基于SEER數據庫構建預測兒童髓母細胞瘤生存率的列線圖及外部驗證

【中圖分類號】 R739.41 【文獻標志碼】 【文章編號】1672-7770（2025）02-0193-08

Abstract： Objective To construct a nomogram model for predicting the survival rate childhood medulloblasma（ MB） based on the Surveillance， Epidemiology and End Results（SEER） database the National Cancer Institute the United States. Methods The data 854 children （ 0-17 years old） with complete clinical information diagnosed in the SEER database from 2000 2021 were collected. Univariate and multivariate COX regression analysis were used determine the independent facrs affecting the prognosis children with MB. The obtained prognostic independent facrs were used construct the nomogram model. The time-dependent receiver operating characteristic（ ROC） curve， consistency index（C-index）， prognostic calibration curve， and clinical decision curve analysis（DCA） were used test the predictive ability the model. A tal 32 children diagnosed with MB in from 2017 2024 were collected as an external validation group， and the ROC curve and C-index were used again test the predictive ability the model. ResultsAmong 854 children with MB selected from the SEER database， 531 were male and 323 were female. Multivariate Cox regression analysis showed that independent prognostic facrs included hispathological type， surgical method， radiotherapy， chemotherapy， and cerebrospinal fluid dissemination or metastasis （ Plt;0.05 ）. The above prognostic facrs were used construct a nomogram model. The area under ROC curve（AUC） for predicting 3-year， 5-year， and 10-year survival MB children in the modeling group was 0.669， 0.660， and 0.611， respectively， and the C-index was 0.647（ 95% CI =0.629 - 0.666）. The AUC for 1-year， 3-year and 5-year survival rates in the external validation group were 0.905， 0.836， and 0.754， respectively. The established nomogram model was verified by ROC curve， C-index， prognostic calibration curve and DCA curve， which proved that it had good predictive effect. Conclusions The hispathological type，surgical method， radiotherapy， chemotherapy and cerebrospinal fluid dissemination in children with MB are independent facrs affecting the prognosis children. The nomogram model constructed based on these independent facrs has good predictive performance and clinical practicality.

Key words： medulloblasma； prediction model； SEER； childrer

髓母細胞瘤（medulloblasma，MB）是一種高度惡性的神經外胚層腫瘤，是兒童最常見的惡性腦腫瘤，占所有兒童中樞神經系統惡性腫瘤的 。雖然采用了多種治療方法（包括手術切除，術后全腦及全脊髓的放化療等），但MB患兒預后仍未見明顯改善，5年無進展生存率和總生存率均低于 。MB 的某些預后特征，如組織學亞型和分子亞型分類等，可能會影響個別患兒的預后[3]。但MB患兒的預后往往受多種因素的影響。因此，建立一種包含多種預后因素的新預測模型以實現對MB患兒的生存時間及生存概率的精確預測，對臨床上指導MB患兒治療方案的制定至關重要。

監測，流行病學和最終結果（Surveillance，EpidemiologyandEndResults，SEER）數據庫是由美國國立癌癥研究所成立的，是臨床上常用的公共數據庫之一，其中包含了大量臨床腫瘤的回顧性研究資料[4]。列線圖是統計學上一種常用的預測模型，它通過圖形的方式將預測事件的發生率進行呈現，從而更加直觀地觀察到患者的預后情況[5]。已有許多研究使用列線圖來預測肝細胞癌[6]、胃癌[7]鼻咽癌[8]和乳腺癌[9]等癌癥的預后。然而，用于預測兒童MB預后的列線圖有限，且大多數基于生物數據庫建立的列線圖只進行了數據庫的內部驗證，并未經臨床樣本的外部驗證，這使得所構建的列線圖的臨床適用性大打折扣。因此，本研究通過回顧性分析SEER數據庫MB患兒的臨床數據，篩選出預后獨立因素構建列線圖，并將構建的列線圖在鄭州大學附屬兒童醫院2017—2024年收集的MB患兒的臨床樣本中進行了外部驗證，證實了列線圖的有效性，為臨床醫生評估MB患兒的預后提供了一種有效工具。……