Abstract 996: Predicting early mortality in a large ethnically diverse phase I clinical trial program

Abstract Background: Various scoring systems are developed to optimize patient selection for phase I clinical trials. We evaluated the Royal Marsden Hospital (RMH), Princess Margaret Hospital Index (PMHI), and MD-Anderson Cancer Center (MDACC) prognostic scores and compared their power to predict 90-day mortality (90DM) in an ethnically diverse population enrolled in the USC Norris Comprehensive Cancer Center (NCCC) Phase I Program. Methods: We queried our database of patients enrolled in phase I clinical trials at the NCCC from 2015-2022 which includes demographics, treatment details and outcomes. The enrollment occurred at all NCCC sites including the Los Angeles General Medical Center, a safety net hospital. Prognostic scores were calculated: RMH (albumin 35 g/L, LDH ULN, 2 metastatic sites), PMHI (albumin 35 g/L, ECOG 0, 2 metastatic sites), MDACC (albumin 35 g/L, LDH ULN, ECOG 0, 2 metastatic sites, GI tumor type). Kaplan-Meier analysis was used to calculate median survival in each group. ROC analysis was performed to assess the 90DM prediction. Results: 611 patients (15 breast, 344 GI, 20 GU, 29 GYN, 110 Lung/H0. 001). The MDACC system had the largest AUC to predict 90DM (AUC: 0. 678, p 0. 001). Score of 4-5 (high risk) per MDACC system corresponds to a specificity of 94. 5% (91. 1-97. 2) and positive likelihood ratio of 3. 6 (1. 8-7. 2). Conclusion: In our highly diverse phase I population, higher scores on the RMH, PMHI, and MDACC scoring systems were associated with worse prognosis and 90DM. Careful consideration should be given to potential patients with higher scores prior to enrollment in phase I trials. Characteristics and outcomes of the study participants (N = 611). Variable Median (IQR) /N (%) Scoring system N (%) Median survival days (95% CI) p-Value Age (yr) 61 54-68 RMH score 0. 001 Female gender 288 (47. 1%) 0 151 (36. 7%) 505 (378-631) Safety net participant 120 (19. 6%) 1 184 (44. 8%) 509 (324-693) Non-English speaking 177 (29%) 2 64 (15. 6%) 176 (107-244) Performance status ECOG 0 153 (33. 3%) 3 12 (2. 9%) 44 (12-76) ECOG 1 302 (65. 7%) Area under ROC curve (95% CI) 0. 650 (0. 579-0. 721) 0. 001 ECOG 2 5 (1. 1%) Number of prior lines 0-2 298 (48. 8%) PMHI score 0. 001 = 3 313 (51. 2%) 0 118 (25. 7%) 603 (405-800) Tumor type Breast 15 (2. 5%) 1 214 (46. 5%) 462 (378-545) GI 344 (56. 8%) 2 115 (25. 0%) 254 (117-390) GU 20 (3. 3%) 3 13 (2. 8%) 87 (52-122) Gyn 29 (4. 8%) Area under ROC curve (95% CI) 0. 629 (0. 560-0. 697) 0. 001 Lung/H 0. 001 Type of trial Chemotherapy 112 (18. 3%) 0 (low risk) 26 (7. 9%) Not estimated Immunotherapy 227 (37. 2%) 1 (low-intermediate risk) 84 (25. 5%) 597 (429-768) Targeted therapy 272 (44. 5%) 2 (intermediate risk) 116 (35. 2%) 471 (115-827) 2 metastatic sites 146 (23. 9%) 3 (high-intermediate risk) 75 (22. 7%) 297 (142-452) 90-day mortality 152 (24. 9%) 4 (high risk) 25 (7. 6%) 119 (108-130) Best response by RECIST Stable disease 225 (41. 5%) 5 (high risk) 4 (1. 2%) Not estimated Partial response 39 (7. 2%) Area under ROC curve (95% CI) 0. 678 (0. 611-0. 745) 0. 001 Complete response 6 (1. 1%) Progressive disease 272 (50. 2%) Citation Format: Hadi Erfani, Sabrina Woll, Austin Luong, Gage Guerra, Anthony Kharrat, Rebecca Umayam, Anastasia Martynova, Laurie Brunette, Lynda Roman, Negeen Izadian, Teshia Bustos, Diana Hanna, Anthony El-Khoueiry, Jacob Thomas. Predicting early mortality in a large ethnically diverse phase I clinical trial program abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts) ; 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84 (6Suppl): Abstract nr 996.