Cost utility analysis of sildenafil compared with papaverine-phentolamine injections

Clinical effects

The study by Goldstein et al is the largest dose escalation study reported.3 It was placebo controlled and the patient population consisted of men with erectile dysfunction due to various causes. Efficacy was assessed with the international index of erectile function.5 This instrument contains questions about the two primary end points of erectile dysfunction treatment as defined by the National Institutes of Health—that is, the ability to penetrate and the ability to maintain an erection sufficient for satisfactory sexual intercourse.6 These end points were used in the trial. Both questions have five response levels, so together they categorise the patients into 25 (5×5) erectile dysfunction states. These erectile dysfunction states were valued in a separate exercise (described below). The elicited utilities were applied to the health states of the patients in the study of Goldstein et al before and after treatment. The difference between the mean utility before and after treatment (controlled for placebo) is the mean gain in utility. Use of disease specific instruments to calculate QALYs is advocated by Brazier and Dixon7 and Drummond et al.8

Because we used previously reported trial data, we had to consider the limitations of these data for use in economic evaluation. Firstly, the trial was designed on an intention to treat basis,3 which meant that patients for whom sildenafil had no or insufficient effect remained in the trial. As we couldnot discriminate between patients with a sufficient or an insufficient response, we used the mean utility gain in the trial to calculate the utility gain of sildenafil. Consequently, we underestimated the utility gain in daily practice because only the utility gain of the successfully treated patients should be taken into account. Secondly, results of the international index of erectile function were not available for injection therapy, nor were any other data that allowed calculation of QALYs. We conservatively assumed that theutility gain of sildenafil and papaverine-phentolamine injections would be the same. Given the low acceptability of injection therapy, 9 10 this assumption probably overestimates the benefits of injection therapy.

Determining utilities for erectile dysfunction states

From a randomly selected sample of 45 000 people obtained from the Rotterdam telephone directory we recruited 354 people to participate in the valuation task. They were invited by telephone to attend a session of health state valuation and were offered about £10 plus travel expenses. In order to avoid selection bias, the invitation was made without referring to erectile dysfunction. Participants were given the opportunity to withdraw from the valuation sessions without financial consequences after they were informed about the subject of the study.

Participants valued 24 erectile dysfunction states on a scale from 0 to 1 using time trade-off.8 The 25th state described normal erectile functioning and was set at a value of 1.0. Time trade-off was measured relative to the life expectancy of the subjects. Before the valuation task, participants gained experience of the time trade-off method using general health states as defined by the EQ-5D questionnaire.11

Time trade-off responses were considered invalid if the participant showed a lexicographic response for the EQ-5D states, had too much missing data either on erectile dysfunction or EQ-5D states, or clearly did not understand the task. A lexicographic response mode means that when a respondent is faced with an option he or she will always choose one particular alternative, no matter how favourable the other might be. Subjects had to value the health states “for a person like yourself.” This means, for example, that older people gave values from their own perspective, and people without a sexually active partner would take this into account when performing the valuation task. The exception was that women were asked to imagine being a man with erectile dysfunction. Values are independent of the sexual activity of the respondents because the descriptions of erectile function referred to the relative number of successful attempts at intercourse. For example, a respondent might be asked the following: “If during the past four weeks, your condition was such that you were sometimes able to attain an erection, and you were (almost) never able to maintain your erection, how many years would you be willing to trade off to restore your erectile function?” This also implies that erectile function is valued the same in patients with different levels of sexual activity.

Because we had decided to obtain social valuations we asked a sample of the general public to value the clinical outcomes.8 12 The reasoning behind this decision relates to issues of equity and medical ethics.13 Some authors, however, claim that healthy people relatively similar to affected patients should value clinical outcomes.14 We therefore explored whether erectile dysfunction is valued differently in different subgroups. We used multivariate analysis of variance to determine whether age, sex, the availability of a partner, having children, sexual activity, and sexual satisfaction influenced the values of the general public.

Costs

All costs are expressed in 1999 British pounds (£1=1.62 euro). We used 1999 datato determine the Dutch cost prices. To determine the medical costs, we estimated resourceuse —for example,consultations and prescription charges (a lump sum charge to refund pharmacy costs and medicines) and multiplied the quantities by the unit prices. We estimated resource use of sildenafil and papaverine-phentolamine injections on the basis of consensus statements on both treatments.15 We refined this estimate by developing a low, baseline, and high cost scenario on the basis of clinical experience in two hospitals (University Medical Centre St Radboud, Nijmegen and Hospital St Antoniushove, Leidschendam). Costs outside the healthcaresector and productivity costs were assumed to be negligible.

The cost of sildenafil was based on observational data from the first quarter that sildenafil was available in the Netherlands.16 A general practitioner or urologist determined the effective dose in an academic or peripheral setting. An appropriate share of the costs of supporting departments was reflected in the cost of a visit to a urologist. The physician's costs were calculated onthe basis of the estimated duration of an outpatient visit. The analysis included all costs related to the hospital, such as costs of salaries and supplies, costs of supporting departments, and overhead costs.

Cost effectiveness

We compared two scenarios: treatment with sildenafil and treatment with papaverine-phentolamine(figure). In the sildenafil scenario, we allowed patients to switch to papaverine-phentolamine injections, as these injections may be effective in patients in whom sildenafil has failed. Since sildenafil has already become the treatment of choice, although its cost is not reimbursed in the Netherlands, patients are unlikely to switch from injections to sildenafil. A switch was therefore not allowed in the papaverine-phentolamine scenario.

We compared the sildenafil and papaverine-phentolamine scenarios assuming use once a week. The maximum recommended frequency of papaverine-phentolamine injections is once a week, but use of sildenafil is not limited for medical reasons. Nevertheless, we believe that the assumption was appropriate as the utility values were elicited independently of the frequency of intercourse. Moreover, the marginal utility gain of increasing intercoursefrequency from once to twice a week is unlikely to be as high as the increase between never being able to have intercourse and being able to have intercourse once a week.

Patient flow for scenarios of treatment of erectile dysfunction with sildenafil or papaverine-phentolamine injections

• Download figure
• Open in new tab
• Download powerpoint

The cost effectiveness of the different scenarios was analysed in a model comprising acceptability of treatment, probability of successful treatment, switching or discontinuation of treatment, and the duration of successful treatment. The patient flows in the model were determined on the basis of secondary data—for example, published clinical trials,9^–20 Dutch observational data,16 and clinical experience in the two participating hospitals. We performed an incremental analysis of the costs and effects of sildenafil compared with papaverine-phentolamine. The results are presented as cost per QALY.

The acceptance rate of papaverine-phentolamine treatment could have been influenced by the factthat erectile dysfunction is no longer a taboo subject. In fact, an acceptance rate of 70% has been suggested as feasible.21 We therefore included this variable in a sensitivity analysis. Other variables included in the sensitivity analysis wereresource use, values, effectiveness of treatment, andfrequency of use. We performed univariate sensitivity analysis to determine which variables have the largest influence on the results. In the multivariate sensitivity analysis we explored to what extent results would change under a (unlikely) worst case scenario

Respondents

A total of 184 subjects (52%) failed to attend the interview sessions. This was probably because of extremely bad weather at the time of interview, which made it difficult for participants to reach the university. One person withdrew from the study after he was informed about the subject. Asample of 169 subjects valued the erectile dysfunction states;89% (150) of the responses were valid. Age ranged from 18 to 80 years (mean age of 45.8 (SD 15.4) years). There were 81 men (54%) and 69 women, which is close to the sex distribution in the general population.

Effects

In Goldstein et al's study the international index of erectile function among men receiving sildenafil rose from 2.0 at baseline to 3.9 at end of treatment for ability to penetrate (placebo group 2.1 to 2.3)and from 1.5 to 3.6 for satisfactory sexual intercourse (placebo group 1.6 to 1.8).3 table 1 gives the mean utilities that were elicited for the 24 erectile dysfunction states described by these two questions. The utilities ranged from 0.74 to 0.94. When these values are combined with trial data, the mean utility increased from 0.807 at baseline to 0.915 at end of treatment for men receiving sildenafil and from 0.819 to 0.821 for men receiving placebo. Therefore, the mean utility gain attributable to sildenafil is 0.11.

We analysed whether the values of the general public were influenced by age, sex, availabilityof a partner, having children, sexual activity, and sexual satisfaction. The only relation we found was that participants with children considered erectile dysfunction less of a problem than subjects without children. Since there were no differences between the values of men and women, we used averaged values in the QALY analysis. More extensive description of this analysis is available.22

Costs

Tables 2 and 3 show the resource use and the costs attributable to treatment of erectile dysfunction with sildenafil or papaverine-phentolamine injections. Papaverine-phentolamine is cheaper per dose, but it has to be prescribed by an urologist and therefore has higher initial costs (£484 versus £407 for sildenafil). Sildenafil has higher running costs: yearly treatment costs are £254 versus £233 for papaverine-phentolamine. The higher initial costs of papaverine-phentolamine are recovered after seven years.

Cost effectiveness

Overall, sildenafil creates more benefits and more costs because more patients are treated (figure). Therefore, the main issue is whether the additional effects of sildenafil are worth the additional costs. This question is addressed in the incremental analysis shown in table 4. The incremental cost utility ratio of sildenafil compared with papaverine-phentolamine is £3639 per QALY in the first year, decreasing to £2630 per QALY after five years.

Sensitivity analysis

The frequency of use influences the outcomes considerably. Doubling the frequency of use of sildenafil increases the cost per additional QALY by 45% in the first year and 85% in each following year. The initial costs are relatively high because the costs of non-responders are added to the costs of responders. Hence, the effect of the frequency of use on the cost per additional QALY is moderated in the first year. In the long term, however, the main cost driver with sildenafil is thedrug.

Assuming a lower utility gain (0.08) than observed in the valuation study, resulted in a37.5% increase in cost per additional QALY.4 Effectiveness and acceptability also influenced the results significantly. The cost per additionalQALY increased 38% with a lower effectiveness of sildenafil (50%), but decreased (1%) in each following year. Changes in acceptability had an opposite effect: when acceptability of papaverine-phentolamine injections is increased to 70%, the incremental cost utility ratio is 25% lower in the first year, but 10% higher from the second year onwards. Uncertainty about resource use did not influence the outcomes significantly; in the analysis of different cost scenarios (based on the number of visits and duration of visits), the high cost scenario increased the costs per additional QALY by only 8%.

When the uncertainty of all variables is combined into a worst case model (low utility gain and effectiveness and high costs, dropout, and acceptability), the incremental cost utility ratio is£ 9343 per QALY in the first year(156% increase), and £4691 in each following year(101% increase).

Validity of assumptions

We made several assumptions that could be viewed as unfavourable to sildenafil. For instance, we underestimated the effects by not including partner satisfaction and we assumed the effects of oral and injection treatment to be equal. Furthermore, we used a relatively low rate of drop out for injection therapy, which results in a more favourable cost effectiveness ratio for injection therapy.9^–20 Although such assumptions might introduce bias, the interpretation of the results is not greatly affected because the assumptions in the economic appraisal of sildenafil were conservative.

The utility values we elicited for erectile dysfunction did not take into account possible comorbidity. As in most cases total disutility is less than the sum of parts,26 we might have overestimated the effect. However, the sensitivity analysis showed that sildenafil remained cost effective with lower utility gains.

The subjective nature of the value of erectile functioning again raises issues about whose values should be used in economic appraisal of health care: the values of the general public or thoseof people at risk (in our case ageing men). However, we found that the utility values for sexual functioning were independent of background variables such as age, sex, and sexual activity. Therefore, neither the limitations in the representativenessof our sample, nor our choice to elicit values from the general public has influenced the results.

Implications

These findings should be interpreted in the light of the discussion about the affordability andvalue of sildenafil to society. Firstly, we have shown that erectile dysfunction limits quality oflife considerably, in the eyes of the general public. Furthermore, our study shows that sildenafilis cost effective, and its reimbursement should therefore be considered. However, as frequency of use greatly affects cost, such reimbursement should not be unconditional.