Figure 1

Structure of the economic model. The economic model was composed of modules for each product, which had a common structure and were combined sequentially to reflect "treatment strategies". The structure of the modules implied, inter alia, that the withdrawal rates "due to intolerable side-effects" should be independent from "no response" to treatment. This requirement was violated because intent-to-treat analyses were used to estimate withdrawal rates, which reported "withdrawals" for many reasons, including lack of efficacy, inevitably leading to double-counting of nonresponders (TAR, p. 230). The impact of this phenomenon was unevenly distributed across the treatments evaluated (TAR, p. 231 and p. 236), resulting in a biased assessment [11]. Graphical representation of model reproduced from King et al. 2006, with kind permission.