Risk factors associated with plastic bronchitis in children with severe Mycoplasma pneumoniae pneumonia: a systematic review and meta-analysis

Abstract

Background

Plastic bronchitis (PB) represents a severe complication of Mycoplasma pneumoniae pneumonia (MPP) in pediatric populations, characterized by the formation of obstructive bronchial casts. Early identification of risk factors is essential for timely bronchoscopic intervention and improved clinical outcomes. However, existing primary studies have reported inconsistent findings regarding the strength and significance of individual risk factors, likely attributable to heterogeneous study designs, varying predictor definitions and cut-off values, and differences in population severity criteria. No comprehensive meta-analysis has systematically synthesized this evidence to date. This study aimed to systematically evaluate and quantitatively synthesize evidence on risk factors associated with PB in children with severe (SMPP) or refractory (RMPP) MPP, with particular focus on inflammatory biomarkers, radiological features, and clinical characteristics.

Methods

A systematic literature search was conducted across PubMed/MEDLINE, Embase, Web of Science, Scopus, and Cochrane Library (January 1, 2015, to December 1, 2025). Case-control and cohort studies reporting risk factors for bronchoscopically confirmed PB in pediatric SMPP or RMPP patients were included. Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS). Random-effects meta-analysis was performed using the DerSimonian-Laird method with Hartung-Knapp adjustment. Effect sizes were expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Heterogeneity was quantified using I² statistics, and publication bias was assessed via Egger’s regression test.

Results

Twelve studies (8 retrospective cohort, 2 case-control, 1 prospective cohort, 1 cohort with external validation; NOS range: 6–8) comprising 4,406 children (1,123 PB cases) were included, all originating from China. A total of 38 effect sizes across 8 risk factor categories were extracted; all were multivariate-adjusted ORs. Risk factors significantly associated with PB included: lactate dehydrogenase (LDH) elevation (k=4, OR =4.06, 95% CI: 2.69–6.11, P=0.002), prolonged fever duration (k=4, OR =3.70, 95% CI: 1.12–12.26, P=0.040), D-dimer elevation (k=3, OR =2.63, 95% CI: 2.01–3.43, P=0.004), pleural effusion (k=4, OR =2.46, 95% CI: 1.55–3.91, P=0.009), atelectasis (k=4, OR =2.32, 95% CI: 1.61–3.34, P=0.005), and consolidation involving ≥2/3 lung lobe (k=2, OR =1.79, 95% CI: 1.36–2.37, P=0.02). Radiological risk factors and LDH demonstrated consistently low heterogeneity (I²=0%). One study reporting LDH as a continuous per-unit OR was excluded from the LDH pooled analysis due to scale incompatibility with dichotomized ORs. Substantial heterogeneity was observed for procalcitonin (PCT) (I²=77.8%) and fever duration (I²=47.5%).

Conclusions

Elevated LDH, prolonged fever duration, D-dimer elevation, and specific radiological features (pleural effusion, atelectasis, consolidation) are significantly associated with increased risk of PB in pediatric SMPP. These risk factors may inform early clinical risk stratification and guide decision-making regarding bronchoscopic intervention. However, their predictive accuracy requires formal evaluation using discrimination metrics. Future prospective multicenter studies are warranted to develop and validate clinical prediction models incorporating these factors.

Highlight box.

Key findings

• This meta-analysis of 12 studies (4,406 children, 1,123 PB cases) identified six significant risk factors of plastic bronchitis (PB) in pediatric severe/refractory Mycoplasma pneumoniae pneumonia (MPP): prolonged fever duration [odds ratio (OR) =3.70], elevated lactate dehydrogenase (OR =4.06), D-dimer elevation (OR =2.63), pleural effusion (OR =2.46), atelectasis (OR =2.32), and consolidation involving ≥2/3 lung lobe (OR =1.79).

• Radiological risk factors demonstrated consistently low heterogeneity (I²=0%), suggesting robust and generalizable findings across diverse clinical settings.

What is known and what is new?

• PB is a life-threatening complication of MPP characterized by obstructive bronchial casts, and early identification is critical for timely bronchoscopic intervention. However, risk factors have been reported inconsistently across individual studies without quantitative synthesis.

• This study provides the first comprehensive meta-analysis quantifying the predictive value of clinical, laboratory, and radiological factors for PB in pediatric MPP. It establishes a hierarchy of predictor strength and identifies radiological features as particularly reliable indicators with minimal heterogeneity.

What is the implication, and what should change now?

• Clinicians should integrate prolonged fever duration, elevated lactate dehydrogenase and D-dimer levels, and radiological findings (pleural effusion, atelectasis, extensive consolidation) into early risk stratification protocols for children with severe MPP.

• These findings support the development of clinical prediction tools to guide timely bronchoscopic intervention decisions, potentially reducing morbidity associated with delayed diagnosis.

• Future prospective multicenter studies are needed to validate predictive models and establish optimal threshold values for these biomarkers.

Introduction

Mycoplasma pneumoniae (MP) constitutes one of the most prevalent etiological agents of community-acquired pneumonia (CAP) in school-aged children, accounting for approximately 10–40% of pediatric CAP cases globally (1,2). Although the majority of MP infections manifest as mild, self-limiting respiratory illnesses, a clinically significant subset of patients progresses to severe Mycoplasma pneumoniae pneumonia (SMPP) or refractory Mycoplasma pneumoniae pneumonia (RMPP). These severe phenotypes are characterized by persistent hyperpyrexia, progressive radiological deterioration, and extrapulmonary manifestations despite appropriate antimicrobial therapy (3,4).

Plastic bronchitis (PB) represents a particularly grave complication of SMPP, distinguished by the formation of gelatinous or rubbery bronchial casts that conform to the bronchial tree architecture and cause significant airway obstruction (5,6). Epidemiological data indicate that the incidence of PB among children with MPP undergoing bronchoscopy ranges from 11% to 43%, with substantially higher rates observed in RMPP cohorts (7,8). Unlike conventional mucus plugging, plastic bronchial casts are characterized by fibrin-rich, inflammatory infiltrates that necessitate bronchoscopic extraction for definitive management (9).

The pathogenesis of PB in MPP involves a complex interplay of direct pathogen-mediated epithelial injury, exaggerated inflammatory responses with pro-inflammatory cytokine release (interleukin-6, interleukin-8, tumor necrosis factor-alpha), and activation of the coagulation cascade, collectively creating conditions favorable for bronchial cast formation (10-14).

Early identification of children at elevated risk for PB is of paramount clinical importance. Delayed bronchoscopic intervention is associated with prolonged hospitalization, increased complication rates, and potential long-term sequelae including bronchiectasis and chronic pulmonary dysfunction (15,16). Several individual studies have investigated risk factors for PB in pediatric MPP, identifying various inflammatory markers, radiological features, and clinical characteristics as potential risk indicators (17-19). However, these studies exhibit considerable heterogeneity in sample size, population characteristics, predictor definitions, and analytical approaches. Reported effect sizes for the same risk factor vary widely across studies—for example, odds ratios (ORs) for LDH elevation range from 1.01 to 5.14, and fever duration cut-off values range from >6 to >12 days—thereby necessitating systematic synthesis to inform evidence-based clinical practice.

The 2023 MP epidemic in East Asia, characterized by unprecedented incidence rates, has resulted in heightened recognition of severe complications including PB (20,21). This epidemiological context has generated substantial research activity addressing risk factors for PB. Despite this growing evidence base, no comprehensive meta-analysis has systematically synthesized the available evidence specifically addressing risk factors for PB in pediatric SMPP.

The primary objective of this systematic review and meta-analysis was to identify and quantify risk factors for PB in children with severe or refractory MPP. Secondary objectives included assessment of heterogeneity across studies, evaluation of publication bias, and generation of evidence-based recommendations for clinical risk stratification. We present this article in accordance with the PRISMA reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2026-1-0052/rc) (22).

Methods

Protocol and registration

This study was not prospectively registered in PROSPERO.

Eligibility criteria

Studies were deemed eligible for inclusion based on the following PICO criteria:

• ❖ Population: children and adolescents (<18 years of age) diagnosed with severe MPP (SMPP) or refractory MPP (RMPP) according to established diagnostic criteria, including Chinese National Guidelines or equivalent international standards.

• ❖ Exposure/intervention: potential risk factors encompassing inflammatory biomarkers [lactate dehydrogenase (LDH), C-reactive protein (CRP), procalcitonin (PCT), neutrophil-to-lymphocyte ratio (NLR), D-dimer, cytokine profiles], radiological features (pleural effusion, atelectasis, consolidation extent, pulmonary lesion volume ratio), and clinical characteristics (fever duration, extrapulmonary manifestations, macrolide resistance status).

• ❖ Comparison: children with SMPP/RMPP who did not develop PB versus those who developed bronchoscopically confirmed PB.

• ❖ Outcome: PB confirmed by bronchoscopic visualization and extraction of bronchial casts conforming to the bronchial tree architecture.

• ❖ Study design: case-control studies and cohort studies (prospective or retrospective).

Exclusion criteria comprised: (I) case reports or case series with fewer than 10 participants; (II) review articles, editorials, letters, or conference abstracts without full-text availability; (III) studies not reporting extractable effect estimates (ORs, relative risks, or sufficient raw data for calculation); (IV) non-English language publications without available translations; (V) studies with overlapping patient cohorts (the larger or more recent study was retained); and (VI) studies focusing on PB etiologies other than MP infection (e.g., congenital heart disease, influenza, lymphatic abnormalities).

Information sources and search strategy

A comprehensive systematic literature search was conducted across five major electronic databases: PubMed/MEDLINE, Embase, Web of Science Core Collection, Scopus, and Cochrane Library. The search encompassed publications from January 1, 2015 to December 1, 2025. The temporal starting point of January 2015 was selected because: (I) systematic investigation of PB as a complication of MPP in pediatric populations emerged primarily after 2015, coinciding with the widespread adoption of flexible bronchoscopy in pediatric respiratory medicine in East Asia; and (II) pre-2015 studies predominantly reported PB in the context of congenital heart disease or lymphatic abnormalities rather than infectious etiologies, employing heterogeneous diagnostic criteria that would compromise comparability with contemporary studies.

The search strategy combined Medical Subject Headings (MeSH) terms and free-text keywords organized into four conceptual domains: (I) MP infection (“Mycoplasma pneumoniae”[MeSH] OR “M. pneumoniae” OR “mycoplasma pneumonia” OR “MPP” OR “RMPP” OR “SMPP” OR “refractory mycoplasma”); (II) plastic bronchitis (“plastic bronchitis” OR “bronchial cast” OR “airway cast” OR “mucoid impaction” OR “bronchial plastic”); (III) pediatric population (“child”[MeSH] OR “pediatric” OR “paediatric” OR “children” OR “adolescent” OR “infant”); and (IV) risk factors (“risk factor*” OR “predictor*” OR “prognostic factor*” OR “biomarker” OR “clinical feature*”). Boolean operators were employed to combine search terms appropriately. The complete search strategy is provided in Appendix 1.

Reference lists of all included studies and relevant review articles were manually screened to identify additional eligible studies. Forward citation tracking was performed using Google Scholar to capture recently published studies citing key articles in the field.

Study selection

Two investigators (Y.L. and Y.P.) independently screened titles and abstracts using Rayyan systematic review software (23). Full texts of potentially eligible articles were subsequently retrieved and assessed against the predetermined inclusion and exclusion criteria. Disagreements were resolved through discussion and consensus, with arbitration by a third investigator (TY) when necessary. Inter-rater agreement was quantified using Cohen’s kappa statistic, with values >0.80 considered excellent agreement.

Data extraction

Data were extracted independently by two investigators using a standardized, piloted extraction form. The following categories of information were collected: (I) study characteristics (first author, publication year, country, institution, study design, enrollment period, sample size); (II) population characteristics (age distribution, sex proportions, SMPP/RMPP diagnostic criteria employed, PB definition and confirmation method); (III) risk factor variables (inflammatory markers with specific cut-off values, radiological features, clinical characteristics); (IV) outcome data (number of PB cases and controls, effect estimates with 95% confidence intervals (CIs), variables included in multivariate models); and (V) quality assessment domains. For studies reporting multiple effect estimates for the same risk factor (e.g., unadjusted and adjusted), multivariate-adjusted estimates were preferentially extracted to minimize confounding bias. All effect estimates included in the meta-analysis were derived from multivariate logistic regression models. For one study [Dong 2025 (24)] that entered LDH as a continuous variable in the regression model (OR per unit increase), the LDH effect estimate was excluded from the LDH pooled analysis due to scale incompatibility with dichotomized ORs from other studies; the study’s other risk factor estimates were retained.

Risk of bias assessment

Methodological quality was assessed using the Newcastle-Ottawa Scale (NOS) for observational studies (25). This validated instrument evaluates three domains: selection of study groups (0–4 points), comparability of groups based on design and analysis (0–2 points), and ascertainment of exposure/outcome (0–3 points), yielding a maximum composite score of 9 points. Studies scoring ≥8 were classified as high quality, 6–7 as moderate quality, and <6 as low quality. Quality assessment was performed independently by two investigators, with discrepancies resolved by consensus. The relationship between study quality and effect estimates was explored through meta-regression where feasible.

Statistical analysis

All meta-analyses were conducted using the meta package (version 7.0-2) and metafor package (version 4.4-0) in R statistical software (version 4.5.0) (26). Effect sizes were expressed as ORs with 95% CIs. For studies reporting continuous risk factor data, standardized mean differences were converted to log-ORs using established formulae. Log-transformed ORs and their standard errors were pooled using the generic inverse variance method.

The DerSimonian-Laird random-effects model was employed to account for anticipated between-study heterogeneity, with the Hartung-Knapp adjustment applied for CI calculation to provide more conservative estimates appropriate for meta-analyses with limited numbers of studies (27). The restricted maximum likelihood (REML) method was used for τ² estimation.

Heterogeneity was assessed using Cochran’s Q statistic (with P<0.10 indicating significant heterogeneity) and quantified using the I² statistic (28). I² values were interpreted as follows: <25% (low heterogeneity), 25–50% (moderate heterogeneity), 50–75% (substantial heterogeneity), and >75% (considerable heterogeneity). Between-study variance (τ²) was reported for all analyses.

Publication bias was evaluated visually using funnel plots and statistically using Egger’s weighted regression test (29) and Begg’s rank correlation test (30). These assessments were performed for risk factor categories with ≥3 contributing studies. Trim-and-fill analysis was conducted when publication bias was detected.

Sensitivity analyses were conducted using the leave-one-out method to assess the robustness of pooled estimates and identify potentially influential studies. Subgroup analyses were performed by study design (prospective vs. retrospective) and external validation status (validated vs. not validated) where sufficient data permitted. Statistical significance was established at P<0.05 for all analyses (two-tailed).

Results

Study selection

The systematic search identified 996 records across five electronic databases and supplementary sources (PubMed: n=312; Embase: n=245; Web of Science: n=198; Scopus: n=156; Cochrane Library: n=42; citation searching and grey literature: n=43). Following removal of 247 duplicate records, 749 unique records underwent title and abstract screening.

Of these, 663 records were excluded for not meeting eligibility criteria. Primary reasons for exclusion included: irrelevance to the research question (n=412), review articles or meta-analyses (n=89), case reports or case series with insufficient sample size (n=76), non-English language without available translation (n=54), and conference abstracts without full-text availability (n=32).

Full texts of 86 potentially eligible articles were sought for detailed evaluation, of which 9 could not be retrieved (full text unavailable: n=8; retracted: n=1). The remaining 77 articles underwent comprehensive eligibility assessment against predetermined criteria. Sixty-five articles were subsequently excluded for the following reasons: no PB-specific outcome data (n=28), no extractable effect estimates or insufficient raw data (n=19), adult population (n=9), different PB definition not meeting bronchoscopic confirmation criteria (n=5), overlapping patient cohorts with other included studies (n=3), and insufficient methodological detail (n=1).

Ultimately, 12 studies fulfilled all inclusion criteria and were incorporated into the meta-analysis (24,31-41). Inter-rater agreement for full-text eligibility assessment was excellent (κ=0.89, 95% CI: 0.82–0.96). The complete study selection process is illustrated in the PRISMA flow diagram (Figure 1).

Figure 1.

PRISMA 2020 flow diagram illustrating the systematic study selection process. A total of 996 records were identified through database searching and supplementary sources. Following duplicate removal and screening, 12 studies met eligibility criteria and were included in the meta-analysis. OR, odds ratio; PB, plastic bronchitis; RR, risk ratio.

Study characteristics

The characteristics of the 12 included studies are comprehensively summarized in Table 1. All studies were conducted in China and published between 2021 and 2025, reflecting the high regional incidence of MP infection and established bronchoscopy practices. Study designs comprised 9 retrospective cohort studies (75.0%), 2 case-control studies (16.7%), and 1 prospective cohort study (8.3%). Two studies (32,36) incorporated external validation cohorts, enhancing the generalizability of their findings.

Table 1. Characteristics of included studies.

Study	Year	Design	Period	Total N	PB, n (%)	Age (y)	NOS
Zhong et al. (31)	2021	Case-control	2018–2020	296	42 (14.2)	5.2	7
Zhao et al. (32)	2022	Retrospective cohort	2019–2021	547	190 (34.7)	5.8	8
Yang et al. (33)	2023	Prospective cohort	2020–2022	1,169	133 (11.4)	4.9	8
Zhang et al. (34)	2023	Retrospective	2019–2022	120	52 (43.3)	5.4	7
Ma et al. (35)	2025	Retrospective cohort	2022–2024	126	32 (25.4)	5.1	7
Mu et al. (36)	2025	Cohort + validation	2021–2023	981	284 (29.0)	5.6	8
Dong et al. (24)	2025	Case-control	2023–2024	115	74 (64.3)	5.0	6
Jiang et al. (37)	2025	Retrospective cohort	2022–2024	273	89 (32.6)	5.3	7
Zheng et al. (38)	2025	Retrospective cohort	2023–2024	62	62 (100.0)	4.8	6
Wang et al. (39)	2025	Retrospective	2022–2024	75	28 (37.3)	5.5	6
Li et al. (40)	2025	Retrospective cohort	2022–2024	361	98 (27.1)	5.2	7
Chen et al. (41)	2025	Retrospective cohort	2022–2024	281	39 (13.9)	5.0	6

All studies were conducted in China with bronchoscopic confirmation of PB. NOS, Newcastle-Ottawa Scale; PB, plastic bronchitis; y, years.

Sample sizes ranged from 62 to 1,169 participants, with a cumulative enrollment of 4,406 children across all studies. The total number of PB cases was 1,123, with study-specific PB incidence ranging from 11.4% to 64.3% (median: 29.0%) among children undergoing bronchoscopy. This variability in PB incidence likely reflects differences in bronchoscopy indication thresholds and population severity across centers.

The mean age of participants ranged from 4.8 to 5.8 years across studies, with male proportions ranging from 50.0% to 55.6%. All studies utilized bronchoscopic visualization and extraction of bronchial casts as the diagnostic criterion for PB, ensuring outcome standardization. SMPP/RMPP definitions varied across studies, with the majority adopting Chinese National Pediatric Guidelines criteria requiring persistent fever ≥39 °C for ≥5 days or fever persistence ≥7 days without improvement despite appropriate macrolide therapy.

Risk of bias assessment

Quality assessment results utilizing the NOS are presented in Figure 2. Four studies (33.3%) attained high-quality ratings (NOS ≥8 points) (32,33,36,37), and eight studies (66.7%) demonstrated moderate quality (NOS 6–7 points) (24,31,34,35,38-41). No study scored below 6 points. Domain-specific NOS scores for all included studies are provided in Table S1. Common methodological strengths included clear PB case definitions with bronchoscopic confirmation (Selection domain) and use of multivariate regression for confounder adjustment (Comparability domain). The principal limitations were retrospective study designs and incomplete reporting of follow-up adequacy (outcome domain).

Figure 2.

Quality assessment of included studies using the NOS. Studies were evaluated across three domains: selection (0–4 points), comparability (0–2 points), and outcome ascertainment (0–3 points). Four studies achieved high quality (≥8 points), seven moderate quality (6–7 points), and one low quality (<6 points). NOS, Newcastle-Ottawa Scale.

Meta-analysis of risk factors

A total of 38 effect sizes were extracted from 12 studies across 8 risk factor categories. The pooled results from random-effects meta-analysis are comprehensively summarized in Table 2 and graphically depicted in the combined forest plot (Figure 3). The distribution of effect sizes across studies is visualized in the risk factor heatmap (Figure S1).

Table 2. Summary of meta-analysis results for risk factors of plastic bronchitis.

Risk factor	k	Pooled OR	95% CI	P value	I2 (%)	τ2
Fever duration	4	3.70	1.12–12.26	0.040	47.5	<0.01
LDH elevation	5	4.06	2.69–6.11	0.002	0.0	0.00
NLR ≥3.30	2	2.70	0.89–8.18	0.056	0.0	0.00
D-dimer elevation	3	2.63	2.01–3.43	0.004	0.0	0.00
Pleural effusion	4	2.46	1.55–3.91	0.009	0.0	0.00
Atelectasis	4	2.32	1.61–3.34	0.005	0.0	0.00
Consolidation ≥2/3	2	1.79	1.36–2.37	0.02	0.0	0.00
PCT elevation	2	1.34	0.04–45.49	0.48	77.8	0.13

Random-effects model with Hartung-Knapp adjustment. CI, confidence interval; k, number of studies; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; OR, odds ratio; PCT, procalcitonin; τ², between-study variance.

Figure 3.

Combined forest plot displaying pooled ORs and 95% CIs for all risk factors of plastic bronchitis. Random-effects model with Hartung-Knapp adjustment. Diamond represents pooled effect estimate; square size proportional to study weight; horizontal lines indicate 95% CIs. Vertical dashed line at OR =1 represents null effect. CI, confidence interval; LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; OR, odds ratio; PCT, procalcitonin; REML, Restricted Maximum Likelihood.

Clinical risk factors

Fever duration: four studies (24,31,34,38) evaluated prolonged fever duration as a risk factor for PB, employing cut-off values ranging from >6 to >12 days. The pooled OR was 3.70 (95% CI: 1.12–12.26, P=0.040), indicating that children with extended fever duration exhibited significantly elevated odds of developing PB (Figure 4). Moderate heterogeneity was observed (I²=47.5%, τ²<0.01, Q P=0.12), likely attributable to variation in fever duration cut-off definitions across studies.

Figure 4.

Forest plot for prolonged fever duration as a risk factor for plastic bronchitis. Pooled OR =3.70 (95% CI: 1.12–12.26), P=0.040. Moderate heterogeneity observed (I2=47.5%). Cut-off definitions ranged from >6 to >12 days across studies. CI, confidence interval; HK, Hartung-Knapp; OR, odds ratio.

Inflammatory biomarkers

LDH

Five studies (24,31,32,34,36) examined the association between LDH elevation and PB. Upon examination of the extracted effect estimates, one study [Dong 2025 (24)] entered LDH as a continuous variable in the multivariate logistic regression model, yielding a per-unit OR (OR =1.006 per U/L increase, 95% CI: 1.000–1.011). This effect estimate operates on a fundamentally different scale from the dichotomized ORs reported by the remaining four studies, which employed specific LDH cut-off values (range, 451–482 U/L or tertile-based categorization). Consequently, the per-unit OR was excluded from the pooled analysis to ensure scale compatibility. The revised pooled analysis of four studies (31,32,34,36) yielded OR =4.06 (95% CI: 2.69–6.11, P=0.002), demonstrating a strong and statistically significant association between elevated serum LDH and PB development (Figure 5). No heterogeneity was detected (I²=0%, τ²=0.00, Q=1.36, P=0.71), confirming that the previously observed considerable heterogeneity (I²=92.6%) in the original five-study analysis was entirely attributable to the inclusion of the scale-incompatible continuous effect estimate.

Figure 5.

Forest plot for LDH elevation as a risk factor for plastic bronchitis. Revised analysis (k=4; one study reporting per-unit continuous OR excluded due to scale incompatibility). Pooled OR =4.06 (95% CI: 2.69–6.11), P=0.002. No heterogeneity detected (I2=0%). CI, confidence interval; HK, Hartung-Knapp; LDH, lactate dehydrogenase; OR, odds ratio; RE, Random-effects; REML, Restricted Maximum Likelihood.

D-dimer

Three studies (35,38,41) examined D-dimer elevation as a risk factor. The pooled OR was 2.63 (95% CI: 2.01–3.43, P=0.004), with no detectable heterogeneity (I²=0%, τ²=0.00) (Figure 6). This homogeneity suggests consistent methodology and cut-off values across contributing studies.

Figure 6.

Forest plot for D-dimer elevation as a risk factor for plastic bronchitis. Pooled OR =2.63 (95% CI: 2.01–3.43), P=0.004. No heterogeneity detected (I²=0%), indicating consistent methodology across contributing studies. CI, confidence interval; HK, Hartung-Knapp; OR, odds ratio.

NLR

Two studies (36,40) reported NLR as a risk factor using a cut-off of ≥3.30. The pooled OR was 2.70 (95% CI: 0.89–8.18, P=0.056), which did not attain statistical significance at the α=0.05 level. However, the point estimate suggests a potentially clinically meaningful association warranting investigation in larger studies.

PCT

Two studies (39,40) evaluated PCT elevation. The pooled OR was 1.34 (95% CI: 0.04–45.49, P=0.48), with considerable heterogeneity (I²=77.8%, τ²=0.13). The extremely wide CI and non-significant result preclude definitive conclusions regarding PCT’s predictive value.

Radiological risk factors

Radiological features demonstrated the most consistent associations with PB across all included studies, uniformly exhibiting low heterogeneity (I²=0%).

• ❖ Pleural effusion: four studies (33-36) reported pleural effusion as a risk factor. The pooled OR was 2.46 (95% CI: 1.55–3.91, P=0.009), with no heterogeneity (I²=0%, τ²=0.00) (Figure 7).

• ❖ Atelectasis: four studies (32,35,36,41) examined atelectasis as a risk factor. The pooled OR was 2.32 (95% CI: 1.61–3.34, P=0.005), with no heterogeneity (I²=0%, τ²=0.00) (Figure 8).

• ❖ Consolidation: two studies (32,36) reported large consolidation (≥2/3 lobe involvement) as a risk factor. The pooled OR was 1.79 (95% CI: 1.36–2.37, P=0.02), with no heterogeneity (I²=0%, τ²=0.00) (Figure 9).

Figure 7.

Forest plot for pleural effusion as a risk factor for plastic bronchitis. Pooled OR =2.46 (95% CI: 1.55–3.91), P=0.009. No heterogeneity detected (I²=0%). CI, confidence interval; HK, Hartung-Knapp; OR, odds ratio.

Figure 8.

Forest plot for atelectasis as a risk factor for plastic bronchitis. Pooled OR =2.32 (95% CI: 1.61–3.34), P=0.005. No heterogeneity detected (I²=0%). CI, confidence interval; HK, Hartung-Knapp; OR, odds ratio.

Figure 9.

Forest plot for consolidation (≥2/3 lobe involvement) as a risk factor for plastic bronchitis. Pooled OR =1.79 (95% CI: 1.36–2.37), P=0.02. No heterogeneity detected (I²=0%). CI, confidence interval; HK, Hartung-Knapp; OR, odds ratio.

Heterogeneity assessment

Heterogeneity varied substantially across risk factor categories, as illustrated in Figure 10. Radiological risk factors (pleural effusion, atelectasis, consolidation) and D-dimer demonstrated excellent consistency across studies with uniformly low heterogeneity (I²=0%). In contrast, inflammatory biomarkers exhibited variable heterogeneity. Following exclusion of a scale-incompatible continuous effect estimate, LDH demonstrated low heterogeneity (I²=0%) consistent with the radiological risk factors. PCT also demonstrated considerable heterogeneity (I²=77.8%), limiting the interpretability of pooled estimates. Fever duration exhibited moderate heterogeneity (I²=47.5%), attributable to varying cut-off definitions (>6 to >12 days).

Figure 10.

Heterogeneity assessment across risk factor categories displaying I² statistics. Horizontal dashed lines indicate thresholds for low (<25%), moderate (25–50%), substantial (50–75%), and considerable (>75%) heterogeneity. Radiological risk factors demonstrated uniformly low heterogeneity. PCT showed considerable heterogeneity (I²=77.8%). Following exclusion of a scale-incompatible effect estimate, LDH demonstrated low heterogeneity (I²=0%). LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; PCT, procalcitonin.

Publication bias assessment

Publication bias was systematically assessed for risk factor categories with ≥3 contributing studies. Funnel plots are presented in Figure 11, and statistical test results are summarized in Table 3.

Figure 11.

Funnel plots for publication bias assessment. Panels show: (A) LDH elevation, (B) fever duration, (C) pleural effusion, (D) atelectasis, (E) D-dimer elevation. Vertical axis represents SE; horizontal axis represents log OR. Borderline asymmetry for revised LDH analysis (Egger’s P=0.09) and fever duration (P=0.056). No significant asymmetry detected for other risk factors. LDH, lactate dehydrogenase; OR, odds ratio; SE, standard error.

Table 3. Publication bias assessment.

Risk factor	k	Egger’s intercept	Egger’s P	Begg’s P	Interpretation
LDH elevation	5	3.43	0.02	0.62	Bias suspected
Fever duration	4	3.02	0.056	0.17	Borderline
Pleural effusion	4	1.55	0.40	>0.99	No bias
Atelectasis	4	2.01	0.29	0.17	No bias
D-dimer	3	−3.83	0.82	–	No bias

Egger’s test P<0.05 indicates potential publication bias. k, number of studies.

Following revision of the LDH analysis (exclusion of scale-incompatible continuous OR), Egger’s regression test for the revised LDH analysis (k=4) yielded borderline results (intercept=−0.054, P=0.09). While this approaches significance, interpretation is limited by the small number of studies. Borderline asymmetry was observed for fever duration (intercept=3.02, P=0.056). No significant publication bias was detected for radiological risk factors (pleural effusion, atelectasis) or D-dimer (all P>0.10).

Sensitivity analysis

Leave-one-out sensitivity analysis was performed for LDH elevation, the risk factor demonstrating the highest heterogeneity. Results are presented in Figure 12 and detailed in Table 4.

Figure 12.

Leave-one-out sensitivity analysis for LDH elevation. Upper panel: original five-study analysis identifying Dong 2025 as the source of heterogeneity (I² drops from 92.6% to 0% upon exclusion). Lower panel: revised four-study analysis demonstrating consistent results across all iterations (I²=0%). LDH, lactate dehydrogenase.

Table 4. Leave-one-out sensitivity analysis for LDH elevation.

Study excluded	Pooled OR	95% CI	I2 (%)	Interpretation
None (all studies)	4.06	2.69–6.11	0.0	Reference estimate
Zhong et al. (31)	2.54	0.95–6.81	94.0	Direction stable
Zhao et al. (32)	3.12	1.08–9.01	93.8	Direction stable
Zhang et al. (34)	2.65	0.93–7.55	93.9	Direction stable
Mu et al. (36)	2.91	0.89–9.52	94.2	Direction stable
Dong et al. (24)	3.01	1.12–8.09	93.5	Direction stable

Leave-one-out analysis excludes each study iteratively and recalculates pooled estimates. CI, confidence interval; LDH, lactate dehydrogenase; OR, odds ratio.

In the original five-study LDH analysis (prior to exclusion of the continuous effect estimate), leave-one-out sensitivity analysis revealed that exclusion of Dong 2025 reduced I² from 92.6% to 0% (OR =4.06, 95% CI: 2.69–6.11), while exclusion of any other study maintained I²>86%. This pattern identified Dong 2025’s scale-incompatible effect estimate as the sole source of heterogeneity, providing the methodological justification for its exclusion from the revised pooled analysis. For the revised LDH analysis (k=4), leave-one-out sensitivity analysis demonstrated robust and consistent results: point estimates ranged from 3.38 to 4.42, and I² remained 0% across all iterations, confirming the stability of the pooled estimate. For fever duration (k=4), leave-one-out analysis identified Zhang 2023 (OR =80.21) as an influential study; its exclusion reduced I² from 47.5% to 0% (OR =3.25, 95% CI: 2.01–5.26). The extreme OR in Zhang 2023 likely reflects the unique definition of “persistent fever” in that study. However, we retained this study in the primary analysis as the definition was clinically valid, and we report this sensitivity finding for transparency.

Subgroup analysis

Subgroup analysis stratified by external validation status was performed for LDH, as presented in Figure 13. Studies incorporating external validation cohorts (32,36) yielded a pooled OR of 3.15 (95% CI: 0.98–10.12), whereas non-validated studies demonstrated a pooled OR of 2.56 (95% CI: 0.87–7.54). The test for subgroup interaction was not statistically significant (Q=0.17, P=0.68), indicating that validation status did not substantially modify the association between LDH and PB risk. Additionally, subgroup analysis by study design was performed for risk factor categories where multiple study designs contributed data (Table S2). For the revised LDH analysis (k=4), retrospective cohort/retrospective studies (k=3) yielded OR =4.00 (95% CI: 2.04–7.86), consistent with the overall pooled estimate. For atelectasis and D-dimer, all contributing studies employed cohort designs, precluding design-based subgroup comparison. Where subgroup analysis was feasible, no significant interaction between study designs was observed, indicating that study design did not substantially modify the estimated associations.

Figure 13.

Subgroup analysis for LDH elevation stratified by external validation status. Studies with validation cohorts (n=2) yielded OR =3.15; non-validated studies (n=3) yielded OR =2.56. Test for subgroup interaction not significant (P=0.68). CI, confidence interval; HK, Hartung-Knapp; LDH, lactate dehydrogenase; OR, odds ratio; SE, standard error.

Summary of evidence

A comprehensive summary integrating effect sizes, heterogeneity metrics, evidence base, and statistical significance across all risk factor categories is presented in the meta-analysis dashboard (Figure 14). The strongest evidence (significant association with low heterogeneity) was observed for D-dimer elevation and radiological features. Strong evidence (significant association with low heterogeneity after exclusion of a scale-incompatible effect estimate) was observed for LDH. Moderate evidence (significant association with moderate heterogeneity) was observed for fever duration. Insufficient evidence was identified for NLR and PCT due to limited study numbers or non-significant pooled estimates.

Figure 14.

Comprehensive meta-analysis summary dashboard integrating: (A) pooled effect sizes with 95% confidence intervals for all risk factors; (B) heterogeneity metrics (I2) by risk factor category; (C) evidence base (number of contributing studies) per risk factor; (D) statistical significance indicators. This figure provides an at-a-glance overview of all meta-analysis results. LDH, lactate dehydrogenase; NLR, neutrophil-to-lymphocyte ratio; NS, not significant; OR, odds ratio; PCT, procalcitonin; SMPP, severe Mycoplasma pneumoniae pneumonia.

Discussion

Summary of principal findings

This systematic review and meta-analysis represents the first comprehensive synthesis of evidence addressing risk factors for PB in children with SMPP. Through rigorous analysis of 12 studies comprising 4,406 pediatric participants, we identified six statistically significant risk factors of PB: elevated LDH (OR =4.06), prolonged fever duration (OR =3.70), elevated D-dimer (OR =2.63), pleural effusion (OR =2.46), atelectasis (OR =2.32), and large consolidation (OR =1.79). Radiological risk factors demonstrated the most consistent associations across studies (uniformly I²=0%), whereas inflammatory biomarkers exhibited variable heterogeneity patterns.

Comparison with existing evidence

Our findings substantially extend and corroborate previous reports regarding risk factors for complicated MPP in pediatric populations. The association between prolonged fever duration and adverse outcomes is well-documented in the literature; fever duration exceeding 7–10 days has been consistently identified as a clinical hallmark of RMPP and is incorporated into established diagnostic criteria (42,43). Our pooled estimate (OR =3.70) quantitatively confirms that extended fever duration specifically predicts PB development, likely reflecting sustained inflammatory activity creating conditions favorable for cast formation. The moderate heterogeneity observed (I²=47.5%) is attributable to variation in fever duration cut-off definitions across studies (range: >6 to >12 days), highlighting the need for standardized thresholds in future research.

The prominent role of LDH as a risk factor warrants particular attention. LDH is a cytoplasmic enzyme released upon cellular membrane disruption, and serum elevations reflect tissue injury severity across diverse pathological conditions (44). In the context of MPP, elevated LDH indicates extensive airway epithelial damage, which provides the substrate for fibrin deposition and subsequent cast formation (45). In the revised analysis (k=4, excluding one study that reported LDH as a continuous per-unit variable), the pooled OR of 4.06 (95% CI: 2.69–6.11) with no heterogeneity (I²=0%) establishes LDH elevation as the strongest and most consistent risk factor for PB identified in this meta-analysis. The absence of heterogeneity across four studies employing different LDH cut-off values (451–482 U/L or tertile-based categorization) and conducted at different centers strengthens confidence in this finding. Notably, two studies incorporating external validation cohorts (32,36) yielded consistent results, further supporting the robustness of this association.

The strong and homogeneous association between radiological features and PB (pleural effusion OR =2.46, atelectasis OR =2.32, consolidation OR =1.79; all I²=0%) constitutes one of the most clinically actionable findings of this meta-analysis. The uniformly low heterogeneity suggests consistent radiological definitions and assessment methods across studies, enhancing the reliability of these estimates. These findings indicate that chest computed tomography performed at admission can meaningfully contribute to early risk stratification. Pathophysiologically, pleural effusion indicates inflammatory exudate accumulation and increased vascular permeability, while atelectasis may signal existing partial airway obstruction from early cast formation or mucus plugging preceding complete cast development (46). The 2023 Chinese National Guidelines for SMPP management emphasize radiological progression as a criterion for severity classification (47), consistent with our quantitative findings.

The association between D-dimer elevation and PB (OR =2.63, I²=0%) provides mechanistic insight into cast pathogenesis. Plastic bronchial casts contain substantial fibrin networks, distinguishing them pathologically from simple mucus plugs (48). Elevated D-dimer reflects ongoing fibrin formation and degradation, suggesting that coagulation cascade activation may be both a biomarker and pathogenic mechanism underlying PB development (49). The absence of heterogeneity across contributing studies strengthens confidence in this finding and suggests that D-dimer may represent a particularly reliable risk factor.

Mechanistic considerations

The identified risk factors can be interpreted within a unified pathophysiological framework. MP infection initiates a cascade of events including direct epithelial injury, robust inflammatory cytokine release, and coagulation activation (10-12). Prolonged fever duration reflects sustained systemic inflammatory response, while elevated LDH indicates extensive cellular damage. D-dimer elevation signals coagulation pathway activation with fibrin deposition. These processes collectively create the pathological conditions—epithelial injury, mucus hypersecretion, fibrin deposition, and inflammatory cell infiltration—conducive to bronchial cast formation (14).

The radiological features identified as risk factors—pleural effusion, atelectasis, and consolidation—represent anatomical manifestations of this inflammatory process. Pleural effusion indicates parapneumonic inflammation with increased vascular permeability. Atelectasis suggests airway compromise potentially from early mucus plugging or partial cast formation. Extensive consolidation reflects severe parenchymal involvement with inflammatory infiltration. The integration of inflammatory biomarkers and radiological features thus provides a comprehensive assessment of disease severity and cast formation risk.

Clinical implications and recommendations

Our findings have several important clinical implications for the management of children with SMPP/RMPP:

Risk stratification

The identified risk factors can be integrated into clinical risk stratification algorithms. A combination of prolonged fever duration (>10–12 days), elevated LDH (>300–450 U/L), elevated D-dimer, and radiological evidence of pleural effusion or atelectasis should alert clinicians to elevated PB risk. Children presenting with multiple risk factors warrant heightened surveillance and consideration of early bronchoscopic evaluation.

Timing of bronchoscopy

Current practice varies regarding bronchoscopy timing in SMPP/RMPP. Our findings support earlier bronchoscopic intervention in high-risk patients, as delayed extraction of bronchial casts is associated with prolonged hospitalization and increased complications (15,16). The presence of atelectasis on imaging, in particular, may indicate existing airway obstruction requiring evaluation.

Immunomodulatory therapy

The role of inflammatory markers in predicting PB supports consideration of anti-inflammatory therapy in high-risk patients. Systemic glucocorticoids are increasingly employed in RMPP management, with emerging evidence suggesting improved outcomes when initiated early (50,51). Children with markedly elevated inflammatory markers may particularly benefit from aggressive immunomodulatory intervention to mitigate the inflammatory milieu conducive to cast formation.

Macrolide resistance

Notably, macrolide resistance was not identified as an independent risk factor for PB in the included studies. While macrolide resistance is highly prevalent in East Asia (exceeding 80–90% in China) (52,53), our findings suggest that disease severity and host inflammatory response, rather than resistance status per se, determine PB risk. This observation is clinically relevant, indicating that PB risk assessment should focus on clinical and laboratory parameters rather than resistance testing alone.

Strengths

This study possesses several methodological strengths. First, it represents the first comprehensive meta-analysis specifically addressing risk factors for PB in pediatric SMPP, filling an important evidence gap. Second, the inclusion of 12 studies with over 4,400 participants provides substantial statistical power for detecting meaningful associations. Third, rigorous adherence to PRISMA and MOOSE guidelines ensures methodological transparency and reproducibility (22,54). Fourth, the use of random-effects models with Hartung-Knapp adjustment provides conservative CIs appropriate for meta-analyses with heterogeneous observational data. Fifth, comprehensive assessment of heterogeneity (I², τ², Q statistics), publication bias (Egger’s and Begg’s tests, funnel plots), and sensitivity analysis (leave-one-out method) strengthens confidence in the robustness of findings.

Limitations

Several limitations warrant consideration when interpreting these findings. First, all included studies originated from China, reflecting the high regional incidence of MP infection and established bronchoscopy practices. While this geographic concentration ensures population homogeneity, generalizability to other populations—particularly those with different MP epidemiology, genetic backgrounds, and healthcare practices—requires confirmation through international studies. Second, the predominance of retrospective study designs (75% of included studies) introduces potential selection and information biases. Prospective studies with standardized data collection protocols would strengthen causal inference.

Third, one study [Dong 2025 (24)] reported LDH as a continuous variable (OR per unit increase) rather than using a dichotomized threshold, necessitating its exclusion from the LDH pooled analysis to ensure scale compatibility. Future studies should clearly report both continuous and dichotomized effect estimates to facilitate meta-analytic synthesis. PCT demonstrated substantial heterogeneity (I²=77.8%), limiting the reliability of its pooled estimate. Fourth, varying definitions of SMPP/RMPP across studies introduce heterogeneity in study populations, although all studies employed bronchoscopic confirmation of PB, ensuring outcome standardization. Fifth, publication bias was detected for LDH (Egger’s P=0.02), suggesting potential overestimation of its predictive value. Sixth, the search was limited to English-language publications, potentially excluding relevant studies published in Chinese or other languages. Finally, the limited number of studies for certain predictors (NLR, PCT, consolidation: k=2) constrains the statistical power and reliability of pooled estimates for these factors. Sixth, the search was limited to five English-language databases and did not include Chinese-language databases such as CNKI or Wanfang Data. Given that all included studies originated from China, relevant studies published exclusively in Chinese may have been missed, potentially introducing language bias. Seventh, this study was not prospectively registered in PROSPERO, which is acknowledged as a limitation of the review protocol. Eighth, while our meta-analysis identifies risk factors associated with PB, the pooled ORs quantify strength of association rather than predictive accuracy. Formal evaluation of discriminative performance [e.g., area under the curve (AUC)/C-statistics] was beyond the scope of this analysis but should be addressed in future prediction model studies.

Future research directions

Several research priorities emerge from this analysis. First, prospective multicenter studies with standardized predictor definitions, cut-off values, and measurement timing would reduce heterogeneity and enhance comparability across studies. Second, development and external validation of clinical prediction models (e.g., nomograms, risk scores) integrating multiple predictors would enhance clinical utility and facilitate risk stratification at the bedside. Third, investigation of cytokine profiles (IL-6, IL-8, TNF-α) and novel biomarkers (such as surfactant proteins, matrix metalloproteinases) in adequately powered studies would provide additional mechanistic insights. Fourth, international studies conducted outside East Asia would establish generalizability across diverse populations. Fifth, research addressing the temporal dynamics of predictors—specifically, how biomarker trajectories evolve during disease progression—may identify optimal timing for risk assessment.

Conclusions

This systematic review and meta-analysis identifies elevated LDH (OR =4.06), prolonged fever duration (OR =3.70), elevated D-dimer (OR =2.63), and specific radiological features—pleural effusion (OR =2.46), atelectasis (OR =2.32), and consolidation (OR =1.79)—as risk factors significantly associated with PB in children with SMPP. LDH and radiological risk factors demonstrated the most consistent associations, with uniformly low heterogeneity (I²=0%) supporting their reliability for clinical risk assessment. The integration of these factors may inform early risk stratification and guide clinical decision-making regarding the timing of bronchoscopic intervention. Future prospective multicenter studies with standardized risk factor definitions are warranted to develop and validate clinical prediction models for implementation in routine practice.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.