A prediction model for disease-free survival following R0/R1 rectal cancer resection: A retrospective longitudinal study based on the Norwegian and Danish colorectal cancer quality registries

BACKGROUND: Prediction models may assist clinicians in communicating prognostic outcomes to cancer survivors. The Norwegian and Danish colorectal cancer registers (NCCR and DCCG) are valuable data sources for developing such models. Although few models exist for rectal cancer, their development and validation remain limited.

METHODS: We assessed data from 21,116 stage I-III rectal cancer patients who underwent R0/R1 surgery. The assessment included clinicopathological predictors of DFS, with the index date being the date of the pathological report. A multivariable Cox regression model with inverse probability weighting was estimated for DFS, as outcome, defined as the time from diagnosis to death, recurrence or end of follow-up with at least 90-day follow-up.

RESULTS: The analysis included 10,234 NCCR and 6691 DCCG patients, with 40·3% and 38·3% being female, mean age of 67·9 (11·7) and 67·9 (10·6) years, median follow-up times of 93·4 (95% CI (91·0-96·0)) and 79·1 (77·4-80·6) months, and 5-year DFS of 29·9% and 24·3%. Sex, age, pathological stage, circumferential resection margin, lymph node yield, malignant lymph nodes, and surgical procedures were significant predictors. The C-index was 0·69 (R2 0·24), and 0·65 (R2 0·07) for the external validation cohort. The calibration slope was 1·22 (SE 0·05). Predicted probabilities for the NCCR and DCCG cohorts closely matched the observed probabilities.

CONCLUSION: Accessible through RECTI.net, the model predicts disease-free survival for up to 12·5 years after surgery, aiding clinicians to provide prognostic outcomes to rectal cancer survivors. To improve the generalisability of the model, further validation in other populations is needed.