Risk factors and clinical correlations of heterotopic ossification following spinal cord injury: a systematic review and meta-analysis

Abstract

Study design

A systematic review and meta-analysis.

Objective

To systematically evaluate the clinical correlates and relative risks of heterotopic ossification (HO) following spinal cord injury (SCI), and to provide an evidence-based framework for clinical HO prevention based on the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system.

Methods

Electronic literature searches were conducted across four databases: PubMed, Web of Science, Embase, and the Cochrane Library. Statistical analyses were performed using Review Manager 5.4 and R 4.5.2. Risk of bias and evidence quality were systematically assessed utilizing the Quality In Prognosis Studies (QUIPS) tool and the GRADE framework.

Results

A total of 11 studies (n = 2,647) were included in the meta-analysis. The results demonstrated that complete SCI, male sex, pressure ulcers, spasticity, and DVT were significantly associated with the occurrence of HO post-SCI (P < 0.05). Initial analysis indicated a positive correlation for UTI (P = 0.02); however, sensitivity analysis revealed insufficient robustness. Cervical versus thoracic injury level, smoking status, and pneumonia exhibited no significant associations with HO. GRADE Assessment: Evidence for male sex, spasticity, pressure ulcers, and DVT was rated as moderate-quality. Complete SCI was downgraded to low-quality evidence due to publication bias (trim-and-fill adjusted OR = 5.45, 95% CI 3.78–7.84, P < 0.0001), and evidence for UTI was rated as very low-quality.

Conclusion

Male sex, spasticity, pressure ulcers, and DVT are reliable predictors for the development of HO following SCI. Furthermore, patients with complete SCI should be monitored as a high-risk cohort. These findings suggest that the aforementioned factors can serve as critical indicators for early screening and intervention of HO. Future high-quality prospective studies are warranted to further validate these potential associations.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s13018-026-06947-5.

Introduction

Heterotopic ossification (HO) is a prevalent complication subsequent to spinal cord injury (SCI), characterized by the abnormal formation of mature lamellar bone within soft tissues, with a reported incidence ranging from 6.3 to 25.0% [1, 2]. Clinical manifestations primarily include restricted joint range of motion, joint stiffness, and neurogenic pain during the acute phase. These sequelae severely diminish patients’ quality of life and exacerbate societal and familial economic burdens [3]. Although traditional interventions, including pharmacological, physical, and radiation therapies, are available [4–7], severe cases with deep, large-scale ossification may necessitate surgical excision, the risks of which depend heavily on the extent of ossification [8, 9]. Clinical detection conventionally relies on radiological imaging, which is contingent upon the radiographic characteristics of ossification and thus exhibits inherent diagnostic lag. This impedes the early identification and prophylactic intervention of high-risk individuals, a step critical for reducing the incidence of HO [10, 11].

Recent clinical research indicates that the pathogenesis of HO is not an isolated event but is closely correlated with patients’ baseline characteristics (e.g., injury type) and acute localized or systemic complications (e.g., pressure ulcers, deep vein thrombosis [DVT], urinary tract infections [UTI]) [12, 13]. However, considerable controversy surrounds these associations. Previous studies have explored the correlations between these factors and post-SCI HO, yet most were restricted to small-sample retrospective designs. The definitive associations between specific clinical conditions (such as sustained spasticity or specific infectious exposures) and HO onset remain equivocal. Furthermore, these prior assessments often failed to account for variations in the sensitivity of diagnostic modalities or the confounding influence of injury severity (e.g., AIS grading), rendering the evidence base for clinical prevention insufficient [14–16].

Given the limitations of the existing evidence, a systematic synthesis and comprehensive evaluation of pertinent risk factors are urgently required. Compared to the 2020 study by Yolcu et al., the present study expands the sample size by 25.1% and notably includes an Asian population for the first time [17]. By incorporating the GRADE framework, this study systematically appraises the quality of evidence (moderate, low, and very low) for each identified factor [18]. In summary, through systematic review and meta-analysis, this study aims to explore and quantify the baseline predictive indicators and clinical factors associated with HO. The objective is to construct a multidimensional risk assessment profile and provide evidence-based recommendations for clinical prophylaxis, ultimately aiming to improve the quality of life for patients with HO secondary to SCI.

Methods

This study was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [19]. The PRISMA checklist encompasses 27 essential items established by an expert panel, serving as the minimum reporting standard for systematic reviews and meta-analyses. To ensure methodological transparency and feasibility, the study protocol was prospectively registered in the PROSPERO database (CRD 42025645215).

Search strategy and PICO construction

Comprehensive electronic literature searches were conducted across PubMed, Web of Science, Embase, and the Cochrane Library from their inception to January 18, 2026. A combination of MeSH terms and free-text words was employed, utilizing core terminology such as “spinal cord injury,” “heterotopic ossification,” “spinal cord trauma,” and “quadriplegia.” The research question was structured using the PICO framework: (P) Adult patients with SCI confirmed by clinical and radiological standards; (I) Exposure to specific baseline characteristics or clinical correlates (e.g., complete SCI, male sex, spasticity, pressure ulcers, DVT, UTI, etc.); (C) SCI patients without exposure to the respective factors; (O) The primary outcome was the development of HO.

Study selection and data extraction

Retrieved literature was imported into EndNote for standardized reference management. Two independent reviewers screened titles and abstracts based on predefined inclusion and exclusion criteria. Articles deemed potentially eligible (reporting cohorts with both SCI and subsequent HO incidence) were retrieved for full-text review. Subsequently, relevant variables were extracted from the included studies: first author, publication year, study design, HO diagnostic criteria, sample size, baseline characteristics (age, sex, injury severity, etc.), and the raw frequencies of exposure factors. Discrepancies during the screening and extraction phases were resolved through discussion with a third independent reviewer.

Inclusion and exclusion criteria

Inclusion criteria: (1) Adult patients diagnosed with traumatic or non-traumatic SCI confirmed by imaging (CT, X-ray, three-phase bone scan, ultrasound, etc.) and clinical standards (ISNCSCI and AIS), without confounding concomitant neurological disorders [20, 21]; (2) Studies utilizing imaging to diagnose HO, stratifying patients into HO and No-HO groups; (3) Studies reporting associations between specific factors and HO development; (4) For duplicate cohorts from the same institution or author, the most recent or comprehensive study was included.

Exclusion criteria: (1) Non-human or cadaveric studies; (2) Non-English literature, reviews, or pathology reports; (3) Studies that did not report HO risk factors or lacked clear incidence rates.

Definition of relevant factors

Given the heterogeneity in the diagnostic window for HO (ranging from 14 days to 2 years), relevant variables were categorized into “baseline characteristics” and “clinical correlates” to address temporal ambiguity and potential causality. Baseline characteristics: Inherent attributes determined at the time of SCI, including sex, age, smoking status, injury level, and injury severity. Due to variations in severity assessment scales across studies, “injury severity” was dichotomized based on motor completeness (AIS-A/AIS-B) to reflect severe immobilization, which is a core physiological barrier. Clinical correlates: Factors such as DVT, UTI, and pneumonia, which predominantly occur during the acute and rehabilitative phases (3–6 months post-SCI), were classified as early inflammatory triggers for HO. Conversely, spasticity and pressure ulcers, presenting as sustained states, were considered concomitant markers reflecting a high-risk microenvironment (hypertonia/localized pressure). To ensure the homogeneity and statistical validity of the pooled data, unadjusted raw frequencies of univariate factors were extracted for analysis.

Data analysis

Data synthesis and statistical analyses were performed using Review Manager 5.4 and R 4.5.2. The raw frequencies were utilized to calculate Odds Ratios (OR) and their corresponding 95% Confidence Intervals (CI). Heterogeneity Assessment: Statistical heterogeneity across studies was evaluated using the Chi-square test and the I² statistic [22]. A fixed-effects model was applied when I² < 50% and P > 0.10; conversely, a random-effects model was employed when I² ≥ 50%, indicating significant heterogeneity [23]. Following Cochrane Handbook guidelines, sensitivity analyses were conducted by sequentially omitting individual studies to identify the primary sources of heterogeneity. Subgroup Analysis: Acknowledging the varying sensitivity of diagnostic modalities, studies utilizing MRI, CT, or ultrasound were categorized into the “sensitive diagnostic subgroup,” while those relying on X-rays or clinical signs were grouped into the “conventional diagnostic subgroup.” This was to assess the impact of diagnostic definitions on the pooled effect sizes. Publication Bias: The metabias and metasens packages in R were used to quantitatively assess funnel plot asymmetry via Egger’s linear regression test (P < 0.05 indicating bias), supplemented by the Lim-Friedrich-König (LFK) index (values < 1 denote no bias, 1–2 minor bias, and ≥ 2 major bias) [24, 25]. To address potential overestimation of effect sizes due to publication bias, Duval and Tweedie’s trim-and-fill method was applied for sensitivity correction, generating an adjusted, robust effect estimate [26].

QUIPS tool and GRADE evidence quality

Methodological quality and evidence grading were independently assessed by two reviewers using the QUIPS tool and the GRADE framework, respectively, with disagreements resolved by a third reviewer [27]. The QUIPS tool evaluates the risk of bias across six domains: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting, rating each domain as “low,” “moderate,” or “high” risk. Subsequently, the GRADE framework was applied to rate the pooled meta-analytic evidence based on study limitations, inconsistency, indirectness, and imprecision. Publication bias was verified using Egger’s test and the LFK index. By synthesizing these metrics, the overall GRADE quality of evidence was categorized as “high,” “moderate,” “low,” or “very low.”

Results

Search results

The initial literature search yielded 2,165 records from PubMed (n = 1,039), Web of Science (n = 436), Embase (n = 668), and the Cochrane Library (n = 22). After removing 612 duplicates, 1,553 articles remained. Title and abstract screening excluded 830 off-topic studies, leaving 723 articles for full-text assessment. Of these, 215 full texts could not be retrieved, resulting in 508 articles evaluated for eligibility. Following manual review, exclusions included conference reports (n = 81), non-relevant studies (e.g., non-human, non-English, basic research; n = 390), and studies with unextractable data (n = 26). Ultimately, 11 studies were included in the meta-analysis and systematic review (see Fig. 1).

Fig. 1.

PRISMA flow diagram of studies selection process

Study characteristics

This review encompasses 11 studies on HO published between 1989 and 2024. Sample sizes ranged from 16 to 547 participants, with ages spanning 21 to 90 years, across multiple countries. The timeframe from injury to HO assessment varied from the acute phase (≤ 14 days) to 2 years post-injury, with follow-up periods ranging from 5 weeks to 1 year. Diagnostic modalities including clinical examination, plain radiography, ultrasound, MRI, and CT were utilized to track the progression of HO post-SCI. Substantial variations were observed across studies regarding sample sizes, participant demographics, assessment timing, and diagnostic techniques (see Table 1).

Table 1.

Details of the included studies

Study	Country	Design	SCI etiology	HO assessment	(HO/No-HO)	Age (years)	Injury level (%)	Neurological severity*	Time to HO assessment	Follow-up
Lal [28]	USA	Case–control	Traumatic (traffic, fall)	Clinical + X-ray	48/50	< 21:25% 21–30:35% 31–40:28%	N R	N R	≤ 14 days	N R
Bravo-Payno [29]	Spain	Case–control	Traumatic (traffic, fall)	Clinical + X-ray	44/44	29.7 ± 1.4/33.3 ± 1.9	C: 38/T: 42/L: 8	66.0% (Frankel A + B)	40.8 ± 45.2 days	Max 180 days
Coelho [30]	Brazil	Case–control	Traumatic (traffic, fall)	X-ray	33/33	N R	C: 50/T: 16/ L: N R	63.6% (AIS A + B)	≤ 2 years	N R
Citak [31]	Germany	Case–control	Traumatic (traffic, fall, sports)	US → MRI/CT	132/132	43.4/49	C: 138/T: 95/L: 31	58.3% (AIS A + B)	≤ 125 days	1 years
Reznik [32]	Australia	Case–control	Traumatic (traffic, fall)	Clinical signs	16/135	31.4 ± 10.9/38.1 ± 17.5	C: 79/T: 54/L: 10	45.0% (Frankel A + B)	N R	N R
Krauss [34]	Germany	Case–control	Traumatic (traffic, fall)	US → MRI	32/547	42.6 (17–79)/48 (14–90)	N R	52.0% (AIS A)	≤ 14 days	5–22 weeks
Razavi [33]	Iran	Case–control	Traumatic (traffic, fall)	X-ray	51/84	36.9 ± 13.7/33.1 ± 12.9	C: 41/T: 84/L: 10	55.0% (AIS A)	62 ± 59 days	N R
Ohlmeier [35]	Germany	Case–control	Traumatic (traffic, fall)	MRI/CT	40/168	N R	N R	60.0% (AIS A + B)	N R	N R
Suero [36]	Germany	Case–control	Traumatic (traffic, fall)	US → CT/MRI	216/320	44.6 ± 17.4/53.0 ± 18.9	N R	61.0% (AIS A + B)	62.8 ± 37.4 days	N R
McKean [37]	UK	Observational	Traumatic (traffic, fall)	MRI	6/34	N R	C: 26/T: 14/ L:3	70.0% (AIS A + B)	81.2 ± 16.8 days	N R
Xie [1]	China	Case–control	Traumatic (traffic, fall)	X-ray + CT	151/331	41 ± 14/48 ± 15	C: 216/T: 205/L: 61	54.0% (AIS A + B)	194 ± 73 days	N R

AIS, ASIA impairment scale; HO, heterotopic ossification; N, number of participants; NR, N R; SCI, spinal cord injury.neurological Severity Harmonization: To ensure consistency across studies utilizing different classification systems (AIS vs. Frankel), injury severity was harmonized based on functional motor completeness. The “Motor Complete” category pools patients with no motor function below the neurological level (defined as AIS A/B or Frankel A/B), reflecting a state of severe immobility. The “Motor Incomplete” category includes those with preserved motor function (AIS C-E or Frankel C-E)

Injury level: Percentages represent the proportion of Cervical (C), Thoracic (T), Lumbar (L), and Sacral (S) injuries within the total study population

QUIPS quality assessment

QUIPS assessment indicated an improvement in methodological quality in more recent publications. Specifically, later studies (e.g., [1], [36], [37]) demonstrated a low risk of bias, characterized by the employment of advanced imaging (CT/MRI) and multivariable regression to adjust for confounders. Conversely, earlier investigations (e.g., [28], [29]) exhibited a higher risk of bias, particularly concerning study confounding and outcome measurement. Older studies frequently relied on univariate analyses unadjusted for injury completeness or spasticity, and utilized diagnostic modalities with lower sensitivity (such as plain radiography or clinical signs), potentially introducing detection bias (see Table 2).

Table 2.

Risk of bias assessment of included studies using the QUIPS tool

Study	Study participation	Study attrition	Prognostic factor measurement	Outcome measurement	Study confounding	Statistical analysis	Overall risk
Lal [28]	Moderate	Low	High	Moderate	High†	Moderate	High
Bravo-Payno [29]	Moderate	Moderate	High	High‡	High†	Moderate	High
Coelho [30]	Low	Low	Moderate	Moderate	Low	Low	Low
Citak [31]	Low	Low	Low	Moderate	Moderate	Low	Moderate
Reznik [32]	Moderate	Low	Moderate	High‡	High	Moderate	High
Krauss [34]	Low	Low	Low	Low	Moderate	Moderate	Moderate
Razavi [33]	High	Low	Moderate	Moderate	High	Moderate	High
Ohlmeier [35]	Low	Low	Low	Low	Moderate	Low	Moderate
Suero [36]	Low	Low	Low	Low	Low	Low	Low
McKean [37]	Low	Low	Low	Low	Low	Low	Low
Xie [1]	Low	Low	Low	Low	Low	Low	Low

QUIPS, quality in prognosis studies; Low: low risk of bias; Moderate: moderate risk of bias; High: high risk of bias. † high risk in confounding: Indicates studies that did not use multivariate analysis to adjust for key covariates (e.g., injury severity, spasticity). ‡ High risk in outcome measurement: Indicates studies relying primarily on clinical signs or plain radiographs rather than sensitive imaging (MRI/CT/Ultrasound)

Evaluation of factors correlated with HO post-SCI

Data regarding specific clinical factors were analyzed using 4 to 10 studies per variable [1, 28–37]. Subgroup analyses based on diagnostic modalities were performed: MRI, CT, and ultrasound formed the sensitive subgroup, whereas X-rays and clinical scales constituted the conventional subgroup. Pooled analysis demonstrated that complete SCI (random-effects: OR = 3.84, 95% CI 2.69–5.48, P < 0.001, I² = 56%), male sex (fixed-effects: OR = 2.46, 95% CI 1.84–3.28, P < 0.001, I² = 18%), pressure ulcers (fixed-effects: OR = 2.11, 95% CI 1.66–2.68, P < 0.0001, I² = 28%), spasticity (random-effects: OR = 1.99, 95% CI 1.40–2.85, P < 0.001, I² = 56%), DVT (fixed-effects: OR = 1.63, 95% CI 1.29–2.06, P < 0.001, I² = 8%), and UTI (random-effects: OR = 2.63, 95% CI 1.15–5.99, P = 0.02, I² = 86%) were significantly associated with the incidence of HO, showing consistent effect directions (all P < 0.05). Significant publication bias was detected solely within the complete SCI subgroup (Egger P = 0.03, LFK index = 1.69).

Cervical versus thoracic

Level Seven studies (n = 1,092) assessed the association between injury level and HO. The HO incidence was 37.0% (203/548) for cervical and 40.0% (218/544) for thoracic injuries. Subgroup analysis showed no statistical difference in either the sensitive (OR = 0.81, 95% CI 0.59–1.10, P = 0.18, I² = 0%) or conventional (OR = 1.29, 95% CI 0.81–2.06, P = 0.28, I² = 0%) groups. The overall pooled analysis revealed no significant association between injury level and HO development (fixed-effects: OR = 0.94, 95% CI 0.72–1.21, P = 0.62, I² = 0%) (see Fig. 2). No significant publication bias was identified (Egger P = 0.22, LFK = -0.93) (See Supplementary S1).

Fig. 2.

Analysis of cervical and thoracic spinal injuries on the risk of heterotopic ossification (HO)

Sex

Eight studies (n = 2,284) evaluated sex differences. HO incidence was 30.1% (549/1,822) in males and 15.4% (71/462) in females. The sensitive diagnosis subgroup showed a significant difference (OR = 2.52, 95% CI 1.87–3.40, P < 0.001, I² = 0%), whereas the conventional subgroup did not (OR = 1.72, 95% CI 0.57–5.17, P = 0.33, I² = 68%). The overall meta-analysis indicated that male sex was significantly positively associated with HO risk (fixed-effects: OR = 2.46, 95% CI 1.84–3.28, P < 0.001, I² = 18%) (see Fig. 3). No significant publication bias was observed (Egger P = 0.85, LFK = − 1.09) (see Supplementary S2).

Fig. 3.

Analysis of gender differences in the risk of heterotopic ossification (HO)

SCI completeness

Ten studies (n = 2,544) explored the association between injury completeness and HO. HO incidence in the complete injury group reached 43.5% (546/1,256), significantly higher than the 13.4% (173/1,288) in the incomplete group. Statistical significance was robust across both the sensitive (OR = 4.83, 95% CI 3.35–6.98, P < 0.001, I² = 51%) and conventional (OR = 2.38, 95% CI 1.43–3.95, P = 0.0008, I² = 0%) subgroups. The overall analysis confirmed that complete SCI is a strong correlate of HO (random-effects: OR = 3.84, 95% CI 2.69–5.48, P < 0.001, I² = 56%) (see Fig. 4). However, publication bias was present (Egger P = 0.03, LFK = 1.68) (see Supplementary S3).

Fig. 4.

Analysis of spinal cord injury (SCI) completeness and its association with heterotopic ossification (HO)

Smoking status

Four studies (n = 963) assessed smoking. HO incidence was 35.0% (139/397) among smokers versus 34.1% (193/566) among non-smokers. Neither the sensitive (OR = 1.92, P = 0.40) nor conventional (OR = 1.88, P = 0.12) subgroups showed significance. The overall analysis found no significant association between smoking and HO (random-effects: OR = 1.84, 95% CI 0.83–4.08, P = 0.13, I² = 73%) (see Fig. 5). No significant publication bias was detected (Egger P = 0.22, LFK = − 1.44) (see Supplementary S4).

Fig. 5.

Analysis of smoking status and its correlation with heterotopic ossification (HO)

Pressure ulcers

Seven studies (n = 1,574) analyzed pressure ulcers. HO incidence was 50.2% (209/416) in patients with pressure ulcers compared to 30.8% (357/1,158) in those without. Significant associations were observed in both sensitive (OR = 1.83, P < 0.001) and conventional (OR = 3.93, P < 0.001) subgroups. The overall pooled effect demonstrated a strong positive association between pressure ulcers and HO (fixed-effects: OR = 2.11, 95% CI 1.66–2.68, P < 0.0001, I² = 28%) (see Fig. 6). No significant publication bias was noted (Egger P = 0.06, LFK = 0.48) (see Supplementary S5).

Fig. 6.

Analysis of pressure ulcers and their relationship to heterotopic ossification (HO) risk

Spasticity

Eight studies (n = 1,914) investigated spasticity. HO incidence was 47.5% (252/530) in patients with spasticity and 31.8% (440/1,384) in those without. Both the sensitive (OR = 1.77, P < 0.001) and conventional (OR = 2.72, P = 0.05) subgroups indicated significance. The overall analysis showed a moderate-to-strong positive association (random-effects: OR = 1.99, 95% CI 1.40–2.85, P < 0.001, I² = 56%) (see Fig. 7). No publication bias was suggested (Egger P = 0.28, LFK = − 0.52) (see Supplementary S6).

Fig. 7.

Analysis of spasticity and its impact on heterotopic ossification (HO) risk

Deep vein thrombosis (DVT)

Seven studies (n = 1,729) evaluated DVT. HO incidence was 43.1% (229/531) in patients with DVT and 34.6% (414/1,198) in those without. Both sensitive (OR = 1.58, P = 0.0004) and conventional (OR = 2.04, P = 0.04) subgroups were statistically significant. The overall analysis revealed a significant positive association between DVT and HO (fixed-effects: OR = 1.63, 95% CI 1.29–2.06, P < 0.001, I² = 8%) (see Fig. 8). No publication bias was indicated (Egger P = 0.71, LFK = − 0.14) (see Supplementary S7).

Fig. 8.

Analysis of deep vein thrombosis (DVT) and its association with heterotopic ossification (HO)

Urinary tract infection (UTI)

Seven studies (n = 1,722) analyzed UTI. HO incidence was 43.1% (467/1,083) in patients with UTIs versus 27.5% (176/639) in those without. High heterogeneity was observed in the sensitive subgroup (OR = 3.93, P = 0.16, I² = 95%), while the conventional subgroup suggested a positive correlation (OR = 1.70, P = 0.03, I² = 0%). Overall, UTI exhibited a positive association with HO (random-effects: OR = 2.63, 95% CI 1.15–5.99, P = 0.02, I² = 86%) (see Fig. 9). No significant publication bias was detected (Egger P = 0.07, LFK = − 1.03) (see Supplementary S8).

Fig. 9.

Analysis of urinary tract infections (UTI) and their relationship to heterotopic ossification (HO) risk

Pneumonia

Four studies (n = 1,490) investigated pneumonia. All utilized sensitive diagnostics, thus precluding subgroup analysis. HO incidence was 40.7% (350/860) with pneumonia and 30.0% (189/630) without. The overall analysis demonstrated no statistically significant association between pneumonia and HO (random-effects: OR = 1.79, 95% CI 0.72–4.40, P = 0.21) (see Fig. 10). No publication bias was evident (Egger P = 0.69, LFK = − 1.23) (see Supplementary S9).

Fig. 10.

Analysis of pneumonia and its correlation with heterotopic ossification (HO) risk

Sensitivity analysis

Given the significant publication bias in the complete SCI subgroup (Egger P = 0.03, LFK = 1.68), Duval and Tweedie’s trim-and-fill method was applied. The funnel plot indicated the omission of five potentially large, positive studies. Following iterative trim-and-fill correction, the pooled OR augmented from 3.84 (95% CI 2.69–5.48) to 5.45 (95% CI 3.78–7.84), with a slight increase in heterogeneity (I² from 56 to 67%). The adjusted result remained highly significant (P < 0.0001). This robust positive association indicates that the literature did not overestimate the effect; rather, the true effect size may have been underestimated due to absent small-sample studies (see Fig. 11).

Fig. 11.

Trim-and-fill analysis of complete spinal cord injury

In the UTI analysis, substantial heterogeneity was observed (I² = 86%). Sensitivity testing identified [31] and [36] as primary sources of heterogeneity, as both included “asymptomatic bacteriuria,” whereas others strictly defined clinical UTI [38]. Upon excluding these two studies, heterogeneity decreased (I² = 48%), but the pooled OR lost significance (fixed-effects: OR = 1.16, 95% CI 0.85–1.57, P = 0.35), indicating the UTI-HO association may be heavily confounded by definitional variations (see Fig. 12).

Fig. 12.

Funnel doi of sensitivity analysis of urinary tract infection

For pneumonia, severe heterogeneity (I² = 93%) was driven primarily by [1], which defined the exposure as “community-acquired pneumonia,” contrasting with the mechanical ventilation/tracheostomy-associated pneumonia defined in other studies. After its exclusion, heterogeneity remained high (I² = 76%), yet the pooled effect became significantly positive (random-effects: OR = 2.14, 95% CI 1.10–4.16, P = 0.02), suggesting that specific pulmonary infections might correlate with HO (see Fig. 13).

Fig. 13.

Funnel doi of sensitivity analysis of pneumonia

In the smoking subgroup (I² = 73%), heterogeneity was eliminated (I² = 0%) upon excluding [1], which had uniquely categorized “former smokers” as exposed. The adjusted pooled effect revealed a significant positive correlation (fixed-effects: OR = 2.88, 95% CI 1.62–5.11, P = 0.0003), implying smoking may be associated with HO, warranting further validation (see Fig. 14).

Fig. 14.

Funnel doi of sensitivity analysis of smoking status

GRADE assessment

The GRADE framework was utilized to assess evidence quality. As all included studies were observational, the baseline rating was “low.” Criteria for downgrading included: “Imprecision” (95% CI lower limit ≤ 0.50 or upper limit ≥ 2.00) or “Severe Imprecision” (sample size < 800 or events < 300, combined with I² > 80%). Publication bias was determined via Egger’s test and the LFK index.

GRADE Results: Evidence for male sex (OR = 2.46), pressure ulcers (OR = 2.11), and DVT (OR = 1.63) was rated as moderate quality, owing to significant effect sizes, acceptable homogeneity (I² < 50%), and lack of publication bias. Spasticity (OR = 1.99) also attained a moderate quality rating due to highly consistent directional effects, despite moderate heterogeneity. Complete SCI (OR = 3.84) was downgraded to low quality due to confirmed publication bias. Cervical/thoracic injury level (OR = 0.94) was downgraded to low quality due to imprecision (CI crossing 1.00). Smoking (OR = 1.84), UTI (OR = 2.63), and pneumonia (OR = 1.79) were all rated as very low quality due to excessively wide confidence intervals and severe statistical heterogeneity (I² ≥ 70%). These structured ratings provide a scientific hierarchy for clinical early warning systems and future research prioritization (see Table 3).

Table 3.

Summary of factors and GRADE of HO after SCI

Factors	Participant	I2(%)	Model	OR (95% CI)	P	Egger P	LFK index	Publication bias	Level
Cervical versus Thoracic	1,092	0	Fixed	0.94 (0.72–1.21)	0.62	0.22	− 0.93	No	Low
Male versus Female	2,284	18	Fixed	2.46 (1.84–3.28)	< 0.001	0.85	− 1.09	No	Moderate
Completeness injury	2,544	56	Random	3.84 (2.69–5.48)	< 0.001	0.03	1.68	Yes	Low
Smoking state	963	73	Random	1.84 (0.83–4.08)	0.13	0.22	− 1.44	No	Very low
Pressure ulcers	1,574	28	Fixed	2.11 (1.66–2.68)	< 0.001	0.06	0.48	No	Moderate
Spasticity	1,914	56	Random	1.99 (1.40–2.85)	< 0.001	0.28	− 0.52	No	Moderate
DVT	1,729	8	Fixed	1.63 (1.29–2.06)	< 0.001	0.71	− 0.14	No	Moderate
UTI	1,722	86	Random	2.63 (1.15–5.99)	0.02	0.07	− 1.03	No	Very low
Pneumonia	1,490	93	Random	1.79 (0.72–4.40)	0.21	0.69	− 1.23	No	Very low

All meta-analytic results presented in this table represent original pooled estimates from subgroup analyses without sensitivity analysis adjustments. GRADE, grading of recommendations assessment, development and evaluation; HO, heterotopic ossification; OR, odds ratio; CI, ionfidence interval; DVT, deep vein thrombosis; UTI, urinary tract infection; I², heterogeneity index; P, P-value; LFK index; Egger, egger’s linear regression; lim-friedrich-könig index

Discussion

This study systematically synthesized 11 clinical trials (n = 2,647) to evaluate factors correlating with HO following SCI. The meta-analysis demonstrated that male sex, complete SCI, spasticity, DVT, pressure ulcers, and UTI significantly augment the risk of HO. Conversely, cervical/thoracic injury, smoking status, and pneumonia yielded no significant statistical correlations. Based on GRADE evaluations, male sex (OR = 2.46), pressure ulcers (OR = 2.11), DVT (OR = 1.63), and spasticity (OR = 1.99) are underpinned by moderate-quality evidence, positioning them as robust criteria for identifying high-risk individuals. Notably, although UTI was positively correlated (OR = 2.63), severe heterogeneity (I² = 86%) reduced its evidence quality to “very low,” demanding cautious clinical interpretation due to definitional discrepancies.

The “complete SCI” subgroup exhibited the most substantial effect size (OR = 3.84, P< 0.001, I² = 56%), despite significant funnel plot asymmetry indicative of publication bias (Egger P = 0.03, LFK = 1.68). After trim-and-fill sensitivity correction, the OR escalated to 5.45 (95% CI 3.78–7.84), a 41.9% amplification over the primary estimate, robustly confirming this association. Recent pathophysiological models elucidate that complete SCI triggers severe autonomic dysregulation and suppression of the HPA neuroimmune axis, disrupting critical homeostasis [39, 40]. Concurrent issues such as DVT, pressure ulcers, and repetitive muscle contractions induced by spasticity generate localized ischemia–reperfusion injury and microtrauma [41, 42]. These microenvironments precipitate inflammatory cascades, ultimately stimulating the osteogenic differentiation of mesenchymal stem cells [43–45]. Consequently, clinical vigilance for early HO signs should prioritize complete SCI patients concurrently presenting with moderate-evidence risk factors (male sex, spasticity, DVT, or pressure ulcers). This mechanistic insight explains why solitary interventions (e.g., blanket prophylactic NSAIDs) often yield marginal efficacy, whereas comprehensive nursing strategies aimed at mitigating spasticity and preventing DVT/pressure ulcers inherently limit the establishment of osteogenic microenvironments [46, 47].

Limitations of this study include the observational nature of the foundational literature, which introduces inherent selection and publication biases, thereby capping the GRADE ratings for certain factors (e.g., complete SCI, UTI). Furthermore, a scarcity of multivariable modeling across the source literature restricted the ability to isolate independent effects from confounders (such as variable rehabilitation intensities or preemptive NSAID usage). Discrepancies in the diagnostic sensitivities of X-ray, ultrasound, and CT/MRI also introduced temporal heterogeneity. Moving forward, large-scale, prospective cohort studies deploying rigorous multivariable adjustments are imperative. Future randomized controlled trials must evaluate whether aggressive modulation of these confirmed high-risk vectors (e.g., intensified spasticity management or targeted early inflammatory blockade) can effectively abrogate the clinical incidence of HO.

Conclusion

This systematic review and meta-analysis establishes the clinical correlates of HO post-SCI. Under the GRADE framework, male sex, spasticity, pressure ulcers, and DVT generated moderate-quality evidence pointing to a consistent escalation in HO risk. While complete SCI originally generated the largest magnitude of risk (OR = 3.84, P < 0.001, I² = 56%) but suffered a downgrade to low-quality evidence due to publication bias, trim-and-fill adjustments significantly reinforced the robust correlation (OR = 5.45, 95% CI 3.78–7.84, P < 0.0001, I² = 67%). These insights afford a crucial evidence base, directing clinicians to prioritize these demographic and clinical phenotypes and implement stratified, prophylactic frameworks against HO.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Yulei Xie: Conceptualization, Methodology, Software, Validation, Formal Analysis, Investigation, Data Curation, Writing—Original Draft, Visualization. Liang Xie: Equal contribution as Yulei Xie. Anli Hu, Chunyu Zhang, Jingjing Liu, Jingyi Lin, Qingsong Ren: Data Curation, Formal Analysis, Investigation. Yinxu Wang, Shan Wang: Supervision, Project Administration, Resources, Writing—Review & Editing.

Funding

This study was supported by the Ministry of Finance’s basic research business project (2023 CZ-10) and High-level talent introduction research start-up fund (2025GC016).

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

Not applicable. This study is a systematic review and meta-analysis based on published literature; no new data from human or animal subjects were collected.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.