Valuing the effects of sildenafil in erectile dysfunction

Sildenafil is a true breakthrough drug in the sense that it provides a potential treatment for a condition for which there was no existing acceptable alternative. This complicates any attempt to describe the cost effectiveness of sildenafil in erectile dysfunction, such as that by Stolk et al in this week's BMJ (p 1165).¹ They compare sildenafil with papaverine-phentolamine injections, which they argue are rationed on “medical grounds” and will not achieve the population benefits that might be achieved by sildenafil. More controversially, they argue that “the incremental cost-effectiveness of sildenafil lies at the favourable end of the scale when compared with interventions in health care for other diseases.”

The comparator therapy Stolk et al refer to, papaverine-phentolamine injections, seems not to have been rigorously evaluated and is not widely used. They used a cost utility approach in which a representative sample of the general population were asked to value the effects of treatment (for a condition that they did not have) to generate a cost per quality adjusted life year (QALY). Why might we question the validity of these findings?

Generating values for a treatment and comparing them with scores for other healthcare interventions requires a method for translating the clinical benefits attributable to treatments into a common metric, in this case the QALY. Stolk et al used a time trade off approach to transform benefits in quality of life to quality adjusted life years.¹ A population sample was asked to trade off the alternatives of being in a less desirable health state for a longer period, followed by death, versus being in a more desirable state for a shorter period followed by death.²

There are several well known assumptions, and many practical problems, associated with generating utility values.³,⁴ Firstly, the QALY depends on an assumption that the trade off between different health states is known, rather than subject to uncertainty and measurement error. Secondly, a constant proportional trade off between risks is assumed—that is, we consider two years at a utility of 0.5 to be worth one year at a utility of 1 (perfect health). Thirdly, we assume that QALY valuations are independent of previous health states. Fourthly, many cost utility models are “black box” analyses where it is hard to disaggregate contributing components even when the methods are clearly written (as in Stolk et al'spaper) so the validity of a model must be taken on trust to some extent. All these assumptions serve to question the validity of generating a single cost utility measure.

In evaluating the cost effectiveness of sildenafil, Stolk et altarget the restoration of sexual function, and do not distinguish between the situation where there is one failed attempt at sexual intercourse in two attempts, or five failures in 10 attempts. In generating estimates of the utility of sildenafil Stolk et aldid not take into account the experiences of men with erectile dysfunction in the trials, but based their estimates on a survey of the general population and thus on the imagination of their sample. The time trade off approach confounds time preferences with patient preferences, thus downgrading the importance of events that are in the distant future,² in this case death, which may make sildenafil appear a relatively valuable treatment when contrasted with a treatment for a condition where the threat to life is more immediate. The time trade off has also only moderate agreement with alternative methods for generating utility measures.²,³

The only convincing argument for conflating cost and utility information into a single summary measure (the QALY) is to compare treatments for a range of conditions. However, since there are so many good reasons to suppose that QALY estimates are derived using strong assumptions, are context specific, and are not comparable across different diseases, we may question whether Stolk et al's methods are the most appropriate.

Like many newly developed drugs, sildenafil is supported by a programme of randomised trials that provide good evidence of its clinical effectiveness. In the pivotal trials sildenafil was associated with a real improvement in sexual function.⁵ A more robust cost effectiveness analysis might focus directly on the trial programme and provide estimates of the costs and effects attributable to sildenafil in the clinical outcomes measured—in other words, unpacking the black box and making explicit the costs and benefits of sildenafil. Estimates of usage and tolerability may be gleaned from the trial programme and open label extension studies. This approach will avoid the need for the strong assumptions required to fulfil the specification of the cost utility approach. Having established and described what the drug may achieve in use and at what cost, it is then a difficult political rather than technical decision whether it is made available.

In the United Kingdom uncertainty remains on whether the National Institute for Clinical Excellence will use QALY methods to redistribute resources for therapeutic interventions and diagnostic techniques.⁶ The alternative is simply to assess each intervention on its merits and make recommendations on the basis of clinical effectiveness and cost considerations. When Professor Sir Michael Rawlins, chair of the National Institute for Clinical Excellence, commented that recommendations will be based on difficult judgments which have “no mathematical quantitative approach”⁷ he appeared to be favouring the latter. This will more honestly reflect the evidence base, enable a broader public debate, and increase public accountability.

Papers p 1165