Predicting Pathological Complete Response in Rectal Cancer Using Machine Learning

This study aims to develop and validate a predictive model based on pre-neoadjuvant clinical, laboratory, and magnetic resonance imaging (MRI) features to estimate the probability of pathological complete response (pCR) in rectal cancer patients after neoadjuvant chemoradiotherapy (nCRT). This retrospective study will enroll patients who received nCRT followed by radical resection at Peking University People's Hospital between December 2017 and October 2025 as the development cohort. Least Absolute Shrinkage and Selection Operator (LASSO) regression will be used for feature selection, and machine learning algorithms will be applied to construct the prediction model. Model performance will be comprehensively evaluated using the receiver operating characteristic (ROC) curve, precision-recall curve, calibration curve, and decision curve analysis (DCA). SHapley Additive exPlanations (SHAP) analysis will be performed to enhance model interpretability. The final model is expected to provide an individualized pCR prediction tool to guide clinical decision-making for rectal cancer patients.

• No interventions (Diagnostic test): No interventions

Primary outcomes

• Pathological Complete Response (pCR) defined by Tumor Regression Grade (TRG) [Evaluated during routine histopathological examination of the resected surgical specimen immediately following radical surgery (typically within 1 to 2 weeks post-surgery).]: The primary endpoint is the occurrence of pCR, assessed by two independent pathologists using the AJCC/CAP Tumor Regression Grade (TRG) system. TRG 0 (no viable cancer cells, only fibrosis or mucin pools) is defined as a positive outcome (pCR). TRG 1 to 3 are combined and defined as a negative outcome (non-pCR). The predictive performance of the model will be evaluated utilizing several metrics including the Area Under the ROC Curve (AUC), Precision-Recall (PR) curve, Calibration curve, and Decision Curve Analysis (DCA).

Secondary outcomes

• Area under the receiver operating characteristic curve (AUC) of the prediction model [At the completion of model development and validation]: To evaluate the discrimination performance of the model for pCR prediction
• Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the prediction model [At the completion of model development and validation]: To evaluate the diagnostic accuracy of the model at the optimal cut-off value
• Calibration curve of the prediction model [At the completion of model development and validation]: To evaluate the consistency between the predicted pCR probability and the actual observed pCR rate
• Net benefit of the model quantified by decision curve analysis (DCA) [At the completion of model development and validation]: To evaluate the clinical utility of the model across different threshold probabilities
• Variable importance quantified by SHapley Additive exPlanations (SHAP) analysis [At the completion of model development and validation]: To interpret the contribution of each predictor to the model prediction

Sex: All, ages: 18 Years to any

Inclusion Criteria:

1. Patients with histopathologically confirmed rectal adenocarcinoma;
2. Clinical stage cT3-4, or cN+, or M1 advanced rectal cancer;
3. Received standardized neoadjuvant chemoradiotherapy or neoadjuvant chemotherapy;
4. Underwent total mesorectal excision (TME) after the completion of neoadjuvant therapy, with complete postoperative pathological data available.

Exclusion Criteria:

1. Previous history of other malignant tumors;
2. Incomplete clinical data;
3. Underwent emergency surgery during nCRT;
4. Complicated with systemic infection or hematological diseases.

• Kong JC, Guerra GR, Warrier SK, Lynch AC, Michael M, Ngan SY, Phillips W, Ramsay G, Heriot AG. Prognostic value of tumour regression grade in locally advanced rectal cancer: a systematic review and meta-analysis. Colorectal Dis. 2018 Jul;20(7):574-585. doi: 10.1111/codi.14106. Epub 2018 May 8. PMID 29582537
• Maas M, Nelemans PJ, Valentini V, Das P, Rodel C, Kuo LJ, Calvo FA, Garcia-Aguilar J, Glynne-Jones R, Haustermans K, Mohiuddin M, Pucciarelli S, Small W Jr, Suarez J, Theodoropoulos G, Biondo S, Beets-Tan RG, Beets GL. Long-term outcome in patients with a pathological complete response after chemoradiation for rectal cancer: a pooled analysis of individual patient data. Lancet Oncol. 2010 Sep;11(9):835-44. doi: 10.1016/S1470-2045(10)70172-8. Epub 2010 Aug 6. PMID 20692872

• Hong-Peng Jiang, docter, Study chair, Peking University People's Hospital