目的 确定小儿髓母细胞瘤（MB）免疫组化结果中是否存在预后预测潜力的标志物,并整合临床特征构建指导MB患儿术后个性化管理的风险分层系统。方法 回顾分析2011年1月1日至2020年8月1日因MB行切除手术患儿的临床资料,构建logistic回归模型预测术后肿瘤残余这一短期结局（STO）。根据术后STO和影像转移特征将患儿分为高风险组和低风险组,采用COX回归模型分析影响复发、再次手术和生存等长期结局（LTO）的独立危险因素,应用R软件构建列线图模型,通过C指数、校准曲线和决策曲线分析（DCA）评价模型。结果 纳入MB患儿111例,男67例、女44例,中位年龄7.0（4.0~9.0）岁。肿瘤直径>5.0 cm（OR=8.07,95% CI：2.62~24.89）和MYC蛋白表达（OR=4.03,95% CI：1.29~12.63）可预测STO（AUC=0.81,95% CI：0.69~0.92,P<0.001）。高风险组（n=50）12个月无LTO生存率（LOFS）为76.0%（95% CI：70.0%~82.0%）,低风险组（n=61）为83.6%（95% CI：78.9%~88.3%）,两组差异有统计学意义（HR=2.77,95% CI：1.53~5.01）。COX回归分析显示,风险分组、年龄≤3岁、肿瘤直径>5.0 cm、MYC阳性是患儿LTO不佳的独立危险因素（P<0.05）,基于以上临床-生物独立危险因素建立列线图可准确预测患儿LOFS,其C指数为0.715,并且在校准曲线中显示列线图预测风险与实际发生风险的一致程度较高,DCA中在较大的预测概率范围内该列线图都有不错的净收益,展示出良好的临床效用。结论 使用风险分组、年龄、肿瘤直径和MYC构建的列线图有利于对小儿MB进行有针对性的监测和管理。

Objective To identify potential prognostic markers of pediatric medulloblastoma (MB) in immunohistochemical results, and to construct a risk stratification system based on the integration of clinical characteristics, so as to guide the postoperative personalized management of MB children. Methods The clinical data of children with MB undergoing resection from January 1, 2011 to August 1, 2020 were retrospectively analyzed, and the logistic regression model was constructed to predict the short-term outcomes (STO) (postoperative tumor residue). The children were divided into high-risk group and low-risk group according to postoperative STO and imaging metastasis characteristics. The COX regression model was used to analyze the independent risk factors for long-term outcome (LTO) including recurrence, reoperation and survival. R software was used to construct the nomogram, and the nomogram was evaluated by C-index, calibration curve and decision curve analysis (DCA). Results A total of 111 children (67 boys and 44 girls) with MB were included, with a median age of 7.0 (4.0-9.0) years. Tumor diameter >5.0 cm (OR=8.07, 95% CI: 2.62-24.89) and MYC protein expression (OR=4.03, 95% CI: 1.29-12.63) can predict STO (AUC=0.81, 95% CI: 0.69-0.92, P<0.001). The 12-month LTO-free survival (LOFS) of the high-risk group (n=50) was 76.0% (95% CI: 70.0%-82.0%), and that of the low-risk group (n=61) was 83.6% (95% CI: 78.9% -88.3%). The difference between the two groups was statistically significant (HR=2.77, 95% CI: 1.53-5.01). COX regression model showed that risk grouping, age≤3 years, diameter>5.0 cm, and MYC positive were independent risk factors for poor LTO (P<0.05). Based on the above clinical-biological independent risk factors, LOFS can be accurately predicted by establishing a nomogram, with a C-index of 0.715. Moreover, the calibration curve showed a high degree of consistency between the predicted risk and the actual risk. The net benefit was good in DCA for a wide range of prediction probabilities, demonstrating a good clinical utility. Conclusions The nomogram constructed based on risk grouping, age, tumor diameter and MYC are conducive to the targeted monitoring and management of pediatric MB.