Development and validation of a histological calculator for anastomotic margins to predict anastomotic failure among rectal cancer patients treated with neoadjuvant chemoradiotherapy

Abstract

Purpose: To identify histological features of anastomotic margins and develop a prediction model for anastomotic failure (AF) in rectal cancer (RC) patients with neoadjuvant chemoradiotherapy (nCRT).

Methods: A total of 350 pairs anastomotic "doughnuts" from RC with nCRT were randomly divided into the primary and validation cohorts at a ratio of 7:3. The histological features were identified and constructed using LASSO (Least absolute shrinkage and selection operator) regression to develop the radiation-induced colorectal injury (RCI) score. An AF prediction mode based on the RCI score was built and evaluated using the area under the receiver operating characteristic curve (AUC), decision curve analysis (DCA), and the DeLong test.

Results: The primary cohort consisted of 245 patients, among whom AF occurred in 26.9% of cases, while the validation cohort comprised 105 patients, with an AF rate of 24.8%. The RCI score of anastomotic margins showed a significant correlation with AF (odds ratio: 2.963; 95% confidence interval [CI]: 2.298-3.822; P < 0.001). Multivariable analysis identified body mass index (BMI) < 18.5, tumor location, long-course radiotherapy, and the RCI score as independent predictors for AF. The nomogram based on the RCI score exhibited good discrimination in both the primary cohort (AUC: 0.886; 95% CI: 0.840-0.931), with a sensitivity of 86.36% (95% CI, 75.7-93.6%) and specificity of 76.54% (95% CI, 69.6-82.5%). Calibration curves revealed satisfactory agreement between the predicted and the observed probabilities.

Conclusions: The comprehensive nomogram incorporating the RCI score could assist physicians in predicting AF and formulating personalized treatment strategies for RC patients with neoadjuvant radiotherapy.

Keywords: Anastomotic failure; Nomogram; Radiation-induced colorectal injury; Radiotherapy; Rectal cancer.

Fig. 1

A A computation framework was used to establish the collagen signature. The hematoxylin-eosin (HE) and Masson’s staining of anastomotic margins were chosen for histological features extraction, including large intestinal glands atrophy, muscularis mucosa integrity, fibrosis, vascular stenosis, inflammatory infiltration, and mucosal edema. Next, the potential predictors were selected using LASSO (least-absolute shrinkage and selection operator) logistic regression. The RCI score can be calculated by these potential predictors. B Recruitment pathways for patients in the primary and validation cohorts. A total of 350 proximal and distal margins from rectal cancer patients who were treated with neoadjuvant chemoradiotherapy (nCRT) and sphincter-preserving surgery at our center from August 2012 and March 2021 were enrolled in this study

Fig. 2

Histological features of the RCI. A Large intestinal glands atrophy (0–3), 0-normal, 1-mild gland atrophy with crypt disarray, 2-moderate mucosal gland atrophy with crypt distortion, and 3-severe mucosal gland atrophy with crypt dropout or loss. B Muscularis mucosa integrity (0–2), 0-normal,1- partial fibrosis or rupture, and 2-complete fibrosis or disappearance. C Intestinal fibrosis 0–4. 0- no fibrosis (normal), 1- obvious fibrosis in the submucosa or subserosa, 2- fibrosis replacing part of the muscular propria, 3- septa through the muscular propria and an increase in subserosa collagen, 4 - significant transmural scar, marked subserosal collagen. D Vascular stenosis (0–3), 0-normal, 1-less than 50%, 2–50–75%, and 3–76%-above. E Inflammatory infiltration (0–2), 0-normal, 1-focal distribution, and 2-extensive distribution. F Mucosal edema (0–1), 0- normal, and 1-mucosal edema. 40× Scale: 200 μm; 100× Scale: 100 μm

Fig. 3

Nomogram indicating the risk of anastomotic failure (AF) in patients with rectal cancer who were undergoing neoadjuvant chemoradiotherapy and sphincter-preserving surgery. For clinical use, the RCI score is determined by drawing a line straight up to the point axis to establish the score associated with the differentiation. Next, this process is repeated for the other three covariates (neoadjuvant radiotherapy, tumor location and BMI). The scores of each covariate are added, and the total score is located on the total score point axis. Finally, a line is drawn straight down to the risk of the AF axis to obtain the probability. F = fractions, BMI = Body Mass Index

Fig. 4

Comparison of prediction models for anastomotic failure (AF). A Receiver operating characteristic (ROC) curves of the nomogram and the clinical model for predicting AF in the training cohort. B ROC curves of the nomogram and the clinical model for predicting AF in the testing cohort. C ROC curves of the nomogram and the clinical model for predicting AF in all 350 patients. D Decision curve analysis for the nomogram and clinical model in all 350 patients