Analysis of risk factors and construction of prediction model of pathological diagnosis upgrading after resection of colorectal laterally spreading tumors
10.3760/cma.j.cn311367-20231229-00234
- VernacularTitle:结直肠侧向发育型肿瘤切除后病理诊断升级的危险因素分析及预测模型构建
- Author:
Erfeng LI
1
;
Jing PANG
;
Libin ZHANG
;
Wenbin ZHANG
;
Feng WANG
;
Bin GUO
Author Information
1. 山西省肿瘤医院(中国医学科学院肿瘤医院山西医院 山西医科大学附属肿瘤医院)内镜中心，太原 030012
- Keywords:
Colorectal laterally spreading tumor;
Biopsy;
Risk factor
- From:
Chinese Journal of Digestion
2024;44(6):391-397
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To investigate the risk factors affecting pathological diagnosis upgrading after resection of colorectal laterally spreading tumor (LST).Methods:From June 2018 to December 2022, the clinical data of 256 patients with LST (297 lesions) admitted to Shanxi Provincial Cancer Hospital were retrospectively included as an modeling group.From January 2023 to January 2024, 125 patients with LST (129 lesions) were collected as an external validation group. The pathological diagnosis of endoscopic forceps biopsy (EFB) samples and the resected LST tissue of modeling group were compared, and the patients were divided into pathological non-difference group and pathological upgrading group. The clinical data such as gender, age, body mass index (BMI), pre-resection carcinoembryonic antigen levels, drinking history, smoking history, family history of colorectal cancer, and whether complicated with underlying diseases as well as endoscopic surface morphological features such as lesion size, morphological features, and lesion location were compared between the two groups. Chi-square test was used for statistical analysis, and multivariate logistic regression analysis was used to identify the risk factors for pathological diagnosis upgrading after resection. Based on the independent risk factors, the prediction models were established and validated by nomogram. The receiver operating characteristic curve (ROC) of repeated samples within the modeling group and external validation growp was plotted, and the area under the curve (AUC) was used to evaluate the predictive value of the model.Results:The proportion of patients with family history of colorectal cancer in the pathological upgrading group was higher than that of the pathological non-difference group (38.7%, 12/31 vs. 22.2%, 50/225), and the difference was statistically significant ( χ2=4.04, P=0.045). There were statistically significant differences in lesion size (63.9% (23/36) and 44.4% (116/261) lesions with long diameter ≥2 cm, respectively), surface morphological characteristics (flat elevated type accounted for 8.3% (3/36) and 22.6% (59/261), granular uniform type accounted for 11.1% (4/36) and 28.0% (73/261), nodular mixed type accounted for 44.4% (6/36) and 24.9% (65/261), pseudo-depressed type accounted for 36.1% (13/36) and 24.5% (64/261)), and lesion location (distal colon accounted for 22.2% (8/36) and 33.3% (87/261), proximal colon accounted for 16.7% (6/36) and 28.7% (75/261), and rectum accounted for 61.1% (22/36) and 37.9% (99/261)) between the pathological upgrading group and the pathological non-difference group ( χ2=4.80, 12.62 and 7.08, all P<0.05). The results of multivariate logistic regression analysis showed that family history of colorectal cancer ( OR=2.211, 95% confidence interval (95% CI) 1.005 to 4.861, P=0.049), lesion length ≥ 2 cm ( OR=2.212, 95% CI 1.074 to 4.555, P=0.031), nodular mixed subtype ( OR=4.841, 95% CI 1.343 to 17.455, P=0.016), pseudo-depressed subtype ( OR=3.995, 95% CI 1.084 to 14.721, P=0.037), and lesion in rectum ( OR=2.417, 95% CI 1.024 to 5.705, P=0.044) were independent risk factors for pathological diagnosis upgrading after LST resection. A nomogram was established based on these four risk factors, with a ROC AUC of 0.833 (95% CI 0.752 to 0.913). The external validation results demonstrated that the ROC AUC was 0.848 (95% CI 0.736 to 0.960), the sensitivity was 0.737, the specificity was 0.972, the maximum Youden index was 0.712, and the overall accuracy was 0.868. Conclusions:Family history of colorectal cancer, lesion length ≥ 2 cm, lesion in rectum, and nodular mixed or pseudo-depressed subtypes may affect the accuracy of pathological diagnosis of LST lesions by EFB, and leading to pathological diagnosis upgrading after resection. The prediction model based on these four factors has good predictive efficacy in pathological diagnosis upgrading after LST resection.
