Citations 2
An explainable transformer model learning from entire treatment timelines for pan-cancer risk profiling across healthcare systems
Keyl, Philipp , Kiermeyer, Niklas , Bosserhoff, Jonah, Lenfers, Tim , Niederhauser, Thibault , Fan, Bowen , Schnake, Thomas , Aubele, Fabio, Kuss, Solveig, Jurmeister, Philipp, Kim, Moon, Bauer, Sebastian , Bechrakis, Nikolaos, Forsting, Michael, Fuehrer-Sakel, Dagmar, Glas, Martin , Gruenwald, Viktor, Hadaschik, Boris, Herrmann, Ken, Kasper, Stefan, Kimmig, Rainer, Lang, Stephan, Pretzell, Ina , Rassaf, Tienush, Roesch, Alexander , Sure, Ulrich, Wichert, Marc, Unger, K, Behr, Juergen, Teupser, Daniel, Stief, Christian G., Mayerle, J, Tufman, Amanda, Ricke, Jens , Lindner, Lars H., Priglinger, Siegfried, Hoeglinger, Guenter, Mahner, Sven, Canis, Martin , Heinzerling, Lucie, Spitzweg, Christine , Tasdogan, Alpaslan , Totzeck, Matthias , Welt, Anja , Wiesweg, Marcel, Westphalen, C. Benedikt, Thasler, Reinhard, Albashiti, Fady , Montavon, Grégoire , Miglino, Nicola, Bergwelt‐Baildon, Michael von , Heinemann, Volker , Belka, Claus , Hartmann, Sylvia, Wicki, A., Nensa, Felix, Teupser, Daniel , Mueller, Klaus-Robert, Kleesiek, Jens , Wicki, Andreas , Wicki, Andreas , Bauer, Sebastian , Bosserhoff, Jonah, Schallenberg, Simon , Aubele, Fabio Paul , Kuss, Solveig , Idaji, Mina Jamshidi , Bechrakis, Nikolaos Emmanouel , Forsting, Michael, Hadaschik, Boris Alexander , Lang, Stephan , Rassaf, Tienush , Siveke, Jens , Guberina, Maja , Sure, Ulrich, Wichert, Marc, Ingrisch, Michael , Behr, Juergen, Mayerle, Julia V. , Harbeck, Nadia , Tufman, Amanda , Lindner, Lars Hartwin , Höglinger, Günter U. , Wiesweg, Marcel , Westphalen, C. Benedikt , Thasler, Reinhard, Miglino, Nicola, Balázs, Zsolt , Hartmann, Sylvia , Nensa, Felix , Schadendorf, Dirk , Schüler, Martin , Klauschen, Frederick , Keyl, Julius
No container title available | July 27, 2026 | Volume: No volume available | Issue: No issue available | Pages: No page available
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.
Evaluation of clinical parameters and biomarkers in older, untreated mantle cell lymphoma patients receiving bendamustine–rituximab
Ramsower, Colleen , Robetorye, Ryan S. , Mwangi, Raphael , Maurer, Matthew J. , Villa, Diego , McDonnell, Tim, Cohen, Jonathon B. , Clot, Guillem , Bühler, Marco M. , Kulis, Marta , Giné, Eva , Cook, James R. , Hill, Brian T. , Beiske, Klaus , Reichart, Alexander, Hartmann, Sylvia , Rimsza, Lisa M. , Ramsower, Colleen A , Maurer, Matthew , Villa, Diego , Feldman, Andrew, Cohen, Jonathon B , Martin-Subero, Jose Ignacio , Cook, James R , Hill, Brian , Raess, Philipp W , Hartmann, Sylvia , Holte, Harald , Rimsza, Lisa M , Rosenthal, Allison , McDonnell, Tim, Feldman, Andrew L. , Habermann, Thomas M. , Campo, Elı́as , Raess, Philipp W. , Holte, Harald , Scott, David W.
British Journal of Haematology | October 26, 2023 | Volume: 204 | Issue: 1 | Pages: 160-170
SummaryMantle cell lymphoma (MCL) is clinically and biologically heterogeneous. While various prognostic features have been proposed, none currently impact therapy selection, particularly in older patients, for whom treatment is primarily dictated by age and comorbidities. Herein, we undertook a comprehensive comparison of clinicopathological features in a cohort of patients 60 years and older, uniformly treated with bendamustine and rituximab, with a median survival of >8 years. The strongest prognostic indicators in this cohort were a high‐risk call by a simplified MCL international prognostic index (s‐MIPI) (HR: 3.32, 95% CI: 1.65–6.68 compared to low risk), a high‐risk call by MCL35 (HR: 10.34, 95% CI: 2.37–45.20 compared to low risk) and blastoid cytology (HR: 4.21, 95% CR: 1.92–9.22 compared to classic). Patients called high risk by both the s‐MIPI and MCL35 had the most dismal prognosis (HR: 11.58, 95% CI: 4.10–32.72), while those with high risk by either had a moderate but clinically relevant prognosis (HR: 2.95, 95% CI: 1.49–5.82). A robust assay to assess proliferation, such as MCL35, along with stringent guidelines for cytological evaluation of MCL, in combination with MIPI, may be a strong path to risk‐stratify older MCL patients in future clinical trials.