Diaphragmatic pacing for spinal cord injury patients: impact on weaning rate – a systematic review and meta-analysis Authors: Jiahao Xiangli, Xuelei Wen, Ying Liu, Yu Liang, Xifang Liu Published: 2026 ABSTRACT AbstractBackgroundSpinal cord injury (SCI), particularly when affecting the high cervical region, frequently results in diaphragmatic paralysis and subsequent reliance on mechanical ventilation. Although life-sustaining, mechanical ventilation presents its own set of limitations and potential complications. A thorough, quantitative evaluation of diaphragmatic pacing (DP) as an alternative intervention, specifically regarding its influence on weaning success rates in SCI patients, has been notably absent. This investigation aimed to address these existing knowledge gaps.MethodsTwo independent researchers conducted a systematic search across PubMed, Embase, the Cochrane Library, and Web of Science databases, encompassing all publications from their inception until March 17, 2025. Eligibility criteria prioritized within-cohort comparative studies involving SCI patients undergoing DP, which reported weaning rate. The risk of bias was evaluated using the ROBINS-I tool. A meta-analysis was subsequently performed utilizing Cochrane Review Manager 5.4, employing a random-effects model to account for clinical and methodological heterogeneity.ResultsThe meta-analysis ultimately incorporated 6 articles. All six studies reported weaning rate after diaphragmatic pacing. The pooled weaning success rate was 75% (95% CI: 69%–81%) with no statistical heterogeneity (, P = 0.24).ConclusionDiaphragmatic pacing appears to be associated with a substantial weaning success rate among individuals with spinal cord injury. Nevertheless, these findings warrant cautious interpretation due to the prevalence of non-randomized study designs, restricted sample sizes, and inherent clinical heterogeneity across the included literature, resulting in a Very Low certainty of evidence as per GRADE assessment. Consequently, further high-quality, large-scale randomized controlled trials are imperative to substantiate these conclusions.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s12890-026-04391-z.Keywords: Diaphragmatic pacing, Spinal cord injury, Ventilator weaning, Systematic review and meta-analysis Introduction Spinal cord injury (SCI) represents a grave form of central nervous system damage. It commonly manifests as a loss of motor and sensory capabilities caudal to the point of injury, thus creating considerable challenges for patients, their families, and the wider societal framework. In recent years, with the expansion of human activities, the incidence of SCI has been gradually increasing, with global incidence rates ranging from 13 to 163.4 cases per million, and the cervical spine has been identified as the most common site of SCI, both in developed and developing countries [1]. Severe high cervical spinal cord injury commonly impairs the transmission of neural signals to the diaphragm, the primary muscle responsible for breathing, leading to phrenic nerve dysfunction, diaphragmatic paralysis, and significant respiratory compromise [2, 3]. Evidence indicates that between 36% and 83% of acute spinal cord injury patients experience respiratory complications. These complications are particularly pronounced in those with cervical spinal cord injuries, with an approximate rate of 84% in C1-C4 level injuries being associated with respiratory issues [4, 5]. These patients often cannot maintain spontaneous breathing and require reliance on invasive mechanical ventilation to sustain respiratory function. For these individuals, successful ventilator weaning is a paramount clinical objective, critical for enhancing functional independence and mitigating long-term morbidity and mortality. While traditional mechanical ventilation strategies are indispensable for life support, their prolonged application poses substantial challenges and limitations in promoting diaphragmatic functional recovery. Patients not only face recurrent problems such as ventilator-associated pneumonia [6, 7], airway injury [8], and septic shock [9], but also experience severe impairment of mobility, verbal communication, and swallowing function due to prolonged bed rest and tracheostomy [10, 11], leading to a significant decline in their quality of life. In response to these challenges, Diaphragmatic Pacing (DP) has emerged as an innovative technique to replace or assist mechanical ventilation, garnering widespread attention in recent decades. This technique involves surgically implanting electrodes to electrically stimulate the phrenic nerve or the diaphragm itself, thereby inducing rhythmic diaphragmatic contractions that simulate the process of spontaneous breathing [12, 13]. Theoretically, diaphragmatic pacing can provide negative-pressure ventilation, which is more consistent with physiological respiratory dynamics, and is expected to reduce adverse reactions caused by positive-pressure ventilation, improve pulmonary gas exchange, and potentially promote the recovery of diaphragmatic function, preventing its disuse atrophy. Furthermore, patients who successfully use diaphragmatic pacing can be freed from the constraints of ventilator tubing, significantly increasing their freedom of movement and restoring speech ability, thereby potentially improving their overall quality of life. Despite the theoretical promise and growing clinical interest, a comprehensive, quantitative assessment of diaphragmatic pacing (DP) as an alternative intervention, specifically regarding its influence on ventilator weaning success rates in SCI patients, has been notably absent in the existing literature. While the clinical effectiveness of diaphragmatic pacing in individuals with diaphragmatic dysfunction has been the subject of several clinical investigations, a recurring observation is that these studies frequently adopt single-center designs, involve small participant cohorts, and are presented as observational studies or case series. These limitations, coupled with differences in study design and inconsistent definitions of outcome indicators, contribute to heterogeneity in conclusions. For example, a previous meta-analysis by Maria et al. [13] provided crucial evidence on DP’s clinical utility for individuals with diaphragmatic dysfunction, systematically presenting outcomes such as mortality, quality of life, and hospital stay in SCI patients. However, this investigation had two significant limitations pertinent to our study: Firstly, it neither statistically analyzed nor reported the “weaning rate” for SCI patients. Secondly, a meta-analysis on SCI patient survival rate could not be performed due to data scarcity, leaving a notable void in this crucial outcome. Furthermore, while existing literature for other conditions (e.g., amyotrophic lateral sclerosis) has suggested a potential link between DP and elevated mortality, the absence of robust weaning rate data in SCI patients has made it difficult to comprehensively quantify the overall clinical benefit-risk balance of this intervention in this specific population. This uncertainty limits the standardized application and promotion of diaphragmatic pacing in clinical practice. Therefore, this systematic review and meta-analysis is precisely designed to address and deepen these aforementioned research gaps by comprehensively retrieving and synthesizing existing clinical research data on diaphragmatic pacing for spinal cord injury patients, with a particular focus on quantifying its impact on weaning rate. Specifically, our PICO-based research question is: In spinal cord injury patients requiring mechanical ventilation (Population), does diaphragmatic pacing (Intervention) compared to conventional mechanical ventilation or no pacing (Comparison) improve ventilator weaning rate. Our main objectives are: (1) to comprehensively evaluate the overall effect of diaphragmatic pacing on improving patients’ successful weaning rate. By integrating and thoroughly analyzing the current body of evidence, our goal is to furnish more robust, evidence-supported medical insights pertaining to the clinical implementation of diaphragmatic pacing in individuals with spinal cord injury, consequently facilitating informed clinical judgments. Data and methods Literature search strategy (A)Searchers: The literature search was executed by two independent researchers, who rigorously followed a pre-established search protocol, operating without any reciprocal influence. Any disagreements that arose were subsequently resolved by a third researcher. Searchers: The literature search was executed by two independent researchers, who rigorously followed a pre-established search protocol, operating without any reciprocal influence. Any disagreements that arose were subsequently resolved by a third researcher. (B)Databases: The comprehensive search encompassed several English databases: PubMed, Embase, the Cochrane Library, and Web of Science. To optimize the breadth and specificity of the search, Medical Subject Heading (MeSH) terms were utilized in conjunction with Boolean operators across all these platforms. Databases: The comprehensive search encompassed several English databases: PubMed, Embase, the Cochrane Library, and Web of Science. To optimize the breadth and specificity of the search, Medical Subject Heading (MeSH) terms were utilized in conjunction with Boolean operators across all these platforms. (C)The following terms were incorporated into the search strategy: (1) MeSH term “Diaphragm Pacing [Mesh]”; (2) “Mechanical Ventilation [Mesh]”; (3) (Title/Abstract): “DP” or “Electric Stimulation Therapy” or “Phrenic Nerve/physiology”; (4) “Respiration, Artificial” or “Ventilators, Mechanical”. The following terms were incorporated into the search strategy: (1) MeSH term “Diaphragm Pacing [Mesh]”; (2) “Mechanical Ventilation [Mesh]”; (3) (Title/Abstract): “DP” or “Electric Stimulation Therapy” or “Phrenic Nerve/physiology”; (4) “Respiration, Artificial” or “Ventilators, Mechanical”. (D)Search Timeframe: The scope of the search encompassed all published literature from its inception through March 17, 2025. Screening Platform: EndNote 21 software served to identify and eliminate duplicate entries, as well as for the blinded review of article titles and abstracts. All rationales for exclusion were comprehensively documented within a PRISMA flow diagram. Search Timeframe: The scope of the search encompassed all published literature from its inception through March 17, 2025. Screening Platform: EndNote 21 software served to identify and eliminate duplicate entries, as well as for the blinded review of article titles and abstracts. All rationales for exclusion were comprehensively documented within a PRISMA flow diagram. (E)Literature Search Approach: Both controlled vocabulary (MeSH terms) and keywords were incorporated into the search strategy. Literature Search Approach: Both controlled vocabulary (MeSH terms) and keywords were incorporated into the search strategy. The protocol for this systematic review was registered in the International Prospective Register of Systematic Reviews (PROSPERO) on August 30, 2025 (last edited September 5, 2025) under registration number CRD420251137007. The registration was retrospective, as data extraction had been completed prior to registration. This limitation is acknowledged in the Discussion. Inclusion and exclusion criteria Inclusion criteria Study Design: “Included study designs comprised cohort studies (with concurrent control groups) and self-controlled before-and-after studies (without concurrent controls). Recognizing the fundamental difference in comparator structure between these designs, we did not pool them under a comparative effect measure (e.g., risk ratio). Instead, we extracted the weaning success rate only from the diaphragmatic pacing group in each study and performed a single-arm proportion meta-analysis. This approach avoids invalid comparisons and does not require a common comparator across studies.” Study Design: “Included study designs comprised cohort studies (with concurrent control groups) and self-controlled before-and-after studies (without concurrent controls). Recognizing the fundamental difference in comparator structure between these designs, we did not pool them under a comparative effect measure (e.g., risk ratio). Instead, we extracted the weaning success rate only from the diaphragmatic pacing group in each study and performed a single-arm proportion meta-analysis. This approach avoids invalid comparisons and does not require a common comparator across studies.” Study Population: The participant group for this study was comprised of spinal cord injury (SCI) patients who required mechanical ventilation. Study Population: The participant group for this study was comprised of spinal cord injury (SCI) patients who required mechanical ventilation. Intervention: Diaphragmatic pacing (DP) was the primary intervention, with baseline measurements taken before DP serving as self-controls for post-DP assessments. Intervention: Diaphragmatic pacing (DP) was the primary intervention, with baseline measurements taken before DP serving as self-controls for post-DP assessments. Outcome Measures: For this meta-analysis, qualifying studies focused on patients with spinal cord injury (SCI). Although no explicit age criteria were established for participant eligibility (implying that investigations involving both adults and children were considered if other conditions were met), all studies ultimately selected predominantly examined adult cohorts. Inclusion necessitated that studies reported the primary outcome: weaning rate. Outcome Measures: For this meta-analysis, qualifying studies focused on patients with spinal cord injury (SCI). Although no explicit age criteria were established for participant eligibility (implying that investigations involving both adults and children were considered if other conditions were met), all studies ultimately selected predominantly examined adult cohorts. Inclusion necessitated that studies reported the primary outcome: weaning rate. Exclusion criteria Studies were deemed ineligible for inclusion if any of the subsequent criteria were met: (1) They lacked crucial data indispensable for meta-analysis (e.g., absent means, standard deviations, or sample sizes pertinent to critical outcomes). (2) They were classified as reviews, conference proceedings/abstracts, correspondence to editors, Randomized Controlled Trials (RCTs), individual case reports, case series, or alternative meta-analyses. (3) They constituted redundant publications stemming from the identical study cohort and dataset (in such instances, only the most exhaustive or most recent publication was kept). (4) Studies involving non-SCI patients (e.g., ARDS or general ICU weakness) and studies using non-implantable/extracorporeal pacing were excluded. Methodological quality and risk of bias assessment The assessment of bias in all included non-randomized intervention studies (NRSI) was conducted using the Cochrane ROBINS-I (Risk of Bias in Non-randomized Studies–of Interventions) tool, as ROBINS-I provides a more detailed and appropriate evaluation for the types of studies now included. All studies, regardless of their initial risk of bias assessment, were included in the primary quantitative analysis. For studies that received an overall rating of “critical” risk of bias, or those with a “high” risk in specific areas deemed essential for interpreting this meta-analysis’s core findings, a qualitative review was conducted to understand their potential influence and limitations. Similarly, studies exhibiting substantial flaws in key domains such as inter-group comparability (where applicable), the precision of intervention descriptions, the reliability of outcome measures, or the appropriateness of statistical analyses, were also included in the primary analysis. To assess the robustness of our findings, sensitivity analyses were subsequently performed by excluding studies identified as having a “critical” or “high” risk of bias, or those failing to meet pre-established minimum quality benchmarks. A detailed ROBINS-I summary table, including domain-level judgments and overall risk of bias ratings for each included study, is provided in Supplementary Table S2. Data collection Data extraction was carried out independently by two reviewers, who utilized a pre-tested Excel spreadsheet that had been customized based on recognized standards. The specific data points gathered included: (1) Study Identifier (Author and Year) (2) Demographic characteristics of participants (age, spinal cord injury level, and time elapsed since injury) (3) Diaphragmatic pacing regimen (details on frequency, intensity, and duration) (4) Weaning success rate. Any disagreements arising between the reviewers were settled either by mutual agreement or, when required, through the mediation of a third reviewer (κ = 0.81). Assessment of risk of bias in included studies The methodological quality and risk of bias for all included observational comparative studies (which encompass cohort studies, prospective observational studies, retrospective cohort studies, and rigorous before-and-after, self-controlled designs) were rigorously assessed using the Cochrane Risk Of Bias In Non-randomized Studies–of Interventions (ROBINS-I) tool. This tool was selected for its comprehensive framework, which is specifically designed for evaluating non-randomized intervention studies and ensures consistency across all included designs. The Cochrane ROBINS-I tool systematically assesses seven key areas of methodological bias: (1) Confounding bias (2) Bias in the selection of study participants (3) Bias in how interventions were classified (4) Bias stemming from deviations from planned interventions (5) Bias due to incomplete outcome data (6) Bias in how outcomes were measured (7) Bias arising from selective reporting of results. Each domain was systematically assigned a judgment of “low risk of bias,” “moderate risk of bias,” “serious risk of bias,” or “no information” (if insufficient data were available), based on predefined criteria. An overall risk of bias judgment was then derived for each study. Specifically, for studies lacking adequate concurrent control groups, failing to perform appropriate multivariate adjustment for known confounders, or exhibiting other critical imbalances in baseline characteristics, the ’Bias due to confounding’ domain was rated as ’serious risk of bias’ or ’high risk of bias’. Crucially, all eligible studies were included in the primary quantitative synthesis, irrespective of their assessed risk of bias. The risk of bias assessment served to inform the interpretation of the meta-analysis findings and to guide subsequent sensitivity analyses. Specifically, sensitivity analyses were planned to explore the impact of studies judged to be at “serious risk of bias” on the overall effect estimates, rather than excluding them from the initial analysis. This approach aligns with best practices in systematic reviews and meta-analyses. Data extraction and bias assessment were carried out independently by two reviewers, with any disagreements resolved through deliberation with a third reviewer. The findings were visually represented in heatmaps (Figs 1 and 2) and subsequently integrated into the GRADE evidence profile. The labels ‘Low risk of bias,’ ‘Unclear risk of bias,’ and ‘High risk of bias’ correspond to a minimal, indeterminate, and significant potential for bias, respectively. Quality assessment outcomes for the included studies “-” denotes high risk; “?” denotes unclear risk; “+” denotes low risk. Appraisal overview of the selected publications Certainty of evidence assessment (GRADE) The certainty of evidence for the primary outcome (weaning rate) was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology. This involved evaluating five domains that can lower certainty: risk of bias, inconsistency, indirectness, imprecision, and publication bias. The overall certainty of evidence was then categorized as high, moderate, low, or very low. A Summary of Findings (SoF) table was generated to present the GRADE assessment results. Outcome measures This study quantified the response of spinal cord injury (SCI) patients to diaphragmatic pacing (DP) using weaning rate as the primary outcome. Only weaning rate was quantitatively synthesized in the meta-analysis. Other outcomes, including pulmonary function parameters (e.g., Maximum Inspiratory Pressure, vital capacity) and survival rate, were not pooled due to insufficient or inconsistent reporting across the included studies. Where such outcomes are mentioned in the Discussion, they are derived from individual studies and are not part of the meta-analytic findings. Statistical inference framework The risk of bias assessment was conducted using Cochrane Review Manager 5.4 (RRID: SCR_003581). The synthesis of evidence was performed using StataMP 17 (64-bit), applying methodologies suitable for binary variable measurements. For the primary outcome of weaning rate, we performed a single-arm proportion meta-analysis. The weaning success rate (proportion) with its 95% confidence interval (CI) was estimated using a random-effects model with restricted maximum likelihood (REML) estimation for the between-study variance (). The metaprop command in StataMP 17 (64-bit) was used for this analysis. Detailed model specification: The binary outcome (weaning success: yes/no) was modeled using a generalized linear mixed-model (GLMM) with a logit link function. The between-study variance () was estimated using restricted maximum likelihood (REML) on the logit scale. The pooled proportion and its 95% confidence interval were then back-transformed to the original proportion scale using the inverse logit transformation. This approach is recommended for random-effects meta-analysis of proportions, particularly when some studies have small sample sizes or event rates close to 0 or 1. Specifically, let be the number of weaning successes in study i with sample size . Assume: where represents the random effect for study i, and is the between-study variance. The pooled proportion is . Justification for single-arm approach despite design heterogeneity: Although the included studies differ in design (cohort vs. self-controlled), all provide a measurable weaning success rate in patients who received diaphragmatic pacing. A traditional comparative meta-analysis (e.g., pooling risk ratios) would be invalid because self-controlled studies lack an independent concurrent control group, violating the assumption of independence required for ratio measures. Therefore, the single-arm proportion meta-analysis is the appropriate synthesis method. However, we acknowledge that this approach does not eliminate the inherent biases of self-controlled designs (e.g., spontaneous recovery over time), which we discuss as a limitation. Inter-study heterogeneity was evaluated using the test, with the statistical significance level fixed at α = 0.1, and the statistic. Given the substantial clinical and methodological heterogeneity across the included studies (e.g., differences in stimulation modality, injury level, study design, and outcome definitions), a random-effects model was consistently utilized for all meta-analyses. This approach accounts for both within-study and between-study variance, providing a more conservative and generalizable estimate of the effect, irrespective of the observed statistical heterogeneity ( value). To ascertain the stability of the aggregated estimates, sensitivity analyses were performed. These analyses comprised: (A) a ’leave-one-out’ validation, to assess the influence of individual studies on the overall proportion; and (B) a stratified sensitivity analysis predicated on methodological quality. For the latter approach, all eligible studies, regardless of their initial risk of bias assessment, were included in the primary quantitative analysis. For sensitivity analysis, studies were categorized based on their assessed risk of bias (e.g., distinguishing between ‘low risk of bias’ and ‘moderate/high/serious risk of bias,’ as determined by ROBINS-I evaluations). The pooled proportion for high-quality study subgroups was then re-calculated, or the effect of study quality on the comprehensive results was determined by excluding studies identified with a high risk of bias or poor methodological quality. For robustness validation, an iterative leave-one-out approach was utilized, where individual datasets were progressively excluded to assess their impact on the composite proportion. Diagnostic tests for publication bias (e.g., Egger’s test) were considered but not performed due to the limited number of included studies (n < 10), as such tests would be unreliable and underpowered. The funnel plot was interpreted solely as a descriptive visual aid. Across all inferential analyses employing two-tailed hypothesis tests, an α threshold of 0.05, adjusted using the Bonferroni method, was consistently applied. Results Literature search and selection Our initial database query generated a total of 1972 records. After a systematic process of duplicate elimination, managed with EndNote 21, and a subsequent relevance assessment of titles and abstracts, 122 publications were provisionally identified as suitable. A comprehensive full-text evaluation then confirmed that 6 articles satisfied the predetermined inclusion criteria, consequently leading to their integration into this meta-analysis (Fig. 3). Schematic representation of the literature screening process (PRISMA flow diagram) Characteristics of included studies Our ultimate analysis incorporated 6 publications, all of which explored how diaphragmatic pacing (DP) influenced weaning rate in individuals with spinal cord injury (SCI). Comprehensive details regarding all selected studies can be found in Tables 1 and 2. Table 1Basic characteristics (Part 1: design and population)StudyRegionQualityDesignN (Total)Kaufman 2020USAmediumProspective10Anthony 2005USAmediumSingle-group prosp.5 (eff.4)Amer 2008USAmediumCohort36 (enr.26)Oo 1999BritainhighRetrospective cohort107Raymond 2008USAhighProspective obs.50Andrew 2020USAhighRetrospective cohort40 (enr.37)NR Not reported, enr. enrolled, eff. effective Basic characteristics (Part 1: design and population) NR Not reported, enr. enrolled, eff. effective Table 2Basic characteristics (Part 2: demographics)StudyAge (mean±SD)Sex (M/F)WeightDisease courseKaufman 202028, NR7/3NRNRAnthony 2005NR5/0210.8 lbs3.4 yearsAmer 200831.5±14.022/4NRNROo 1999NRNRNRNRRaymond 200836±NR37/13NR5.6 yearsAndrew 202045±1629/11NRNR Basic characteristics (Part 2: demographics) Quality assessment of included studies After the initial search and subsequent screening phase, 122 articles advanced to a full-text evaluation. A thorough quality assessment then resulted in the exclusion of 116 manuscripts (details of exclusion reasons are provided in Fig. 3). This process culminated in the final selection of 6 methodologically sound studies for quantitative synthesis, as visually depicted in the PRISMA flow diagram (Fig. 3). The ROBINS-I assessment results are summarized in Supplementary Table S2. Overall, all six studies were rated as having moderate risk of bias, with no study rated as low, serious, or critical risk. Meta-analysis findings Impact of DP on weaning rate in spinal cord injury The six studies reported weaning success rates ranging from 40% (Kaufman 2020, 95% CI: 10%–70%) to 83% (Raymond 2008, 95% CI: 54%–113%, with the upper bound exceeding 100% due to back-transformation from the logit scale) [14–20]. The pooled weaning success rate was 75% (95% CI: 69%–81%), with no statistical heterogeneity (, P = 0.24). Interpretation of these results warrants caution. The observed reduction in weaning failure is substantial, but it reflects observational associations rather than definitive causal effects. The high uncertainty associated with these findings stems from the limited number of studies, relatively small sample sizes within individual studies, and the inherent clinical and methodological heterogeneity across the included literature, including differences in stimulation modality, injury level, study design, and the absence of concurrent control groups in some instances. Therefore, while DP appears to be associated with an improved weaning outcome, these findings should be interpreted prudently. Further details are presented in Fig. 4 and Table 3. Forest plot of the weaning success rate (proportion) after diaphragmatic pacing in spinal cord injury patients. A random-effects model (REML) was used. The pooled proportion was 75% (95% CI: 69%–81%), with I2 = 0% Table 3Study quality and intervention protocolsStudyPacing MethodOutcomeRegimenParametersKaufman 2020ImplantedWeaning rateNRNRAnthony 2005ImplantedWeaning rateNR20 Hz; 0.1-0.15 ms; 24-25 mAAmer 2008ImplantedWeaning rateNR0.1 ms; 20 HzOo 1999ImplantedWeaning rateNR1 HzRaymond 2008ImplantedWeaning rateNR25 mA; 100 µs; 50 ms; 20 HzAndrew 2020ImplantedWeaning rateNRNR Study quality and intervention protocols Caution regarding the narrow confidence interval: The 95% confidence interval (69%–81%) is narrow due to the very low statistical heterogeneity () and the consistent weaning rates across the six studies. However, this statistical precision should not be misinterpreted as high certainty of evidence. As shown in the GRADE assessment (Table 3), the overall certainty is very low because of the observational designs, small sample sizes (total N=230), and lack of concurrent control groups in some studies. The narrow CI reflects consistency across studies, not a large evidence base or robust causal inference (Table 4). Table 4Summary of Findings (SoF) table for diaphragmatic pacing on weaning rate in SCI patients (GRADE Assessment)OutcomeNumber of participants (studies)Pooled weaning success rate (95% CI)Certainty of evidence (GRADE)Weaning rate230 (6 studies)75% (69%–81%)Very LowGRADE assessment details: Risk of bias: Serious (all six studies non-randomized, moderate risk of bias). Inconsistency: Serious (low statistical heterogeneity I2=0%, but substantial clinical heterogeneity). Indirectness: Serious (variability in outcome definitions and comparator approaches). Imprecision: Serious (limited number of studies and small sample sizes). Publication bias: Not assessable (n=6) Summary of Findings (SoF) table for diaphragmatic pacing on weaning rate in SCI patients (GRADE Assessment) GRADE assessment details: Risk of bias: Serious (all six studies non-randomized, moderate risk of bias). Inconsistency: Serious (low statistical heterogeneity I2=0%, but substantial clinical heterogeneity). Indirectness: Serious (variability in outcome definitions and comparator approaches). Imprecision: Serious (limited number of studies and small sample sizes). Publication bias: Not assessable (n=6) Publication bias assessment To evaluate publication bias for outcome measures pertaining to weaning rate, funnel plots were generated using StataMP 17 (64-bit) and are displayed in Supplementary Figures S1. Diagnostic tests for publication bias, such as Egger’s test, were considered but were not performed due to the limited number of included studies (n = 6), as such tests would be unreliable and underpowered with fewer than 10 studies. Therefore, publication bias could not be reliably assessed, and the funnel plot was interpreted solely as a descriptive visual aid. This limitation is explicitly acknowledged in the Discussion. Leave-One-Out Analysis: To assess the influence of individual studies on the overall proportion, a ’leave-one-out’ validation was performed. Stratified Analysis by Study Quality: An additional stratified sensitivity analysis was undertaken, contingent on study quality. Following the exclusion of investigations identified with a “high” or “critical” risk of bias, the aggregated estimates were re-calculated. For the primary outcomes, both the direction and the statistical significance of the observed effect were maintained. Sensitivity analysis (excluding high-risk studies) To assess the robustness of the primary findings, we performed an additional sensitivity analysis excluding the two studies with the highest methodological concerns (Andrew 2020 and Kaufman 2020), which had multiple domains rated as “unclear” in the ROBINS-I assessment (see Supplementary Table S2). After excluding these two studies, the pooled weaning success rate among the remaining four studies (Amer 2008, Anthony 2005, Oo 1999, and Raymond 2008) was 78% (95% CI: 73%–83%), with . This result was consistent with the primary analysis (75% [69%–81%]), indicating that the overall estimate was not unduly influenced by these studies. Discussion The cervical spinal cord is the most frequent site of spinal cord injury (SCI). Serving as the primary muscle for respiration, the diaphragm receives its innervation from the phrenic nerves, which originate from the C3-C5 segments of the cervical spinal cord. Severe cervical SCI (C1-C3) can precipitate complete respiratory paralysis, thereby requiring affected individuals to depend on long-term mechanical ventilation [21, 22]. Lesions affecting the C4-C5 spinal cord segments can lead to a diminished capacity of the diaphragm and less effective respiration. Subsequent to SCI, respiratory impairment constitutes a leading contributor to adverse health outcomes and fatalities [23]. After SCI, immune function can be compromised, increasing the risk of respiratory infections, which further impairs diaphragmatic function [24]. Furthermore, prolonged mechanical ventilation itself can have detrimental effects on the diaphragm, leading to diaphragmatic atrophy and functional decline [25]. Previous studies have suggested that diaphragmatic pacing may be associated with improved weaning rates, but survival and respiratory function outcomes were not quantitatively synthesized in this meta-analysis. The meta-analysis results show that DP intervention is associated with a weaning success rate of 75% (95% CI: 69%–81%) in SCI patients. However, due to the observational nature of the included studies, causal inference cannot be made. A pertinent question is how DP achieves this improvement in the rate of ventilator liberation for SCI patients. Although not quantitatively synthesized in this meta-analysis due to limited data, several studies have reported secondary outcomes related to respiratory function and survival. Successful weaning is contingent upon the pacing system’s ability to deliver adequate tidal volumes and ultimately permit liberation from mechanical ventilation [26, 27]. Research by Onders et al. [18] demonstrated that all patients achieved or exceeded their baseline required tidal volumes after diaphragmatic pacing. Diaphragmatic pacing (DP) facilitates prolonged independence from mechanical ventilation for individuals with spinal cord injury (SCI). This capability arises from the fact that, in cases of high cervical SCI, the phrenic nerves and diaphragm, which are essential for breathing, maintain their physiological soundness and can be electrically stimulated. In essence, the pacemaker constructs a synthetic neural route to reactivate this vital bodily process [17]. Simultaneously, diaphragmatic pacing (DP) enables individuals to utilize natural negative-pressure ventilation, in stark contrast to positive-pressure mechanical ventilation. This physiologically advantageous method of breathing primarily targets the ventilation of the posterior lung lobes and may enhance pulmonary compliance. Some authors have suggested that such a mechanism could contribute to successful weaning from mechanical support in SCI patients [18], but this interpretation is speculative and not directly tested in our meta-analysis. Onders et al. [28]’s study indicated that in SCI patients with long-term ventilator dependence, the diaphragm undergoes atrophy, and muscle fibers convert from fatigue-resistant Type I to fatigue-prone Type IIb. Some authors have hypothesized that diaphragmatic pacing might reverse diaphragmatic atrophy and maintain Type I muscle fibers, but this mechanistic interpretation is speculative and not directly tested in our meta-analysis. Moreover, neuroplastic responses are evident with functional electrical stimulation, such as diaphragmatic pacing. This intervention possesses the ability to adjust the dimensions of phrenic motor neurons and subsequently alter synaptic resistance, a process that might encourage the development of alternative neural circuits around the spinal cord lesion, thereby reinstating the elaborate connections crucial for respiratory regulation [28]. Although some observational studies have suggested potential survival benefits associated with diaphragmatic pacing [27, 29–31], survival outcome could not be quantitatively synthesized in this meta-analysis due to insufficient data. Therefore, any interpretations regarding survival remain speculative and should be considered hypothesis-generating rather than conclusive. For spinal cord injury populations exhibiting similar characteristics, the imperative of mechanical ventilation is linked to a notable decrease in survival, with figures falling from 84% in non-ventilated cohorts to 33% in ventilated ones [32]. On the other hand, diaphragmatic weakness is considered highly correlated with poor prognosis; patients with the weakest diaphragms require longer mechanical ventilation and exhibit significantly higher mortality rates than those with better diaphragmatic strength [33, 34]. However, these outcomes were not quantitatively synthesized in our meta-analysis. Research by Cavka K et al. showed that brief daily diaphragmatic pacing can increase the expression of satellite cells in diaphragm muscle. Diaphragmatic pacing, through electrical stimulation, can increase diaphragmatic activity, thereby enhancing diaphragmatic strength and endurance [34]. Nevertheless, certain researchers argue that diaphragmatic pacing is not without its inherent hazards. These perils chiefly stem from the operative dangers of general anesthesia and the laparoscopic implantation of stimulating electrodes, with subcutaneous emphysema and pneumothorax being frequently encountered. Additionally, the external positioning of electrodes introduces a slight possibility of infection and potential dislodgement [35]. Concurrently, despite the evident efficacy of diaphragmatic pacing devices, they have no impact on underlying neurological damage. In such cases, if nerve damage fails to improve, patients will still experience severe motor and sensory deficits [36]. For un-improved neurological damage, rehabilitation training can be implemented after SCI patients are weaned from the ventilator, tailored to the patient’s condition, to gradually address motor and sensory issues. Although the statistical heterogeneity was low (), clinical heterogeneity across the included studies warrants careful consideration. The six studies varied substantially in several key aspects. Pacing protocols: Stimulation parameters differed across studies, including pulse frequency (ranging from 1 Hz to 20 Hz), pulse width (0.1 ms to 0.15 ms), and stimulation amplitude (24 mA to 25 mA), with some studies not reporting these parameters at all (e.g., Kaufman 2020, Andrew 2020). Patient characteristics: The level of spinal cord injury (ranging from C1 to C5), time since injury (from acute to chronic), and age distribution varied across studies, which may influence diaphragmatic responsiveness to pacing. Study design: The included studies comprised both self-controlled before-and-after designs and cohort studies with concurrent control groups, which differ in their inherent risk of bias and ability to establish causality. Outcome definition: Although all studies reported weaning success, the criteria for successful weaning were not uniformly defined across studies. These sources of clinical heterogeneity could potentially influence the observed weaning success rates, yet they could not be formally explored through subgroup analysis or meta-regression due to the limited number of studies (n=6). Therefore, while the low value indicates that the observed effect sizes were statistically consistent, the pooled estimate should be interpreted with caution, as it represents an average across clinically diverse settings. Future large-scale studies with standardized protocols and detailed reporting of patient and intervention characteristics are needed to better understand how these factors moderate the effectiveness of diaphragmatic pacing. The GRADE assessment for the primary outcome of weaning rate indicated a Very Low certainty of evidence (Table 3). This low certainty is primarily attributable to the inherent limitations of the included observational study designs, which contribute to a high risk of bias, particularly confounding. Although statistical heterogeneity was low (), the clinical heterogeneity across studies regarding pacing protocols and patient characteristics (as discussed in the limitations) further contributes to concerns about inconsistency and indirectness. Furthermore, the limited number of studies (n=6) and the inability to perform reliable statistical tests for publication bias led to concerns about imprecision and potential publication bias. These factors underscore the need for cautious interpretation of our findings and highlight the imperative for future high-quality randomized controlled trials. The narrow 95% confidence interval of the pooled weaning rate reflects the low statistical heterogeneity () and the consistency of reported rates across the six studies. However, this precision should not be misinterpreted as high certainty of the evidence. As reflected in the GRADE assessment (very low certainty), the observational design, small sample sizes, and lack of concurrent control groups in some studies substantially limit the confidence in the exact magnitude of the weaning success rate. This investigation is subject to several limitations: (1) Firstly, the primary studies incorporated into this meta-analysis were exclusively non-randomized comparative designs, which inherently carry a higher risk of bias, particularly confounding and selection bias. Furthermore, the limited number of included studies (n=6) inherently limits the statistical power and generalizability of our findings. (2) Secondly, clinical heterogeneity existed across different studies regarding diaphragmatic pacing protocols (e.g., pacing frequency, intensity, duration, implantation method) and patient baseline characteristics (e.g., level of spinal cord injury, time since injury), which could not be fully explored through subgroup analysis due to the limited number of studies. (3) For survival rate assessments, a unique methodology was employed in this investigation. This was necessitated by the absence of documented survival rate information within the control arms of the included studies. The approach involved utilizing the experimental cohort’s pre-diaphragmatic pacing period as its own control. It is imperative to highlight that this methodology deviates from conventional meta-analysis control setups, which carries the risk of introducing confounding variables and precludes the generation of standard inter-group comparative insights. (4) The included studies comprised both self-controlled before-and-after designs and cohort studies with concurrent control groups. To avoid the methodological issue of pooling these heterogeneous designs under a risk ratio framework, we adopted a single-arm proportion meta-analysis, which appropriately synthesizes the weaning success rate from all studies without requiring concurrent comparators or violating independence assumptions. (5) The relatively small number of included studies (n=6) for the primary outcome might reduce the reliability of the evidence. Moreover, this limited number precluded reliable statistical assessment of publication bias; formal tests such as Egger’s regression could not be performed, and the funnel plot served only as a descriptive visual aid. Therefore, the potential for publication bias cannot be ruled out. (6) Although we used a single-arm proportion meta-analysis to avoid inappropriate pooling of cohort and self-controlled designs, this method cannot fully correct for time-related biases (e.g., regression to the mean, natural history of recovery) present in before-after studies. Consequently, the pooled weaning rate represents an association, not a causal effect, and should be interpreted with caution. (7) The narrow confidence interval of the pooled weaning rate (69%–81%) arises from low statistical heterogeneity () rather than from large sample sizes or high methodological quality. This statistical precision does not imply clinical or evidential certainty, as reflected by the GRADE rating of very low certainty. (8) The PROSPERO registration was completed after data extraction had begun (retrospective registration), which may introduce the risk of post hoc changes to the analysis plan. However, the primary outcome and analysis methods remained unchanged from the original protocol. Strengths and contributions: This study has several strengths. To our knowledge, this is the first systematic review and meta-analysis to quantitatively synthesize the weaning success rate of diaphragmatic pacing specifically in spinal cord injury patients. By adopting a single-arm proportion meta-analysis, we were able to include both self-controlled and cohort studies, maximizing the available evidence while avoiding the methodological pitfalls of inappropriately pooling heterogeneous designs under a risk ratio framework. The use of the ROBINS-I tool for risk of bias assessment, the GRADE framework for certainty of evidence, and comprehensive sensitivity analyses (including leave-one-out and exclusion of high-risk studies) strengthens the transparency and robustness of our findings. Despite the limitations inherent to the available literature, this study provides a quantitative benchmark (75% weaning success rate) that may inform clinical decision-making and sample size calculations for future trials. Recommendations for future research: Randomized controlled trials with concurrent control groups are urgently needed to establish causality. Standardized weaning criteria should be adopted across studies to reduce clinical heterogeneity. Long-term outcomes (e.g., survival, quality of life, pulmonary function) should be systematically collected and reported. Subgroup analyses (by injury level, time since injury, pacing parameters) require larger sample sizes to be informative. Prospective registration of protocols and publication of negative results are necessary to mitigate publication bias. To summarize, existing evidence suggests that diaphragmatic pacing may be associated with weaning success, but interpretations regarding survival, mechanisms of action, and long-term outcomes are based on limited observational data and remain speculative. Nevertheless, owing to constraints such as the reliance on self-controlled designs in the analyzed studies and limited participant numbers, a prudent approach to interpreting these findings is necessary. These aspects require confirmation in future well-designed studies. Subsequent rigorous, large-scale, multicenter, randomized, double-blind, controlled clinical trials remain essential to confirm the robustness of these conclusions. Conclusion Diaphragmatic pacing appears to be associated with a substantial weaning success rate among individuals with spinal cord injury. Nevertheless, these findings warrant cautious interpretation due to the prevalence of non-randomized study designs, restricted sample sizes, and inherent clinical heterogeneity across the included literature, resulting in a Very Low certainty of evidence as per GRADE assessment. Consequently, further high-quality, large-scale randomized controlled trials are imperative to substantiate these conclusions. Supplementary Information Supplementary Material 1. Acknowledgements The authors extend their gratitude to the library staff at Xi’an Sport University for their assistance in retrieving databases and accessing full-text articles. Appreciation is also extended to the Clinical Research Center of Xi’an Honghui Hospital for providing statistical consultation. Authors’ contributions Jiahao Xiangli, Xuelei Wen, and Xifang Liu made equivalent contributions to this study and are thus designated as co-first authors. Jiahao Xiangli: Conducted investigation, managed data, contributed to writing (review & editing, methodology, original draft). Xuelei Wen: Conducted investigation, managed data, contributed to writing (review & editing, methodology, original draft). Xifang Liu: Conducted investigation, managed data, contributed to writing (review & editing, methodology, original draft), and served as the corresponding author. Ying Liu: Managed data, conducted investigation, contributed to writing (original draft). Yu Liang: Contributed to writing (review & editing, methodology). Funding No financial support was received by the author(s) for the execution of this research or the publication of this manuscript. Data availability The primary data and analytical outputs underpinning this investigation are contained within the main article or its Supplementary Material. For any additional information, direct inquiries to the corresponding author. Declarations Ethics approval and consent to participate Not applicable. This study is a systematic review and meta-analysis based on previously published literature and does not involve any new human participants or animals. Therefore, ethical approval and consent to participate were not required. Consent for publication Not applicable. This study is a systematic review and meta-analysis of previously published data, and no individual patient data, images, or videos are included. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The primary data and analytical outputs underpinning this investigation are contained within the main article or its Supplementary Material. For any additional information, direct inquiries to the corresponding author.