paule–Mandel estimator and Hartung–Knapp adjustment. The pooled analysis shows no statistically significant reduction in mortality (RR 0.66, 95% CI 0.35–1.25), with no observed statistical heterogeneity (I² = 0%).

Effects of methylprednisolone on secondary outcomes

Figure 4 illustrates the effect of methylprednisolone on postoperative infectious complications. Five trials with a combined sample size of 1,621 patients were included in this analysis. The pooled effect estimate showed no statistically significant difference in postoperative infection rates between the methylprednisolone and control groups (RR 1.06, 95% CI 0.68–1.66). No heterogeneity was detected (I² = 0.0%). The prediction interval (0.64–1.77) also encompassed the null value. Across individual studies, point estimates were close to unity, with none showing a statistically significant increase or reduction in infection risk. These results suggest that perioperative methylprednisolone does not materially affect the incidence of postoperative infections in pediatric cardiac surgery.

Fig. 4.

Forest plot of postoperative infection risk. Forest plot illustrating the pooled effect of methylprednisolone versus control on postoperative infections in pediatric cardiac surgery. Risk ratios (RR) and 95% confidence intervals were calculated using a random-effects model. The meta-analysis demonstrates no significant difference between groups (RR 1.06, 95% CI 0.68–1.66), with no evidence of statistical heterogeneity (I² = 0%).

Figure 5 summarizes the results of the meta-analysis of postoperative hyperglycemia. Five randomized trials contributed data, including both small single-center studies and a large multicenter trial. The pooled analysis demonstrated a statistically significant increase in the risk of hyperglycemia among patients receiving methylprednisolone compared with controls (RR 2.28, 95% CI 1.38–3.78). Moderate heterogeneity was observed (I² = 45.0%). Despite heterogeneity, all studies showed point estimates favoring increased hyperglycemia in the methylprednisolone group. The prediction interval was wide (0.80–6.33). Collectively, these findings provide consistent evidence that perioperative methylprednisolone substantially increases the risk of postoperative hyperglycemia.

Fig. 5.

Forest plot of postoperative hyperglycemia. Forest plot showing the effect of perioperative methylprednisolone on the incidence of postoperative hyperglycemia. Effect estimates are presented as risk ratios (RR) with 95% confidence intervals. The pooled analysis indicates a significantly increased risk of hyperglycemia among patients receiving methylprednisolone compared with controls (RR 2.28, 95% CI 1.38–3.78), with moderate heterogeneity observed across studies (I² = 45%).

Figure 6 depicts the pooled analysis of mechanical ventilation duration as a continuous outcome. Five trials reporting the mean ventilation duration were included. The random-effects meta-analysis showed a statistically significant reduction in ventilation duration associated with methylprednisolone administration (MD − 0.27, 95% CI − 0.46 to − 0.09). No heterogeneity was detected (I² = 0.0%). The prediction interval narrowly crossed the null (− 0.68 to 0.03). Although the absolute reduction was modest, this estimate was strongly influenced by Dalili 2015, which contributed 49.8% of the pooled weight owing to unusually narrow reported standard deviations. In a prespecified leave-one-out analysis, exclusion of Dalili 2015 shifted the pooled mean difference to − 0.55 days (95% CI − 1.31 to 0.21), with the 95% CI crossing the null; comparable null-crossing intervals were obtained when DerSimonian–Laird, REML, and sample-size weighting were applied (Supplementary Table S2). The apparent reduction in mechanical-ventilation duration should therefore be regarded as exploratory and clinically uncertain.

Fig. 6.

Forest plot of mechanical ventilation duration. Forest plot depicting the pooled mean difference in the duration of postoperative mechanical ventilation between methylprednisolone and control groups. Results are presented as mean differences (MD) with 95% confidence intervals using a random-effects model. The analysis shows a statistically significant reduction in ventilation duration associated with methylprednisolone (MD − 0.27 days, 95% CI − 0.46 to − 0.09), with no heterogeneity detected (I² = 0%).

Figure 7 presents the meta-analysis of the length of stay in the cardiac intensive care unit (ICU). Five trials contributed data to this outcome. The pooled effect estimate showed no statistically significant difference between methylprednisolone and control groups (MD − 0.13 days, 95% CI − 0.51 to 0.25). Heterogeneity was absent (I² = 0.0%), and the prediction interval (− 0.99 to 0.63) encompassed both modest benefit and harm. Individual study estimates were centered around zero, with overlapping confidence intervals. These results indicate that perioperative methylprednisolone does not meaningfully reduce ICU length of stay despite its association with a slight reduction in mechanical ventilation duration.

Fig. 7.

Forest plot of cardiac intensive care unit (ICU) length of stay. Forest plot summarizing the effect of methylprednisolone on the length of stay in the cardiac intensive care unit. Mean differences (MD) with 95% confidence intervals were pooled using a random-effects model. The meta-analysis indicates no statistically significant difference between treatment and control groups (MD − 0.13 days, 95% CI − 0.51 to 0.25), with no statistical heterogeneity (I² = 0%).

In a prespecified sensitivity analysis excluding the historical trial of Toledo-Pereyra 1980 — whose effect estimate (RR 0.26) is implausibly large in the context of contemporary surgical and intensive-care practice — the pooled mortality estimate moved closer to the null (RR 0.79, 95% CI 0.43–1.45; I² = 0%; Supplementary Table S2). The primary pooled estimate should therefore be interpreted as an upper bound of any potential mortality benefit, not as evidence of effectiveness.

Certainty of evidence assessment (GRADE)

The certainty of evidence for the main outcomes was assessed using the GRADE framework, and the results are summarized in the Summary of Findings table (Table 3).

Table 3.

Summary of findings, GRADE certainty of evidence.

Outcomes	No. of participants (Studies)	Certainty of the evidence (GRADE)	Relative effect (95% CI)	Anticipated absolute effects
				Risk with placebo	Risk difference with Methylprednisolone
Methylprednisolone for Heart Surgery in pediatric patients
All-cause mortality	1,586 (6 RCTs)	⊕⊕◯◯ Lowa	RR 0.66 (0.35–1.25)	43 per 1,000	15 fewer per 1,000 (from 28 fewer to 11 more)
Postoperative infections	1,621 (5 RCTs)	⊕⊕◯◯ Lowa	RR 1.06 (0.68–1.66)	68 per 1,000	4 more per 1,000 (from 22 fewer to 45 more)
Hyperglycemia	1,588 (6 RCTs)	⊕⊕◯◯ Lowb	RR 2.28 (1.38–3.78)	84 per 1,000	108 more per 1,000 (from 32 more to 234 more)
Duration of mechanical ventilation (days)	390 (5 RCTs)	⊕⊕◯◯ Low c	MD -0.27 (-0.56 to -0.09)	—	0.27 days shorter (from 0.56 to 0.09 shorter)
Length of stay in cardiac ICU (days)	390 (5 RCTs)	⊕⊕⊕◯ Moderated	MD -0.13 (-0.51 to 0.25)	—	0.13 days shorter (from 0.51 shorter to 0.25 longer)

CI: confidence interval; RCT: Randomized Controlled Trial; RR: risk ratio; MD: mean difference; ^aDowngraded two levels due to high risk of bias in the dominant study (Hill 2022) and imprecision (95% CI crosses the null).^bDowngraded two levels due to predominance of high-risk-of-bias studies and moderate heterogeneity (I² = 45.0%).^cDowngraded one level due to concerns in the “Selection of the reported result” domain. ^dDowngraded one level due to imprecision (95% CI crosses the null). Assessment of publication bias was limited by the small number of studies per outcome; therefore, no formal downgrading was applied.

For the primary outcome of all-cause mortality, the certainty of evidence was rated as low. This downgrading was mainly due to concerns about risk of bias in the dominant trial and imprecision in the pooled estimate, as the 95% confidence interval crossed the line of no effect.

Similarly, the certainty of evidence for postoperative infections was judged to be low, primarily because of methodological limitations in the contributing studies and imprecision in the pooled effect estimate.

For the safety outcome of hyperglycemia, the certainty of evidence was also rated as low. This assessment reflected concerns about the predominance of studies with methodological limitations and moderate statistical heterogeneity across trials.

The certainty of evidence on the duration of mechanical ventilation was rated as low. Downgrading was applied for serious imprecision (small total sample size [n = 390], absence of the largest contemporary trial [STRESS], and a 95% prediction interval that crossed the null), and for serious indirectness (the pooled estimate was largely driven by a single small trial with implausibly narrow standard deviations, raising doubts that the available evidence is representative of modern pediatric cardiac-surgery populations).

Discussion

This systematic review and meta-analysis of 8 randomized controlled trials involving 1,735 pediatric patients undergoing cardiac surgery with cardiopulmonary bypass provides a comprehensive contemporary evaluation of the efficacy and safety of perioperative methylprednisolone. Our principal findings demonstrate that methylprednisolone does not significantly reduce overall mortality but shows potential benefit in contemporary trials, modestly reduces mechanical ventilation duration, and substantially increases the risk of hyperglycemia. These findings have important implications for clinical practice and challenge the routine use of prophylactic corticosteroids in this population.

The present synthesis advances the existing evidence base in five specific ways. First, it is restricted to methylprednisolone — the most widely administered agent in pediatric cardiac surgery — eliminating the drug-class heterogeneity that has limited the interpretability of broader corticosteroid syntheses. Second, it is updated to 30 November 2025 and incorporates post-STRESS re-analyses that have not been included in prior trial-level meta-analyses. Third, it applies a conservative inferential framework (inverse-variance with the Paule–Mandel estimator, Hartung–Knapp–Sidik–Jonkman confidence intervals, and explicit prediction intervals) that improves coverage when the number of trials is small. Fourth, it uses proportional treatment-arm continuity correction for sparse-event outcomes, avoiding the upward bias associated with fixed 0.5 corrections. Fifth, it provides a complete GRADE Summary of Findings table with anticipated absolute effects per 1,000 patients, supporting direct use by clinical-practice-guideline panels. The net result is a more conservative, drug-specific, and clinically actionable estimate of the efficacy and safety of methylprednisolone in this population.

The absence of an overall mortality benefit observed in our analysis aligns closely with findings from the STRESS trial¹², the most extensive and methodologically rigorous study to date. This multicenter trial of 1,200 infants found no significant improvement in a composite outcome of death, heart transplantation, or major complications with methylprednisolone compared with placebo (adjusted OR 0.86, 95% CI 0.71–1.05)¹². Similarly, a recent meta-analysis by Pesonen et al.²¹ including 1,263 patients reported no mortality reduction with corticosteroids (RR 0.83, 95% CI 0.59–1.17). Our findings extend this evidence by incorporating additional contemporary trials and conducting prespecified subgroup analyses to explore potential sources of heterogeneity. Although no definitive mortality reduction was demonstrated, the biological rationale for perioperative corticosteroid therapy remains plausible, given the inflammatory response triggered by cardiopulmonary bypass²².

A reduction in mechanical ventilation duration was consistently observed across studies and represents one of the most clinically meaningful findings of this meta-analysis. This modest yet statistically significant reduction likely reflects methylprednisolone’s anti-inflammatory properties, which attenuate pulmonary injury from cardiopulmonary bypass^15,16. Keski-Nisula et al.^15,16, consistently demonstrated that high-dose methylprednisolone (30 mg/kg) suppresses pro-inflammatory cytokine release, particularly interleukin-6 and interleukin-8, and reduces neutrophil activation and endothelial glycocalyx degradation—mechanisms directly relevant to postoperative respiratory function. However, the clinical relevance of this reduction is uncertain. The pooled estimate was largely driven by a single small trial (Dalili 2015) with implausibly narrow standard deviations, and prespecified sensitivity analyses excluding this study or using alternative variance estimators consistently produced 95% confidence intervals that crossed the null (Supplementary Table S2). Accordingly, the suggestion that perioperative methylprednisolone shortens mechanical-ventilation duration must be considered hypothesis-generating rather than definitive. This effect did not translate into meaningful reductions in intensive care unit length of stay (MD − 0.13 days, 95% CI − 0.51 to 0.25), suggesting that factors beyond pulmonary recovery determine ICU discharge readiness in this population. The disconnect between surrogate biochemical markers and hard clinical outcomes has been a persistent theme in corticosteroid research, underscoring the importance of patient-centered outcomes in clinical trials²³.

The significantly increased risk of hyperglycemia with methylprednisolone (RR 2.28, 95% CI 1.38–3.78) represents a critical safety concern that must be weighed against potential benefits. This finding was consistent across six trials, with all studies showing point estimates favouring increased risk of hyperglycemia, despite moderate statistical heterogeneity (I² = 45.0%).The STRESS trial¹² reported that 19% of methylprednisolone-treated infants required insulin therapy compared with only 7% of placebo-treated patients (p < 0.001). Postoperative hyperglycemia in pediatric cardiac surgery has been associated with increased infection risk, prolonged mechanical ventilation, and longer intensive care unit stay in observational studies^24,25. Although we found no significant increase in postoperative infection rates with methylprednisolone in our meta-analysis (RR 1.06, 95% CI 0.68–1.66), the immunosuppressive effects of corticosteroids combined with hyperglycemia may create a vulnerable metabolic state. Furthermore, the long-term neurodevelopmental implications of perioperative hyperglycemia in neonates and infants remain inadequately characterized²⁶.

Three further analyses inform the interpretation of our findings. Wang and colleagues (BMC Anesthesiol 2024) restricted their synthesis to infant trials and reported no mortality benefit and an increased risk of hyperglycemia, both directly concordant with our pooled estimates. The 2023 systematic review by Elbadawi and colleagues, which pooled all corticosteroid classes (including dexamethasone-only trials), found a non-significant trend toward reduced mortality and a more confident reduction in mechanical-ventilation duration; the more conservative estimates obtained in the present, methylprednisolone-restricted synthesis are consistent with the hypothesis that the apparent corticosteroid effect in broader analyses is partially driven by drug-class heterogeneity. Finally, the recent subpopulation re-analysis of the STRESS trial (Crit Care Med 2025) demonstrated heterogeneous effects across neonatal subgroups — including potential reductions in postoperative hydrocortisone use and thrombosis among term neonates, balanced against increases in acute kidney injury and insulin exposure — emphasizing that meaningful refinement of the risk–benefit profile will require individual patient-level data rather than further trial-level pooling²⁷.

Our findings contrast somewhat with previous meta-analyses that suggested more favorable benefit–risk profiles for perioperative corticosteroids. Pasquali et al.¹³ in 2012 pooled data from 13 trials (n = 1,087) and reported a trend toward reduced mortality (OR 0.54, P = 0.07) and significantly shorter intensive care unit stay (MD − 0.36 days, P = 0.003) with corticosteroids. However, that analysis included heterogeneous corticosteroid regimens (dexamethasone, methylprednisolone, hydrocortisone) with varied dosing and timing, predating the STRESS trial and other extensive contemporary studies. More recent meta-analyses by Gibbison et al.²² and Pesonen et al.²¹, which incorporated newer evidence, similarly found no mortality benefit but confirmed modest reductions in mechanical ventilation duration and increased hyperglycemia, concordant with our findings. The consistency of these updated analyses strengthens confidence that routine methylprednisolone provides minimal clinical benefit at the cost of metabolic derangement.

Several biological mechanisms may explain why the anti-inflammatory effects of methylprednisolone fail to translate into improved survival. First, while methylprednisolone potently suppresses cytokine-mediated inflammation, cardiopulmonary bypass induces multifaceted injury involving complement activation, ischemia–reperfusion injury, endothelial dysfunction, and coagulopathy—pathways that corticosteroid therapy only partially addresses^22,23. Second, the immunosuppressive effects of corticosteroids may impair host defense mechanisms necessary for postoperative recovery, potentially offsetting anti-inflammatory benefits²⁶. Third, observational evidence suggests corticosteroids may be more harmful in lower-risk patients, possibly through metabolic side effects that disproportionately affect less critically ill children²⁵. Finally, the optimal timing and dosing of corticosteroid administration remain unclear, with studies employing varied strategies, ranging from preoperative administration to cardiopulmonary bypass prime supplementation to repeated postoperative dosing^6,15–17.

The mechanistic disconnect between surrogate inflammatory markers and clinical outcomes warrants emphasis. Keski-Nisula and colleagues^15–17 demonstrated that methylprednisolone consistently reduces interleukin-6, interleukin-8, and other inflammatory mediators, preserves endothelial glycocalyx, and reduces neutrophil activation. Despite these favorable biochemical effects, clinical outcomes remain essentially unchanged. This pattern suggests that either: [1] the measured inflammatory markers are not on the critical causal pathway to morbidity and mortality; [2] compensatory inflammatory responses are necessary for appropriate tissue healing and recovery; or [3] the optimal therapeutic window or target population for anti-inflammatory therapy has not yet been identified. The emphasis on surrogate biochemical endpoints in early corticosteroid trials may have contributed to the premature adoption of this practice before adequate evaluation of patient-centered outcomes.

From a clinical standpoint, our findings do not support routine administration of perioperative methylprednisolone to all pediatric patients undergoing cardiac surgery with cardiopulmonary bypass. We acknowledge, however, that contemporary practice in many high-volume centres is to administer corticosteroids selectively to neonates and to patients undergoing complex high-STAT category repairs, and that recent subpopulation re-analyses of the STRESS trial have suggested heterogeneous effects across neonatal subgroups (e.g., reductions in postoperative hydrocortisone use and thrombosis among term neonates, set against increases in acute kidney injury and insulin requirements). Because individual patient-level data were not available for the present synthesis, we cannot adjudicate the neonatal question with confidence. Our trial-level findings should therefore not be interpreted as a recommendation to discontinue selective neonatal use, but rather as a call for adequately powered, neonate-focused randomized trials and individual patient-level data meta-analyses.

Clinical implications and limitations

Our findings have direct implications for clinical practice and future research. First, routine prophylactic methylprednisolone cannot be recommended for all pediatric patients undergoing cardiac surgery based on current evidence. Clinicians should weigh the modest reduction in mechanical ventilation duration against the significantly increased risk of hyperglycemia and the lack of mortality benefit when making individualized treatment decisions. Second, if methylprednisolone is used, rigorous glucose monitoring protocols must be implemented, given the consistent hyperglycemia signal across trials. Third, our findings highlight the need for adequately powered randomized trials specifically designed to identify patient subgroups most likely to benefit from corticosteroid therapy, incorporating modern risk stratification tools and evaluating long-term neurodevelopmental outcomes.

Several limitations warrant consideration. First, our analysis included only 8 trials, all of which exhibited substantial heterogeneity in patient populations, surgical procedures, cardiopulmonary bypass techniques, and perioperative management protocols. Although statistical heterogeneity was low for most outcomes, clinical heterogeneity limits generalizability. Second, subgroup analyses were not specified in the original protocol but were added as an amendment before data extraction. While biologically plausible and hypothesis-driven, these analyses should be considered exploratory and hypothesis-generating rather than definitive. Third, individual patient-level data were not available, precluding more sophisticated subgroup analyses based on surgical complexity scores, cardiopulmonary bypass duration, or baseline patient characteristics. Fourth, the dominant influence of the STRESS trial (78% of the total sample) on pooled estimates raises concerns about generalizability, particularly given that trial’s unique design features, including rescue corticosteroid use in controls. Fifth, publication bias cannot be definitively excluded, although funnel plot asymmetry testing was not possible given the small number of included studies (n = 8). Sixth, outcome definitions varied across trials, particularly for infections and hyperglycemia, potentially introducing measurement heterogeneity. Finally, long-term outcomes including neurodevelopmental function, growth, and metabolic sequelae were not evaluated in most trials, representing an essential gap in the evidence base.

Despite these limitations, our systematic review has several strengths. We employed rigorous PRISMA 2020–compliant methodology, conducted comprehensive searches without language restrictions, used robust statistical methods appropriate for sparse data (Paule-Mandel estimator with Hartung-Knapp adjustment), and incorporated the most contemporary high-quality evidence. The consistency of our findings with recent independent meta-analyses strengthens confidence in the results^28–30.

Seventh, our principal mortality estimate includes Toledo-Pereyra 1980, a trial conducted under surgical and intensive-care conditions that no longer reflect contemporary practice; sensitivity analysis restricted to post-2000 trials yielded a pooled estimate closer to the null (RR 0.79, 95% CI 0.43–1.45).

Conclusion

In this contemporary trial-level meta-analysis of randomized controlled trials, perioperative methylprednisolone was not associated with a reduction in all-cause mortality among pediatric patients undergoing cardiac surgery with cardiopulmonary bypass. A modest reduction in mechanical-ventilation duration was observed, but this estimate was driven by a single small trial and was not robust to sensitivity analysis. Methylprednisolone was consistently associated with a clinically meaningful increase in postoperative hyperglycemia. Taken together, current evidence does not support the routine prophylactic use of methylprednisolone in unselected pediatric cardiac surgical populations. The question of whether selective neonatal use — currently practised in many centres — confers a net clinical benefit cannot be answered by trial-level synthesis and will require adequately powered neonate-focused trials and individual patient-level data meta-analyses, with explicit assessment of long-term neurodevelopmental and metabolic outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization: J.J.B., O.R.-L., C.B.-A., J.A.M.-O., R.F.-P., and J.F.Á.Methodology: J.J.B., O.R.-L., C.B.-A., and J.A.M.-O.Software: J.J.B.Validation: J.J.B., O.R.-L., C.B.-A., and J.A.M.-O.Formal analysis: J.J.B., O.R.-L., and C.B.-A.Investigation: J.J.B., O.R.-L., C.B.-A., J.A.M.-O., R.F.-P., and J.F.Á.Resources: J.J.B.Data curation: J.J.B.Writing—original draft preparation: J.J.B., O.R.-L., and C.B.-A.Writing—review and editing: J.J.B., O.R.-L., C.B.-A., J.A.M.-O., R.F.-P., and J.F.Á.Visualization: J.J.B.Supervision: J.J.B.

Funding

This research received no external funding. The APC was funded by Universidad Señor de Sipan–Vicerrectorado de investigacion.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files. The full electronic search strategy is provided in Supplementary Table S1.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval and consent to participate

This study was conducted as a systematic review and meta-analysis of previously published randomized controlled trials. As the study did not involve the collection of primary data from human participants and relied exclusively on publicly available data from published studies, ethical approval and informed consent were not required in accordance with institutional and international research ethics guidelines.

Consent for publication

All authors have read and approved the final version of the manuscript and consent to its publication.

Use of artificial intelligence

The authors used a large language model (Anthropic Claude) as a writing-assistance tool to support copy-editing, language polishing, and the organization of methodological and discussion paragraphs of this manuscript. All study design, search strategy, eligibility decisions, data extraction, statistical analyses, risk-of-bias judgements, GRADE certainty assessments, and clinical interpretation were conceived, performed, and verified by the human authors. The authors reviewed and edited all AI-assisted text and take full responsibility for the content, accuracy, and integrity of the manuscript.