宫颈锥切术后创面修复不良的危险因素及Nomogram预测模型的构建

Risk factors for unfavorable wound healing after cold knife conization and construction of a nomogram prediction model

摘要:

目的 探讨宫颈锥切术（CKC）后创面修复不良的危险因素，并构建Nomogram预测模型。方法 回顾性选取2020年1月—2025年1月在中国人民解放军南部战区总医院完成CKC术的567例宫颈上皮内瘤变（CIN）患者的临床资料，依据患者术后创面修复情况分为不良组和良好组，分别有51、516例。分析两组患者基线资料，影响因素的分析采用多因素一般Logistic回归模型；构建列线图；绘制受试者工作特征（ROC）曲线、临床决策曲线和校准曲线，评估该列线图模型的区分度及拟合情况。结果 两组年龄、孕次构成、产次构成、体质量指数、CIN分级构成、绝经率、吸烟史率、高血压史率、初次性生活年龄构成、病变面积占比、病变累及象限、病变累及腺体、手术时间、术中出血量比较，差异均无统计学意义（P >0.05）。不良组糖尿病史率、术前阴道微环境异常率、术后4周内性生活恢复率、锥切组织高度、人乳头瘤病毒（HPV）感染率均高于良好组（P <0.05），血红蛋白水平低于良好组（P <0.05）。多因素一般Logistic回归分析结果显示：合并糖尿病［O^R=4.737（95% CI：1.540，14.564）］、术前阴道微环境异常［O^R=2.753（95% CI：1.206，6.281）］、锥切组织高度高［O^R=218.344（95% CI：29.775，1601.141）］、术后4周内恢复性生活［O^R=34.968（95% CI：14.173，86.273）］及HPV感染［O^R=8.467（95% CI：2.044，35.065）］均为CKC术后创面修复不良的独立危险因素（P <0.05），血红蛋白水平高［O^R=0.946（95% CI：0.919，0.973）］为CKC术后创面修复不良的保护因素（P <0.05）。基于多因素一般Logistic回归模型结果，构建CKC术后创面修复不良的风险预测列线图模型，P =1/［1+e-（-16.703+1.555×糖尿病+1.031×术前微环境-0.056×血红蛋白+5.386×锥切组织高度+3.554×术后4周恢复性生活+2.136×HPV感染）］。ROC曲线分析结果显示，CKC术后创面修复不良的曲线下面积为0.932（95% CI：0.907，0.964），特异性、敏感性分别为91.5%（95% CI：0.708，0.938）、84.8%（95% CI：0.886，0.939）。模型进行Hosmer-Lemeshow拟合优度检验，该模型拟合度良好（P >0.10）。临床决策曲线结果示，模型可提供临床净收益。结论 合并糖尿病、术前阴道微环境异常、锥切组织高度高、术后4周内恢复性生活及HPV感染均是CKC术后创面修复不良的独立危险因素，血红蛋白水平高为CKC术后创面修复不良的保护因素，基于以上影响因素构建的列线图预测模型具有较高的预测效能。

Abstract:

Objective To analyze the risk factors for poor wound healing after cold knife conization (CKC) and construct a nomogram prediction model.Methods The clinical data of 567 patients with cervical intraepithelial neoplasia (CIN) who underwent CKC at General Hospital of Southern Theater Command from January 2020 to January 2025 were retrospectively collected. Based on postoperative wound healing status, patients were divided into the unfavorable wound-healing group (n = 51) and the favorable wound-healing group (n = 516). Baseline data of the two groups were analyzed, and a generalized multivariate logistic regression model was used to identify influencing factors. A nomogram was constructed, and ROC curves, clinical decision curves, and calibration curves were plotted to evaluate discrimination and fit of the nomogram model.Results Comparison of two groups in terms of age, gravidity, parity, BMI, CIN grade, menopause rate, smoking history rate, hypertension history rate, age at first intercourse, lesion area proportion, number of involved quadrants, glandular involvement, surgery time, and intraoperative blood loss showed no statistically significant differences (P > 0.05). The unfavorable wound-healing group had a significantly higher prevalence of diabetes, preoperative abnormal vaginal microenvironment, resumption of sexual intercourse within 4 weeks after surgery, greater cone height, and a higher prevalence of HPV infection than the favorable wound-healing group (P < 0.05), whereas the hemoglobin level was significantly lower than that in the favorable wound-healing group (P < 0.05). Multivariate binary logistic regression analysis showed that diabetes mellitus [O^R = 4.737 (95% CI: 1.540, 14.564) ], preoperative abnormal vaginal microenvironment [O^R = 2.753 (95% CI: 1.206, 6.281) ], greater cone height [O^R = 218.344 (95% CI: 29.775, 1601.141) ], resumption of sexual intercourse within 4 weeks post-surgery [O^R = 34.968 (95% CI: 14.173, 86.273) ], and HPV infection [O^R = 8.467 (95% CI: 2.044, 35.065) ] were independent risk factors for unfavorable wound healing after CKC (P < 0.05), while higher hemoglobin level [O^R = 0.946 (95% CI: 0.919, 0.973) ] was a protective factor against unfavorable wound healing after CKC (P < 0.05). Based on the results of the multivariate binary logistic regression model, a nomogram prediction model for unfavorable wound healing after CKC was constructed: P = 1 / [1 + e-(-16.703 + 1.555 × diabetes + 1.031 × preoperative vaginal microenvironment - 0.056 × hemoglobin + 5.386 × cone height + 3.554 × resumption of sexual intercourse within 4 weeks post-surgery + 2.136 × HPV infection) ]. ROC curve analysis showed that the area under the curve for unfavorable wound healing after CKC was 0.932 (95% CI: 0.907, 0.964), with a specificity of 91.5% (95% CI: 0.708, 0.938) and sensitivity of 84.8% (95% CI: 0.886, 0.939). The Hosmer-Lemeshow goodness-of-fit test indicated that the model had a good fit (P > 0.10). The clinical decision curve showed that the model provided a net clinical benefit.Conclusion Factors associated with unfavorable wound healing after CKC include diabetes, an abnormal preoperative vaginal microenvironment, a low hemoglobin level, greater cone height, resumption of sexual life within 4 weeks postoperatively, and HPV infection. The nomogram prediction model constructed based on these factors demonstrates high predictive efficacy.