Prophylactic mesh reinforcement at emergency midline laparotomy: a systematic review and meta-analysis with pre-specified subgroup analysis of contaminated and dirty abdominal wounds (CDC Class III–IV) Authors: Ricardo Antonio Gonzalez Jaramillo, Lorena I. Avendaño, Andrea Melissa Briceño González, Brandon Durán Suárez, Jorge Luis Rosales Becerra, Eduardo Jordan Garcia, Carlos Javier Mata Quintero Published: 2026 ABSTRACT PurposeTo evaluate the efficacy and safety of prophylactic mesh reinforcement (PMR) at the closure of emergency midline laparotomy (EML), with a pre-specified subgroup analysis of contaminated and dirty wounds (CDC class III–IV).MethodsPRISMA 2020-compliant review (PROSPERO CRD420261324766; OSF 10.17605/OSF.IO/ZX2BG; amendments registered before data extraction). PubMed/MEDLINE, Embase, Cochrane CENTRAL, Scopus and Web of Science were searched (January 2015–March 2026). RCTs and comparative cohort studies reporting incisional hernia (IH) at ≥ 12 months were eligible; bias was assessed with ROB-2/ROBINS-I. Random-effects meta-analysis used the REML estimator with Hartung–Knapp–Sidik–Jonkman correction. A pre-specified Borenstein interaction test compared CDC class I–II with III–IV; trial sequential analysis (TSA) and GRADE were applied.ResultsFive studies (4 RCTs; 1 cohort; n = 617) were included. PMR was associated with a reduction in IH (pooled OR 0.221, 95% CI 0.122–0.402 [Knapp–Hartung]; 0.101–0.483 under the conservative truncated variant; I2 = 0%); because control-arm event rates exceeded 10%, this corresponds to an approximate RR of ~ 0.36 (0.28–0.38). The RCT-only analysis was concordant (OR 0.179, 95% CI 0.074–0.435). TSA gave an accrued information size of 617 versus a required ~ 151 (409%), interpreted only as numerical sufficiency, not firm evidence. The CDC interaction was non-significant and underpowered (z = 0.327, p = 0.743). Seroma was significantly increased (k = 4; OR 3.980, 95% CI 1.924–8.234; 1.06–14.94 under the conservative variant), whereas deep SSI was non-significant and underpowered (OR 0.991). GRADE: LOW to VERY LOW.ConclusionPMR at EML closure was associated with a consistent IH reduction across CDC classes; however, contaminated-wound evidence is sparse, the subgroup interaction is exploratory, non-significant and underpowered, and the safety profile — particularly deep SSI — remains uncharacterized. Seroma was significantly increased. Given LOW–VERY LOW certainty, these hypothesis-generating findings should not change practice and support an adequately powered multicentre RCT in CDC class III–IV wounds.Trial registration: PROSPERO CRD420261324766 (registered 23 February 2026); OSF Registrieshttps://doi.org/10.17605/OSF.IO/ZX2BG.Supplementary InformationThe online version contains supplementary material available at 10.1007/s10029-026-03807-3. Introduction Incisional hernia (IH) is one of the most common long-term complications after midline laparotomy. Cumulative incidence ranges from 10–30% in unselected surgical populations and exceeds 40% in patients with established risk factors [1, 2]. Emergency surgery is one of the highest-risk settings for IH: rates after emergency midline laparotomy (EML) range from 18 to 54%, reflecting the combined effects of wound contamination, hemodynamic instability, nutritional impairment, and the technical limitations imposed by the emergency setting on fascial closure [3, 4]. Emergency laparotomy is also an independent risk factor for fascial dehiscence, which carries a mortality of up to 20% [5, 6]. Prophylactic mesh reinforcement (PMR) at primary fascial closure is now a well-established strategy for IH prevention in elective high-risk surgery. The five-year follow-up of the PRIMA trial showed IH rates of 24.7–29.8% with PMR versus 53.4% with primary suture [7]. Two large meta-analyses confirmed this benefit: Jairam et al. (k = 12 RCTs; RR 0.35; 95% CI 0.21–0.57) [8] and Aiolfi et al. (k = 14 RCTs; RR 0.38) [9]. On this basis, the updated EHS/AHS guidelines now conditionally recommend PMR in high-risk elective scenarios [2]. The emergency and contaminated-wound setting, however, falls outside current guideline recommendations, representing a fundamental gap in the surgical evidence [2, 10]. Whether PMR can be used safely in contaminated and dirty wounds (CDC class III–IV) is a clinically urgent question. These patients have the highest absolute baseline risk of IH, so the absolute benefit of mesh placement could be greatest precisely in the population where infectious concerns about a prosthetic implant are also greatest. No prior meta-analysis has tested this question with a formal subgroup interaction analysis. The 2023 WSES ECLAPTE guidelines [11] specifically named it as the principal unresolved question in the management of EML closure [10]. Following a prospectively registered protocol (PROSPERO CRD420261324766; OSF 10.17605/OSF.IO/ZX2BG), we performed a PRISMA 2020-compliant systematic review and meta-analysis comparing PMR with suture-only closure after EML (January 2015 to March 2026). The primary IH analysis pooled all wound classes, and a pre-specified interaction test then compared CDC class I–II with class III–IV. One additional eligible RCT (Lima et al., 2020) [12] was included only for 30-day safety outcomes. We expanded the lower bound of the search from the originally registered 2020 to January 2015 (an amendment made before any data extraction), so that the dataset would cover the entire post-STITCH era [13] and capture the full contemporary evidence base. Methods Study design and registration This systematic review and meta-analysis were designed, conducted, and reported in accordance with the PRISMA 2020 statement [14], the Cochrane Handbook v6.3 [15], and AMSTAR-2 [16]. The protocol was prospectively registered in PROSPERO (CRD420261324766; 23 February 2026) and in OSF Registries (10.17605/OSF.IO/ZX2BG), and last amended on 23 March 2026 to (i) pre-specify CDC class III–IV wound contamination as the primary subgroup of interest and (ii) extend the search lower bound from 2020 back to January 2015. Both amendments were made before any data extraction. CONSORT 2010 [17] guided the methodological appraisal of each included RCT. The completed PRISMA 2020 checklist is provided as Online Resource ESM_2. In line with PRISMA 2020 transparency requirements, the registered PROSPERO title ("Prophylactic Mesh Reinforcement in High-Risk Midline Laparotomy: A Systematic Review and Meta-analysis [2020–Present]") differs from the final manuscript title in three respects: (i) "Emergency" replaces "High-Risk" to describe more precisely the operative setting we studied; (ii) the CDC class III–IV subgroup analysis was pre-specified in the 23 March 2026 amendment, before any data extraction; and (iii) the search lower bound was extended from 2020 back to January 2015 in the same amendment. No data had been extracted at the time of registration or amendment. To address the single-reviewer limitation noted in earlier versions of this work, title/abstract screening and full-text eligibility assessment were performed independently by RAG-J, BDS, and JLRB; discrepancies were resolved by consensus, with arbitration by CJMQ when required. Data extraction was carried out by RAG-J and verified cell-by-cell against the original published tables by BDS, JLRB, LIA-V, and CJMQ. RAG-J and CJMQ independently reviewed risk-of-bias assessments. All authors read and approved every extracted value and every pooled estimate. Eligibility criteria (PICO framework) Population Adults (≥ 18 years) undergoing emergency or urgent open midline laparotomy for any abdominal indication. Studies confined to hernia repair or to pediatric populations were excluded. Mixed elective/emergency studies were eligible only if emergency-setting data were separately extractable. Intervention PMR at the time of primary fascial closure, in any anatomical plane (onlay, retromuscular/sublay, retrorectus, preperitoneal, or intraperitoneal) and with any mesh material (permanent synthetic, absorbable synthetic, biosynthetic, or biologic). Comparator Primary fascial closure with suture-only technique, without mesh reinforcement. Outcomes Primary: IH at ≥ 12 months (clinical or radiological). Secondary: deep and superficial surgical site infection (SSI; ≤ 30 days), seroma (≤ 30 days), fascial dehiscence, mesh explantation, and 30-day mortality. Studies reporting only short-term outcomes (≤ 90 days) without IH at ≥ 12 months were excluded from the primary analysis but were eligible for the 30-day secondary analyses. Study designs RCTs, prospective cohort studies with a concurrent control arm, and retrospective comparative cohort studies. Non-comparative designs and conference abstracts without extractable full-text data were excluded. Date and language January 2015 – March 2026, all languages. The 2015 lower bound was selected to coincide with the publication of the STITCH trial [13] and the consequent widespread adoption of the small-bites fascial-closure technique. As noted, this lower bound was amended from the originally registered 2020, before data extraction. Information sources and search strategy Electronic searches were conducted in PubMed/MEDLINE, Embase, Cochrane CENTRAL, Scopus, and Web of Science. ClinicalTrials.gov and the WHO ICTRP were searched for ongoing trials. Reference lists of all included studies and of key systematic reviews were screened by hand. Eight journals were hand-searched for the period 2015–2026: Annals of Surgery, British Journal of Surgery, Hernia, Surgical Endoscopy, JAMA Surgery, Journal of the American College of Surgeons, World Journal of Surgery, and Langenbeck's Archives of Surgery. PubMed Boolean string (adapted for each database): ("prophylactic mesh" OR "preventive mesh" OR "mesh reinforcement" OR "mesh augmentation") AND ("midline laparotomy" OR "abdominal wall closure" OR "emergency laparotomy") AND ("incisional hernia" OR "hernia prevention") AND (2015:2026[pdat]) AND (humans[MeSH Terms]). Study selection Title and abstract screening was performed independently in Rayyan by RAG-J, BDS, and JLRB. Full-text eligibility was assessed independently by the same reviewers against all pre-specified PICO criteria; disagreements were resolved through discussion and, when needed, arbitrated by CJMQ. All full-text exclusion decisions were documented with explicit reasons (Fig. 1; Online Resource ESM_2).Fig. 1PRISMA 2020 flow diagram. Records identified through database searching (PubMed/MEDLINE: 487; Embase: 312; Cochrane CENTRAL: 156; Scopus: 234; Web of Science: 198; total 1,387) plus 47 additional records from hand-searching and trial registries (n = 1,434 total; 892 after deduplication). After title/abstract screening, 51 full texts were assessed for eligibility; 45 were excluded for various reasons. Six studies were included in the qualitative synthesis: five for the primary IH meta-analysis and Lima 2020 for secondary 30-day outcomes only PRISMA 2020 flow diagram. Records identified through database searching (PubMed/MEDLINE: 487; Embase: 312; Cochrane CENTRAL: 156; Scopus: 234; Web of Science: 198; total 1,387) plus 47 additional records from hand-searching and trial registries (n = 1,434 total; 892 after deduplication). After title/abstract screening, 51 full texts were assessed for eligibility; 45 were excluded for various reasons. Six studies were included in the qualitative synthesis: five for the primary IH meta-analysis and Lima 2020 for secondary 30-day outcomes only Data extraction and verification A standardized, pre-piloted extraction form was used. RAG-J extracted the data, and BDS, JLRB, LIA-V, and CJMQ verified every value, cell by cell, against the corresponding tables in the original published manuscripts. No values were imputed or estimated. When intention-to-treat (ITT) and per-protocol analyses were both reported, ITT was selected as the primary analysis set. A continuity correction of 0.5 was applied to all cells of studies with zero events in one arm (Mäkäräinen 2025, IH outcome). Lima et al. 2020 [12] reported only the 30-day SSI and seroma analyses; these were excluded from the primary IH analysis because the primary endpoint was fascial dehiscence at 30 days, and IH at ≥ 12 months was not reported. Analysis sets necessarily differed across studies and outcomes (e.g., ITT for Pizza and Ulutas, the 2-year clinically evaluated set for Mäkäräinen, and the as-analysed cohort for Bravo-Salva), reflecting differing follow-up windows and reporting; each extracted cell is documented in ESM_1, and the RCT-only and leave-one-out sensitivity analyses confirmed that this did not materially alter the pooled estimates. Risk of bias Risk of bias in RCTs was assessed using the revised Cochrane ROB-2 tool [18], which evaluates five domains: D1 randomization process; D2 deviations from intended interventions; D3 missing outcome data; D4 measurement of the outcome; D5 selection of the reported result. Bravo-Salva 2021 was assessed using ROBINS-I [19] across its bias domains (summarised in Table 2). Lima 2020 was assessed using ROB-2 for the secondary outcomes it included. Assessments were performed independently by RAG-J and CJMQ; disagreements were resolved by consensus. Statistical analysis All meta-analyses used random-effects models with the REML estimator. The Hartung–Knapp–Sidik–Jonkman (HKSJ) variance correction was applied to all confidence intervals, using t-distribution critical values with df = k − 1 [20]. Prediction intervals (PIs) were computed with t(df = k − 2) [21]. Effect estimates are reported as odds ratios (OR), computed with the Woolf log-OR formula; a 0.5 continuity correction was applied to zero-event arms. Heterogeneity was quantified by Cochran's Q and I2, and interpreted with caution at k < 10 [22]. Because the Hartung–Knapp adjustment can produce confidence intervals narrower than the conventional random-effects interval when between-study heterogeneity is low (Q/(k − 1) < 1), all primary and safety estimates were additionally computed with the truncated (modified) Hartung–Knapp variant as a sensitivity analysis, and both intervals are reported. A pre-specified RCT-only sensitivity analysis (k = 4; excluding Bravo-Salva 2021) was performed for the primary IH outcome. The pre-specified subgroup interaction comparing CDC class I–II with class III–IV was tested with the Borenstein z-test for the difference in subgroup log-ORs (fixed-effect pooling within each subgroup; normal-distribution z for the interaction) [23]; this approach was chosen a priori over the Cochran Q-between test because of its better small-sample behavior at k ≤ 5. A pre-specified exploratory meta-regression of log-OR on follow-up duration is presented in Supplementary Fig. S8. Influence diagnostics included leave-one-out analysis, Cook's-distance approximation, and the Baujat plot (Supplementary Figs. S1–S4). A funnel plot was generated for completeness; formal asymmetry testing was not performed at k < 10 [22]. Trial sequential analysis (TSA) was performed using the method of Wetterslev et al. [24], with α = 0.05 (two-sided), power = 80%, control-arm event rate (CER) = 26%, and relative risk reduction (RRR) = 65%. All analyses were performed in R v4.3.0, using the meta (v6.5–0) and metafor (v4.4–0) packages [Online Resource ESM_1]. The certainty of evidence was assessed with the GRADE approach [25]. Results Study selection The study selection process is depicted in Fig. 1 (PRISMA 2020 flow diagram). After deduplication, 892 records were screened by title and abstract. Fifty-one full-text articles were assessed for eligibility; 45 were excluded with reasons documented in Online Resource ESM_2. Lima et al. 2020 [12] met eligibility criteria only for 30-day safety outcomes and were therefore excluded from the primary IH analysis. Five studies [26–30] met all eligibility criteria for the primary IH meta-analysis. Study characteristics Characteristics of the included studies are summarised in Table 1. All five primary-analysis studies enrolled adults undergoing EML. Four were RCTs [27–30], and one was a retrospective comparative cohort [26]. Wound-contamination profile is the central clinical axis of the analysis. Three studies enrolled patients with CDC class III–IV wounds: Bravo-Salva 2021 [26] (retrospective cohort; mixed emergency indications, Spain; median follow-up 64.4 months), Mäkäräinen 2025 (PREEMER) [27] (RCT; mixed contamination; Finland; 26.3 months), and Ulutas 2023 [30] (RCT; mixed contamination, including dirty wounds; Turkey; 13 months). Two enrolled exclusively clean or clean-contaminated procedures (CDC class I–II): Pizza 2021 (PROMETHEUS) [28] (RCT; CDC II only; Italy; 24 months) and Nageeb 2021 [29] (RCT; CDC I–II; Egypt; 24 months).Table 1Study characteristics — primary IH analysis (k = 5) and Lima 2020 (secondary outcomes)Study (ref.)DesignCountryN mesh/ctrlCDC classMesh & planeFollow-upAnalysis setBravo-Salva 2021 [26]Retrospective cohortSpain56/131Mixed (III–IV)Onlay; Ultrapro PP/absorbableMedian 64.4 mo (SD 35.2)As analysedMäkäräinen 2025 PREEMER [27]RCTFinland35/37 (22/28 at 2 y)Mixed (~ 34% CDC III + IV; ~ 17% CDC IV)Retrorectus; Progrip 8 cm polyester26.3 ± 3.6 moEvaluated at 2 yPizza 2021 PROMETHEUS [28]RCTItaly100/100CDC II onlySublay; Parietex ProGrip polyester24 mo (fixed)ITT (n = 100/arm)Nageeb 2021 [29] ††RCTEgypt38/34CDC I–II onlyOnlay: Prolene polypropylene24 mo (fixed)Analysed cohortUlutas 2023 [30]RCTTurkey54/54 (50/51 PP)Mixed (incl. III–IV; dirty ~ 28%)Onlay; Ultrapro 10 × 10 cm PP13.1 ± 5.6 moITTLima 2020 [12] †RCTBrazil63/52Mixed emergency (77% colorectal)Onlay; polypropylene30 dSecondary outcomes only† Lima 2020: excluded from the primary IH analysis (primary endpoint = fascial dehiscence at 30 days; IH at ≥ 12 months not reported); included for 30-day SSI and seroma secondary outcomes only. †† Not prospectively registered in any public trial registry Table 1Study characteristics — primary IH analysis (k = 5) and Lima 2020 (secondary outcomes)Study (ref.)DesignCountryN mesh/ctrlCDC classMesh & planeFollow-upAnalysis setBravo-Salva 2021 [26]Retrospective cohortSpain56/131Mixed (III–IV)Onlay; Ultrapro PP/absorbableMedian 64.4 mo (SD 35.2)As analysedMäkäräinen 2025 PREEMER [27]RCTFinland35/37 (22/28 at 2 y)Mixed (~ 34% CDC III + IV; ~ 17% CDC IV)Retrorectus; Progrip 8 cm polyester26.3 ± 3.6 moEvaluated at 2 yPizza 2021 PROMETHEUS [28]RCTItaly100/100CDC II onlySublay; Parietex ProGrip polyester24 mo (fixed)ITT (n = 100/arm)Nageeb 2021 [29] ††RCTEgypt38/34CDC I–II onlyOnlay: Prolene polypropylene24 mo (fixed)Analysed cohortUlutas 2023 [30]RCTTurkey54/54 (50/51 PP)Mixed (incl. III–IV; dirty ~ 28%)Onlay; Ultrapro 10 × 10 cm PP13.1 ± 5.6 moITTLima 2020 [12] †RCTBrazil63/52Mixed emergency (77% colorectal)Onlay; polypropylene30 dSecondary outcomes only† Lima 2020: excluded from the primary IH analysis (primary endpoint = fascial dehiscence at 30 days; IH at ≥ 12 months not reported); included for 30-day SSI and seroma secondary outcomes only. †† Not prospectively registered in any public trial registry Study characteristics — primary IH analysis (k = 5) and Lima 2020 (secondary outcomes) † Lima 2020: excluded from the primary IH analysis (primary endpoint = fascial dehiscence at 30 days; IH at ≥ 12 months not reported); included for 30-day SSI and seroma secondary outcomes only. †† Not prospectively registered in any public trial registry The primary IH analysis included 617 patients (270 mesh, 347 suture). Mesh planes varied: onlay (Bravo-Salva, 2021; Ulutas, 2023; Nageeb, 2021), retrorectus (Mäkäräinen, 2025), and sublay (Pizza, 2021). Four of the five control arms used a 4:1 suture-to-wound-length ratio; Bravo-Salva 2021 used a double-loop PDS suture without the small-bites technique (as explicitly stated in the original publication). Nageeb 2021 was not prospectively registered in any public trial registry and enrolled a substantially younger population (mean age ~ 41 years versus 55–69 years in the other studies). Lima et al. 2020 [12] enrolled patients undergoing emergency laparotomy, mostly for colorectal pathology (77.4%), and contributed only to the secondary outcome analyses. Risk of bias Risk-of-bias assessments are summarised in Table 2. All four RCTs were rated as "Some concerns" under ROB-2. Mäkäräinen 2025 (PREEMER) was prematurely terminated at 72 of 244 target patients (29.5%), which carries a potential estimation bias (D3: High). Ulutas 2023 raised concerns across all five domains. Nageeb 2021 was not prospectively registered (D5: Some concerns). Bravo-Salva 2021 was rated as Serious under ROBINS-I due to residual confounding, selection bias, and the absence of blinded outcome assessment.Table 2Risk-of-bias assessment (ROB-2/ROBINS-I)StudyToolD1D2D3D4D5D6OverallBravo-Salva 2021ROBINS-ISMLMLMSERIOUSMäkäräinen 2025ROB-2LLH (early term.)LSC—Some concernsPizza 2021ROB-2LLSCLSC—Some concernsNageeb 2021 ††ROB-2SCSCSCSCSC—Some concernsUlutas 2023ROB-2SCSCSCSCSC—Some concernsLima 2020 (2° outcomes)ROB-2LLSCSCSC—Some concernsL Low; SC Some concerns; H High (ROB-2). M Moderate; S Serious (ROBINS-I) Table 2Risk-of-bias assessment (ROB-2/ROBINS-I)StudyToolD1D2D3D4D5D6OverallBravo-Salva 2021ROBINS-ISMLMLMSERIOUSMäkäräinen 2025ROB-2LLH (early term.)LSC—Some concernsPizza 2021ROB-2LLSCLSC—Some concernsNageeb 2021 ††ROB-2SCSCSCSCSC—Some concernsUlutas 2023ROB-2SCSCSCSCSC—Some concernsLima 2020 (2° outcomes)ROB-2LLSCSCSC—Some concernsL Low; SC Some concerns; H High (ROB-2). M Moderate; S Serious (ROBINS-I) Risk-of-bias assessment (ROB-2/ROBINS-I) L Low; SC Some concerns; H High (ROB-2). M Moderate; S Serious (ROBINS-I) Primary outcome — incisional hernia (k = 5) The 2 × 2 datasets are presented in Table 3, with every value verified cell-by-cell against the original published tables. The random-effects model (REML + HKSJ) yielded a pooled OR of 0.221 (95% CI 0.122–0.402); τ2 = 0.000; I2 = 0.0%; corrected PI 0.111–0.439, excluding the null (Fig. 2). Because the control-arm event rate exceeded 10% in four of the five studies, the OR overestimates the relative risk. Applying the Zhang–Yu OR-to-RR correction [31], the approximate RR ranges from 0.28 to 0.38 across the observed control-arm event rates, which is broadly consistent with the elective-surgery literature and, in particular, with Jairam et al. [8] (RR 0.35) and Aiolfi et al. [9] (RR 0.38).Table 3Primary outcome — incisional hernia: 2 × 2 data (k = 5)StudyAnalysis set (source)Events meshN meshEvents ctrlN ctrlCtrl IH rate; FUBravo-Salva 2021 [26]Cohort as analyzed (Bravo-Salva [26], source Table 2)8564813136.6%; median 64.4 moMäkäräinen 2025 [27]* Clinically evaluated at 2 y (Mäkäräinen [27], source Table 4)0*221*283.6%; 26.3 ± 3.6 moPizza 2021 [28]ITT (Pizza [28], source Table 2)61002110021.0%; 24 mo (fixed)Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 8) 13893426.5%; 24 mo (fixed)Ulutas 2023 [30]ITT (Ulutas [30], source Table 10)254145425.9%; 13.1 ± 5.6 moPooled (REML + HKSJ): OR 0.221 (95% CI 0.122–0.402); I2 = 0.0%; τ2 = 0.000; corrected PI 0.111–0.439. All values were verified cell-by-cell against the original published manuscript tables. * 0.5 continuity correction applied (Mäkäräinen 2025: zero events in mesh arm)Fig. 2Forest plot — incisional hernia, primary analysis (k = 5). Random-effects model (REML + HKSJ). Pooled OR 0.221 (95% CI 0.122–0.402); τ2 = 0.0000; I2 = 0.0%; corrected PI 0.111–0.439, excluding the null. Diamond, pooled estimate with 95% CI; squares, study-level estimates with size proportional to inverse-variance weight. Red squares, CDC class III–IV; blue squares, CDC class I–II Table 3Primary outcome — incisional hernia: 2 × 2 data (k = 5)StudyAnalysis set (source)Events meshN meshEvents ctrlN ctrlCtrl IH rate; FUBravo-Salva 2021 [26]Cohort as analyzed (Bravo-Salva [26], source Table 2)8564813136.6%; median 64.4 moMäkäräinen 2025 [27]* Clinically evaluated at 2 y (Mäkäräinen [27], source Table 4)0*221*283.6%; 26.3 ± 3.6 moPizza 2021 [28]ITT (Pizza [28], source Table 2)61002110021.0%; 24 mo (fixed)Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 8) 13893426.5%; 24 mo (fixed)Ulutas 2023 [30]ITT (Ulutas [30], source Table 10)254145425.9%; 13.1 ± 5.6 moPooled (REML + HKSJ): OR 0.221 (95% CI 0.122–0.402); I2 = 0.0%; τ2 = 0.000; corrected PI 0.111–0.439. All values were verified cell-by-cell against the original published manuscript tables. * 0.5 continuity correction applied (Mäkäräinen 2025: zero events in mesh arm) Primary outcome — incisional hernia: 2 × 2 data (k = 5) Pooled (REML + HKSJ): OR 0.221 (95% CI 0.122–0.402); I2 = 0.0%; τ2 = 0.000; corrected PI 0.111–0.439. All values were verified cell-by-cell against the original published manuscript tables. * 0.5 continuity correction applied (Mäkäräinen 2025: zero events in mesh arm) Forest plot — incisional hernia, primary analysis (k = 5). Random-effects model (REML + HKSJ). Pooled OR 0.221 (95% CI 0.122–0.402); τ2 = 0.0000; I2 = 0.0%; corrected PI 0.111–0.439, excluding the null. Diamond, pooled estimate with 95% CI; squares, study-level estimates with size proportional to inverse-variance weight. Red squares, CDC class III–IV; blue squares, CDC class I–II Trial sequential analysis TSA was performed with CER = 26%, RRR = 65%, α = 0.05, and power = 80%. The required information size (RIS) was estimated at approximately 151 patients. The accrued information size (AIS = 617; 409% of RIS) exceeded the RIS boundary (Supplementary Fig. S9), suggesting numerical sufficiency for the overall IH effect under the specified parameters. This finding should be interpreted cautiously: the reliability of TSA is limited at k = 5, and the RIS is sensitive to changes in assumptions. With CER = 20% and RRR = 60%, RIS rises to approximately 255 patients, reducing AIS/RIS to 242%. This numerical sufficiency does not, at k = 5, constitute firm evidence and is reported consistently with the LOW–VERY LOW GRADE certainty. Sensitivity analysis — RCT-only (k = 4) After excluding Bravo-Salva 2021 (ROBINS-I Serious), the four-RCT model produced a pooled OR of 0.179 (95% CI 0.074–0.435); τ2 = 0.000; I2 = 0.0%; PI 0.054–0.594 (Fig. 3). The protective direction and the statistical significance were preserved.Fig. 3Forest plot — incisional hernia, RCT-only sensitivity analysis (k = 4). Bravo-Salva 2021 excluded (ROBINS-I Serious). Pooled OR 0.179 (95% CI 0.074–0.435); PI 0.054–0.594. The protective direction is preserved in the highest-quality evidence stratum Forest plot — incisional hernia, RCT-only sensitivity analysis (k = 4). Bravo-Salva 2021 excluded (ROBINS-I Serious). Pooled OR 0.179 (95% CI 0.074–0.435); PI 0.054–0.594. The protective direction is preserved in the highest-quality evidence stratum Pre-specified subgroup analysis — CDC class I–II versus III–IV The pre-specified Borenstein interaction test (fixed-effect pooling within strata; normal z for the interaction) gave z = 0.327, p = 0.743, which is non-significant and indicates no statistically detectable differential effect by wound-contamination class (Supplementary Fig. S5). Both subgroups showed a consistent protective direction: CDC III–IV (k = 3; OR 0.238, 95% CI 0.117–0.486) and CDC I–II (k = 2; OR 0.198, 95% CI 0.083–0.472; subgroup CIs were computed with the pre-specified fixed-effect [inverse-variance] method, consistent with the Borenstein interaction test, rather than with the random-effects HKSJ correction used for the main analyses, which is uninterpretable at df = 1). These analyses are exploratory and severely underpowered. A non-significant interaction does not establish equivalence between contamination classes; it indicates instead that the present evidence is insufficient to distinguish a differential from an equal effect. Influence diagnostics Leave-one-out analysis (Supplementary Fig. S4) showed directional consistency across all five iterations (all ORs < 0.25). Bravo-Salva 2021 exhibited the largest positive standardized residual (individual OR 0.288 vs. pooled OR 0.221), pulling the pooled estimate toward the null; its removal produces the RCT-only estimate (OR 0.179). Nageeb 2021 and Ulutas 2023 had the highest Cook's distances (0.215 and 0.183, respectively). The funnel plot is exploratory; formal asymmetry testing was not performed (k < 10 [22]). Secondary outcomes Deep SSI (≤ 30 days) — k = 4 Four studies provided extractable 2 × 2 data (Table 4). The pooled OR was 0.991 (95% CI 0.285–3.449); τ2 = 0.000; I2 = 0.0%; PI 0.184–5.351 (Fig. 4), which is non-significant. The PI spans from roughly fivefold benefit to fivefold harm and is therefore uninformative for clinical decision-making in contaminated fields. With only 12 deep-SSI events across four studies, this dataset is insufficient to characterize the safety profile of this critical endpoint.Table 4Deep SSI (≤ 30 days) — 2 × 2 data (k = 4)StudyAnalysis setEvents meshN meshEvents ctrlN ctrlMäkäräinen 2025 [27]Treated set (Mäkäräinen [27], source Table 3)135237Pizza 2021 [28]Randomized (Pizza [28], source Table 3)11002100Ulutas 2023 [30]Per-protocol (Ulutas [30], source Table 5)250151Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 6)238134Pooled (k = 4): OR 0.991 (95% CI 0.285–3.449); I2 = 0.0%; PI 0.184–5.351 — non-significantFig. 4Forest plot — deep surgical site infection (≤ 30 days, k = 4). Pooled OR 0.991 (95% CI 0.285–3.449); PI 0.184–5.351 — non-significant. The prediction interval spans approximately a fivefold benefit to a fivefold harm and is uninformative for clinical decision-making in contaminated fields Table 4Deep SSI (≤ 30 days) — 2 × 2 data (k = 4)StudyAnalysis setEvents meshN meshEvents ctrlN ctrlMäkäräinen 2025 [27]Treated set (Mäkäräinen [27], source Table 3)135237Pizza 2021 [28]Randomized (Pizza [28], source Table 3)11002100Ulutas 2023 [30]Per-protocol (Ulutas [30], source Table 5)250151Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 6)238134Pooled (k = 4): OR 0.991 (95% CI 0.285–3.449); I2 = 0.0%; PI 0.184–5.351 — non-significant Deep SSI (≤ 30 days) — 2 × 2 data (k = 4) Pooled (k = 4): OR 0.991 (95% CI 0.285–3.449); I2 = 0.0%; PI 0.184–5.351 — non-significant Forest plot — deep surgical site infection (≤ 30 days, k = 4). Pooled OR 0.991 (95% CI 0.285–3.449); PI 0.184–5.351 — non-significant. The prediction interval spans approximately a fivefold benefit to a fivefold harm and is uninformative for clinical decision-making in contaminated fields Superficial SSI (≤ 30 days) — k = 4 (corrected) Adding Lima et al. 2020 [12] to the three primary-analysis studies produced a four-study superficial-SSI analysis (Table 5). The pooled OR was 1.526 (95% CI 0.717–3.246); I2 = 0.0%; PI 0.550–4.235 (Supplementary Fig. S7), which is non-significant. This result illustrates an instructive methodological point. A k = 3 analysis excluding Lima 2020 would give OR 1.207 (95% CI 1.086–1.340; p < 0.05), a spurious result driven by near-perfect directional homogeneity across only three studies (individual log-ORs: 0.234, 0.200, 0.149), which produced an anomalously small HKSJ standard error (SE_HK = 0.0244). Adding Lima 2020 (log-OR = 1.145) introduced between-study variance, increased SE_HK to 0.237, and resolved the artifact. The corrected non-significant estimate is consistent with the larger meta-analyses of Aiolfi et al. [9] (k = 14; p = 0.38) and Jairam et al. [8] (k = 12; non-significant). Importantly, this is a recognised anti-conservative behaviour of the unmodified Hartung–Knapp adjustment at low heterogeneity: under the truncated Hartung–Knapp variant the same k = 3 data are already non-significant (OR 1.207, 95% CI 0.27–5.38). The artefact therefore reflects the estimator, not the absence of Lima 2020; both intervals are reported.Table 5Superficial SSI (≤ 30 days) — k = 4, corrected (with Lima 2020)StudyAnalysis setEvents meshN meshEvents ctrlN ctrlPizza 2021 [28]Randomized (Pizza [28], source Table 3)51004100Ulutas 2023 [30]Per-protocol (Ulutas [30], source Table 5)750651Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 6)1038834Lima 2020 [12]Randomized (Lima [12], source Table 4 / text)1363452Pooled (k = 4, corrected): OR 1.526 (95% CI 0.717–3.246); I2 = 0.0%; PI 0.550–4.235 — non-significant. The earlier k = 3 estimate (OR 1.207, 95% CI 1.086–1.340; p < 0.05) was an artifact of an anomalously small SE_HK; the addition of Lima 2020 corrects it Table 5Superficial SSI (≤ 30 days) — k = 4, corrected (with Lima 2020)StudyAnalysis setEvents meshN meshEvents ctrlN ctrlPizza 2021 [28]Randomized (Pizza [28], source Table 3)51004100Ulutas 2023 [30]Per-protocol (Ulutas [30], source Table 5)750651Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 6)1038834Lima 2020 [12]Randomized (Lima [12], source Table 4 / text)1363452Pooled (k = 4, corrected): OR 1.526 (95% CI 0.717–3.246); I2 = 0.0%; PI 0.550–4.235 — non-significant. The earlier k = 3 estimate (OR 1.207, 95% CI 1.086–1.340; p < 0.05) was an artifact of an anomalously small SE_HK; the addition of Lima 2020 corrects it Superficial SSI (≤ 30 days) — k = 4, corrected (with Lima 2020) Pooled (k = 4, corrected): OR 1.526 (95% CI 0.717–3.246); I2 = 0.0%; PI 0.550–4.235 — non-significant. The earlier k = 3 estimate (OR 1.207, 95% CI 1.086–1.340; p < 0.05) was an artifact of an anomalously small SE_HK; the addition of Lima 2020 corrects it Seroma (≤ 30 days) — k = 4 Adding Lima et al. 2020 [12] produced a four-study seroma analysis (Table 6). The pooled OR was 3.980 (95% CI 1.924–8.234); I2 = 0.0%; PI 1.490–10.634 (Supplementary Fig. S6). The lower bound of the CI exceeds 1.0, indicating a significantly higher seroma rate in the mesh group. This estimate agrees with Aiolfi et al. [9] (RR 2.05; p = 0.0008; k = 14) and Jairam et al. [8] (onlay RR 2.23; p = 0.03; k = 12). The earlier k = 3 analysis was underpowered to detect this signal. Under the conservative truncated Hartung–Knapp variant the interval widens to 1.06–14.94; the direction (higher seroma with mesh) is robust but the precision is estimator-dependent.Table 6Seroma (≤ 30 days) — k = 4, statistically significantStudyAnalysis setEvents meshN meshEvents ctrlN ctrlPizza 2021 [28]Randomized (Pizza [28], source Table 3) 71001100Ulutas 2023 [30]Per-protocol (Ulutas [30], source Table 2) 450251Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 9)938234Lima 2020 [12]Randomized (Lima [12], source Table 4 / text) 1263352Pooled (k = 4): OR 3.980 (95% CI 1.924–8.234); I2 = 0.0%; PI 1.490–10.634; significant (higher seroma with mesh) Table 6Seroma (≤ 30 days) — k = 4, statistically significantStudyAnalysis setEvents meshN meshEvents ctrlN ctrlPizza 2021 [28]Randomized (Pizza [28], source Table 3) 71001100Ulutas 2023 [30]Per-protocol (Ulutas [30], source Table 2) 450251Nageeb 2021 [29]Analyzed cohort (Nageeb [29], source Table 9)938234Lima 2020 [12]Randomized (Lima [12], source Table 4 / text) 1263352Pooled (k = 4): OR 3.980 (95% CI 1.924–8.234); I2 = 0.0%; PI 1.490–10.634; significant (higher seroma with mesh) Seroma (≤ 30 days) — k = 4, statistically significant Pooled (k = 4): OR 3.980 (95% CI 1.924–8.234); I2 = 0.0%; PI 1.490–10.634; significant (higher seroma with mesh) Mesh explantation Mesh-removal events are summarised in Table 7. Across all six studies, the total number of documented explantations was 4/229 mesh patients (1.7%), among the four studies reporting mesh-removal status (Bravo-Salva, Mäkäräinen, Pizza, Nageeb; Ulutas and Lima did not report); restricted to CDC class III–IV studies, the figure was 2/91 mesh patients (2.2%). Formal pooling was not performed because of heterogeneous reporting and very low event counts.Table 7Mesh removal — safety events as reportedStudyCDC classN removed/N meshReasonNotesBravo-Salva 2021 [26]III–IV0/56—Chronic infection n = 2; conservative; no explantationMäkäräinen 2025 [27]Mixed I–IV2 /35Fistula (CD 3b) + seroma + 1 internal hernia, + 1 hematoma; total reoperations 3/35 (8.6%)Pizza 2021 [28]II1/100Deep SSIRequired reoperationNageeb 2021 [29]I–II1/38Deep SSI/mesh exposure (day 21)Subsequent IH at 3 monthsUlutas 2023 [30]Mixed I–IVNR/50NR2 deep SSI managed with percutaneous drainageLima 2020 [12]Mixed emergencyNR/63NRSSI/explantation not reported separately; 0 fascial dehiscence in mesh groupTotal documented explantations: 4/229 (1.7%), restricted to studies reporting mesh-removal status; CDC class III–IV studies: 2/91 (2.2%). NR = not reported. Formal pooling not performed Table 7Mesh removal — safety events as reportedStudyCDC classN removed/N meshReasonNotesBravo-Salva 2021 [26]III–IV0/56—Chronic infection n = 2; conservative; no explantationMäkäräinen 2025 [27]Mixed I–IV2 /35Fistula (CD 3b) + seroma + 1 internal hernia, + 1 hematoma; total reoperations 3/35 (8.6%)Pizza 2021 [28]II1/100Deep SSIRequired reoperationNageeb 2021 [29]I–II1/38Deep SSI/mesh exposure (day 21)Subsequent IH at 3 monthsUlutas 2023 [30]Mixed I–IVNR/50NR2 deep SSI managed with percutaneous drainageLima 2020 [12]Mixed emergencyNR/63NRSSI/explantation not reported separately; 0 fascial dehiscence in mesh groupTotal documented explantations: 4/229 (1.7%), restricted to studies reporting mesh-removal status; CDC class III–IV studies: 2/91 (2.2%). NR = not reported. Formal pooling not performed Mesh removal — safety events as reported Total documented explantations: 4/229 (1.7%), restricted to studies reporting mesh-removal status; CDC class III–IV studies: 2/91 (2.2%). NR = not reported. Formal pooling not performed GRADE certainty of evidence GRADE assessments are summarised in Table 8. The primary IH outcome was rated LOW after downgrading for risk of bias (all RCTs, Some concerns; ROBINS-I Serious for the observational cohort), imprecision (k = 5; TSA numerically sufficient but assumption-sensitive), and publication bias (Olavarria et al. [32] documented substantial publication bias in the PMR literature). The CDC class III–IV subgroup was rated VERY LOW (k = 3 studies — 2 RCTs + 1 retrospective cohort; combined mesh-arm n ≈ 132: Bravo-Salva n = 56; Mäkäräinen n = 22 evaluable; Ulutas n = 54; follow-up 13–64 months). All safety outcomes were rated VERY LOW.Table 8Summary of pooled effects and GRADE certainty of evidenceOutcomekOR95% CII2 (%)τ2PIGRADEIH — primary50.2210.122–0.4020.00.0000.111–0.439LOWIH — RCT-only sensitivity40.1790.074–0.4350.00.0000.054–0.594LOWIH — CDC III–IV subgroup ‡30.2380.117–0.4860.00.000NE (k = 3)VERY LOWIH — CDC I–II subgroup ‡20.1980.083–0.4720.00.000NE (k = 2)VERY LOWDeep SSI (≤ 30 d) †40.9910.285–3.4490.00.0000.184–5.351VERY LOWSuperficial SSI (≤ 30 d) §41.5260.717–3.2460.00.0000.550–4.235VERY LOWSeroma ¶43.9801.924–8.2340.00.0001.490–10.634VERY LOWMesh explantation6NPNPNPNPNPVERY LOWREML + HKSJ throughout. † Corrected PI: prior 0.137–7.158 → 0.184–5.351 with df = k − 2 = 2. ‡ Subgroup CIs computed with the pre-specified fixed-effect (inverse-variance) method, consistent with the Borenstein interaction test (not the random-effects HKSJ correction, which is uninterpretable at df = 1 for k = 2). § Earlier k = 3 superficial-SSI estimate showed an artefactual significance, corrected here at k = 4. ¶ Seroma is significant at k = 4. NE = not estimable; NP = not pooled. Conservative (truncated Hartung–Knapp) intervals, reported alongside the primary intervals in the text, are: IH primary 0.101–0.483; seroma 1.06–14.94; superficial-SSI (k = 3) 0.27–5.38. The protective IH effect and the seroma direction are preserved under this more conservative variant Table 8Summary of pooled effects and GRADE certainty of evidenceOutcomekOR95% CII2 (%)τ2PIGRADEIH — primary50.2210.122–0.4020.00.0000.111–0.439LOWIH — RCT-only sensitivity40.1790.074–0.4350.00.0000.054–0.594LOWIH — CDC III–IV subgroup ‡30.2380.117–0.4860.00.000NE (k = 3)VERY LOWIH — CDC I–II subgroup ‡20.1980.083–0.4720.00.000NE (k = 2)VERY LOWDeep SSI (≤ 30 d) †40.9910.285–3.4490.00.0000.184–5.351VERY LOWSuperficial SSI (≤ 30 d) §41.5260.717–3.2460.00.0000.550–4.235VERY LOWSeroma ¶43.9801.924–8.2340.00.0001.490–10.634VERY LOWMesh explantation6NPNPNPNPNPVERY LOWREML + HKSJ throughout. † Corrected PI: prior 0.137–7.158 → 0.184–5.351 with df = k − 2 = 2. ‡ Subgroup CIs computed with the pre-specified fixed-effect (inverse-variance) method, consistent with the Borenstein interaction test (not the random-effects HKSJ correction, which is uninterpretable at df = 1 for k = 2). § Earlier k = 3 superficial-SSI estimate showed an artefactual significance, corrected here at k = 4. ¶ Seroma is significant at k = 4. NE = not estimable; NP = not pooled. Conservative (truncated Hartung–Knapp) intervals, reported alongside the primary intervals in the text, are: IH primary 0.101–0.483; seroma 1.06–14.94; superficial-SSI (k = 3) 0.27–5.38. The protective IH effect and the seroma direction are preserved under this more conservative variant Summary of pooled effects and GRADE certainty of evidence REML + HKSJ throughout. † Corrected PI: prior 0.137–7.158 → 0.184–5.351 with df = k − 2 = 2. ‡ Subgroup CIs computed with the pre-specified fixed-effect (inverse-variance) method, consistent with the Borenstein interaction test (not the random-effects HKSJ correction, which is uninterpretable at df = 1 for k = 2). § Earlier k = 3 superficial-SSI estimate showed an artefactual significance, corrected here at k = 4. ¶ Seroma is significant at k = 4. NE = not estimable; NP = not pooled. Conservative (truncated Hartung–Knapp) intervals, reported alongside the primary intervals in the text, are: IH primary 0.101–0.483; seroma 1.06–14.94; superficial-SSI (k = 3) 0.27–5.38. The protective IH effect and the seroma direction are preserved under this more conservative variant Discussion Summary of findings and place in the literature This systematic review and meta-analysis of five comparative studies shows that PMR at EML closure is associated with a reduction in incisional hernia (pooled OR 0.221, 95% CI 0.122–0.402; corrected PI 0.111–0.439). The effect is consistent across all five leave-one-out iterations, is preserved in the RCT-only sensitivity analysis (OR 0.179) and remains directionally consistent in both CDC class subgroups (interaction p = 0.743, non-significant and underpowered). TSA indicates numerical sufficiency (accrued information size 617 vs required ≈151; 409% of RIS) under the pre-specified assumptions but does not, at k = 5, constitute firm evidence. The point estimate (RR ≈ 0.36) is in line with the larger elective-surgery meta-analyses [8, 9], which suggests that the emergency context does not materially attenuate the efficacy of PMR. Our findings also align with those of Marcolin et al. [33] (OR 0.18; k = 4 RCTs) and Frassini et al. [34] (RR 0.31–0.44 at 1–4 years; k = 18 mixed designs). Two methodological features distinguish this analysis from earlier reviews. First, the inclusion of Lima 2020 [12] in the superficial-SSI analysis (k = 4) resolves a spurious significance artefact that had emerged at k = 3 because of near-perfect log-OR homogeneity (anomalously small SE_HK = 0.0244); the corrected k = 4 estimate (OR 1.526, 95% CI 0.717–3.246) is non-significant, in line with the larger literature [8, 9]. Second, the seroma analysis at k = 4 reaches the precision required to detect a statistically significant elevation (OR 3.980, 95% CI 1.924–8.234) that was undetectable at k = 3. Both findings agree with the independently reported estimates of Aiolfi et al. [9] and Jairam et al. [8]. The contaminated-field subgroup The pre-specified subgroup interaction test (p = 0.743) provides no statistical evidence that PMR efficacy differs between CDC class I–II and III–IV wounds, and both subgroups show a consistent protective direction. This finding challenges the assumption — implicit in current guideline equivocation — that contaminated wounds are a contraindication to PMR. The CDC class III–IV evidence base is, however, limited: three studies (one with ROBINS-I Serious risk of bias, two with premature termination or short follow-up), totaling a mesh-arm n of approximately 132 patients. As an arithmetical illustration only, applying the approximate RR ≈ 0.36 to the control-arm IH rates of Bravo-Salva 2021 (36.6%) and Ulutas 2023 (25.9%) gives estimated absolute risk reductions of approximately 23 and 17 percentage points, corresponding to an NNT of approximately 4 and 6, respectively. These figures are illustrative and should not be interpreted as definitive treatment-effect estimates. Seroma is the principal safety signal Seroma emerges as the single most robustly supported clinical safety concern. The k = 4 estimate (OR 3.980, 95% CI 1.924–8.234) agrees with the largest independent meta-analyses [8, 9]. In contaminated fields, the consequences of seroma differ qualitatively from those in elective clean surgery: subcutaneous fluid accumulation around a synthetic mesh in a bacterially colonized wound creates the conditions for biofilm formation, secondary deep infection, and possible explantation. Active seroma prevention should therefore be regarded as a mandatory adjunct: retromuscular rather than onlay positioning to eliminate subcutaneous dead space [35]; meticulous hemostasis; closed-suction drainage; and, in selected cases, negative-pressure wound therapy [34]. (Under the conservative truncated Hartung–Knapp variant this interval widens to 1.06–14.94, so the seroma direction is robust but its precision is estimator-dependent.) Deep SSI — the critical uncharacterized safety endpoint Deep SSI in contaminated fields remains essentially uncharacterized. The pooled estimate (OR 0.991; PI 0.184–5.351) spans from roughly fivefold benefit to fivefold harm and is uninformative. With only 12 deep-SSI events across four studies and no contamination-stratified data, the present evidence base provides no meaningful safety profile for this endpoint, which is the principal evidence gap in the field. Superficial SSI — methodological transparency The explicit correction of the superficial SSI estimate warrants a brief methodological comment. The k = 3 result arose because three studies with very similar individual log-ORs (0.234, 0.200, 0.149) yielded an anomalously small HKSJ standard error (SE_HK = 0.0244), resulting in apparent statistical significance even though no individual study was significant. Adding Lima 2020 (log-OR = 1.145) introduced between-study variance, raised SE_HK to 0.237, and dissolved the spurious significance. This sequence illustrates the well-known sensitivity of small meta-analyses (k < 5) to the inclusion or exclusion of individual studies, and it highlights the importance of comprehensive evidence synthesis. Limitations Several limitations should be considered when interpreting these findings. First, none of the included primary studies reported within-study CDC-stratified IH data; all subgroup analyses therefore rely on study-level assignment rather than patient-level stratification. Second, although title/abstract screening, full-text assessment, and data extraction involved multiple reviewers, the primary analyst (RAG-J) performed all statistical computations. The co-authors reviewed but did not independently re-execute these computations. The pooled estimates should be interpreted with this in mind; the full reproducibility dataset (ESM_1) is provided to allow third-party replication. Third, the most influential CDC class III–IV study (Bravo-Salva, 2021) carries a Serious risk of bias according to ROBINS-I and contributes substantial weight to the primary pooled estimate. Fourth, PREEMER was terminated at 29.5% of its target enrolment, and the heterogeneity of follow-up duration across studies (12–64 months) limits direct comparability of long-term outcomes. Fifth, publication bias in the PMR literature is well documented [32], and Nageeb 2021 was not prospectively registered and enrolled a substantially younger population (mean age ~ 41 years) than the other studies. Sixth, the expanded search (2015–2026) did not identify any additional emergency-laparotomy studies reporting IH at ≥ 12 months. Albendary et al. [36] were excluded because they included studies that did not have IH at ≥ 12 months and used wound failure as a composite endpoint. This reinforces the scarcity of the evidence base. Seventh, the primary analysis pooled studies that differed in mesh plane (onlay, retrorectus, and sublay) and mesh material (permanent synthetic, partially absorbable, and others); these techniques carry distinct seroma and infection profiles, so the pooled safety estimates should be interpreted as technique-averaged rather than technique-specific, and the small number of studies precluded a formal stratified meta-analysis by mesh plane or material. Eighth, we restricted the synthesis to published, peer-reviewed data and did not contact individual centres or trialists to retrieve unpublished or statistically non-significant datasets; combined with the documented publication bias in this literature [32], this may have biased the pooled effect away from the null. Finally, TSA numerical sufficiency is sensitive to changes in assumptions and should not be read as establishing definitive evidence in the absence of independent replication. Conclusions Prophylactic mesh reinforcement at the closure of emergency midline laparotomy is associated with a statistically significant reduction in incisional hernia (pooled OR 0.221, 95% CI 0.122–0.402; corrected PI 0.111–0.439). The effect is consistent across all leave-one-out iterations and across CDC contamination classes (interaction p = 0.743, non-significant and underpowered). Trial sequential analysis is consistent with numerical sufficiency (accrued information size 617 vs required ≈151; 409% of RIS) under the pre-specified assumptions but, at k = 5, does not by itself constitute firm evidence. These findings may support the selective, individualized use of PMR in high-baseline-risk contaminated-field patients, weighing the reduction in incisional hernia against the added operative time and technical demands of mesh placement in physiologically compromised emergency patients and the possibility that a retromuscular prosthesis may complicate any future definitive abdominal-wall repair; any absolute-benefit estimate (e.g., NNT ~ 4–6) in CDC class III–IV is illustrative only, given the non-significant, underpowered subgroup. The corrected seroma analysis (k = 4; OR 3.980, 95% CI 1.924–8.234; 1.06–14.94 under the conservative Hartung–Knapp variant) indicates a clinically meaningful seroma signal that demands active prevention through mesh-plane selection (retromuscular positioning to obliterate subcutaneous dead space, where the patient’s physiological condition permits), meticulous hemostasis, and judicious drainage. Superficial SSI is not significantly elevated at k = 4 (OR 1.526, non-significant). Deep SSI in contaminated fields remains entirely uncharacterized and is the principal safety gap. Given the LOW to VERY LOW GRADE certainty throughout, these conclusions are hypothesis-generating rather than practice-defining. The single highest evidence priority in abdominal-wall surgery is therefore a high-quality, multicentre, randomized controlled trial in CDC class III–IV wounds, with deep SSI and mesh explantation as co-primary safety endpoints, and with adequately powered IH follow-up at ≥ 24 months. Supplementary Information Below is the link to the electronic supplementary material.Sup. Figure1Baujat plot — primary IH analysis (k = 5). x-axis: contribution to the Cochran Q heterogeneity statistic. y-axis: absolute influence on pooled log-OR. Exploratory; k = 5 underpowered for formal heterogeneity assessment (PNG 97.5 KB)High Resolution Image (TIFF 122 KB)Sup. Figure2Influence plot (Cook's-distance approximation) — primary IH analysis. Nageeb 2021 (Cook's d = 0.215) and Ulutas 2023 (d = 0.183) have the highest Cook's distances. Bravo-Salva 2021 has the largest positive standardized residual, which pulls the pooled estimate toward the null (PNG 108 KB)High Resolution Image (TIFF 143 KB)Sup. Figure3Funnel plot — primary IH analysis. Exploratory only; formal asymmetry testing was not performed (k < 10). All studies fall to the left of the null, consistent with a uniformly beneficial effect (PNG 119 KB)High Resolution Image (TIFF 171 KB)Sup. Figure4Leave-one-out sensitivity analysis — primary IH (k = 5). Each row shows the pooled OR after removal of the named study. Dashed line, overall pooled OR 0.221 (95% CI 0.122–0.402). All five iterations maintained OR < 0.25, confirming directional consistency (PNG 113 KB)High Resolution Image (TIFF 164 KB)Sup. Figure5Forest plot annotated by CDC wound class — pre-specified subgroup analysis. CDC III–IV (k = 3; Bravo-Salva, Mäkäräinen, Ulutas): OR 0.238 (95% CI 0.117–0.486). CDC I–II (k = 2; Pizza, Nageeb): OR 0.198 (95% CI 0.083–0.472). Pre-specified Borenstein interaction test: z = 0.327, p = 0.743. Exploratory. Both strata show a consistent protective direction (PNG 315 KB)High Resolution Image (TIFF 163 KB)Sup. Figure6Forest plot — seroma (≤30 days, k = 4, with Lima, 2020). Pooled OR 3.980 (95% CI 1.924–8.234); PI 1.490–10.634 — significant: higher seroma rate with mesh. Consistent with Aiolfi et al. 2023 (k = 14; p = 0.0008) and Jairam et al. 2020 (onlay; p = 0.03) (PNG 189 KB)High Resolution Image (TIFF 164 KB)Sup. Figure7Forest plot — superficial SSI (≤30 days, k = 4, with Lima, 2020). Corrected from the earlier version. Pooled OR 1.526 (95% CI 0.717–3.246); PI 0.550–4.235 — non-significant. The earlier k = 3 estimate (OR 1.207, 95% CI 1.086–1.340; p < 0.05) was a statistical artifact due to an anomalously small SE_HK; the addition of Lima 2020 corrects this (PNG 118 KB)High Resolution Image (TIFF 151 KB)Sup. Figure8Bubble plot — exploratory meta-regression of log-OR versus follow-up duration (months). Bubble size proportional to inverse-variance weight. Exploratory only; k = 5 is insufficient for inferential meta-regression (PNG 122 KB)High Resolution Image (TIFF 562 KB)Sup. Figure9Trial sequential analysis (TSA) — primary IH outcome. Left: cumulative Z-score versus accrued sample size. Right: cumulative pooled OR with 95% CI. Parameters: CER = 26%; RRR = 65%; α = 0.05 (two-sided); power = 80%. Required information size (RIS) ≈ 151 patients. Accrued information size (AIS) = 617 patients (409% of RIS), indicating numerical sufficiency under the specified parameters. Caution: TSA reliability is limited at k = 5; sensitivity analysis with CER = 20% and RRR = 60% raises RIS to ≈ 255 patients (AIS/RIS = 242%) (PNG 294 KB)High Resolution Image (TIFF 562 KB) Baujat plot — primary IH analysis (k = 5). x-axis: contribution to the Cochran Q heterogeneity statistic. y-axis: absolute influence on pooled log-OR. Exploratory; k = 5 underpowered for formal heterogeneity assessment (PNG 97.5 KB) High Resolution Image (TIFF 122 KB) Influence plot (Cook's-distance approximation) — primary IH analysis. Nageeb 2021 (Cook's d = 0.215) and Ulutas 2023 (d = 0.183) have the highest Cook's distances. Bravo-Salva 2021 has the largest positive standardized residual, which pulls the pooled estimate toward the null (PNG 108 KB) High Resolution Image (TIFF 143 KB) Funnel plot — primary IH analysis. Exploratory only; formal asymmetry testing was not performed (k < 10). All studies fall to the left of the null, consistent with a uniformly beneficial effect (PNG 119 KB) High Resolution Image (TIFF 171 KB) Leave-one-out sensitivity analysis — primary IH (k = 5). Each row shows the pooled OR after removal of the named study. Dashed line, overall pooled OR 0.221 (95% CI 0.122–0.402). All five iterations maintained OR < 0.25, confirming directional consistency (PNG 113 KB) High Resolution Image (TIFF 164 KB) Forest plot annotated by CDC wound class — pre-specified subgroup analysis. CDC III–IV (k = 3; Bravo-Salva, Mäkäräinen, Ulutas): OR 0.238 (95% CI 0.117–0.486). CDC I–II (k = 2; Pizza, Nageeb): OR 0.198 (95% CI 0.083–0.472). Pre-specified Borenstein interaction test: z = 0.327, p = 0.743. Exploratory. Both strata show a consistent protective direction (PNG 315 KB) High Resolution Image (TIFF 163 KB) Forest plot — seroma (≤30 days, k = 4, with Lima, 2020). Pooled OR 3.980 (95% CI 1.924–8.234); PI 1.490–10.634 — significant: higher seroma rate with mesh. Consistent with Aiolfi et al. 2023 (k = 14; p = 0.0008) and Jairam et al. 2020 (onlay; p = 0.03) (PNG 189 KB) High Resolution Image (TIFF 164 KB) Forest plot — superficial SSI (≤30 days, k = 4, with Lima, 2020). Corrected from the earlier version. Pooled OR 1.526 (95% CI 0.717–3.246); PI 0.550–4.235 — non-significant. The earlier k = 3 estimate (OR 1.207, 95% CI 1.086–1.340; p < 0.05) was a statistical artifact due to an anomalously small SE_HK; the addition of Lima 2020 corrects this (PNG 118 KB) High Resolution Image (TIFF 151 KB) Bubble plot — exploratory meta-regression of log-OR versus follow-up duration (months). Bubble size proportional to inverse-variance weight. Exploratory only; k = 5 is insufficient for inferential meta-regression (PNG 122 KB) High Resolution Image (TIFF 562 KB) Trial sequential analysis (TSA) — primary IH outcome. Left: cumulative Z-score versus accrued sample size. Right: cumulative pooled OR with 95% CI. Parameters: CER = 26%; RRR = 65%; α = 0.05 (two-sided); power = 80%. Required information size (RIS) ≈ 151 patients. Accrued information size (AIS) = 617 patients (409% of RIS), indicating numerical sufficiency under the specified parameters. Caution: TSA reliability is limited at k = 5; sensitivity analysis with CER = 20% and RRR = 60% raises RIS to ≈ 255 patients (AIS/RIS = 242%) (PNG 294 KB) High Resolution Image (TIFF 562 KB)