Citations

Abstract Cancer outcomes vary widely between individual patients, each accumulating an irregular record of treatments, diagnoses, measurements, and complications. Current prognostic models reduce this complexity into a single snapshot, focus on narrow clinical settings, and rarely generalize across hospitals. Here we introduce Chronicle , an explainable transformer that learns from entire patient trajectories to predict diverse clinical outcomes throughout the disease course while capturing both short- and long-term temporal dependencies. Trained on 53.7 million longitudinal data points from 51,711 patients spanning 67 cancer types, Chronicle operates natively on irregular data without imputation and jointly predicts eight endpoints within a flexible framework adaptable to additional outcomes. Chronicle outperformed cross-sectional models for overall survival prediction (C-index 0.84 vs 0.76-0.79), stratified patients more accurately than established prognostic systems, including TNM stage, and predicted seven adverse event and transfusion endpoints (AUC 0.80-0.92). Applied without retraining to 69,341 patients in Germany, Switzerland, and the United States, Chronicle generalized across healthcare systems and improved further with local fine-tuning. Integrated explainability traced each risk update to patient-specific clinical factors, revealing distinct temporal persistence of prognostic information, with relevance half-lives ranging from weeks for therapies to nearly one year for baseline characteristics. These findings demonstrate that learning from hospital-wide patient trajectories enables interpretable and continuously updated predictions, providing a scalable framework to support individualized treatment decisions.

Abstract Anti–PD-1 therapy has improved clinical outcomes in advanced melanoma, but most patients experience intrinsic resistance. Responding patients can develop acquired resistance to anti–PD-1. We retrospectively reviewed 488 patients treated with anti–PD-1 from three academic centers and identified 36 patients with acquired resistance, defined as disease progression following objective response. The incidence, timing, disease sites, post-progression survival (PPS), and outcomes were evaluated descriptively. The acquired resistance cohort consisted of 67% with more than 1 feature of poor prognosis (stage M1c, elevated LDH, or brain metastasis), and 67% had previously received ipilimumab. Partial and complete responses were achieved in 89% (n = 32) and 11% (n = 4) of patients, respectively, and median time to resistance (progression-free survival; PFS) was 11.1 months (range 4.3–32.8 months). Most progression was isolated (78% of patients, n = 28) and occurred while receiving therapy (78%, n = 28). The median PPS was 12.8 months (range 0.1–51.8 months), and the median overall survival was 33.7 months. Among isolated progressors, 15 received localized therapy (12 with surgery, 3 with radiation). Patients with isolated versus systemic progression exhibited a trend for improved PPS (P = 0.081), and patients with an initial PFS ≥ 15 months showed significant PPS improvement (P = 0.036). Two patients experienced subsequent responses to anti–PD-1 resumption. In conclusion, acquired resistance to anti–PD-1 was frequently associated with excellent clinical outcomes and often presented as isolated progression amenable to localized therapy (surgery or radiation) or systemic progression sensitive to therapy resumption. Cancer Immunol Res; 5(5); 357–62. ©2017 AACR.