PURPOSE The application of Cox proportional hazards (CoxPH) models to survival data and the derivation of hazard ratio (HR) are well established. Although nonlinear, tree-based machine learning (ML) models have been developed and applied to the survival analysis, no methodology exists for computing HRs associated with explanatory variables from such models. We describe a novel way to compute HRs from tree-based ML models using the SHapley Additive exPlanation values, which is a locally accurate and consistent methodology to quantify explanatory variables’ contribution to predictions. METHODS We used three sets of publicly available survival data consisting of patients with colon, breast, or pan cancer and compared the performance of CoxPH with the state-of-the-art ML model, XGBoost. To compute the HR for explanatory variables from the XGBoost model, the SHapley Additive exPlanation values were exponentiated and the ratio of the means over the two subgroups was calculated. The CI was computed via bootstrapping the training data and generating the ML model 1,000 times. Across the three data sets, we systematically compared HRs for all explanatory variables. Open-source libraries in Python and R were used in the analyses. RESULTS For the colon and breast cancer data sets, the performance of CoxPH and XGBoost was comparable, and we showed good consistency in the computed HRs. In the pan-cancer data set, we showed agreement in most variables but also an opposite finding in two of the explanatory variables between the CoxPH and XGBoost result. Subsequent Kaplan-Meier plots supported the finding of the XGBoost model. CONCLUSION Enabling the derivation of HR from ML models can help to improve the identification of risk factors from complex survival data sets and to enhance the prediction of clinical trial outcomes.

In nature, load‐bearing polymeric materials that do not present significant trade‐off between elongation and elastic modulus are often found, but are rare in synthetic systems. One mechanism for emulating these natural systems is with functionally graded materials (FGMs). The development of synthetic FGM systems with varying moduli gradient using tuned poly(meth)acrylate multilayers is shown. Such localized tuning of crosslink density is shown as a mechanism to increase rigidity without significant compromise to maximum strain. The toughest FGM shows 43% higher tensile strength and 9% higher stiffness than copolymer elastomers with similar maximum strain (εmax ≈ 110%), increasing toughness by 25%. These improved tensile mechanical properties can be due to beneficial interlayer crack deflection and delamination. This gradient approach provides potential to improve toughness of various load‐bearing polymeric materials.

Anticipation of clinical decompensation is essential for effective emergency and critical care. In this study, we develop a multimodal machine learning approach to predict the onset of new vital sign abnormalities (tachycardia, hypotension, hypoxia) in ED patients with normal initial vital signs. Our method combines standard triage data (vital signs, demographics, chief complaint) with features derived from a brief period of continuous physiologic monitoring, extracted via both conventional signal processing and transformer-based deep learning on ECG and PPG waveforms. We study 19,847 adult ED visits, divided into training (75%), validation (12.5%), and a chronologically sequential held-out test set (12.5%). The best-performing models use a combination of engineered and transformer-derived features, predicting in a 90-minute window new tachycardia with AUROC of 0.836 (95% CI, 0.800-0.870), new hypotension with AUROC 0.802 (95% CI, 0.747–0.856), and new hypoxia with AUROC 0.713 (95% CI, 0.680-0.745), in all cases significantly outperforming models using only standard triage data. Salient features include vital sign trends, PPG perfusion index, and ECG waveforms. This approach could improve the triage of apparently stable patients and be applied continuously for the prediction of near-term clinical deterioration.

Purpose The exposure–response relationships for efficacy and safety of ipatasertib, a selective AKT kinase inhibitor, were characterized using data collected from 1101 patients with metastatic castration-resistant prostate cancer in the IPATential150 study (NCT03072238). Methods External validation of a previously developed population pharmacokinetic model was performed using the observed pharmacokinetic data from the IPATential150 study. Exposure metrics of ipatasertib for subjects who received ipatasertib 400 mg once-daily orally in this study were generated as model-predicted area under the concentration–time curve at steady state (AUCSS). The exposure–response relationship with radiographic progression-free survival (rPFS) was evaluated using Cox regression and relationships with safety endpoints were assessed using logistic regression. Results A statistically significant correlation between ipatasertib AUCSS and improved survival was found in patients with PTEN-loss tumors (hazard ratio [HR]: 0.92 per 1000 ng h/mL AUCSS, 95% confidence interval [CI] 0.87–0.98, p = 0.011). In contrast, an improvement in rPFS was seen in subjects receiving ipatasertib treatment (HR: 0.84, 95% CI 0.71–0.99, p = 0.038) but this effect was not associated with ipatasertib AUCSS in the intention-to-treat population. Incidences of some adverse events (AEs) had statistically significant association with ipatasertib AUCSS (serious AEs, AEs leading to discontinuation, and Grade ≥ 2 hyperglycemia), while others were associated with only ipatasertib treatment (AEs leading to dose reduction, Grade ≥ 3 diarrhea, and Grade ≥ 2 rash). Conclusions The exposure–efficacy results indicated that patients receiving ipatasertib may continue benefiting from this treatment at the administered dose, despite some variability in exposures, while the exposure–safety results suggested increased risks of AEs with ipatasertib treatment and/or increased ipatasertib exposures.