Pediatric Hematology and Oncology, Department of Pediatrics III, University Hospital Essen and the University of Duisburg-Essen, Essen, Germany.
Center for Evidence-based Healthcare, University Hospital and Faculty of Medicine Carl Gustav Carus, TU Dresden, Germany.
Cancer Epidemiol. 2021 Feb;70:101873. doi: 10.1016/j.canep.2020.101873. Epub 2020 Dec 24.
The potential existence of spatial clusters in childhood cancer incidence is a debated topic. Identification of such clusters may help to better understand etiology and develop preventive strategies. We evaluated widely used statistical approaches to cluster detection in this context.
Incidence of newly diagnosed childhood cancer (140/1,000,000 children under 15 years) and nephroblastoma (7/1,000,000) was simulated. Clusters of defined size (1-50) were randomly assembled on the district level in Germany. Each cluster was simulated with different relative risk levels (1-100). For each combination 2000 iterations were done. Simulated data was then analyzed by three local clustering tests: Besag-Newell method, spatial scan statistic and Bayesian Besag-York-Mollié with Integrated Nested Laplace Approximation approach. The operating characteristics (sensitivity, specificity, predictive values, power and correct classification) of all three methods were systematically described.
Performance varied considerably within and between methods, depending on the simulated setting. Sensitivity of all methods was positively associated with increasing size, incidence and RR of the high-risk area. Besag-York-Mollié showed highest specificity for minimally increased RR in most scenarios. The performance of all methods was lower in the nephroblastoma scenario compared with the scenario including all cancer cases.
This study illustrates the challenge to make reliable inferences on the existence of spatial clusters based on single statistical approaches in childhood cancer. Application of multiple methods, ideally with known operating characteristics, and a critical discussion of the joint evidence seems recommendable when aiming to identify high-risk clusters.
儿童癌症发病率中潜在的空间聚集存在争议。识别这些聚集可能有助于更好地了解病因并制定预防策略。我们评估了在这种情况下广泛使用的聚类检测统计方法。
模拟了新诊断的儿童癌症（140/1,000,000 名 15 岁以下儿童）和肾母细胞瘤（7/1,000,000）的发病率。在德国的区一级上随机组合大小为 1-50 的聚类。用不同的相对风险水平（1-100）模拟每个聚类。对每种组合进行了 2000 次迭代。然后使用三种局部聚类检验方法（Besag-Newell 方法、空间扫描统计和贝叶斯 Besag-York-Mollié 与集成嵌套 Laplace 逼近方法）分析模拟数据。系统地描述了所有三种方法的操作特性（敏感性、特异性、预测值、功效和正确分类）。
在不同的模拟环境中，三种方法的性能在方法内部和方法之间都有很大差异。所有方法的敏感性都与高风险区域的大小、发病率和 RR 的增加呈正相关。Besag-York-Mollié 在大多数情况下对最小增加的 RR 具有最高的特异性。与包括所有癌症病例的情景相比，肾母细胞瘤情景中所有方法的性能都较低。
本研究说明了在儿童癌症中，基于单一统计方法对空间聚集的存在做出可靠推断所面临的挑战。应用多种方法，理想情况下具有已知的操作特性，并对联合证据进行批判性讨论，当旨在识别高风险聚集时，似乎是值得推荐的。
