Improving Patient Selection for Phase I Trials in Hematology: An External Validation of the Gustave Roussy Scoring System at Toulouse University Cancer Institute, Oncopole (IUCT-O)

Abstract

Background Phases I trials are generally offered to patients with hematological malignancies after failure of standard therapies, the alternative being often non-evidence based treatments or palliative care. Unluckily, not all patients find benefit in participation, either because of disease progression or worsening general condition that translates in survival of less than 3 months after inclusion. Gustave Roussy prognostic scoring system (Benajiba, 2016) was developed to predict overall survival and death within 3 months after enrollment in phase 1 trial for patients with hematological malignancies. This score which relies on performance status (PS), albumin and histology, and segregates 3 groups (low, intermediate and high risk). The objective of this study was to evaluate the performance of the Gustave Roussy prognostic scoring system in an independent cohort of hematological patients included in phase 1 trials at IUCT-O. Patients and methods We reviewed medical records of patients with hematological malignancies consecutively enrolled in phase 1 trials between 2013 and 2019 at IUCT-O. Performance of the prognostic scoring system for predicting OS have been evaluated using monotonocity of gradients and discriminatory ability (Harrell's C). Sensitivity and specificity for predicting death within 3 months were estimated using time dependent ROC curve. Results A total of 96 patients were included and classified by the scoring system as low (L) (n=5, 5.2%), intermediate (I) (n=44, 45.8%) and high (H) risk (n=47; 49%). Median age was 66 years (range 26-80). Main hematological malignancies were multiple myeloma (n=32, 33.3%), acute myeloid leukemia (n=30, 31.3%) and lymphoma (n=30, 31.3%). The median of prior lines was 3 (Range 1-9). In total, 38 (40%) and 6 (6%) patients previously received autologous and allogeneic transplantation, respectively. At a median follow-up time of 57.9 months, 79 patients (L/I: 34; H: 45) were deceased (L/I: 34; H: 45). Patients defined as L/I risk had superior OS (27.9 months; 95% CI, 19 to 51.6 m) compared with patients classified as H risk (5.0 months; 95% CI, 3 to 7.2 months) (Hazard Ratio=3.46 95% CI, 2.18 to 5.50; p<0.001). The C'index was estimated to 0.67. The 90-days mortality rate was 2% in the L/I risk and 38.3% in H risk patients. The sensitivity and specificity for predicting early death were 94.7% and 62.3%, respectively. Conclusions The utilization of this score could allow a better patients selection in early trials, especially during the critical periods of dose escalation and proof-of-concept expansion cohorts, with standard markers such as PS and albumin. Figure 1View largeDownload PPTFigure 1View largeDownload PPT Close modal