Overview
The SURGUT study aims to assess how well a computer model predicts bowel leaks after surgery. It will also examine whether adding surgeons' intuition improves these predictions.
The study will include adults having surgery to remove part of the colon or rectum. During the operation, the remaining bowel ends are joined together. Sometimes, this connection does not heal properly and a leak develops. This is called an anastomotic leak and can cause serious complications.
The research questions are:
- How well does a previously developed computer model predict a leak within 30 days after surgery?
- How accurately do surgeons estimate the risk of a leak?
- Does combining the computer model with surgeons' assessments improve prediction compared with the model alone?
This observational study will take place at several hospitals in different countries. Researchers will collect information about participants' health, surgery, and presence of complications. Surgeons will complete a pre-operative and an immediately post-operative questionnaire about the estimated risk of a leak, surgical difficulty, and concerns about the bowel connection.
Researchers will record whether a leak occurs within 30 days after surgery. They will compare these outcomes with the predictions from the computer model, surgeons' assessments, and the combination of both.
The findings may help improve how clinicians identify people at higher risk of a leak after bowel surgery.
Description
Background : Anastomotic leakage (AL) is a disruption of bowel-wall integrity at the anastomosis that creates communication between the intestinal lumen and the extraluminal space. A pelvic abscess adjacent to the anastomosis is also considered AL. AL is a major complication after colorectal surgery. Despite established risk factors and existing prediction models, reliable prediction in individual patients remains challenging.
Rationale: SURGUT aims to prospectively validate the machine learning model developed by Ochs, Taha-Mehlitz, and colleagues and determine whether adding structured assessments of surgical intuition improves prediction. Surgeons' preoperative and immediately post-operative judgments may capture clinically relevant information not represented by the model's variables.
Methods: This prospective, international, multicenter observational cohort study will enroll adults undergoing colorectal resection with an anastomosis in two cohorts: colon and rectal resections. Clinical and operative data, surgeons' assessments recorded before and immediately after surgery, and 30-day postoperative outcomes will be collected using REDCap. All anastomotic leaks diagnosed within 30 days will be included, regardless of severity (ISREC grades A-C). The study will not alter standard clinical care.
The primary outcome is model discrimination for AL within 30 days, measured by the area under the receiver operating characteristic curve (AUC) with a 95% confidence interval. Secondary outcomes are AL incidence, discrimination of surgeons' risk estimates, and the additional predictive value of surgical intuition, assessed using ΔAUC and net reclassification improvement. Analyses will include paired AUC comparisons and multivariable logistic regression.
Eligibility
Inclusion Criteria:
- Age 18 years or older.
- Undergoing colorectal surgery with creation of an anastomosis.
- Indication for surgery: malignancy or suspected malignancy, diverticulitis, ischemia, perforation, bleeding, volvulus, stenosis or obstruction, iatrogenic injury, or inflammatory bowel disease.
- Availability of general informed consent.
Exclusion Criteria:
- Age younger than 18 years.
- Refusal to participate in the study.
- Incomplete datasets
