Related Experiment Video
Updated: Oct 19, 2026
Assessment of Resistance to Tyrosine Kinase Inhibitors by an Interrogation of Signal Transduction Pathways by Antibody Arrays
Published on: September 19, 2018
Ensemble signal detection reveals target-specific temporal risk of hypokalemia after immune checkpoint inhibition: a
Shengwang Chen1, Ziyi Dai2, Zihan Guo3
1Department of Neurosurgery, The Third Hospital of Xiamen, Xiamen City, Fujian, China.
Introduction:
Immune checkpoint inhibitors (ICIs) have revolutionized cancer treatment, but traditional pharmacovigilance lacks a translational computational framework capable of resolving target-specific nonlinear temporal risks of adverse events such as hypokalemia.
Methods:
We integrated 3,243 ICI-related hypokalemia reports from the FDA Adverse Event Reporting System (FAERS) and VigiBase (2014-2024). An ensemble signal-detection strategy cross-validated four disproportionality algorithms (Reporting Odds Ratio [ROR], Proportional Reporting Ratio [PRR], Empirical Bayesian Geometric Mean [EBGM], and Information Component [IC]). Flexible nonlinear survival analysis was performed using restricted cubic spline (RCS) Cox regression to model time-to-event dynamics.
Results:
Programmed Death-Ligand 1 (PD-L1) inhibitors triggered the strongest signal (ROR = 3.20, 95% CI: 2.95-3.48), with ultra-early onset (median 23 days) and the highest case fatality rate (23.7%). An asymmetric biphasic mortality pattern showed a nadir at 60 days: early-onset (≤30 days) hypokalemia peaked at a hazard ratio (HR) of ∼1.9, whereas late-onset (>90 days) risk gradually rose to HR ∼1.8 within 350 days. Programmed Death-1 (PD-1) inhibitors significantly increased severe hypokalemia risk (odds ratio [OR] = 28.32), with male predominance across PD-1 and CTLA-4 subgroups.
Discussion:
This interpretable ensemble framework reveals target-specific temporal toxicity signatures from multi-source real-world data, providing a computationally grounded translational pharmacology foundation for precision electrolyte monitoring and mechanism-informed safety prediction.
