Predictors of early mortality and validation of novel prognostic models in Asian patients on phase I trials (PIT): A single-center experience.
6618 Background: Despite the growing number of clinical trials conducted in Asia and the unique pt demographic and molecular profile of endemic cancers, there is still lacking of studies examining the outcomes of pts on PIT. We examined the characteristics and outcomes of pts enrolled in phase I trials in National Cancer Center Singapore (NCCS). Methods: We reviewed 296 pts enrolled in PIT from 2004-2012 to identify factors that predicted for 90-day mortality and OS. Logistic regression model and Cox proportional hazard model assessed factors associated with 90DM and OS, respectively. The discriminative ability of the RMH prognostic score (LDH, no of met sites, albumin) and the final multivariate Cox model derived in Asian pts was evaluated using Harrell’s concordance index (c-index), and that for the logistic model was based on area under ROC curve (AUC). Internal validation was performed based on simulated data via bootstrapping. Results: Median age 60 (23-85), M:F (65/35%). The commonest cancer type thoracic (56%), GI (24%), head and neck (12%). 57% pt had comorbidities, commonest diabetes (16.6%) and hepatitis B (11.1%). Median no of lines of treatment 2. 20% of pts had known mut status. 15% pts survived less than 90 days from start of trial. Median OS was 9.6 m (95CI 7.8-11.2 m). Based on the RMH prognostic model, pts with score of 0 to 1 had a superior OS (median OS = 12.9 m) compared to those with score of 2 to 3 (OS = 5.7 m) (p < 0.001). The c-index for the RMH prognostic score was 0.64. On univariate analysis, alb<35,LDH>ULN, ≥3 met sites, low BMI, ECOG>0, Na<135,plt>440, Hb<12 and ≥3 lines of chemo were negative prognostic factors. On multivariate analysis, reduced model selection techniques identified alb<35, LDH>ULN, ≥3 met sites, and ≥ 3 prior lines of chemo as independent negative predictors of OS with a bias-corrected of 0.69. For 90DM, we developed a risk normogram based on ECOG, WBC, Na and Hb as continuous predictors with bias-corrected AUC 0.78. Conclusions: We have developed a novel NCCS normogram to predict for 90DM for PIT. The RMH prognostic score can be improved upon with the addition of number of prior lines of chemotherapy, underscoring the importance of early access to phase I trials.