Single-center versus multi-center data sets for molecular prognostic modeling: a simulation study

摘要

Oncological treatment is based on surgery, radiotherapy, chemotherapy and immunotherapy for reduction of tumor burden and for improvement of local control of the tumor. Of particular importance is radiotherapy, which has been shown in numerous studies to improve local control and overall survival of patients [ , ]. Radiation oncology treatment strives to optimize the reduction of tumor cells while preserving the surrounding non-tumor tissue. Effectiveness is influenced by a number of factors such as radiation sensitivity, the anatomical borders and immunogenic constitution of the tumor, and its environment [ ]. The interplay between these factors is complex and prediction of the radiation response and overall clinical performance requires detailed measurement of the underlying molecular state of the tissue. This is increasingly attempted through the use of systemic multi-level omics biology approaches [ , ]. The complexity of the interplay is consistently reflected in the heterogeneous risks of subgroups of cancer patients in terms of local and distant control and overall survival, e.g. in head and neck cancer or glioblastoma [ , ]. This heterogeneity is a great challenge in oncology since it means that only a subgroup of treated patients is likely to benefit from standard therapy. Hence, the need for prognostic factors predicting individual response is great and a lot of research effort has been invested in the past decade to identify molecular prognostic markers from multi-level omics data generated from clinical patient samples. Examples that have reached clinical practice are the diagnostic assays OncotypeDX and Mammaprint, which predict the risk of recurrence or metastasis in breast cancer [ , ]. For locally advanced head and neck cancer and glioblastoma, prognostic gene and miRNA signatures predicting local and distant control or overall survival have been recently identified and are promising markers with the potential to allow substratification of standard-therapy treated patients for alternative treatment strategies [ – ].