A Predictive Score of Antitumour Activity of Novel Agents in Cancer Patients Treated in Early Phase Studies

Abstract

Introduction: Phase I trials aim to determine the maximum-tolerated dose of a particular drug while minimizing the number of patients exposed to either sub-therapeutic doses or severe toxicity. Thus, patient selection for phase I trials is a key component of any clinical trial design. Though several studies have been made to address this issue, patient selection still represents a major clinical challenge that needs further investigation. Methods: Twenty-nine baseline clinical and analytical characteristics of 773 consecutive patients treated in phase I trials between 2008 and 2016 in START Madrid-CIOCC were analysed and correlated to objective response (OR), progression-free survival, median overall survival, toxicity, and treatment type. The ones associated to OR in the univariate analysis were included in the stepwise logistic regression multivariate and Cox analysis. The statistically significant ones were included in a predictive score (named here as the Madrid score) of antitumour activity. Results: Body mass index (BMI) >25 (p = 0.027), two or less previous lines of treatment (p = 0.007), and normal levels of alkaline phosphatase (ALP) (p = 0.007) were found to positively correlate to radiological response. A Madrid score was generated using these three factors as predictive parameters: compared to a score of 2–3 (where 2 or 3 of these variables are altered), a score of 0–1 is associated with longer survival time (11.6 vs. 8.6 months; p = 0.005) and overall response (17 vs. 7.6%; p = 0.003). Conclusion: The predictive Madrid score, based on the BMI, number of prior lines of treatment, and ALP levels, might be helpful to accurately select patients who would benefit from oncology phase I clinical trials.