Choudhury Ananya, Theophanous Stelios, Lønne Per-Ivar, Samuel Robert, Guren Marianne Grønlie, Berbee Maaike, Brown Peter, Lilley John, van Soest Johan, Dekker Andre, Gilbert Alexandra, Malinen Eirik, Wee Leonard, Appelt Ane L
MAASTRO (Dept of Radiotherapy), GROW School of Oncology and Developmental Biology, Maastricht University & Maastricht University Medical Centre+, Limburg, The Netherlands.
Leeds Institute of Medical Research at St James's, University of Leeds, United Kingdom.
Radiother Oncol. 2021 Jun;159:183-189. doi: 10.1016/j.radonc.2021.03.013. Epub 2021 Mar 20.
Predicting outcomes is challenging in rare cancers. Single-institutional datasets are often small and multi-institutional data sharing is complex. Distributed learning allows machine learning models to use data from multiple institutions without exchanging individual patient-level data. We demonstrate this technique in a proof-of-concept study of anal cancer patients treated with chemoradiotherapy across multiple European countries.
atomCAT is a three-centre collaboration between Leeds Cancer Centre (UK), MAASTRO Clinic (The Netherlands) and Oslo University Hospital (Norway). We trained and validated a Cox proportional hazards regression model in a distributed fashion using data from 281 patients treated with radical, conformal chemoradiotherapy for anal cancer in three institutions. Our primary endpoint was overall survival. We selected disease stage, sex, age, primary tumour size, and planned radiotherapy dose (in EQD2) a priori as predictor variables.
The Cox regression model trained across all three centres found worse overall survival for high risk disease stage (HR = 2.02), male sex (HR = 3.06), older age (HR = 1.33 per 10 years), larger primary tumour volume (HR = 1.05 per 10 cm) and lower radiotherapy dose (HR = 1.20 per 5 Gy). A mean concordance index of 0.72 was achieved during validation, with limited variation between centres (Leeds = 0.72, MAASTRO = 0.74, Oslo = 0.70). The global model performed well for risk stratification for two out of three centres.
Using distributed learning, we accessed and analysed one of the largest available multi-institutional cohorts of anal cancer patients treated with modern radiotherapy techniques. This demonstrates the value of distributed learning in outcome modelling for rare cancers.
预测罕见癌症的预后具有挑战性。单机构数据集通常较小，多机构数据共享又很复杂。分布式学习使机器学习模型能够使用来自多个机构的数据，而无需交换个体患者层面的数据。我们在一项针对多个欧洲国家接受放化疗的肛管癌患者的概念验证研究中展示了这项技术。
atomCAT是利兹癌症中心（英国）、马斯特里赫特诊所（荷兰）和奥斯陆大学医院（挪威）之间的三中心合作项目。我们使用来自三个机构接受根治性适形放化疗的281例肛管癌患者的数据，以分布式方式训练并验证了Cox比例风险回归模型。我们的主要终点是总生存期。我们预先选择疾病分期、性别、年龄、原发肿瘤大小和计划放疗剂量（以等效剂量2表示）作为预测变量。
在所有三个中心训练的Cox回归模型发现，高危疾病分期（HR = 2.02）、男性（HR = 3.06）、年龄较大（每10年HR = 1.33）、原发肿瘤体积较大（每10 cm HR = 1.05）和放疗剂量较低（每5 Gy HR = 1.20）的患者总生存期较差。验证期间平均一致性指数为0.72，各中心之间差异有限（利兹 = 0.72，马斯特里赫特 = 0.74，奥斯陆 = 0.70）。全局模型在三个中心中的两个中心的风险分层中表现良好。
通过使用分布式学习，我们获取并分析了接受现代放疗技术治疗的肛管癌患者中最大的可用多机构队列之一。这证明了分布式学习在罕见癌症预后建模中的价值。
