Comparison of Different Corticosteroid Regimens for Refractory Mycoplasma pneumoniae Pneumonia in Children: A Retrospective Cohort Study

Abstract

Background

Refractory Mycoplasma pneumoniae pneumonia (RMPP) poses a significant therapeutic challenge in pediatric practice, and the optimal corticosteroid regimen remains controversial. This study aimed to compare the clinical efficacy and safety of three corticosteroid regimens in children with RMPP.

Methods

In this retrospective cohort study, 216 children with RMPP hospitalized between January 2023 and June 2025 were categorized by corticosteroid regimen: Group A (low-dose methylprednisolone, 1–2 mg/kg/day intravenously for 3–5 days, n=73), Group B (high-dose methylprednisolone pulse therapy, 10–30 mg/kg/day for 3 days, n=70), and Group C (step-down regimen: pulse therapy for 3 days followed by an oral prednisolone taper, n=73). The primary outcome was time to fever resolution after corticosteroid initiation. Multivariable logistic and Cox regression, inverse probability of treatment weighting (IPTW), and E-value analyses were performed.

Results

Baseline characteristics were largely comparable, although white blood cell counts differed (P=0.013). Defervescence was faster in Groups B (2.40 ± 1.00 days) and C (2.00 ± 0.78 days) than in Group A (4.25 ± 1.71 days; both P<0.001); the Group C-versus-Group B difference (adjusted P=0.047) occurred largely before the oral taper began. Group C had the shortest hospital stay (8.26 ± 1.74 days) and the highest treatment effectiveness (89.04% vs 82.86% vs 64.38%; P=0.001), but the adjusted Group C-versus-Group B comparison was not significant (OR 2.17; 95% CI 0.77–6.12). Both high-dose regimens were independently associated with treatment effectiveness versus Group A (Group C: OR 5.26, 95% CI 2.08–13.32; E-value 2.16) and with faster defervescence; IPTW analyses yielded consistent results. Thirty-day readmission did not differ significantly (exact P=0.130).

Conclusion

High-dose corticosteroid regimens were associated with faster defervescence and higher treatment effectiveness than low-dose therapy in children with RMPP. Whether an oral taper after pulse therapy provides additional benefit remains uncertain and warrants prospective validation.

Introduction

Mycoplasma pneumoniae (MP) remains one of the most prevalent etiological agents of community-acquired pneumonia (CAP) in children aged 5 years and older, accounting for approximately 10–40% of all pediatric CAP cases globally.^1,2 Although MPP predominantly affects school-aged children, infections are increasingly recognized in younger children and infants.³ The resurgence of MP infections following the COVID-19 pandemic has been particularly noteworthy, with several countries reporting unprecedented outbreaks in 2023–2024, characterized by increased disease severity and higher rates of macrolide resistance.^4–6 This epidemiological shift has reinvigorated clinical and academic interest in optimizing therapeutic strategies for MP-associated respiratory illness.

While the majority of MP pneumonia (MPP) cases respond favorably to macrolide antibiotics, a subset of patients develops refractory MPP (RMPP), defined as persistent fever and clinical or radiological deterioration despite appropriate macrolide therapy for ≥7 days.^7,8 The pathogenesis of RMPP extends beyond direct microbial injury and involves a complex interplay of host immune dysregulation, characterized by excessive proinflammatory cytokine release, T-helper cell imbalance, and complement activation.^9,10 Host factors such as atopy may further modulate disease severity and extrapulmonary manifestations.¹¹ These immunopathological mechanisms provide the theoretical rationale for corticosteroid use as adjunctive therapy in RMPP, and serum biomarkers such as lactate dehydrogenase have been proposed to identify candidates for corticosteroid therapy.¹² Moreover, the increasing prevalence of macrolide-resistant MP has complicated antimicrobial management, and tetracyclines or fluoroquinolones have been used as alternative agents in macrolide-unresponsive cases.^13,14

Current evidence supports the use of corticosteroids in RMPP management, yet substantial heterogeneity persists regarding optimal dosing, timing, and duration.^15–17 Low-dose oral or intravenous corticosteroids (1–2 mg/kg/day of methylprednisolone equivalent) have been traditionally employed, demonstrating modest clinical benefit in mild-to-moderate cases.¹⁸ In contrast, high-dose intravenous methylprednisolone pulse therapy (10–30 mg/kg/day) has gained traction for severe RMPP, with several studies reporting more rapid defervescence and radiological improvement.^8,16,19 Recent comparative studies, including propensity score-matched analyses, have reported that high-dose regimens achieve faster defervescence and shorter hospitalization than low-dose therapy, although differences in overall efficacy and adverse event burden remain inconsistent.^20,21 The timing of corticosteroid initiation may further influence clinical outcomes and late pulmonary sequelae.^22,23 A step-down approach combining initial pulse therapy with a subsequent oral taper has been proposed to couple rapid inflammatory control with sustained immunomodulation, but direct comparative evidence remains limited.²⁴

The 2023 evidence-based guidelines from the Chinese Medical Association acknowledged the role of corticosteroids in RMPP but highlighted the absence of high-quality comparative studies to guide regimen selection.²⁴ Although recent studies have compared low- and high-dose regimens, direct comparisons of low-dose, high-dose pulse, and step-down regimens within a single cohort remain scarce.^20,21 This knowledge gap hinders the development of standardized treatment protocols and may contribute to inconsistent clinical outcomes observed across institutions.

Therefore, we conducted this retrospective cohort study to compare the clinical efficacy and safety of low-dose, high-dose pulse, and step-down corticosteroid regimens in children with RMPP. We hypothesized that high-dose-based regimens would be associated with more rapid clinical improvement than low-dose therapy, and we exploratively compared the step-down regimen with high-dose pulse monotherapy.

Methods

Study Design and Setting

This retrospective cohort study was conducted at the three campuses of Shiyan People’s Hospital, which operate as a single institution under a unified institutional protocol, and the study is therefore reported as a single-center study. Medical records of children hospitalized with a diagnosis of RMPP between January 2023 and June 2025 were systematically reviewed. The study protocol was approved by the institutional review boards of Shiyan People’s Hospital (approval numbers: 20250502), and the requirement for individual informed consent was waived given the retrospective nature of the investigation. This study was conducted in accordance with the Declaration of Helsinki and reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.²⁵ The completed STROBE checklist is provided as Supplementary Checklist 1.

Participants

Eligible participants were children aged 1–14 years who met the diagnostic criteria for RMPP. The diagnosis of MPP required: (1) clinical features consistent with pneumonia (fever, cough, and abnormal auscultatory findings); (2) radiological evidence of pulmonary infiltrates on chest imaging; and (3) laboratory confirmation of MP infection by a positive serum MP-IgM antibody (colloidal gold immunochromatographic assay) or a positive MP DNA test by polymerase chain reaction (PCR) from nasopharyngeal swab specimens.^24,26 Macrolide-resistance genotyping was not routinely performed during the study period. RMPP was defined as persistent fever (≥38.5°C) and/or progressive clinical or radiological deterioration after ≥7 days of appropriate macrolide antibiotic therapy (azithromycin or erythromycin), including both pre-admission and in-hospital treatment.^7,8 For each patient, the index time (time zero) was defined as the day of the first corticosteroid dose, and exposures, covariates, and outcomes were measured relative to this time point.

Exclusion criteria included: (1) coexisting bacterial or viral co-infection confirmed by culture or molecular testing; (2) underlying immunodeficiency or immunosuppressive therapy; (3) chronic pulmonary disease (excluding well-controlled asthma); (4) corticosteroid use within 2 weeks prior to admission; and (5) incomplete medical records. A total of 216 patients who met all eligibility criteria were included in the final analysis (Figure 1).

Figure 1.

Study flow diagram. A total of 892 patients with Mycoplasma pneumoniae pneumonia were identified between January 2023 and June 2025. After excluding 480 patients who did not meet refractory MPP criteria and 196 patients who met other exclusion criteria, 216 patients were enrolled and allocated to three corticosteroid regimen groups: Group A (low-dose methylprednisolone, n=73), Group B (high-dose pulse methylprednisolone, n=70), and Group C (step-down regimen, n=73). All patients completed the study follow-up and were included in the final analysis. Key outcomes for each group are summarized. LOS, length of stay; MP, methylprednisolone.

Treatment Groups

Patients were categorized into three groups based on the corticosteroid regimen administered. The choice of regimen was determined by the treating physician according to institutional protocols and clinical judgment, and group assignment reflected the regimen planned and documented at the time of corticosteroid initiation. Group A (low-dose regimen, n=73) received intravenous methylprednisolone at 1–2 mg/kg/day for 3–5 days. Group B (high-dose pulse regimen, n=70) received intravenous methylprednisolone at 10–30 mg/kg/day (maximum 1 g/day) for 3 consecutive days. Group C (step-down regimen, n=73) received initial intravenous methylprednisolone pulse therapy (10–30 mg/kg/day) for 3 days, followed by oral prednisolone at 1–2 mg/kg/day with gradual tapering over 5–10 days. For Group C patients discharged before completion of the oral taper, the remaining doses were prescribed at discharge, with adherence confirmed through outpatient or telephone follow-up. All patients concurrently received macrolide antibiotics throughout the treatment course (intravenous azithromycin 10 mg/kg/day, transitioned to oral administration upon clinical improvement, in accordance with the 2023 Chinese guideline;²⁴ alternative agents such as doxycycline were considered for children aged ≥8 years with suspected macrolide resistance at the discretion of the treating physician).

Data Collection

Clinical data were extracted from electronic medical records using a standardized case report form. The following variables were collected: demographic characteristics (age, sex, weight, height, body mass index); pre-admission clinical features (duration of fever, cough, and macrolide use before hospitalization); vital signs at admission (temperature, heart rate, respiratory rate, oxygen saturation); laboratory parameters at admission and post-treatment (complete blood count, C-reactive protein [CRP], erythrocyte sedimentation rate [ESR], procalcitonin [PCT], lactate dehydrogenase [LDH], ferritin, D-dimer, liver function tests, interleukin-6 [IL-6], and interleukin-10 [IL-10]); chest computed tomography findings; treatment details; and clinical outcomes. Supplemental oxygen therapy and bronchoscopy were recorded if performed at any time during hospitalization; the exact timing of oxygen therapy relative to corticosteroid initiation was not systematically documented.

Outcome Measures

The primary outcome was time to fever resolution after initiation of corticosteroid therapy, defined as the interval from the first dose of corticosteroid to sustained defervescence (body temperature <37.5°C for ≥24 hours without antipyretic use). Secondary outcomes included: (1) total length of hospital stay; (2) overall treatment effectiveness, a composite defined as fever resolution within 72 hours of corticosteroid initiation, improvement in respiratory symptoms (cough, tachypnea, and increased work of breathing) assessed daily by the treating team, and improvement or stability on follow-up chest imaging performed between hospital days 7 and 10 (or earlier if clinically indicated); (3) incidence of corticosteroid-related adverse events (hyperglycemia, hepatotoxicity, secondary infection, insomnia/agitation), assessed throughout hospitalization and at discharge; (4) 30-day readmission rate, ascertained from outpatient records and telephone follow-up; and (5) changes in inflammatory biomarkers (CRP, LDH, IL-6) between admission and post-treatment assessment performed on hospital days 5–7, or on the day of discharge for patients discharged before day 5.

Statistical Analysis

Continuous variables were assessed for normality using the Shapiro–Wilk test and expressed as mean ± standard deviation (SD) or median (interquartile range [IQR]) as appropriate. Categorical variables were presented as frequencies and percentages. Comparisons among the three groups were performed using one-way analysis of variance (ANOVA) for normally distributed continuous variables, the Kruskal–Wallis test for non-normally distributed variables, and the chi-square test for categorical variables; the Fisher-Freeman-Halton exact test was used for contingency tables with expected cell counts below 5. Post hoc pairwise comparisons were conducted with Bonferroni correction. Time to fever resolution was additionally analyzed as a time-to-event outcome using Cox proportional hazards regression adjusted for prespecified covariates (age, sex, pre-admission fever duration, standardized baseline CRP and LDH, consolidation, and pleural effusion), and the proportional hazards assumption was assessed using Schoenfeld residuals. Multivariable logistic regression was used to identify factors associated with treatment effectiveness; because 46 treatment-failure events were observed (events-per-variable ratio 5.1), the model was internally validated using 500 bootstrap resamples (optimism-corrected area under the receiver operating characteristic curve [AUC]), and calibration was assessed with the Hosmer-Lemeshow test. An adjusted direct contrast between Groups C and B was obtained by refitting an identical model with Group B as the reference category. To address confounding by indication, inverse probability of treatment weighting (IPTW) was performed as a sensitivity analysis, with propensity scores estimated by multinomial logistic regression including the prespecified covariates plus baseline white blood cell count, oxygen saturation, and admission temperature; weighted comparisons used heteroskedasticity-consistent standard errors. An E-value was calculated for the adjusted association between the step-down regimen and treatment effectiveness to assess robustness to unmeasured confounding. Adjusted linear models with heteroskedasticity-consistent (HC3) standard errors were fitted for fever resolution time and length of hospital stay. A sensitivity analysis restricted to patients with ≥7 days of pre-admission macrolide therapy was performed to assess eligibility-related robustness. Within-group biomarker changes were compared using the Wilcoxon signed-rank test, between-group differences in the magnitude of change were compared using the Kruskal–Wallis test, and the group-by-time interaction was assessed using a linear mixed-effects model on log-transformed values. Subgroup analyses stratified by baseline CRP (>40 vs ≤40 mg/L) included a formal treatment-by-subgroup interaction test. The primary outcome was prespecified before data analysis; secondary and subgroup analyses were considered exploratory, and no formal adjustment for multiplicity across secondary outcomes was applied. All statistical analyses were performed using Python (version 3.11) with scipy, statsmodels, and lifelines packages. A two-sided P value <0.05 was considered statistically significant. All numerical data were reported to two decimal places, and P values to three decimal places.

Results

Baseline Demographics and Clinical Characteristics

A total of 216 children with RMPP were included in this study: 73 in Group A (low-dose), 70 in Group B (high-dose pulse), and 73 in Group C (step-down). The patient enrollment and study flow are presented in Figure 1. The mean age of the entire cohort was 6.72 ± 2.30 years, with 120 (55.56%) male patients. Baseline demographic characteristics, pre-admission clinical features, vital signs, laboratory parameters, and imaging findings were largely comparable across the three groups (Table 1); however, white blood cell (WBC) counts differed significantly among groups (9.36 ± 3.16, 9.01 ± 3.03, and 10.55 ± 3.55 ×10⁹/L in Groups A, B, and C, respectively; P=0.013) and were therefore included as a covariate in the propensity score model. The mean fever duration before admission was 9.55 ± 2.46 days. The mean duration of pre-admission macrolide therapy was 7.21 ± 2.02 days (range 3–12 days); 81 patients (37.50%) had received fewer than 7 days of macrolides before admission and fulfilled the ≥7-day criterion when in-hospital macrolide therapy administered before corticosteroid initiation was included. All patients had laboratory-confirmed MP infection; serum MP-IgM was positive in 163 patients (75.46%) and weakly positive in 53 patients (24.54%), with no significant difference among groups.

Table 1.

Baseline Demographics, Laboratory Parameters, and Imaging Findings

Variable	Group A (n=73)	Group B (n=70)	Group C (n=73)	P value
Age (years)	6.73 ± 2.11	6.62 ± 2.57	6.82 ± 2.24	0.866
Male, n(%)	43 (58.90%)	34 (48.57%)	43 (58.90%)	0.360
Weight (kg)	24.77 ± 5.67	24.41 ± 6.66	25.06 ± 5.53	0.811
BMI (kg/m2)	19.24 ± 1.65	18.96 ± 1.59	19.62 ± 1.79	0.069
Fever pre-admission (d)	9.49 ± 2.44	9.84 ± 2.71	9.34 ± 2.23	0.466
Macrolide pre-admission (d)	7.03 ± 2.09	7.20 ± 1.97	7.41 ± 2.01	0.519
Temperature (°C)	38.76 ± 0.56	38.90 ± 0.65	38.77 ± 0.53	0.309
SpO2 (%)	95.23 ± 1.80	95.07 ± 2.10	95.00 ± 2.02	0.762
WBC (×109/L)	9.36 ± 3.16	9.01 ± 3.03	10.55 ± 3.55	0.013
CRP (mg/L)	21.39 (11.39–34.58)	18.75 (12.82–30.68)	18.00 (10.55–40.02)	0.849
ESR (mm/h)	47.93 ± 19.07	46.04 ± 18.95	46.67 ± 18.84	0.831
PCT (ng/mL)	0.52 (0.28–1.09)	0.52 (0.26–0.95)	0.56 (0.28–1.02)	0.912
LDH (U/L)	466.23 ± 160.48	434.61 ± 151.80	446.36 ± 138.39	0.446
Ferritin (ng/mL)	160.35 (97.62–221.37)	142.17 (91.23–207.41)	116.74 (75.28–196.95)	0.218
D-dimer (mg/L)	1.18 (0.82–1.70)	1.46 (0.87–1.93)	1.18 (0.74–1.96)	0.231
ALT (U/L)	24.52 ± 12.84	22.45 ± 15.06	22.01 ± 13.79	0.509
AST (U/L)	29.86 ± 15.39	24.82 ± 11.79	30.53 ± 19.24	0.065
IL-6 (pg/mL)	20.43 (11.40–36.05)	22.88 (13.03–45.75)	21.05 (14.11–36.07)	0.850
Consolidation, n(%)	61 (83.56%)	58 (82.86%)	68 (93.15%)	0.128
Pleural effusion, n(%)	29 (39.73%)	18 (25.71%)	28 (38.36%)	0.154
Atelectasis, n(%)	19 (26.03%)	27 (38.57%)	25 (34.25%)	0.267
Mucus plug, n(%)	30 (41.10%)	29 (41.43%)	38 (52.05%)	0.320
Bilateral, n(%)	23 (31.51%)	20 (28.57%)	21 (28.77%)	0.911

Notes: Data are presented as mean ± SD, median (IQR), or n (%). Group A: low-dose MP; Group B: high-dose pulse MP; Group C: step-down regimen. D-dimer, ALT, and AST were added per the Methods section.

At admission, the median CRP level was 19.28 mg/L (IQR 11.63–34.29), and the mean LDH was 449.27 ± 150.35 U/L, reflecting a moderate inflammatory burden consistent with RMPP. Chest CT findings demonstrated consolidation in 86.57% of cases, mucus plug in 44.91%, atelectasis in 32.87%, and pleural effusion in 34.72%. Bilateral lung involvement was observed in 29.63% of patients. These clinical features were comparable to previously reported RMPP cohorts.^27,28

Treatment Characteristics

The treatment parameters differed significantly among groups by design (Table 2). The mean methylprednisolone dose was 1.46 ± 0.27 mg/kg/day in Group A, 19.84 ± 5.39 mg/kg/day in Group B, and 20.15 ± 5.70 mg/kg/day (initial pulse dose) in Group C (P<0.001). The total duration of corticosteroid therapy was 3.90 ± 0.87 days in Group A, 3.00 ± 0.00 days in Group B, and 10.16 ± 2.15 days in Group C (P<0.001); in Group C, the oral taper was completed after discharge in most patients, as reflected in the difference between corticosteroid duration and length of stay. Bronchoscopy was performed in 96 patients (44.44%), with similar rates across groups (P=0.760). The proportion of patients who received supplemental oxygen at any time during hospitalization was higher in Group A (45.21%) than in Group B (28.57%) and Group C (26.03%; P=0.029); because the timing of oxygen therapy relative to corticosteroid initiation was not systematically recorded, this variable was not treated as a baseline characteristic nor included as a confounder in adjusted analyses.

Table 2.

Treatment Characteristics

Variable	Group A (n=73)	Group B (n=70)	Group C (n=73)	P value
Steroid dose (mg/kg/day)	1.46 ± 0.27	19.84 ± 5.39	20.15 ± 5.70	<0.001
Steroid duration (days)	3.90 ± 0.87	3.00 ± 0.00	10.16 ± 2.15	<0.001
Bronchoscopy, n(%)	35 (47.95%)	30 (42.86%)	31 (42.47%)	0.760
Oxygen therapy during hospitalization, n(%)	33 (45.21%)	20 (28.57%)	19 (26.03%)	0.029

Primary Outcome: Fever Resolution Time

The primary outcome demonstrated significant differences among the three groups (Figure 2A and Table 3). The mean time to fever resolution after corticosteroid initiation was 4.25 ± 1.71 days in Group A, 2.40 ± 1.00 days in Group B, and 2.00 ± 0.78 days in Group C (Kruskal–Wallis test, P<0.001). Pairwise comparisons with Bonferroni correction revealed that both Group B (adjusted P<0.001) and Group C (adjusted P<0.001) achieved significantly faster defervescence than Group A, whereas the difference between Groups C and B was of borderline statistical significance (adjusted P=0.047). Notably, fever resolved within the 3-day pulse phase—that is, before the oral taper distinguishing Group C had begun—in 60 of 70 patients (85.71%) in Group B and 71 of 73 patients (97.26%) in Group C. In adjusted Cox proportional hazards analysis, both high-dose regimens were associated with faster defervescence compared with Group A (Group B: hazard ratio [HR] 4.49, 95% CI 3.03–6.67; Group C: HR 7.83, 95% CI 5.11–11.98), and the adjusted Group C-versus-Group B hazard ratio was 1.74 (95% CI 1.23–2.46; P=0.002). The proportional hazards assumption was satisfied for all covariates (all P>0.05 by Schoenfeld residuals). An adjusted linear model yielded consistent results (Group C vs Group B: β=−0.40 days, 95% CI −0.72 to −0.09; P=0.011).

Figure 2.

Comparison of primary and secondary outcomes across corticosteroid regimen groups. (A) Time to fever resolution after corticosteroid initiation. (B) Length of hospital stay. Box plots show the median and interquartile range, with whiskers extending to 1.5 times the interquartile range; all individual observations are displayed as jittered points. (C) Treatment effectiveness rates. (D) Overall complication rates. Brackets indicate Bonferroni-corrected pairwise comparisons: *P<0.05, **P<0.01, ***P<0.001, ns = not significant (P≥0.05). The P value in (D) is from the overall chi-square test. Group A: low-dose methylprednisolone (1–2 mg/kg/day); Group B: high-dose pulse methylprednisolone (10–30 mg/kg/day); Group C: step-down regimen (pulse followed by oral taper).

Table 3.

Primary and Secondary Outcomes

Outcome	Group A (n=73)	Group B (n=70)	Group C (n=73)	P value
Fever resolution (days)	4.25 ± 1.71	2.40 ± 1.00	2.00 ± 0.78	<0.001
Hospital stay (days)	11.92 ± 2.87	9.26 ± 2.51	8.26 ± 1.74	<0.001
Treatment effective, n(%)	47 (64.38%)	58 (82.86%)	65 (89.04%)	0.001
30-day readmission, n(%)	8 (10.96%)	3 (4.29%)	2 (2.74%)	0.130
Any complication, n(%)	13 (17.81%)	25 (35.71%)	20 (27.40%)	0.054
Hyperglycemia	4 (5.48%)	8 (11.43%)	8 (10.96%)	–
Hepatotoxicity	4 (5.48%)	5 (7.14%)	4 (5.48%)	–
Secondary infection	5 (6.85%)	6 (8.57%)	3 (4.11%)	–
Insomnia/Agitation	0 (0.00%)	6 (8.57%)	5 (6.85%)	–
CRP post-treatment (mg/L)	6.20 (3.02–10.48)	5.68 (3.56–9.92)	5.73 (2.85–13.14)	0.925
LDH post-treatment (U/L)	313.92 ± 124.98	283.57 ± 111.47	306.41 ± 109.68	0.267
IL-6 post-treatment (pg/mL)	5.48 (3.01–12.84)	6.85 (3.93–16.25)	6.34 (3.55–12.33)	0.377

Notes: Data are presented as mean ± SD, median (IQR), or n (%). Fever resolution: time from steroid initiation to sustained defervescence. The P value for 30-day readmission is from the Fisher-Freeman-Halton exact test; the P value for any complication is from the chi-square test.

Secondary Outcomes

Length of hospital stay followed a similar pattern, with Group C exhibiting the shortest duration (8.26 ± 1.74 days), followed by Group B (9.26 ± 2.51 days) and Group A (11.92 ± 2.87 days; P<0.001; Figure 2B). Pairwise analyses confirmed significant differences between Group A and both Group B (adjusted P<0.001) and Group C (adjusted P<0.001), as well as between Group B and Group C (adjusted P=0.036). Adjusted linear models yielded consistent results: versus Group A, the adjusted coefficients were −2.58 days (95% CI −3.51 to −1.65) for Group B and −3.59 days (95% CI −4.40 to −2.79) for Group C, and the adjusted Group C-versus-Group B difference was −1.01 days (95% CI −1.76 to −0.27; P=0.008).

Treatment effectiveness rates were 64.38% in Group A, 82.86% in Group B, and 89.04% in Group C (P=0.001; Figure 2C). Pairwise comparisons with Bonferroni correction demonstrated significant differences between Group A and Group C (adjusted P=0.001) and between Group A and Group B (adjusted P=0.037), whereas the difference between Groups B and C was not statistically significant (adjusted P=0.860). Fever resolution within 72 hours—the principal component of the composite outcome—was achieved by 30.14%, 85.71%, and 97.26% of patients in Groups A, B, and C, respectively.

The 30-day readmission rate was numerically lower in Group C (2.74%) and Group B (4.29%) compared with Group A (10.96%), but this difference did not achieve statistical significance (Fisher-Freeman-Halton exact test, P=0.130), and the analysis was limited by the small number of events (13 readmissions in total).

Changes in Inflammatory Biomarkers

Post-treatment laboratory data (hospital days 5–7, or at discharge for patients discharged before day 5) were available for all 216 patients (Figure 3). Median CRP levels decreased from 21.39 (IQR 11.39–34.58) to 6.20 (3.02–10.48) mg/L in Group A, from 18.75 (12.82–30.68) to 5.68 (3.56–9.92) mg/L in Group B, and from 18.00 (10.55–40.02) to 5.73 (2.85–13.14) mg/L in Group C; within-group reductions were statistically significant in all three groups (Wilcoxon signed-rank test, all P<0.001). The magnitude of the CRP reduction did not differ among groups (ΔCRP: 19.50 ± 16.06, 18.24 ± 15.64, and 21.40 ± 21.30 mg/L; P=0.839), and a linear mixed-effects model on log-transformed CRP showed no significant group-by-time interaction (both interaction P>0.4). Median IL-6 levels likewise declined significantly in all groups (all within-group P<0.001), with no between-group difference in the magnitude of change (P=0.953). LDH levels decreased by 152.32 ± 68.69 U/L, 151.04 ± 66.82 U/L, and 139.95 ± 61.16 U/L in Groups A, B, and C, respectively (P=0.505).

Figure 3.

Changes in inflammatory biomarkers from admission (pre-treatment) to post-treatment assessment (hospital days 5–7, or at discharge for patients discharged earlier). (A) C-reactive protein levels. (B) Interleukin-6 levels. Box plots show the median and interquartile range; filled boxes represent pre-treatment values and lighter boxes represent post-treatment values. Within-group changes were assessed using the Wilcoxon signed-rank test (***P<0.001 for all three groups). The P value within each panel refers to the between-group comparison of the magnitude of change (Kruskal–Wallis test).

Adverse Events

Corticosteroid-related adverse events were observed in 58 patients (26.85%) across the cohort (Figure 2D). The overall complication rate was 17.81% in Group A (13/73), 35.71% in Group B (25/70), and 27.40% in Group C (20/73) (chi-square test, P=0.054). Hyperglycemia was the most common adverse event (9.26%, 20/216), followed by secondary infection (6.48%, 14/216), hepatotoxicity (6.02%, 13/216), and insomnia/agitation (5.09%, 11/216). Insomnia/agitation was exclusively observed in Groups B and C, consistent with the higher initial steroid doses in these groups. No serious adverse events requiring treatment discontinuation or intensive care unit admission were reported; however, given the sample size and the 30-day follow-up, rare or delayed adverse events cannot be excluded.

Multivariate Analysis

Multivariable logistic regression analysis identified factors associated with treatment effectiveness (Figure 4 and Table 4). Compared with Group A, both the step-down regimen (Group C: OR 5.26, 95% CI 2.08–13.32; P<0.001) and the high-dose pulse regimen (Group B: OR 2.42, 95% CI 1.05–5.57; P=0.037) were independently associated with higher odds of treatment effectiveness. Younger age was also associated with higher effectiveness (OR per SD 0.63, 95% CI 0.44–0.91; P=0.014). In the adjusted direct contrast, Group C did not differ significantly from Group B (OR 2.17, 95% CI 0.77–6.12; P=0.143). The model showed adequate calibration (Hosmer-Lemeshow test, P=0.518), an apparent AUC of 0.741, and a bootstrap optimism-corrected AUC of 0.685; the events-per-variable ratio was 5.1, indicating a risk of overfitting that warrants cautious interpretation. The E-value for the adjusted Group C-versus-Group A association was 2.16 (1.75 for the lower confidence limit), meaning that an unmeasured confounder would need to be associated with both regimen selection and treatment effectiveness by a risk ratio of at least 2.16, beyond the measured covariates, to fully explain the association. IPTW sensitivity analysis yielded consistent results: weighted effectiveness rates were 65.2%, 82.8%, and 92.2% in Groups A, B, and C, with adjusted weighted odds ratios of 2.67 (95% CI 1.22–5.86; P=0.014) for Group B versus Group A, 4.49 (95% CI 1.87–10.80; P=0.001) for Group C versus Group A, and 1.68 (95% CI 0.64–4.40; P=0.290) for Group C versus Group B; weighted analyses of fever resolution time were likewise consistent (Group B vs A: β=−1.89; Group C vs A: β=−2.40; both P<0.001). In the sensitivity analysis restricted to the 135 patients with ≥7 days of pre-admission macrolide therapy, fever resolution times (4.32 ± 1.81, 2.38 ± 1.10, and 1.98 ± 0.76 days in Groups A, B, and C; P<0.001) and treatment effectiveness rates (58.54%, 87.50%, and 87.04%; P=0.001) were consistent with the primary analysis.

Figure 4.

Forest plot depicting the multivariable logistic regression analysis for treatment effectiveness. Odds ratios with 95% confidence intervals are shown on a logarithmic scale. Blue diamonds indicate statistically significant associations (P<0.05); gray diamonds indicate non-significant associations. The dashed vertical line represents an odds ratio of 1 (null effect). Continuous variables were standardized (per SD). Group A (low-dose) served as the reference category except where indicated. ^†The Group C-versus-Group B contrast was obtained from an identical model with Group B as the reference category.

Table 4.

Multivariate Logistic Regression for Treatment Effectiveness

Variable	OR	95% CI	P value
Age (years)	0.63	0.44–0.91	0.014
Male sex	0.62	0.29–1.29	0.199
Group B (vs Group A)	2.42	1.05–5.57	0.037
Group C (vs Group A)	5.26	2.08–13.32	<0.001
Group C (vs Group B)†	2.17	0.77–6.12	0.143
Fever duration pre-admission	1.05	0.73–1.50	0.785
CRP (mg/L)	0.76	0.55–1.06	0.104
LDH (U/L)	0.98	0.69–1.40	0.932
Consolidation	1.42	0.53–3.82	0.489
Pleural effusion	0.48	0.23–1.00	0.051

Notes: Group A (low-dose) served as reference. Continuous variables were standardized (per SD). The model was recalculated using statsmodels; calibration was adequate (Hosmer-Lemeshow P=0.518), with an apparent AUC of 0.741 and a bootstrap optimism-corrected AUC of 0.685. ^†The Group C-versus-Group B contrast was obtained from an identical model with Group B as the reference category.

Abbreviations: OR, odds ratio; CI, confidence interval.

Subgroup Analysis by Inflammatory Burden

Subgroup analysis stratified by baseline CRP levels is shown in Figure 5. Among patients with high inflammatory burden (CRP >40 mg/L, n=47), treatment effectiveness rates were 50.0% in Group A, 85.7% in Group B, and 78.9% in Group C (P=0.076). In the lower inflammation subgroup (CRP ≤40 mg/L, n=169), effectiveness rates were 67.8% in Group A, 82.1% in Group B, and 92.6% in Group C (P=0.004). Formal interaction testing showed no statistically significant treatment-by-CRP interaction (interaction P=0.339 for Group B and P=0.616 for Group C), providing no statistical evidence that the relative effectiveness of the regimens differs by baseline inflammatory burden. These subgroup findings should therefore be considered exploratory and hypothesis-generating.

Figure 5.

Subgroup analysis of treatment outcomes stratified by baseline C-reactive protein levels. (A) Patients with CRP >40 mg/L (n=47, high inflammatory burden). (B) Patients with CRP ≤40 mg/L (n=169). Solid bars indicate treatment effectiveness rates, with error bars showing 95% Wilson confidence intervals; open bars indicate mean fever resolution time (right axis). P values within panels refer to the across-group chi-square test within each stratum. The treatment-by-CRP interaction was not statistically significant (P>0.05).

Discussion

This retrospective cohort study compared three commonly used corticosteroid regimens for RMPP in children. The principal finding is that both high-dose-based regimens—high-dose pulse therapy and the step-down regimen—were associated with faster defervescence, shorter hospitalization, and higher treatment effectiveness than low-dose methylprednisolone, and these associations remained consistent across multivariable adjustment, IPTW, and sensitivity analyses. In contrast, the added value of the oral taper that distinguishes the step-down regimen is uncertain: the adjusted direct comparison between Groups C and B was not statistically significant for treatment effectiveness, and the modest difference in fever resolution time occurred largely before the taper began.

The association between high-dose regimens and more rapid clinical improvement observed in our study aligns with accumulating evidence. Chen et al¹⁹ demonstrated that early methylprednisolone pulse therapy in children with severe RMPP was associated with reduced fever duration and improved radiological outcomes, and a systematic review and meta-analysis concluded that high-dose pulse therapy accelerates clinical recovery in RMPP.²⁹ More recent comparative studies have directly addressed corticosteroid dosing: in a propensity score-matched analysis, Okumura et al reported faster defervescence and shorter hospitalization with high-dose versus low-dose corticosteroids without an increase in steroid-related adverse events,²⁰ whereas a 2025 propensity-matched study of children with severe MPP found faster clinical and radiological improvement but more adverse reactions with high-dose regimens, without a significant difference in overall efficacy.²¹ Our findings are broadly consistent with this literature. In addition, Yan et al identified prolonged fever and elevated CRP and LDH levels as predictors of corticosteroid-resistant RMPP, supporting early risk stratification when selecting corticosteroid intensity.³⁰ The timing of corticosteroid initiation also appears relevant: initiation approximately 5–10 days after disease onset has been associated with better outcomes and fewer late sequelae,²² and inadequately controlled inflammation in RMPP is associated with long-term pulmonary complications such as bronchiolitis obliterans.²³ The dose-response relationship is biologically plausible, because corticosteroids at conventional doses primarily exert genomic anti-inflammatory effects, whereas supratherapeutic doses additionally engage rapid non-genomic pathways.³¹

The interpretation of the step-down regimen requires particular caution because of the temporal relationship between treatment components and outcomes. Groups B and C received an identical 3-day pulse regimen, and fever resolved during this pulse phase in most patients of both groups (85.71% and 97.26%, respectively). The 0.40-day difference in mean fever resolution time therefore cannot be causally attributed to the oral taper, which had not yet begun. The numerically lower 30-day readmission rate in Group C (2.74% vs 4.29% in Group B and 10.96% in Group A) was not statistically significant (exact P=0.130) and was based on only 13 events; whether a taper reduces late relapse—for example, through more sustained immunomodulation during the subacute phase of RMPP^10,32—remains a hypothesis that our data cannot confirm. The observed differences may also reflect residual confounding by indication, because clinicians selected regimens partly on the basis of baseline severity and early clinical response.

Post-treatment inflammatory markers declined substantially in all three groups, and neither the magnitude of change nor the group-by-time interaction differed among regimens. This suggests that conventional biomarkers such as CRP, IL-6, and LDH, while useful for monitoring treatment response, may not capture the immunological processes that differentiate clinical outcomes among regimens. Serum biomarkers, particularly LDH, have also been proposed as predictors of refractory disease and delayed radiographic resolution^12,33 and may help identify patients who require intensified anti-inflammatory therapy. Whether biomarker-guided regimen selection improves outcomes requires prospective study.

The safety findings also warrant cautious interpretation. The overall complication rate tended to be higher with the high-dose regimens (35.71% in Group B and 27.40% in Group C vs 17.81% in Group A; P=0.054), a pattern consistent with the dose-dependent nature of corticosteroid adverse effects.³⁴ Insomnia and agitation occurred only in the high-dose groups, compatible with the neuropsychiatric effects of supraphysiologic steroid exposure reported in the pediatric literature.³⁵ However, the between-group difference did not reach statistical significance, no serious adverse events were observed, and follow-up was limited to 30 days; these data therefore cannot establish the comparative safety of the three regimens, and longer-term monitoring—particularly for growth, skeletal, and adrenal effects not assessed in this study—is warranted.

The subgroup analysis by baseline CRP levels suggested higher effectiveness of high-dose regimens irrespective of inflammatory burden; however, the treatment-by-CRP interaction was not statistically significant, and no preferential benefit of either high-dose regimen within a specific inflammatory stratum can be claimed. These exploratory findings should not be used to guide regimen selection until confirmed in prospective studies.

Antimicrobial management also merits discussion. Macrolide resistance among MP isolates has risen sharply in China and was a hallmark of the 2023–2024 resurgence.^4,5 In macrolide-unresponsive cases, tetracyclines and fluoroquinolones have shown efficacy in a systematic review and meta-analysis and in recent real-world studies,^13,14 and the 2023 Chinese guideline recommends considering doxycycline in children aged ≥8 years.²⁴ Because macrolide-resistance genotyping was not routinely performed in our cohort, we could not assess the interplay between antimicrobial resistance, antibiotic switching, and corticosteroid response; this remains an important question for future research. Other adjunctive therapies used in RMPP, including intravenous immunoglobulin and bronchoscopic clearance of mucus plugs,^15,36 were not evaluated in the present study.

This study has several important limitations. First, the retrospective, non-randomized design is subject to confounding by indication: regimen selection was influenced by physician judgment, baseline severity, and early clinical response. Although we performed multivariable adjustment, IPTW, and E-value analysis, residual and unmeasured confounding cannot be excluded. Second, the treatment groups were defined by a regimen whose distinguishing component—the oral taper—began after the primary outcome had already occurred in most Group C patients, which limits causal attribution; classification according to the intended regimen documented at corticosteroid initiation mitigates but does not eliminate this temporal ambiguity. Third, baseline WBC counts and the use of oxygen therapy differed among groups, and the timing of oxygen therapy relative to corticosteroid initiation was not recorded, precluding its inclusion as a confounder. Fourth, 81 patients had received fewer than 7 days of macrolides before admission; their eligibility rested on the total duration of macrolide therapy before corticosteroid initiation, including in-hospital treatment, and although a sensitivity analysis restricted to patients with ≥7 pre-admission days yielded consistent results, eligibility misclassification cannot be excluded. Fifth, 122 patients were excluded for incomplete records, introducing potential selection bias. Sixth, the logistic model included 9 predictors with 46 events (events-per-variable ratio 5.1), below conventional thresholds; bootstrap validation suggested moderate overfitting, and the regression results should be considered exploratory. Seventh, multiple outcomes and subgroup comparisons were analyzed without formal multiplicity adjustment beyond pairwise Bonferroni correction, and the composite effectiveness outcome incorporated clinician-assessed components without blinded adjudication. Eighth, macrolide-resistance testing was unavailable, and antimicrobial co-interventions were not fully standardized. Ninth, post-treatment biomarkers were measured on hospital days 5–7 or at discharge, so measurement timing varied with the treatment course and may not represent identical treatment stages across groups. Finally, as a single-center study with 30-day follow-up, this study cannot assess long-term pulmonary outcomes such as bronchiolitis obliterans, and its findings may not be generalizable to other institutions or healthcare settings. Collectively, these limitations substantially constrain causal interpretation, and our results should be viewed as associations requiring confirmation in prospective randomized trials.

Conclusions

In this retrospective cohort of children with RMPP, high-dose corticosteroid regimens—whether pulse monotherapy or a step-down regimen—were associated with faster defervescence, shorter hospitalization, and higher treatment effectiveness than low-dose methylprednisolone. The step-down regimen was not significantly superior to pulse monotherapy in adjusted direct comparisons, and because fever resolved in most patients before the oral taper began, the added value of the tapering component remains uncertain. These findings describe adjusted associations rather than established efficacy and should not, on their own, be used to define an optimal treatment strategy. Prospective randomized controlled trials with standardized regimens, longer follow-up, and assessment of long-term pulmonary outcomes are warranted to validate these observations.

Funding Statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data Sharing Statement

All data generated or analyzed in this study are included in the present manuscript.

Ethics Approval and Consent to Participate

The study protocol was approved by the institutional review boards of Shiyan People’s Hospital (approval numbers: 20250502), and the requirement for individual informed consent was waived given the retrospective nature of the investigation. Patient data were de-identified before analysis and handled confidentially in accordance with institutional data protection requirements; only aggregated, anonymized data are reported. This study was conducted in accordance with the Declaration of Helsinki and reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas. They took part in drafting, revising, or critically reviewing the article. All authors have given final approval of the version to be published, have agreed on the journal to which the article has been submitted, and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that there are no conflicts of interest regarding this study.