Abstract Background: Safety analysis in first-in-human dose escalation oncology trials is typically limited to descriptive tabulation of adverse events and clinically relevant laboratory abnormalities. Such an approach may not fully capture the temporal dynamics and the combined contribution of patients’ baseline characteristics and escalating doses. Here we utilized data analytical techniques including machine learning to extract deeper insights from the trial of LP-184 (NCT05933265). LP-184 is a tumor site activated acylfulvene pro-drug that alkylates DNA after bioactivation by the intracellular oxidoreductase prostaglandin reductase1. In this trial, 63 patients with advanced solid tumors were enrolled across 12 dose cohorts. LP-184 was infused over 30 minutes on days 1 and 8 of every 21-day cycle until disease progression or unacceptable toxicity. The most common treatment-related adverse events that required dose modification were for platelet count (PLT) decrease and alanine aminotransferase (ALT) increase. Methods: Longitudinal PLT and ALT data were analyzed to characterize temporal trends and inter-patient variability. Correlation analysis was used to explore the relationships among the greatest PLT and ALT changes, baseline characteristics, exposure, and time. Linear regression, logistic regression, and decision tree modeling were applied to identify predictors of clinically relevant PLT and ALT abnormalities. Results: PLT typically reached nadir in cycle 2 with greater frequency observed in patients treated with higher doses. Correlation analysis and linear regression identified baseline PLT and total single-infusion dose as critical predictors of PLT nadir. A logistic regression model predicting the occurrence of ≥ grade 2 decrease in PLT using baseline PLT and dose achieved an area under the receiving operating characteristic curve of 0. 9. A decision tree model (kappa = 0. 82) further indicated that patients with baseline PLT below 199 K/µL and doses above 10 mg (approximately dose level 7+) were at a high risk (0. 82) of developing ≥ grade 2 decrease in PLT. ALT levels generally peaked in cycle 1 and were moderately associated with dose, with no evident relationship to baseline liver lesions. ALT elevations at ≥ grade 2 were more frequent in patients previously treated for glioblastoma. Machine learning models showed lower predictive performance for ALT compared with PLT. Conclusions: Data-driven analysis in the LP-184 trial identified risk windows for PLT decrease (cycle 2) and ALT increase (cycle 1). Baseline PLT and dose arekey predictors, enabling modeling of ≥ grade 2 decrease in PLT. Our study supports the integration of in-depth data analytics in Phase 1 studies to facilitate earlier safety signal detection and inform more proactive safety monitoring and dose selection in later studies. Citation Format: Jianli Zhou, Daruka Mahadevan, Jay Parekh, Marc Chamberlain, Kishor Bhatia, Reginald Ewesuedo. Data-driven characterization of platelet count and alanine aminotransferase dynamics in the first-in-human LP-184 dose escalation trial abstract. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts) ; 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86 (8Suppl): Abstract nr CT116.
Zhou et al. (Fri,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: