With the ability to assess the possible causes for radioresistance of a tumor on a pre-treatment sample, we would be able to allocate the best fitting radiotherapy schedule and biological agent combination, eventually leading to better survival and/or less toxicity. Therapeutic options consist, apart from surgery, of various radiotherapy doses and fractionation schedules, dose painting, as well as the addition of cisplatin, hypoxia sensitizers, EGFR-inhibitors, hopefully soon to be expanded with for example immune checkpoint- (108, 109), DNA repair- (110, 111) or CD44-inhibitors (112, 113). Having multiple therapy options is an asset, but only if we know when to use which treatment.
Data-driven analyses on small patient series to find prognostic gene sets are not the way forward. Preferably we should focus on finding predictive markers in (randomized) studies, controlled for known factors. To move forward to the point where we know exactly which patient should get which treament(-s) we would have to study complete clinical and biological data from large numbers of patients that have been treated with different treatment alternatives (114). Within such a large cohort it would be possible to study subgroups and interactions between different biological factors and come up with the best predictive model (1). Furthermore, a subgroup could possibly be isolated that does not respond to any of the available treatments. Knowing the biological profile of these tumors could help design new ways to improve their treatment outcome. To achieve this, a database should be set-up across multiple countries, possibly combining already available data, with the ability to add data from new trials.