Longitudinal assessment of biopsy-resection histological discordance in digestive tract epithelial neoplasms and development of a colorectal upgrading prediction model

Background Biopsy-based grading guides the management of digestive tract epithelial neoplasms. However, its concordance with post-endoscopic resection (ER) histology remains suboptimal, leading to clinically significant discrepancies. Large-scale real-world studies comparing different organs and examining long-term trends remain scarce. Methods In this single-center retrospective cohort study, we included consecutive patients who underwent ER for digestive tract epithelial neoplasms at Chinese PLA General Hospital from 2010 to 2024 and had available preoperative endoscopic forceps biopsy (EFB) results. The primary outcomes were histologic concordance, upgrading, and downgrading between EFB and ER histology across esophageal, gastric, and colorectal lesions. Temporal trends were evaluated by calendar year and predefined study periods with multivariable adjustment. Multinomial logistic regression was used to identify factors associated with pathological discordance. A nomogram for predicting upgrading in colorectal lesions was developed and temporally validated in an independent cohort in 2025. Results Among 7460 patients, the overall concordance rate was 69.0% (5149/7460), with upgrading in 24.1% (1794/7460) and downgrading in 6.9% (517/7460). Concordance rates were 56.7% (795/1401) in the esophagus, 55.1% (1444/2619) in the stomach, and 84.6% (2910/3440) in the colorectum. Concordance was significantly higher in the colorectum than in the esophagus (84.6% vs 56.7%; χ 2 = 429.88, df = 1; P < 0.001) or stomach (84.6% vs 55.1%; χ 2 = 638.08, df = 1; P < 0.001). Concordance increased across the three study periods, from 59.1% (554/937) in Period 1 to 65.1% (1571/2413) in Period 2 and 73.6% (3024/4110) in Period 3 ( P < 0.001). After multivariable adjustment, concordance remained high in Period 2 (odds ratio [OR], 1.193; 95% confidence interval [CI] 1.016–1.401; P = 0.031) and Period 3 (OR, 1.561; 95% CI 1.334–1.825; P < 0.001) than in Period 1. In multinomial analyses, upgrading was independently associated with older age (relative risk ratio [RRR], 1.008; 95% CI 1.003–1.014; P = 0.004), larger lesion size (RRR, 1.112; 95% CI, 1.071–1.155; P < 0.001), flat morphology (RRR, 1.204; 95% CI, 1.054–1.375; P = 0.006), depressed morphology (RRR, 1.361; 95% CI, 1.090–1.699; P = 0.007), nodularity (RRR, 1.249; 95% CI, 1.008–1.547; P = 0.042), ulceration (RRR, 1.428; 95% CI, 1.049–1.945; P = 0.024), hyperemia (RRR, 1.309; 95% CI, 1.127–1.520; P < 0.001), esophageal location (RRR, 3.357; 95% CI, 2.811–4.010; P < 0.001), and gastric location (RRR, 3.576; 95% CI, 3.089–4.140; P < 0.001). Downgrading was associated with larger lesion size (RRR, 1.087; 95% CI, 1.022–1.155; P = 0.008), esophageal location (RRR, 4.218; 95% CI, 3.143–5.661; P < 0.001), and gastric location (RRR, 4.296; 95% CI, 3.350–5.510; P < 0.001). The colorectal nomogram showed area under the receiver operating characteristic curves (AUCs) of 0.728 (95% CI, 0.698–0.758), 0.706 (95% CI, 0.659–0.753), and 0.703 (95% CI, 0.631–0.775) in the training, validation, and temporal validation cohorts, respectively. Conclusions Histological discordance between EFB and ER was frequent and was predominantly caused by upgrading. Although concordance improved from 2010 to 2024, discrepancies remained common, particularly in upper gastrointestinal lesions. These findings highlight the need for careful targeted biopsies and risk-based pre-resection assessments, particularly for upper gastrointestinal lesions with higher risks of pathological upgrading.