Estimating the Causal Effect of Realistic Treatment Strategies Using Longitudinal Observational Data.

Area of Science:

• Health Services Research
• Biostatistics
• Clinical Epidemiology

Background:

• Real-world data and causal inference methods are crucial for evaluating complex, adaptive treatment strategies.
• Longitudinal Targeted Minimum Loss-based Estimation (LTMLE) offers a robust approach for analyzing time-varying confounding in observational studies.
• Myelodysplastic syndromes (MDS) management requires nuanced treatment strategies to optimize patient outcomes.

Purpose of the Study:

• To apply the LTMLE method to evaluate the causal effects of static versus dynamic treatment regimes for erythropoiesis-stimulating agents (ESAs) in low to intermediate-1 risk MDS.
• To assess the impact of these treatment strategies on patient mortality and health-related quality of life (EQ-5D).

Main Methods:

• Longitudinal registry data from patients with low to intermediate-1 risk MDS were analyzed.
• Dynamic treatment regimes (DTRs) were defined based on clinical decision rules, adapting ESA treatment over time.
• LTMLE was employed to estimate the causal effects of DTRs compared to static regimes (always or never administering ESAs), accounting for time-varying confounding.

Main Results:

• The static regime of never administering ESAs was associated with declining EQ-5D scores and increased mortality risk.
• Both continuous ESA administration and dynamic regimes improved EQ-5D scores and showed a trend towards reduced mortality compared to the no-ESA regime.
• While dynamic regimes demonstrated potential benefits, differences in mortality were not statistically significant in this study.

Conclusions:

• Dynamic treatment strategies for ESAs show promise in improving quality of life and potentially survival for MDS patients.
• The LTMLE method is effective for evaluating realistic, adaptive treatment policies in settings with time-varying confounding, even with small sample sizes and long follow-up.
• Findings support personalized treatment adaptation in MDS management and provide methodological insights for health technology assessment and policy making.