CONSEQUENCES OF WITHDRAWING THE SECOND TUMOUR NECROSIS FACTOR α ANTAGONIST IN SEQUENTIAL TREATMENTS OF ACTIVE ANKYLOSING SPONDYLITIS: A HEALTH-ECONOMICAL PERSPECTIVE USING A POPULATION DYNAMICS SIMULATION MODEL

Purpose:    Ankylosing spondylitis (AS) patients with inadequate response to the first tumour necrosis factor-alpha antagonist (anti-TNF) are treated with a second one in clinical practice. However, the second anti-TNF was shown to have somewhat lower efficacy. Considering high costs of anti-TNF therapies, the objectives of this study were to (1) develop a model that can quantify cost-of-illness and effectiveness of sequential treatments of AS with anti-TNFs for a specific society taking population dynamics into account, and (2) analyze the simulation outcomes for the Dutch society from scenarios with and without withdrawing the second anti-TNF.

Methods:    Dynamics of the AS population are characterized by temporal changes in the population size and patients' attributes, including age, gender, symptom duration, work status, disease activity (BASDAI) and function (BASFI). Let 1JanX denote the first of January of year X, PP_X denote the AS population on 1JanX, and IP_X denote the incident AS population within a one-year period from 1JanX to 1Jan(X+1).  Given a simulation length of n years starting from 1JanY, the model tracks individually all AS patients appearing during this period, including (1) PP_Y, and (2) IP_Y, IP_Y+1, …, IP_Y+n-1. For each population, the model creates a number of virtual AS patients equalling the population size, simulates disease progression of each patient and generates data on patient attributes, cumulative costs and effectiveness at discrete time points using a discrete event simulation approach. The model was parameterized using the Dutch cohort data. Two scenarios were simulated for the period 1Jan2012-1Jan2032. In Scenario 1, five nonsteroidal anti-inflammatory drugs (NSAIDs) and one anti-TNF were available. In Scenario 2, five NSAIDs and two anti-TNFs were available. The model was developed using the Delphi and R languages.

Results:    Mean BASDAI in Scenario 1 slightly increased with increasing time and in Scenario 2 was almost constant. Means of BASFI followed the same trend with differences between the two scenarios increased with increasing time (Figure 1). Saved costs per QALY lost in Scenario 1 compared to Scenario 2 on 1Jan2022 and 1Jan2032 were €86980 and €75683, respectively.

Conclusions:    Decision on withdrawing one anti-TNF should be based on the willingness to accept a loss of one QALY to save about €80000. The modelling framework is novel and flexible, which can be used for different societies.

 

Figure 1