A Mechanistic Framework Integrating Renal QSP-PK-PD and Machine Learning for Baseline-Informed Stratification of Diuretic Resistance

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Not quantified in abstract

• Primary Outcome: Identification of baseline fractional excretion of sodium and glomerular filtration rate as integrative biomarkers for diuretic resistance stratification

• Nnt Or Sensitivity: Sensitivity and specificity of biomarker stratification not reported

• Confidence Interval: Not reported

Clinical Application

Requires baseline measurement of fractional sodium excretion and GFR - feasible in hospital settings but may need point-of-care development Relevant to Australian heart failure management under RACGP guidelines and PBS-subsidised furosemide therapy. Could inform TGA-approved treatment algorithms if validated Heart failure patients requiring loop diuretic therapy, particularly those at risk of diuretic resistance

Abstract

Diuretic resistance represents a major source of heterogeneity in loop diuretic response and remains a key barrier to effective decongestion in heart failure. A key clinical challenge is the early identification of patients at high risk of an inadequate response to standard-dose furosemide in order to inform timely treatment intensification or alternative decongestive strategies. However, current approaches remain reactive and rely on post-treatment response, limiting prospective risk stratification and treatment decision-making. To address this problem, we integrated a mechanistic renal quantitative systems pharmacology (QSP) model with furosemide pharmacokinetics and pharmacodynamics to enable baseline-informed risk stratification of diuretic resistance. The model was calibrated using published clinical pharmacokinetic and renal response data and applied to generate a physiologically constrained virtual patient population capturing heterogeneity in renal function, tubular sodium handling, and neurohormonal activation. Machine learning methods were incorporated as complementary analytical tools to identify and validate baseline physiological determinants that define risk categories of diuretic response within the mechanistic simulation framework. Model-based analyses indicated that diuretic resistance arises from the combined effects of reduced filtration capacity, enhanced tubular sodium avidity, diminished pharmacodynamic sensitivity and sustained neurohormonal activation. Across analytical approaches, baseline fractional excretion of sodium and glomerular filtration rate emerged as integrative biomarkers that stratify patients into distinct risk categories of diuretic response, reflecting a mechanism-informed reduction of underlying physiological variability. These findings provide a mechanistically grounded framework for baseline-informed risk stratification and may inform earlier identification of patients at risk of inadequate response, supporting model-informed evaluation of treatment intensification or alternative decongestive strategies.

References

1. 1.Zhou, J., Lou, V., Zhao, Y., Tang, L., Hu, Z., Sun, J., Chen, J., Cheng, P., Cheng, Y., & He, H. (2026). A mechanistic framework integrating renal QSP-PK-PD and machine learning for baseline-informed stratification of diuretic resistance. The AAPS Journal. https://doi.org/10.2215/CJN.04750422