Objective: To explore the risk factors of colorectal advanced adenomas (AA) and construct a nomogram to predict the risk of colorectal AAs. Methods: Clinical data of patients were retrospectively collected who underwent their first colonoscopy from January 2017 to December 2020 in the First Affiliated Hospital of Nanjing Medical University and were pathologically confirmed harboring colorectal polyps. A credible random split-sample method was used to divide data into training and validation cohorts (split ratio=7∶3). Univariate and multivariate logistic regression analysis were used to identify the predictors of colorectal advanced adenomas, and a nomogram was developed based on the above results. The validation cohort was used for internal validation of the nomogram. The discriminatory value of the nomogram was evaluated using the area under the receiver operating characteristic (ROC) curve (AUC). The consistency between actual outcomes and predicted probabilities was evaluated by the calibration curve. The clinical validity of the model was evaluated by the decision analysis curve (DCA). Results: A total of 1 936 patients with colorectal polyps were eligible. Including 1 356 patients in the training cohort (840 males and 516 females), and 580 patients in the validation cohort (379 males and 201 females), with the mean ages of (57.4±9.8) and (57.6±9.7) years, respectively. There were 1 502 (77.6%) patients without AAs and 434 [22.4%,1-9 mm 73(16.8%) cases、>9-<20 mm 271(62.5%) cases、≥20 mm 90(20.7%) cases] patients with AAs. The regression analysis found that age (OR=1.018, 95%CI:1.003-1.033), fatty liver (OR=1.870, 95%CI:1.274-2.744), low-density lipoprotein (LDL) (OR=1.378, 95%CI:1.159-1.637), fecal occult blood test (FOBT) (OR=2.597, 95%CI:1.857-3.631), and location of adenomas [proximal (OR=2.869, 95%CI:1.727-4.764), distal (OR=2.791, 95%CI:1.721-4.527)] were identified as predictors of colorectal AAs. The AUC of the nomogram was 0.664 (95%CI:0.630-0.698) in the training cohort and 0.640 (95%CI:0.587-0.693) in the validation cohort. The calibration curve showed good consistency between the predicted and actual risk, and the Hosmer-Lemeshow (H-L) test P value was 0.830 and 0.150 in the training cohort and the validation cohort. DCA demonstrated that the nomogram had a better clinical application value. Conclusions: A nomogram with five predictors, including age, fatty liver, LDL, FOBT, and location of adenomas, helped predict the risk of colorectal AAs in patients with polyps and implemented colorectcal cancer stratified screening strategy for colonoscopy in the high-risk population.
目的： 通过分析结直肠进展性腺瘤（AA）发生的危险因素，构建预测结直肠AA发生风险的列线图模型。 方法： 回顾性收集2017年1月至2020年12月在南京医科大学第一附属医院行首次结肠镜检查并病理证实为结直肠息肉的患者临床资料，将整个队列随机分成训练集和验证集（7∶3）。通过单因素和多因素logistic回归分析结直肠AA发生的危险因素，并基于以上因素构建列线图模型。运用验证集进行模型内部验证，通过受试者工作特征（ROC）曲线下面积（AUC）评价模型的区分度，Calibration校准曲线评估模型预测概率与实际结果的一致性，决策分析曲线（DCA）评估模型的临床有效性。 结果： 共有1 936例结直肠息肉患者纳入分析，其中1 356例患者纳入训练集（男840例、女516例），580例患者纳入验证集（男379例、女201例），年龄分别为（57.4±9.8）和（57.6±9.7）岁；其中无AA患者1 502例（77.6%）、有AA患者434例［22.4%，其中1~9 mm 73例（16.8%）、>9~<20 mm 271例（62.5%）、≥20 mm 90例（20.7%）］。回归分析发现，年龄［OR=1.018，95%可置信区间（CI）：1.003~1.033］、脂肪肝（OR=1.870，95%CI：1.274~2.744）、低密度脂蛋白（LDL）（OR=1.378，95%CI：1.159~1.637）、粪便隐血试验（FOBT）（OR=2.597，95%CI：1.857~3.631）、息肉部位［近端（OR=2.869，95%CI：1.727~4.764）、远端（OR=2.791，95%CI：1.721~4.527）］与结直肠AA发生有关。该模型的AUC在训练集和验证集分别为0.664（95%CI：0.630~0.698）、0.640（95%CI：0.587~0.693），Calibration校准曲线表明预测和实际风险一致性良好，Hosmer-Lemeshow（H-L）检验P值在训练集和验证集分别为0.830、0.150，DCA表明该列线图具有临床应用价值。 结论： 基于年龄、脂肪肝、LDL、FOBT、息肉部位这5个预测因素构建的列线图模型有助于预测结直肠息肉患者发生AA的风险，从而实现针对高危人群进行结直肠癌精准分层筛查策略。.