Presentation (CTI)

OBJECTIVES: In chronic relapsing diseases, relapse risk varies over time, with phases of higher and lower disease activity, yet many health economic models assume a constant relapse risk throughout the disease course. This assumption may inadequately reflect real-world dynamics and bias long-term clinical and economic outcomes. This study assessed the impact of incorporating time-varying relapse risk, compared with constant relapse assumptions, on long-term cost-effectiveness outcomes.
METHODS: A cohort-based Markov model simulated three health states (relapse, no relapse, and death), capturing relapse events, healthcare costs, and QALYs over a lifetime horizon with monthly cycles. Relapse was defined as a clinically meaningful worsening event associated with increased healthcare use and disutilities. Two approaches were compared: a time-varying relapse risk estimated from longitudinal relapse patterns using flexible smoothing and applied dynamically across cycles, and a constant relapse risk applied uniformly throughout the model.
RESULTS: Under a long-term treatment effect, relapse risk is higher in the initial years and decreases with continued treatment. Accordingly, under the time-varying approach, the annual relapse rate declined over time, producing relapse patterns that differed substantially from the constant hazard assumption. Over 10 years, the constant relapse model predicted 7.08 relapses (annual rate 70.8%), compared with 3.29 relapses (annual rate 33%) under the time-varying model. Over 20 years, the constant model projected 13.99 relapses (annual rate 70%) versus 5.21 (annual rate 26%) under the time-varying approach. In contrast, the constant model assumes a fixed relapse risk across all years, resulting in no temporal decline. These differences meaningfully affected long-term costs and ICERs, especially over longer horizons.
CONCLUSIONS: Assuming a constant relapse risk may oversimplify disease dynamics and bias long-term cost-effectiveness results. Incorporating dynamic relapse risk may improve the relevance of economic evaluations for healthcare decision-making. In future, additional methods, such as parametric approaches, can be explored further to model time-varying relapse risk.