Circular stapler platform and the risk of anastomotic leakage after elective left colectomy and high anterior resection.

Francesco Masedu: Conceptualization; methodology; software; data curation; investigation; validation; formal analysis; supervision; writing – original draft; writing – review and editing. Stefano Guadagni: Conceptualization; methodology; data curation; investigation; validation; supervision; visualization; writing – original draft; writing – review and editing. Marco Catarci: Conceptualization; methodology; investigation; validation; data curation; supervision; visualization; project administration; resources; writing – original draft; writing – review and editing.

All data were prospectively collected, and the study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, with approval from the relevant ethics committees (Marche Regional Ethics Committee [CERM] 2018/334 for iCral 2; CERM 2020/192 for iCral 3; Lazio 2 Ethics Committee 0054579/2022 for iCral 4).

This research received no specific grants from any funding agency in the public, commercial or not‐for‐profit sectors.

This pre‐planned post hoc retrospective analysis used a prospectively collected multicentre database including patients who underwent colorectal resection with anastomosis for malignant or benign disease. Individual patient‐level data were pooled from three consecutive studies of the Italian ColoRectal Anastomotic Leakage (iCral) study group (iCral‐2, iCral‐3 and iCral‐4), all of which predefined anastomotic leakage as a study endpoint. The aim was to evaluate the association between circular stapler platform and the risk of anastomotic leakage using advanced confounding control methods. Data from the published iCral‐2 [ 19 ] and iCral‐3 [ 20 ] studies and from the iCral‐4 study, whose primary results are currently under peer review, were combined to address a research question that was not specified as a primary or secondary endpoint in the original study protocols. Eligibility criteria for the three‐iCral studies are summarized in the Supporting Information . Differences among studies in American Society of Anesthesiologists (ASA) class distribution, urgent surgery and protective stoma use were harmonized in the present analysis. To facilitate assessment of the pooled dataset, the Supporting Information includes descriptive characteristics of the three constituent cohorts (Table  S1 ) and a sensitivity analysis restricted to the published iCral‐2/iCral‐3 cohorts (Table  S2 ). Among the 11,849 patients in the overall cohort, 7,394 (62.4%) underwent circular stapled anastomosis. Both malignant and benign indications were included because the stapled end‐to‐end anastomotic technique was considered technically comparable despite differences in resection extent and vascular control. Potential confounding by underlying pathology was addressed by including surgery for malignancy as a balancing covariate in the Generalized Boosted Models (GBM). To minimize procedural and anatomical confounding and to ensure a technically homogeneous comparison across circular stapler platforms—prioritizing internal over external validity—additional exclusion criteria were applied. Patients with incomplete data, urgent surgery, ASA class IV–V or protective stoma proximal to the anastomosis were excluded. Procedures with substantially different surgical anatomy or anastomotic configuration (i.e. resections other than left colectomy or anterior resection and anastomoses other than end‐to‐end) were also excluded. Conditions known to exert a disproportionate subgroup‐specific effect on anastomotic leakage, namely anastomotic height <5 cm from the external anal verge and receipt of neoadjuvant therapy, were excluded. Circular staplers with a 25‐mm diameter were excluded because this size was unavailable across all compared platforms. After application of all exclusion criteria, 3,653 patients undergoing standardized end‐to‐end circular stapled anastomosis for left colectomy (i.e. anastomosis >18 cm from the external anal verge) or high anterior resection (i.e. anastomosis 5–18 cm from the external anal verge) constituted the final study population. When a circular stapler was used, manufacturer, model and diameter were prospectively recorded, allowing precise platform classification. Stapler selection was neither randomized nor protocolized; therefore, multivariable adjustment and propensity‐based methods were applied to mitigate confounding related to patient‐, procedure‐, and centre‐level characteristics. Patients were categorized into five circular stapler platforms based on manufacturer‐specific groupings encompassing devices with shared staple‐line architecture and mechanical design: Platform 1 (3‐row circular stapler, n  = 830), Platform 2 (2‐row circular stapler, n  = 1,068), Platform 3 (2‐row circular stapler, n  = 1,082), Platform 4 (2‐row powered circular stapler, n  = 367) and Platform 5 (2‐row circular stapler, n  = 306). The study flow diagram is shown in Figure  1 . Flow diagram of patient selection from the pooled iCral 2–4 cohorts. Of 11,849 enrolled patients, 7,394 underwent circular stapled anastomosis. After application of predefined exclusion criteria, 3,653 patients undergoing standardized end‐to‐end circular stapled anastomosis for left colectomy or high anterior resection comprised the final study cohort. A total of 3,653 patients were enrolled across 87 Italian surgical centres between January 2019 and January 2024. The median enrolment was 19 patients per centre; six centres enrolled only one patient, 27 enrolled fewer than 10 and 39 enrolled fewer than 20. Five circular stapler platforms were used across the study population, although platform availability and preference varied by centre. Six centres used a single platform, whereas five used all five platforms. Baseline characteristics stratified by circular stapler platform are reported in Table  1 . Twenty‐one descriptive variables are summarized in Figure  2 . Continuous variables were dichotomized at their median values. Centres were classified as high‐volume (≥4 cases/month) or low‐volume (<4 cases/month) based on the median monthly enrolment. Each centre was additionally characterized by a binary variable indicating use of ≥2 circular stapler platforms. Adherence to ERP protocols was evaluated using 20 items common to all three iCral study protocols, encompassing the preoperative, intraoperative and postoperative phases (Table  S3 ). Open surgery was included as an individual covariate because of its established clinical relevance. The complementary variable, minimally invasive surgery, also contributed to the composite intraoperative ERP adherence score, which was incorporated as a summary measure of overall adherence to the perioperative care pathway. Baseline characteristics of the study population according to circular stapler platform. Abbreviations: ASA, American Society of Anesthesiologists.; BT, blood transfusion(s); CS, circular stapler; ERP, Enhanced recovery pathway. Chi square independence test with four degrees of freedom. Major: any event grade > II according to Clavien‐Dindo. Platforms were defined according to staple‐line architecture and mechanical design; detailed manufacturer and model information is provided in Table  2 . Love plot of covariate balance before and after adjustment using generalized boosted models. Love plot displaying Kolmogorov–Smirnov statistics for baseline covariates across pairwise comparisons of the five circular stapler platforms included in the study. Platform 1 (3‐row circular stapler), labelled as group “0” in the figure, served as the reference category and was compared with Platform 2 (2‐row circular stapler, “1”), Platform 3 (2‐row circular stapler, “2”), Platform 4 (2‐row powered circular stapler, “3”) and Platform 5 (2‐row circular stapler, “4”). Red dots indicate unadjusted comparisons, whereas blue dots indicate comparisons after generalized boosted model weighting. Improved covariate balance after adjustment is reflected by lower Kolmogorov–Smirnov statistics. ERP, Enhanced Recovery Pathway. Detailed information on circular stapler platforms, manufacturers and models used for end‐to‐end colorectal anastomosis is provided in Table  2 . Platforms were classified according to staple‐line architecture and mechanical design. Manufacturer and model information is reported solely for descriptive and reproducibility purposes. Circular stapler platforms, manufacturers and models used for end‐to‐end colorectal anastomosis. Platform 1–3‐row circular stapler (n. = 830) Platform 2–2‐row circular stapler (n. = 1,068) Platform 3–2‐row circular stapler (n. = 1,082) Platform 4–2‐row powered circular stapler (n. = 367) Platform 5–2‐row circular stapler (n. = 306) Note : Platforms were defined according to staple‐line architecture and mechanical design. Manufacturer and model information is provided for descriptive and reproducibility purposes only. Details of data collection procedures and ethical approval are provided in the Supporting Information . All data were prospectively collected, and the study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, with approval from the relevant ethics committees. All enrolled patients were followed for 60 days after surgery. Adverse events were recorded and graded according to the Clavien‐Dindo classification [ 21 ], the Japanese Clinical Oncology Group (JCOG) extended criteria [ 22 ] and included any unplanned readmission, reoperation or death within 60 days after surgery. Anastomotic leakage (AL) was defined according to the international consensus [ 23 ], and anastomotic bleeding (AB) was defined as persistent rectal bleeding associated with a ≥ 20 g/L decrease in haemoglobin concentration [ 24 ]. All outcomes were assessed 60 days after surgery. The primary endpoint was Overall AL (any AL). Secondary endpoints were Major AL (any AL grade > II), Overall AB (any AB), Major AB (any AB grade > II), Global morbidity (any adverse event), Major morbidity (any adverse event grade > II) and mortality (all‐cause death). No missing data were present in the final analytic dataset of 3,653 patients. The estimand of interest was the average treatment effect on the treated (ATT), representing the expected difference in outcomes among patients receiving the reference circular stapler platform had they instead received an alternative platform. The reference platform was selected based on findings from a previously published study [ 8 ]. Confounding was addressed using propensity score weighting in pairwise comparisons between the reference platform and each alternative platform. Propensity scores were estimated using GBM, a machine learning approach that flexibly captures complex, nonlinear relationships between treatment assignment and covariates while minimizing model misspecification [ 25 ]. Analyses were performed using the twang package in R (version 4.5.0). Twenty‐one covariates reflecting patient‐, procedure‐ and centre‐level characteristics were selected a priori and corresponded to the descriptive variables (Figure  2 ). Effective sample sizes after weighting were evaluated to assess the stability of the weighted comparisons. Treatment effects were estimated using doubly robust weighted outcome models adjusted for the same covariates included in the propensity score models. Weighted event rates represent marginal risk estimates in the weighted pseudo‐population. Sensitivity to unmeasured confounding was assessed using E‐values [ 26 ]. Additional methodological details are provided in the Supporting Information .

Among the 3,653 patients included in the final analytic cohort, overall anastomotic leakage occurred in 2.7% of patients receiving the reference 3‐row circular stapler platform. Before adjustment, higher crude rates of overall AL were observed for 2‐row platforms. After GBM weighting and doubly robust adjustment, Platforms 3 and 4 remained independently associated with a significantly increased risk of overall AL compared with the reference platform (Platform 3: weighted event rate 5.7%, OR 2.34, 95% CI 1.16–4.71; Platform 4: 6.6%, OR 2.81, 95% CI 1.16–6.80). No statistically significant differences were observed for Platforms 2 or 5 after weighting. Among the measured covariates, residual imbalance after weighting was greatest for circular stapler diameter, particularly in comparisons involving Platforms 3 and 5, reflecting limited overlap in diameter distributions across commercially available stapler platforms (Table  S4 ). This finding indicates constrained covariate overlap for certain platform–diameter combinations, consistent with the practical limitations of achieving full exchangeability in these comparisons. Major anastomotic leakage was infrequent across all platforms. After weighting, no platform was associated with a statistically significant increase in major AL compared with the reference platform. Overall, anastomotic bleeding was significantly more frequent with several 2‐row platforms. After GBM weighting, Platforms 2, 3 and 5 were associated with a markedly increased risk of overall anastomotic bleeding compared with the reference platform (ORs ranging from 5.65 to 7.23), with large E‐values indicating robustness to potential unmeasured confounding. Similar patterns were observed for major anastomotic bleeding, particularly for Platforms 3 and 4, which showed substantially higher adjusted odds despite the low absolute event rates. Given the small number of events, particularly for major bleeding, these estimates should be interpreted cautiously because of limited statistical stability and precision, and are better interpreted as indicating the direction rather than the exact magnitude of the association. Overall morbidity did not differ significantly across stapler platforms after adjustment. In contrast, major morbidity remained significantly higher for Platforms 3 (weighted event rate 7.5%, OR 2.66, 95% CI 1.31–5.39) and 5 (8.4%, OR 2.95, 95% CI 1.30–6.72) compared with the reference platform. Thirty‐ and 60‐day mortality rates were low (<1%) and did not differ significantly across platforms. Adjusted weighted event rates, odds ratios, 95% confidence intervals and E‐values for all primary and secondary endpoints are presented in Table  3 . Adjusted logistic multiple regression analysis for primary and secondary endpoints before and after generalized boosted model (GBM) weighting. Major anastomotic bleeding Note: The bold values indicate the statistically significant p values. Abbreviation: CS: circular stapler. GBM weighting: Generalized Boosted Model weighting. OR (95%CI): odds ratio and 95% confidence intervals. GBM‐weighted event rate (%): GBM‐weighted event rates represent adjusted marginal risk estimates in the ATT‐weighted pseudo‐population. E‐values are only reported for statistically significant associations. Their absence indicates that the corresponding comparison was not statistically significant after GBM weighting and doubly robust adjustment. Platform 1–3‐row CS: Covidien EEA™ circular stapler with Tri‐Staple™ Technology TRIEEA28MT, TRIEEA28XT, TRIEEA31MT, TRIEEA31XT, TRIEEA33MT, TRIEEA33XT. Platform 2–2‐row CS: Covidien EEA™ circular stapler with DST™ Technology DSTEEA28, DSTEEA28XL, DSTEEA31, DSTEEA31XL, DSTEEA33, DSTEEA33XL. Platform 3–2‐row CS: Ethicon circular stapler CDH29A, CDH33A, ECS29A, ECS33A. Platform 4–2‐row powered CS: Ethicon Echelon Circular™ Powered Stapler CDH29P, CDH31P. Platform 5–2‐row CS: Touchstone circular stapler CSC29A, CSC33A, ECSC29, ECSC33. Platforms were defined according to staple‐line architecture and mechanical design. Manufacturer and model information is reported for descriptive and reproducibility purposes only.

The present multicentre analysis provides a comparative evaluation of circular stapler platforms in left colectomy and high anterior resection using an advanced multi‐treatment propensity‐based framework. After extensive adjustment for patient‐, procedure‐ and centre‐level characteristics, clinically relevant differences in anastomotic outcomes were observed across stapler platforms. Compared with the reference platform, Platforms 3 and 4 were associated with a significantly increased risk of overall anastomotic leakage, whereas several platforms were associated with an increased risk of anastomotic bleeding. These associations persisted after GBM weighting, were supported by high E‐values and were accompanied by higher rates of major postoperative morbidity, whereas no differences in short‐term mortality were observed. This analysis addresses a device‐specific comparative effectiveness question that was not a primary or secondary endpoint of the iCral studies. Although part of the dataset derives from the ongoing iCral‐4 study, the present analysis was conceived independently to address a distinct clinical question using advanced causal inference methods applied to pooled individual patient‐level data. The impact of circular stapler technology on anastomotic outcomes has been extensively investigated in observational studies and meta‐analyses, but the evidence remains inconsistent. Most comparative studies evaluating two‐row versus three‐row circular staplers have reported either lower anastomotic leakage rates or a numerical trend favouring three‐row devices [ 4 , 5 , 8 , 10 ], although these findings have varied according to study design and methods of confounding adjustment. In contrast, large administrative database analyses have reported neutral results, further highlighting the heterogeneity of the available evidence [ 7 ]. Recent meta‐analyses have attempted to synthesize this literature but have produced partly divergent conclusions depending on how stapler technology was defined and grouped. While some analyses suggest a reduction in anastomotic leakage associated with three‐row or powered devices [ 15 , 16 , 18 ], others [ 17 ] have reported no significant differences between three‐row and two‐row staplers. These divergent conclusions largely reflect differences in the studies included, which varied substantially in design, patient populations and statistical methodology. Importantly, most available evidence evaluates individual device characteristics (e.g. staple‐row configuration or powered firing) or selected pairwise comparisons rather than complete commercial stapler platforms as used in routine clinical practice. Consequently, the relative contribution of individual engineering features cannot be disentangled from the combined effects of device design, surgical technique and perioperative care. Within this context, the present study addresses this gap by comparing multiple commercially available stapler platforms within a unified analytical framework, thereby reflecting real‐world clinical practice rather than isolated technical attributes. In the present study, Platforms 3 and 4 remained associated with more than a twofold increase in overall anastomotic leakage after GBM weighting and doubly robust adjustment. Effect estimates were attenuated after weighting, indicating that part of the crude association was explained by baseline imbalances. Nevertheless, the persistence of the association after extensive adjustment suggests robustness to measured confounding, although residual unmeasured confounding, limited covariate overlap (particularly for stapler diameter) and variation in postoperative surveillance across centres may still have influenced the results. The increased risk was confined to overall rather than major anastomotic leakage, a finding that warrants cautious interpretation because it may reflect differential ascertainment of Grade I leaks, differences in statistical power, leak severity or postoperative detection rather than a true dissociation between overall and major AL. E‐value analyses indicated that modest unmeasured confounding could attenuate the observed associations, whereas substantially stronger confounding would be required to fully explain them. Potential sources of residual confounding include surgeon‐specific technical expertise and intraoperative decision‐making, such as vascular ligation level, splenic flexure mobilization, rectal stump management, intraoperative assessment of anastomotic integrity and reinforcement strategies, particularly if unevenly distributed across stapler platforms. Because these factors are closely interrelated, they may function as confounders, mediators or proxies of surgical quality depending on their temporal relationship to stapler platform selection and anastomotic construction. Although individual intraoperative variables are recognized determinants of anastomotic integrity, their contribution to residual confounding cannot be directly inferred because these practices were not uniformly adopted across participating centres. Moreover, in centres using multiple stapler platforms, these interdependent intraoperative factors may influence stapler platform selection differently according to local practice and surgeon preference, precluding assumptions about the direction of any resulting residual confounding. From a statistical standpoint, covariate balance improved substantially after GBM weighting, although limited covariate overlap persisted for Platform 3, resulting in the smallest effective sample size. Consequently, although the association with overall anastomotic leakage remained statistically significant, the corresponding effect estimate should be interpreted cautiously because of its reduced precision and confirmed in independent datasets. After GBM weighting, several 2‐row platforms remained associated with an increased risk of overall and major anastomotic bleeding. However, these estimates were based on few events, particularly for major bleeding, resulting in wide confidence intervals. Accordingly, they should be interpreted cautiously as indicating the direction rather than the precise magnitude of the observed associations. Experimental evidence suggests that graduated‐height staple configurations may improve staple‐line microvascular perfusion [ 27 ], providing a plausible biological explanation for the observed associations. However, the present study was not designed to identify which engineering characteristics, if any, account for these associations. Consequently, the findings should be regarded as hypothesis‐generating rather than as evidence of a specific causal biological mechanism. Although overall postoperative morbidity did not differ significantly across stapler platforms after adjustment, major morbidity remained significantly higher for Platforms 3 and 5, supporting the clinical relevance of the observed differences in anastomotic complications. This study has several limitations. First, the analysis is based on pooled individual patient‐level data from multiple iCral studies conducted over different time periods, with approximately three quarters of the cohort derived from the published iCral‐2 and iCral‐3 studies and the remainder from the iCral‐4 study, whose primary results are currently under peer review. This approach was adopted to provide adequate statistical power for a comparative effectiveness analysis of relatively infrequent but clinically important outcomes, such as anastomotic leakage and bleeding. Comparative characteristics of the constituent cohorts and a sensitivity analysis restricted to the published iCral‐2 and iCral‐3 datasets are provided in the Supporting Information , showing consistent findings despite the smaller sample size. Second, this post hoc comparative observational analysis is based on non‐random allocation of stapler platforms and a highly selected study population. Exclusion of urgent surgery, ASA class IV–V patients, protective stomas, ultra‐low rectal anastomoses and neoadjuvant therapy was intended to improve internal validity by reducing major sources of clinical heterogeneity and confounding. Consequently, the findings may not be generalizable to higher‐risk colorectal surgery populations. Third, inclusion of both benign and malignant cases represents a potential source of confounding. However, surgery for malignancy was included as a prespecified covariate in the GBM, allowing comparisons between stapler platforms to be adjusted for the underlying disease indication. Fourth, despite the use of advanced causal inference techniques, residual confounding due to unmeasured variables, including surgeon‐specific technical expertise and intraoperative decision‐making, cannot be completely excluded. Relevant intraoperative practices, such as vascular ligation level, splenic flexure mobilization, rectal stump management, perfusion assessment and air leak testing, were not uniformly adopted across participating centres and therefore could not be consistently incorporated into the adjusted analyses. In addition, although weighting substantially improved balance across most measured covariates, residual imbalance persisted for selected variables, particularly stapler diameter, reflecting limited covariate overlap between commercially available stapler platforms and their predominant diameter configurations. Consequently, the estimated effects should be interpreted as platform‐level associations observed in real‐world clinical practice rather than as diameter‐independent mechanical effects. Fifth, the effective sample size after weighting was smallest for Platform 3, reflecting limited covariate overlap with the reference platform. Although the association with overall anastomotic leakage remained statistically significant, the corresponding effect estimate should be interpreted cautiously because of its reduced precision and confirmed in independent datasets.

All patients enrolled in the prospective studies gave full written consent.

Anastomotic leakage (AL) after left colectomy and anterior resection remains one of the most severe complications in colorectal surgery, with substantial clinical, oncological and economic consequences [ 1 ]. Risk factors for AL include patient‐, disease‐ and technical‐related determinants [ 2 ]. Among these, technical factors are particularly attractive because they are potentially modifiable. Within this context, the circular stapler platform represents a potentially modifiable technical determinant of anastomotic integrity. Since the introduction of mechanical circular staplers [ 3 ], multiple platforms with distinct design features have been developed, each intended to improve anastomotic integrity. However, whether these design differences translate into clinically meaningful differences in anastomotic outcomes remains controversial. The impact of circular stapler platforms on AL has been explored mainly through retrospective observational studies, including propensity score–matched analyses [ 4 , 5 , 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 ] and meta‐analyses [ 14 , 15 , 16 , 17 , 18 ]. While meta‐analyses synthesize available evidence, their conclusions remain influenced by differences in study selection, device classification, comparator definitions and statistical methodology. Randomized controlled trials comparing multiple stapler platforms are unlikely to be feasible because of limited equipoise, device heterogeneity and logistical constraints. Consequently, propensity score–based methods have become increasingly common in comparative effectiveness research. However, conventional pairwise matching is suboptimal for multiple treatment comparisons because it may produce heterogeneous reference groups, reduce sample size and limit generalizability. Comparisons across studies are further complicated by methodological differences, including target estimands (average treatment effect vs. average treatment effect on the treated), covariate selection, matching algorithms, calliper width, sample size and eligibility criteria. Accordingly, comparative analyses that address real‐world heterogeneity while maintaining methodological rigour are needed. The present study addresses this gap by comparing multiple circular stapler platforms in a large multicentre cohort using an advanced multi‐treatment propensity‐based analytical framework.

The authors declare no conflicts of interest.

Table S1. Comparison of the constituent iCral cohorts. To improve transparency regarding the contribution of the previously unpublished iCral‐4 dataset, the principal characteristics of the published iCral‐2/iCral‐3 cohorts are compared with those of the iCral‐4 cohort, including study period, sample size, overall anastomotic leakage rate, circular stapler platform distribution, participating centres and Enhanced Recovery Pathway (ERP) adherence. Table S2 . Sensitivity analysis restricted to the published iCral‐2 and iCral‐3 cohorts ( n  = 2,751). Treatment effect estimates for overall anastomotic leakage obtained after exclusion of the previously unpublished iCral‐4 cohort, compared with the primary pooled analysis. Table S3 . Definitions of the 20 Enhanced Recovery Pathway (ERP) items adopted in the present study. Table S4 . Post‐weighting Kolmogorov–Smirnov (KS) statistics and standardized mean differences (SMDs) for all covariates in each pairwise comparison between the reference Platform 1 and the alternative circular stapler platforms.