Selection of appropriate end-points (pCR vs 2yDFS) for tailoring treatments with prediction models in locally advanced rectal cancer

Purpose Personalized treatments based on predictions for patient outcome require early characterization of a rectal cancer patient's sensitivity to treatment. This study has two aims: (1) identify the main patterns of recurrence and response to the treatments (2) evaluate pathologic complete response (pCR) and two-year disease-free survival (2yDFS) for overall survival (OS) and their potential to be relevant intermediate endpoints to predict. Methods Pooled and treatment subgroup analyses were performed on five large European rectal cancer trials (2795 patients), who all received long-course radiotherapy with or without concomitant and/or adjuvant chemotherapy. The ratio of distant metastasis (DM) and local recurrence (LR) rates was used to identify patient characteristics that increase the risk of recurrences. Findings The DM/LR ratio decreased to a plateau in the first 2 years, revealing it to be a critical follow-up period. According to the patterns of recurrences, three patient groups were identified: 5-15% had pCR and were disease free after 2 years (excellent prognosis), 65-75% had no pCR but were disease free (good prognosis) and 15-30% had neither pCR nor 2yDFS (poor prognosis). Interpretation Compared with pCR, 2yDFS is a stronger predictor of OS. To adapt treatment most efficiently, accurate prediction models should be developed for pCR to select patients for organ preservation and for 2yDFS to select patients for more intensified treatment strategies.