Predictive performance of a simplified colon leakage score for anastomotic leakage after left-sided colorectal surgery

Abstract

BackgroundAnastomotic leakage is a serious complication after colorectal surgery associated with significant morbidity, mortality, and prolonged hospitalization. The Colon Leakage Score (CLS) was developed to estimate leakage risk after left-sided colorectal surgery, but its performance may vary across clinical settings. AimThis study evaluated the predictive performance of a simplified CLS-based score for clinically relevant anastomotic leakage after left-sided colorectal surgery with primary anastomosis. MethodsAdult patients undergoing left-sided colorectal resection with primary anastomosis at a tertiary referral center were included. A simplified CLS-based score was calculated from predefined patient-related and operative variables. Predictive performance was assessed using CLS risk categories, logistic regression, and receiver operating characteristic curve analysis. ResultsOf 167 patients, 10 (6.0%) developed anastomotic leakage. Median CLS scores were similar between leakage and non-leakage groups (4.0 [3.0–8.0] vs. 5.0 [3.0–7.0], P = .881). Leakage rates did not increase across CLS risk categories: 6.4% (low-risk), 5.6% (moderate-risk), and 0% (high-risk; P = .919). The AUC was 0.514, and each 1-point CLS increase was not associated with leakage risk (OR 1.03, 95% CI 0.82–1.30, P = .808). No patient exceeded the predefined >500 mL blood-loss threshold. ConclusionThe simplified CLS-based score did not meaningfully predict anastomotic leakage in this cohort. These findings indicate limited transportability and support the need for recalibration or alternative prediction tools before clinical use.

Keywords

Introduction

In left-sided colorectal surgery, the decision to construct a primary anastomosis with or without diversion still depends on an imperfect estimation of leakage risk before the complication becomes clinically evident. Anastomotic leakage remains one of the most consequential complications after colorectal surgery, with reported rates varying according to patient selection, tumor location, anastomotic level, operative technique, and leakage definition.^1,2 Although advances in surgical technique, perioperative care, and enhanced recovery pathways have improved postoperative management, leakage continues to be associated with increased morbidity, mortality, prolonged hospitalization, reoperation, and higher healthcare burden.^1-4
The clinical consequences of leakage are not limited to the early postoperative period. In colorectal cancer surgery, anastomotic leakage may delay recovery and adjuvant treatment and has been associated with worse oncological outcomes, including increased local recurrence and reduced survival.⁵ For this reason, surgeons often integrate patient-related and operative factors when estimating leakage risk, including male sex, obesity, smoking, steroid exposure, poor physiological status, low anastomosis, prolonged operative duration, and intraoperative blood loss.^1,6-8
Because individual risk factors do not provide a sufficiently structured estimate of risk, composite prediction tools have been proposed to support perioperative decision-making. The Colon Leakage Score (CLS) was developed for left-sided colorectal surgery by combining patient-related and operative variables into a numerical score intended to estimate the risk of anastomotic leakage.⁹ Subsequent work described the original CLS as a multi-parameter model with a total score ranging from 0 to 43 and proposed modified CLS-based approaches to improve discrimination in selected left-sided colorectal cancer cohorts.¹⁰
However, the clinical value of a leakage score depends on whether it remains discriminative when applied outside the population and scoring structure in which it was developed. Prediction-model literature emphasizes that apparent performance in a development or selected validation cohort does not automatically translate into clinical usefulness, because discrimination, calibration, external validation, and transportability across case mixes must be considered.^11,12 For a simplified CLS-based score, this concern is especially relevant because local case mix, event prevalence, operative practice, and the distribution of individual score components may alter risk stratification.
The present study aimed to assess the predictive performance of a simplified CLS-based score for clinically relevant anastomotic leakage in patients undergoing left-sided colorectal surgery with primary anastomosis.

Materials and Methods

Study Design and Patient SelectionThis retrospective cohort study was conducted at a tertiary referral center. All data were analyzed retrospectively and anonymized before analysis, and patient confidentiality was maintained throughout the study.
Adult patients who underwent left-sided colorectal resection with primary anastomosis were eligible for inclusion. Left-sided procedures included left hemicolectomy, anterior resection, low anterior resection, transanal total mesorectal excision, and intersphincteric resection. Patients without primary anastomosis, including Hartmann procedure and abdominoperineal resection, were excluded. Right-sided colectomy, transverse colectomy, total/subtotal colectomy, and total proctocolectomy were not included in the primary analysis because CLS was originally proposed for left-sided colorectal surgery.^9,10Data CollectionClinical and operative data were retrieved from institutional medical records. Patient-related variables included age, sex, body mass index, American Society of Anesthesiologists physical status, smoking status, alcohol consumption, steroid use, preoperative chemotherapy, and preoperative radiotherapy. Operative variables included type of surgery, emergency status, operative duration, intraoperative blood loss, anastomotic level from the anal verge, and diverting stoma formation.Simplified CLS-Based ScoreA simplified CLS-based score was calculated using predefined patient-related and operative variables. One point was assigned for age >65 years, male sex, ASA score ≥3, body mass index ≥30 kg/m², smoking, alcohol use, steroid use, preoperative chemotherapy, operative duration >180 minutes, and intraoperative blood loss >500 mL. Two points were assigned for preoperative radiotherapy and emergency surgery. Three points were assigned for anastomotic level ≤6 cm from the anal verge.
The total score was calculated as the sum of all component points. Patients were categorized as low risk (0–5 points), moderate risk (6–10 points), or high risk (≥11 points). Because this was a simplified CLS-based score rather than the original 0–43-point CLS model, the analysis was interpreted as an institutional performance assessment of a simplified CLS-based model.Outcome DefinitionThe primary outcome was clinically relevant anastomotic leakage. Anastomotic leakage was defined according to the International Study Group of Rectal Cancer as a defect of the intestinal wall at the anastomotic site leading to communication between the intra- and extraluminal compartments.¹³ Diagnosis was based on clinical findings, radiologic imaging, endoscopic evaluation, surgical exploration, or their combination.Ethical ApprovalThis retrospective cohort study was approved by the Çukurova University Faculty of Medicine Clinical Research Ethics Committee (Meeting No: 166, Date: 08.05.2026, Decision No: 46). The study was conducted in accordance with the 1964 Declaration of Helsinki and its later amendments. Informed consent was waived by the ethics committee owing to the retrospective nature of the study and the use of anonymized data.Statistical AnalysisContinuous variables were reported as median with interquartile range or mean with standard deviation, as appropriate. Categorical variables were reported as frequencies and percentages. Patients were compared according to the presence or absence of anastomotic leakage. Continuous variables were compared using the Mann–Whitney U test or Student's t-test, and categorical variables using the chi-square test or Fisher's exact test. The association between CLS and anastomotic leakage was assessed using univariable logistic regression. Multivariable regression was not performed because of the limited number of leakage events. Discriminative performance was evaluated using receiver operating characteristic curve analysis, and the area under the curve was interpreted according to standard ROC methodology. Statistical analyses were performed using Python 3.13.5 with pandas 2.2.3, scipy, statsmodels 0.14.6, and scikit-learn packages. A two-sided P value <0.05 was considered statistically significant.Reporting GuidelinesThis study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines for cohort studies.

Results

Patient Characteristics A total of 462 patients were screened during the study period. After excluding patients without primary anastomosis, non-left-sided colorectal procedures, and patients with unavailable key score variables, 167 patients were included in the final analysis (Figure 1). The median age of the cohort was 63.0 [56.0–68.5] years, and 112 patients (67.1%) were male. The median BMI was 26.0 [24.0–28.4] kg/m², and 23 patients (13.8%) had an ASA score ≥3. Preoperative chemotherapy and radiotherapy were recorded in 67 (40.1%) and 61 (36.5%) patients, respectively. Anastomotic level ≤6 cm from the anal verge was present in 91 patients (54.5%). The median simplified CLS score was 5.^3-7 Baseline and operative characteristics are summarized in Table 1. Anastomotic Leakage Anastomotic leakage occurred in 10 patients, corresponding to an overall leakage rate of 6.0%. The median simplified CLS score was not significantly different between patients with and without leakage (4 ^3-8 vs. 5 ,^3-7 P = .881). Intraoperative blood loss was higher in patients with leakage than in those without leakage (150 mL vs. 50 mL, P < .001). No patient exceeded the predefined >500 mL blood-loss threshold used in the simplified CLS scoring rule. Comparisons between patients with and without leakage are shown in Table 2. CLS Risk Categories and Leakage Patients were stratified according to predefined simplified CLS risk categories. Anastomotic leakage occurred in 6 of 94 patients (6.4%) in the low-risk group, 4 of 71 patients (5.6%) in the moderate-risk group, and 0 of 2 patients (0%) in the high-risk group. Leakage rates did not increase progressively across CLS risk categories (P = .919) (Table 3). Predictive Performance of the Simplified CLS-Based Score Receiver operating characteristic analysis showed poor discriminative performance of the simplified CLS-based score for anastomotic leakage (Figure 2). The AUC was 0.514, with a bootstrap 95% confidence interval of 0.302–0.732. The best dataset-derived CLS threshold was ≥8, yielding 40.0% sensitivity, 77.1% specificity, 10.0% positive predictive value, and 95.3% negative predictive value. In univariable logistic regression, each 1-point increase in CLS was not associated with anastomotic leakage (OR 1.03, 95% CI 0.82–1.30, P = .808) (Table 3).

Discussion

The main finding of this study was that the simplified CLS-based score did not meaningfully discriminate clinically relevant anastomotic leakage after left-sided colorectal surgery. The score failed at both levels expected from a clinically useful risk tool: it did not separate leakage from non-leakage phenotypes, and its predefined categories did not translate into a usable risk gradient. The overall discriminative performance was poor, and a higher CLS was not associated with increased leakage risk in univariable analysis. Although intraoperative blood loss differed between groups, this signal was not captured by the predefined blood-loss component because no patient exceeded the >500 mL threshold. This weak performance frames the present analysis as a negative institutional assessment of simplified CLS-based risk stratification rather than a validation of its clinical utility in this cohort.
The limited discrimination observed in this study should be interpreted in the context of how CLS was originally developed and subsequently evaluated. The CLS was originally proposed by Dekker et al. for left-sided colorectal surgery, and Yang et al. later described the original CLS as a numerical model composed of multiple clinical and operative parameters, with a total score ranging from 0 to 43.^9,10 Yang et al. also emphasized that the clinical validity of CLS had not been widely studied and reported better discrimination after deriving a modified CLS model for left-sided colorectal cancer surgery.¹⁰ These data indicate that CLS performance is not a fixed property of the score itself, but may depend on how closely the target cohort resembles the population in which the model was developed or recalibrated.
Several cohort-specific factors may help explain the weak performance of the simplified CLS-based score in the present analysis. First, the evaluated score was a simplified derivative of the CLS domains rather than the original 0–43-point model, which limits direct comparability with the original score framework. Second, the number of leakage events was small, reducing the stability of risk estimates and leaving very few patients in the highest predefined risk category. Third, the distribution of individual score components was uneven: although intraoperative blood loss was higher among patients with leakage, no patient crossed the >500 mL threshold used in the simplified score, so this component did not contribute to risk separation. Because anastomotic leakage reflects multiple patient-related, anatomical, and operative factors rather than a single dominant variable, a simplified additive score may lose discrimination when key components are rare, absent, or unevenly distributed in a local cohort.^1,6-8
Beyond statistical performance, the present findings also limit the clinical role of this simplified score. A leakage risk score is useful only if it separates patients into clinically meaningful risk strata that can support decisions such as intensified surveillance or diversion planning. Reviews of surgical prediction models and reporting guidance emphasize that apparent numerical association is insufficient for clinical use; discrimination, calibration, external validation, and transportability across case mixes must also be considered.^11,12 This distinction is clinically relevant because protective diversion may reduce the clinical consequences of leakage in selected low rectal anastomoses, but it also introduces stoma-related morbidity and requires separate judgment regarding patient selection.^14-16 In the present cohort, predefined CLS categories did not produce a clinically interpretable risk gradient, and the highest-risk category contained too few patients to guide decision-making. Therefore, this simplified CLS-based score should not be used as a stand-alone basis for protective stoma decisions in similar cohorts.
The discordance between intraoperative blood loss as a continuous variable and the predefined blood-loss component of the simplified score deserves particular attention. In the original CLS framework, blood loss is one of the operative domains included among the score components.¹⁰ In the present cohort, the blood-loss finding exposed a calibration problem: the raw operative variable differed between leakage groups, but the simplified binary threshold was too high to contribute to the score. This means that a potentially informative operative signal was present in the raw data but was functionally silent within the simplified scoring system. This interpretation is consistent with the broader surgical principle that anastomotic integrity depends not only on systemic risk factors but also on local perfusion and technical conditions at the anastomotic site.^17-19 In contemporary left-sided colorectal surgery cohorts with relatively low absolute blood loss, high-threshold binary components may weaken score discrimination by failing to capture variation in recorded blood loss that remains clinically visible at the group level.

Limitations

The negative result should be interpreted within several methodological constraints. The retrospective single-center design may limit generalizability and makes the analysis vulnerable to documentation variability in operative and perioperative records. The small number of leakage events limited statistical power, prevented reliable multivariable modeling, and reduced the stability of threshold-based performance estimates, as reflected by the wide bootstrap confidence interval for the AUC. Another important limitation is that the evaluated score was a simplified CLS-based model rather than the original 0–43-point CLS, so the results should not be read as a definitive external validation of the original Dekker score. In addition, diverting stoma was performed in 83 of 162 patients with available data (51.2%), and the simplified score does not incorporate diversion status; because protective diversion may attenuate the clinical manifestation of leakage in high-risk anastomoses, this may have introduced additional misclassification and contributed to the weak observed risk gradient. Finally, some clinically relevant domains, such as operative judgment and intraoperative perfusion assessment, could not be fully incorporated into the score-based assessment. These limitations narrow the interpretation of the negative result but do not negate the central observation that this simplified score failed to provide useful risk separation in this cohort.

Conclusion

The simplified CLS-based score did not provide meaningful discrimination for clinically relevant anastomotic leakage in this left-sided colorectal surgery cohort. The absence of a risk gradient across predefined categories and the poor ROC performance indicate that this simplified implementation should not be assumed to be clinically transferable without local assessment. Future work should focus on validating the full original CLS and recalibrated models in larger, multicenter cohorts with sufficient leakage events and broader variation in operative risk components.

Declarations

References

1. McDermott FD, Heeney A, Kelly ME, et al. Systematic review of preoperative, intraoperative and postoperative risk factors for colorectal anastomotic leaks. Br J Surg. 2015;102(5):462-479. doi:10.1002/bjs.9697
   Article · PubMed · Google Scholar
2. Sciuto A, Merola G, De Palma GD, et al. Predictive factors for anastomotic leakage after laparoscopic colorectal surgery. World J Gastroenterol. 2018;24(21):2247-2260. doi:10.3748/wjg.v24.i21.2247
   Article · PubMed · Google Scholar
3. Hyman N, Manchester TL, Osler T, Burns B, Cataldo PA. Anastomotic leaks after intestinal anastomosis: it's later than you think. Ann Surg. 2007;245(2):254-258. doi:10.1097/01.sla.0000225083.27182.85
   Article · PubMed · Google Scholar
4. Gustafsson UO, Scott MJ, Hubner M, et al. Guidelines for perioperative care in elective colorectal surgery: ERAS Society recommendations: 2018. World J Surg. 2019;43(3):659-695. doi:10.1007/s00268-018-4844-y
   Article · PubMed · Google Scholar
5. Mirnezami A, Mirnezami R, Chandrakumaran K, Sasapu K, Sagar P, Finan P. Increased local recurrence and reduced survival from colorectal cancer following anastomotic leak: systematic review and meta-analysis. Ann Surg. 2011;253(5):890-899. doi:10.1097/sla.0b013e3182128929
   Article · PubMed · Google Scholar
6. Pommergaard HC, Gessler B, Burcharth J, Angenete E, Haglind E, Rosenberg J. Preoperative risk factors for anastomotic leakage after resection for colorectal cancer: a systematic review and meta-analysis. Colorectal Dis. 2014;16(9):662-671. doi:10.1111/codi.12618
   Article · PubMed · Google Scholar
7. Frasson M, Flor-Lorente B, Rodríguez JLR, et al. Risk factors for anastomotic leak after colon resection for cancer: multivariate analysis and nomogram from a multicentric, prospective, national study with 3193 patients. Ann Surg. 2015;262(2):321-330. doi:10.1097/sla.0000000000000973
   Article · PubMed · Google Scholar
8. Sparreboom CL, van Groningen JT, Lingsma HF, et al. Different risk factors for early and late colorectal anastomotic leakage in a nationwide audit. Dis Colon Rectum. 2018;61(11):1258-1266. doi:10.1097/dcr.0000000000001202
   Article · PubMed · Google Scholar
9. Dekker JWT, Liefers GJ, de Mol van Otterloo JCA, Putter H, Tollenaar RAEM. Predicting the risk of anastomotic leakage in left-sided colorectal surgery using a colon leakage score. J Surg Res. 2011;166(1):e27-e34. doi:10.1016/j.jss.2010.11.004
   Article · PubMed · Google Scholar
10. Yang SU, Park EJ, Baik SH, Lee KY, Kang J. Modified colon leakage score to predict anastomotic leakage in patients who underwent left-sided colorectal surgery. J Clin Med. 2019;8(9):1450. doi:10.3390/jcm8091450
    Article · PubMed · Google Scholar
11. Souwer ETD, Bastiaannet E, Steyerberg EW, et al. Risk prediction models for postoperative outcomes of colorectal cancer surgery in the older population: a systematic review. J Geriatr Oncol. 2020;11(8):1217-1228. doi:10.1016/j.jgo.2020.04.006
    Article · PubMed · Google Scholar
12. Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ. 2015;350:g7594. doi:10.1136/bmj.g7594
    Article · PubMed · Google Scholar
13. Rahbari NN, Weitz J, Hohenberger W, et al. Definition and grading of anastomotic leakage following anterior resection of the rectum: a proposal by the International Study Group of Rectal Cancer. Surgery. 2010;147(3):339-351. doi:10.1016/j.surg.2009.10.012
    Article · PubMed · Google Scholar
14. Hüser N, Michalski CW, Erkan M, et al. Systematic review and meta-analysis of the role of defunctioning stoma in low rectal cancer surgery. Ann Surg. 2008;248(1):52-60. doi:10.1097/sla.0b013e318176bf65
    Article · PubMed · Google Scholar
15. Montedori A, Cirocchi R, Farinella E, Sciannameo F, Abraha I. Covering ileo- or colostomy in anterior resection for rectal carcinoma. Cochrane Database Syst Rev. 2010;(5):CD006878. doi:10.1002/14651858.cd006878.pub2
    Article · PubMed · Google Scholar
16. Phatak UR, Kao LS, You YN, et al. Impact of ileostomy-related complications on the multidisciplinary treatment of rectal cancer. Ann Surg Oncol. 2014;21(2):507-512. doi:10.1245/s10434-013-3287-9
    Article · PubMed · Google Scholar
17. Blanco-Colino R, Espin-Basany E. Intraoperative use of ICG fluorescence imaging to reduce the risk of anastomotic leakage in colorectal surgery: a systematic review and meta-analysis. Tech Coloproctol. 2018;22(1):15-23. doi:10.1007/s10151-017-1731-8
    Article · PubMed · Google Scholar
18. Jafari MD, Wexner SD, Martz JE, et al. Perfusion assessment in laparoscopic left-sided/anterior resection (PILLAR II): a multi-institutional study. J Am Coll Surg. 2015;220(1):82-92.e1. doi:10.1016/j.jamcollsurg.2014.09.015
    Article · PubMed · Google Scholar
19. Vignali A, Gianotti L, Braga M, Radaelli G, Malvezzi L, Di Carlo V. Altered microperfusion at the rectal stump is predictive for rectal anastomotic leak. Dis Colon Rectum. 2000;43(1):76-82. doi:10.1007/bf02237248
    Article · PubMed · Google Scholar

Additional Information

Publisher’s Note
Bayrakol MP remains neutral with regard to jurisdictional and institutional claims.

Rights and Permissions