组织工程与重建外科杂志

组织工程与重建外科杂志 ›› 2024, Vol. 20 ›› Issue (2): 190-.

• • 上一篇    下一篇

增生性瘢痕的影响因素分析及风险预测模型研究

  

  • 出版日期:2024-04-30 发布日期:2024-05-14

Analysis of influencing factors and risk prediction model of hyperplastic scar

  • Online:2024-04-30 Published:2024-05-14

PDF (PC)

52

摘要：

目的　探究增生性瘢痕的影响因素并建立风险预测模型。方法　选择2021年4月至2023年4月我院收治的
102例增生性瘢痕患者作为增生性瘢痕组，102例伤口痊愈后未转化为增生性瘢痕的患者作为对照组。比较两组临床
资料，采用随机森林算法筛选变量，并通过多因素 Logistic回归分析影响增生性瘢痕形成的因素；相关性 E值法对研究
结果的灵敏度进行分析；将多因素分析结果中的 β值代入回归方程 y=1－1/（1+e-z），建立预测模型并评价其预测效能。
结果　随机森林算法筛选出9个变量，多因素Logistic回归分析结果显示，年龄≤30岁、瘢痕史、家族瘢痕史、创面愈合不
良史、辛辣饮食习惯、伤口类型为烧伤、转化生长因子 β1（TGF-β1）水平升高等，是导致增生性瘢痕形成的危险因素
（P＜0.05），采取瘢痕预防措施、肿瘤坏死因子α（TNF-α）水平升高为保护因素（P＜0.05）；E=1.984，95%CI下限为1.216，
研究结果的灵敏度较高；当模型预测增生性瘢痕形成概率为 0.85 时，约登指数最高（74.38），预测效果最好，预测准确
度、灵敏度和特异度分别为 89.03%、85.73%、88.65%。预测模型的 ROC 曲线下面积为 0.847（95%CI：0.782~0.913，P<
0.001），区分度较好；十字交叉试验结果显示，训练集和验证集模型参数的拟合度较高（Nagelkerke R2 =0.602），模型较稳
定。结论　年龄、瘢痕史、饮食习惯、血清炎症因子水平等与增生性瘢痕的形成有关，临床上应采取个性化干预措施，以
降低增生性瘢痕形成率。

关键词:

Abstract:

Objective To explore the influencing factors of hyperplastic scar and establish a risk prediction model. Methods From April 2021 to April 2023， 102 patients with hyperplastic scar treated in our hospital were selected as hyperplastic scar group， and 102 patients without hyperplastic scar were selected as control group. The clinical data of the two groups were compared， random forest algorithm was used to screen variables， and multivariate Logistic regression analysis was used to analyze the factors affecting the formation of hyperplastic scar. The sensitivity of the research results was analyzed by correlation E-value method. The value of β in the results of multifactor analysis was substituted into the regression equation y=1－1 / （ 1+e-z）， and the prediction model was established and its prediction efficiency was evaluated. Results A total of 9 variables were selected by random forest algorithm. Multivariate Logistic regression analysis showed that age ≤ 30 years old， scar history， family scar history， history of poor wound healing， spicy eating habits， wound type of burn， and elevated level of transforming growth factor β1（ TGF-β1） were risk factors for hyperplastic scar formation. Scar prevention measures and the increase of tumor necrosis factor-α（ TNF-α） level were protective factors. E=1.984， with a 95% CI lower limit of 1.216， the sensitivity of the research results was relatively high. When the probability of hypertrophic scar formation predicted by the model was 0.85， the Youden's J statistic was the highest （74.38）， and the prediction effect was the best. The prediction accuracy， sensitivity and specificity were 89.03%，85.73% and 88.65% respectively. The area under the ROCcurve of the subjects was 0.847（95% CI：0.782-0.913， P<0.001）， indicating good differentiation. The results of the cross over experiment showed that the fitting degree of the model parameters in the training and validation sets was high  （Nagelkerke R2=0.602）， and the model was relatively stable. Conclusion Age， scar history， dietary habits， and serum  inflammatory factor levels are related to the formation of hypertrophic scars. Personalized intervention measures should be  taken in clinical practice to reduce the rate of hypertrophic scar formation.

Key words: