Multi-institutional prognostic factor analysis of patients (pts) enrolled in phase I (Ph I) oncology trials: Can pts selection be improved?

Abstract

2518 Background: The selection of suitable pts for Ph I oncology trials represents an important challenge as opportunities for new drug development increase. In our previous single-center studies (including 290 pts) we described a prognostic score comprising 3 variables (RMH score): number of metastatic (MTS) organs, albumin, and LDH; this identified pts with the shortest overall survival (OS) and highest 90 days (90d) mortality (JCO 2009; 27: 2692-6). Our aim here was to re-examine this issue in a larger multi-institutional population. Methods: Pts treated in Ph I trials from 14 leading European units between 2005 and 2007 were included. We collated baseline characteristics and outcomes. In addition we derived the clinical tumor growth rate (CTGR) (a log ratio of time from advanced disease to Ph I entry and number of treatment lines). Pts with PS ≤ 1, albumin, LDH and number of MTS organs were included with other factors to derive a predictive model using logistic regression in multiple training/test sets. Results: We identified a total of 2,232 pts. Median age was 58 years (range 13-86) and male/female ratio was 1.3. Median time on treatment was 51 days (range 1-1,342). The most common causes of Ph I discontinuation were progression in 75% and toxicity in 12%. The toxicity death rate was 0.4% (CI-95% 0.1-0.6). The estimated median OS was 8.8 months (CI-95% 8.3-9.4) with a 90d mortality of 16% (CI-95% 15-18). 10% discontinued Ph I during the first 30 days mainly due to progression. Initial analysis with 1,780 pts using predictive modeling identified 7 variables associated with early death: low albumin, lymphocytes, and CTGR; elevated WCC, LDH, and ALP; and higher number of MTS organs. Estimated sensitivity and specificity for this model were 48% and 78% respectively. Conclusions: Selection of pts using a 7 variables score was not superior to a 3 variables score (sensitivity 50%, specificity 80%); both would reduce the 90d mortality from 16% (unselected pts) to 10%. However either of these scores would reduce the number of eligible pts by 20% and may therefore impact on accrual; thus routine use needs careful consideration. Finally, this European collaboration provides a valuable tool for future analysis. No significant financial relationships to disclose.