26 ICER Thresholds <strong>KCE</strong> Reports 100 The same applies to a social security system, be it that the ICER threshold value approach would in this system not only impact on health inequality but also on income inequality. It is g<strong>en</strong>erally acknow<strong>le</strong>dged that <strong>de</strong>cision makers take other aspects into account besi<strong>de</strong>s the ICER of an interv<strong>en</strong>tion, to <strong>de</strong>ci<strong>de</strong> whether or not the interv<strong>en</strong>tion is worth its cost. Because these factors differ across interv<strong>en</strong>tions, a sing<strong>le</strong> threshold value for the ICER below which an interv<strong>en</strong>tion is consi<strong>de</strong>red value for money, is not consist<strong>en</strong>t with how society chooses to make <strong>de</strong>cisions (see also chapter 3). If it would be possib<strong>le</strong>, however, to inclu<strong>de</strong> these additional consi<strong>de</strong>rations in the ICER, e.g. by weighting the QALYs of populations the society wants to protect more heavily, the ICER threshold value approach might still be applicab<strong>le</strong>. 75 The ICER threshold value would in this case be <strong>de</strong>fined in terms of a cost-per-weighted QALY. The objective is no longer ‘health outcome maximisation’ but ‘weighted health outcome maximisation’. This weighted QALY approach has three pot<strong>en</strong>tial drawbacks. First, a prerequisite for this approach is transpar<strong>en</strong>cy of the composition of the weights for QALYs: which criteria are <strong>de</strong>terminant, what value is assigned to each <strong>de</strong>terminant and how are these values combined to obtain a unique weight? Obviously, this is not an easy requirem<strong>en</strong>t. A second pot<strong>en</strong>tial prob<strong>le</strong>m of including additional <strong>de</strong>cision criteria in the costeffectiv<strong>en</strong>ess ratio is the se<strong>le</strong>ction of the appropriate cost-effective comparator. If <strong>de</strong>cisions are not purely inspired by a pursuit of maximal health, it might happ<strong>en</strong> that interv<strong>en</strong>tions that are not cost-effective according to the theoretical ICER threshold value approach (without QALY weighting) are neverthe<strong>le</strong>ss reimbursed. For examp<strong>le</strong>, suppose that, <strong>de</strong>spite a high ICER, a specific interv<strong>en</strong>tion is reimbursed because there is no alternative treatm<strong>en</strong>t for treating a specific serious disease and pati<strong>en</strong>ts would otherwise be <strong>le</strong>ft untreated. Suppose that the conv<strong>en</strong>tional ICER of this programme is higher than the ICER threshold value that would imply maximal health outcomes but that the outcomes have be<strong>en</strong> giv<strong>en</strong> a higher weight in or<strong>de</strong>r to stay below the ICER threshold value. If after the <strong>de</strong>cision is tak<strong>en</strong> a new interv<strong>en</strong>tion for this pati<strong>en</strong>t population is <strong>de</strong>veloped, it might have a low ICER wh<strong>en</strong> the existing treatm<strong>en</strong>t is used as the comparator in the CEA. Its ICER might be below the ICER threshold value and h<strong>en</strong>ce it might be conclu<strong>de</strong>d that it is cost-effective. However, the existing reimbursed treatm<strong>en</strong>t might not be the re<strong>le</strong>vant comparator, as the reason for initial reimbursem<strong>en</strong>t (i.e. non-exist<strong>en</strong>ce of an alternative treatm<strong>en</strong>t for the pati<strong>en</strong>ts) no longer holds. Previous <strong>de</strong>cision might have to be revised in the light of new <strong>de</strong>velopm<strong>en</strong>ts. This <strong>le</strong>ads to the conclusion that including additional consi<strong>de</strong>rations in the ICER and including such ‘weighted’ ICER subsequ<strong>en</strong>tly in a <strong>le</strong>ague tab<strong>le</strong> might complicate their interpretation and their practical usability. Moreover, the risk for misuse or errors in the choice of the appropriate comparator increases. A third weakness of the weighted QALY approach, if used in combination with a threshold value, is the remaining requirem<strong>en</strong>ts of perfect divisibility of health programmes and constant returns to sca<strong>le</strong>, two requirem<strong>en</strong>ts that may not hold in real life (see 2.7.3). A few attempts have be<strong>en</strong> ma<strong>de</strong> to <strong>de</strong>rive an ICER threshold value from past health policy <strong>de</strong>cisions. 65, 93-95 The exercises showed that in<strong>de</strong>ed there is no sing<strong>le</strong> threshold value above which the <strong>de</strong>cision is always negative and below which it is always positive. Rather, a range of acceptab<strong>le</strong> ICERs has be<strong>en</strong> id<strong>en</strong>tified. This can mean differ<strong>en</strong>t things: (1) the <strong>de</strong>cision maker does not know the true ICER threshold value that would maximise health b<strong>en</strong>efits from a giv<strong>en</strong> budget; 61, 64, 96 (2) other consi<strong>de</strong>rations than health maximisation <strong>de</strong>termine the acceptability of an interv<strong>en</strong>tion with an ICER that is, strictly speaking, above the ICER threshold value 65, 67 , (3) differ<strong>en</strong>t methods are used to obtain the ICER estimates as a consequ<strong>en</strong>ce of which they are not always comparab<strong>le</strong>, (4) the <strong>le</strong>vel of uncertainty around the ICER estimates <strong>de</strong>termines their acceptability and (5) <strong>de</strong>cision makers do not ‘trust’ all ICER estimates to the same ext<strong>en</strong>t. The differ<strong>en</strong>t reasons probably all apply to some ext<strong>en</strong>t. 85 An empirically id<strong>en</strong>tified range of ICER threshold values should therefore be interpreted as the range of societal willingness to pay for an additional QALY or LYG at that time, in that specific budgetary and societal context and for those specific interv<strong>en</strong>tions rather than as an ICER threshold value in 65, 96 the purely theoretical meaning of an absolute criterion for health maximisation.
<strong>KCE</strong> reports 100 ICER Thresholds 27 2.7.6 Additional caveats There is a serious risk of bias towards the ICER threshold value, if one is <strong>de</strong>fined. 97 Once a threshold value is set, there is the danger that ICERs of new technologies will converge towards this threshold value by inducing commercial companies to adapt their prices in or<strong>de</strong>r to ‘satisfy’ the cost-effectiv<strong>en</strong>ess criterion (ICER