Exploring preoperative prediction of conversion in minimally invasive colorectal cancer surgery using Swedish and Danish national register data
Year: 2026
Type: article
Study design: Observational Study
Abstract: PURPOSE: Conversion of minimally invasive surgery often adds operative time and may lead to complications. The aim was to develop a preoperative prediction model for conversion and identify predictive risk factors in minimally invasive colorectal cancer resections, to be used in consideration and counselling preoperatively. METHODS: An observational study based on data from national Swedish registries (CRCBaSE, 2019-2021). Logistic regression models were used to predict conversion, with variable selection based on statistical performance. Model discrimination, calibration, and uncertainty were evaluated using the Brier score, AUC, calibration plots, and bootstrap resampling. External validation was performed in the Danish Colorectal Cancer Registry (2016-2020). RESULTS: Minimally invasive surgery was performed on 7,811 patients. Seven preoperative predictors were identified as risk factors of conversion: body mass index (BMI), clinical T stage, age, previous abdominal surgery, sex, robotic approach, and left-sided tumour location. Risk of conversion increased with advanced T stage, left-sided tumours, prior abdominal surgery, extremes of BMI, and age up to 80 years, whereas robotic surgery and female sex reduced the risk of conversion. A 7-variable model was created and achieved an AUC of 0.730 in the Swedish test set but demonstrated reduced discrimination and calibration on internal validation (AUC 0.681). External validation in a Danish national cohort confirmed consistent directions of the predictors but showed further reduced predictive performance (AUC 0.608). CONCLUSIONS: Internal validation did not maintain adequate discrimination or calibration, and neither did the external validation. Several variables showed consistent associations with conversion and may be regarded as candidate predictors.
Language: English
Subfield: Oncology
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
Estimated APC paid: $3,790
OpenAlex ID: W7236592172
