Preoperative (neoadjuvant) chemoradiotherapy (CRT) and total mesorectal excision is the standard treatment for rectal cancer patients. Approximately one third of patients treated with CRT have a pathological complete response (pCR). These patients might be spared from surgery and the associated morbidity and mortality, and might be better served by a watch and wait strategy. Arguably, that would require a robust classifier of pCR. However, reliably identifying pCR based on clinical or imaging parameters remains challenging. We generated gene expression profiles of pretreatment biopsies from 175 patients with rectal cancer enrolled in the CAO/ARO/AIO-94 and CAO/ARO/AIO-04 clinical trials. 161 samples were used for building, training and validating a predictor of pCR using a machine-learning algorithm. The performance of the classifier was validated in three independent cohorts, comprising patients from (i) the CAO/ARO/AIO-94 and -04 trial not included in the previous analysis (n=14), (ii) a dataset by Milini et al. (n=38), and (iii) in 24 prospectively collected samples from the TransValid A trial. A 42-transcript signature resulted in the best classification of pCR. The classifier remained robust when applied to the independent datasets. Our classifier reproducibly identified 30% of rectal cancer patients with a pCkenR. Importantly, we never predicted pCR in patients when that was not the case. Therefore, this classifier could augment clinical tests to assign rectal cancer patients to an evidence-based personalized treatment. This is a collaboration between the laboratory of Dr. Eytan Ruppin at the University of Maryland and Tel Aviv University, supported through the NCI/University of Maryland partnership program (Noam Auslander), the University of Goettingen, supported by the Deutsche Forschungsgemeinschaft (Dr. Georg Emons), and the University of Luebeck, supported by the Deutsche Krebshilfe (Dr. Ruediger Meyer). We are planning to expand this project by including the Walter Reed National Military Medical Center.
Showing the most recent 10 out of 29 publications
