# Clinical Prediction Models for Contact X‐Ray Brachytherapy in Managing Rectal Cancers: A Scoping Review

## TL;DR

This review evaluates existing models for predicting rectal cancer response to radiotherapy, aiming to guide the development of a model specific to Contact X-ray brachytherapy.

## Contribution

The study identifies gaps in current models and proposes a framework for a new CXB-specific prediction model.

## Key findings

- Models predicting response based on TRG achieved a pooled AUC of 0.82.
- Models predicting pCR had a pooled AUC of 0.76.
- Common predictive factors included age, tumour grade, and T stage.

## Abstract

Currently, there are no clinically predictive models that can prognosticate the response of rectal cancers to Contact X‐ray brachytherapy (CXB). This review aims to critically evaluate existing models that have attempted to predict the response of rectal cancer to external beam radiotherapy, with the objective of laying the foundation for the development of a CXB‐specific prediction model.

A random‐effects meta‐analysis was employed to calculate pooled estimates of the discriminative ability of published models. Using the Prediction Model Risk Of Bias Assessment Tool (PROBAST), each model was evaluated for its risk of bias and applicability. Additionally, the frequency of commonly utilised predictive factors was documented.

Twelve papers discussed fifteen models based on pre‐treatment factors. Models predicting response based on the Tumour regression grade (TRG) classified responders as patients who achieved a complete response or near complete response and achieved a pooled AUC of 0.82 (95% CI 0.74–0.89). Models that predicted pathologic complete response (pCR) had a pooled AUC of 0.76 (95% CI 0.71–0.82). The most utilised predictive parameters were age, tumour grade and T stage. However, these models were prone to significant risk of bias and had limited applicability to the general population.

Although the existing models were statistically robust, they lacked broad applicability. This was primarily due to a lack of external validation, which limits their clinical utility. A future CXB‐specific model should prioritise dedicated data collection based on pre‐calculated sample size and include the predictive factors identified in this review.