Population based cost utility study of interferon beta-1b in secondary progressive multiple sclerosis

Introduction

Interferon beta is the first treatment to alter the natural course of relapsing-remitting multiple sclerosis.1^–3 In a recent trial involving 718 people the drug was also shown to modify the secondary progressive form of the disease, reducing relapse rates over a 30 month follow up.4 Furthermore, the proportion of interferon beta-1b recipients who became wheelchair dependent was reduced by 32%, and this benefit lasted for nine months.4 A lack of effect on progression of disability was mooted as a reason for not making interferon beta-1b (which costs over £9600 a year for each patient) more widely available to patients with relapsing-remitting multiple sclerosis in the United Kingdom.5 The disease course is different in secondary progressive multiple sclerosis,6 so evidence from trials in relapsing-remitting multiple sclerosis cannot be extrapolated to patients with secondary progressive disease.

The results of the recent trial in secondary progressive multiple sclerosis4 will increase demand for interferon beta-1b, placing further pressure on limited budgets. As resources are limited, spending on treatments such as interferon beta means foregoing benefits from other forms of care—this is known as the opportunity cost.7 To allow comparison of the relative merits of alternative options we conducted a cost utility analysis of treatment with interferon beta-1b in secondary progressive multiple sclerosis based on the benefits shown in the recent trial (box).4

Methods

We compared the effect of treating a cohort of people with secondary progressive multiple sclerosis with interferon beta for 30 months against existing best practice without interferon beta. Ideally, we would have used effectiveness data derived from a meta-analysis of clinical trials, but only one completed trial exists with data in the public domain. Therefore, our model represents a best case scenario based on current knowledge (box). We estimated the costs and benefits of a programme of treatment in sufficient people to postpone wheelchair dependence in one person Although we took a health service perspective, we allowed for a societal perspective through use of sensitivity analysis.

Study population

We aimed to identify all patients with multiple sclerosis in Tayside in order to have a representative cohort. Residents of Tayside region (population 395 600) with multiple sclerosis were identified from four sources (neurology department records, visual evoked response requests, Scottish morbidity records, and a survey of the region's general practitioners). Diagnosis was confirmed from hospital or primary care case records.8 Capture-recapture methods indicate that our point estimate of prevalence on 1 September 1996 was 94% complete.9 Using accepted definitions,6 we identified people with secondary progressive disease on the basis of the most recent information in their case records.

Patients who had diagnosed secondary progressive disease on 1 January 1993 formed the cohort for this study (population cohort). To explore the effect of limiting prescription to ambulatory patients with more active disease, a subset of the population cohort who were either admitted to the neurology unit or referred to neurology outpatient clinics at Tayside hospitals during January 1993 to June 1993 were identified (neurology subset). In Tayside, people with secondary progressive multiple sclerosis are not routinely followed up but are referred to the outpatient clinic for specialist management of new problems such as relapses or other chronic symptoms. All of the cohort was followed up until 31 December 1995

Relapse rates and wheelchair dependence

We used the same definitions of relapse as the clinical trial.4 Admissions for treatment of a relapse and the year in which patients became wheelchair dependent were identified from hospital case records. Ideally, we would have calculated the proportion who became wheelchair dependent within 30 months (the same follow up as the trial4), but the case records lacked detail. Thus the number needed to treat to delay time to wheelchair dependence could be an underestimate.

We identified episodes of oral corticosteroid prescription from the University of Dundee's medicines monitoring unit prescribing database using a record linkage method.10 We assumed that each corticosteroid prescription represented a community treated relapse. Prescriptions associated with bronchodilators were assumed to be for asthma and were excluded. We calculated the rate of relapse requiring hospital admission and the rate of community treated relapse and estimated confidence intervals using a method for proportions.11 A previous study of relapse frequency in secondary progressive multiple sclerosis showed a reduction in relapse from the first year after onset to the second year.12 Although we could have used these data in our model, our population cohort contained people who had had the disease for more than two years and were likely to have a lower relapse rate. A recent survey of British neurologists indicated that 47% of neurologists treated at least half of multiple sclerosis relapses with corticosteroids.13 Therefore, we assumed that for every relapse treated (hospital or community) there was another relapse which was not treated.

Costs saved by delaying time to wheelchair dependence and prevention of relapses

We estimated the costs saved by delaying time to wheelchair dependence by nine months using data from a cost analysis based on information from 672 members of the Multiple Sclerosis Society of Great Britain and Northern Ireland (table 1).14 Other estimates of the cost of multiple sclerosis to the United Kingdom did not have sufficient detail for our model. 15 16 We identified resources used during hospital treated relapses in Tayside and valued them using cost data from local NHS trust finance directorates (Robert Hudson, personal communication). We distinguished between fixed and non-fixed costs when costing relapses, but this was not possible for wheelchair dependence as the study14 lacked that information. We assumed that each community treated relapse required 20 minutes' consultation with a general practitioner17 and that untreated relapses did not incur health service costs (although general practitioner costs may have been incurred). The cost of corticosteroids was taken from the 1995 British National Formulary.18 All costs were adjusted to a 1995 price base using the hospital and community health services price index (Department of Health, London).

Calculation of quality adjusted life years

To determine current health we conducted a postal survey of people with multiple sclerosis in Tayside in Spring 1998 (three years after the end of follow up) after obtaining permission from general practitioners This survey created a baseline from which to estimate the effect of interferon beta. Respondents indicated their current level of ambulation on a postal ambulation scale (based on the scale of McAlpine and Compston19) and completed the EuroQol (EQ-5D) health related quality of life instrument.20 Responses from people with secondary progressive multiple sclerosis were selected. We obtained values for EQ-5D health states from published social tariffs using mean valuations based on 10 year trade off values.21 The EQ-5D was chosen as it is simple to complete and has been extensively tested in UK populations. In an unpublished survey of 67 Tayside residents we had previously established that that increasing postal ambulation scale scores accurately identified people with increasing severity of multiple sclerosis and that appropriate changes in utility values were occurring.

The number of QALYs gained by postponing wheelchair dependence by nine months was equal to the change in EQ-5D value associated with a change in postal ambulation scale score from 4 to 5 (table 2) multiplied by 0.75 years. The number of QALYs gained by preventing a one month hospital relapse was equal to the change in EQ-5D value associated with a change in postal ambulation scale score from 4 to 5 multiplied by 0.083 years. For each community treated and untreated relapse, postal ambulation scale scores changed from 3 to 4 for one month. We assumed that 40% of QALYs were gained in years 1 and 2 and 20% were gained in the final six months of the 30 month programme. A discount rate of 6% was applied to QALYs gained in years 2 and 3.

Cost utility ratios

Assuming a 32% relative reduction in the proportion of people who become wheelchair dependent during follow up, we calculated the number needed to treat22 for 30 months to postpone wheelchair dependence by nine months for one person; 95% confidence intervals for the numbers needed to treat were calculated from the 95% confidence limits for the absolute risk reduction and expressed as number needed to treat (harm) to number needed to treat (benefit).23 In our cost utility model, interferon beta-1b cost £800 per patient per month.18 Other costs associated with interferon beta treatment were not included as we had no reliable estimate from our local practice Using trial data we estimated that 12% of hospital relapses would be prevented, that corticosteroid use would decrease by 21%, and that 31% of untreated relapses would be prevented.4 Treatment costs in the second and third years were discounted at a rate of 6% (UK treasury rate). Upper and lower limits of the cost utility ratios were estimated using the 95% confidence intervals of the number needed to treat.

Sensitivity and threshold analyses

A sensitivity analysis explored the effect of (a) increasing the cost of each additional nine months of wheelchair dependence (taking account of a societal perspective), (b) changes in the unit cost of interferon beta-1b, (c) changes in the number of, and discount rate applied to, QALYs gained, or (d) assuming that the relative reduction in hospital admission and corticosteroid use was 31% (not 12% or 21%). Assuming that an acceptable threshold of cost utility would be £10 000 per QALY gained,24 we determined the changes required to model variables to breach this threshold. We also performed threshold analyses at £20 000, £30 000, and £50 000 per QALY gained.

Results

We identified 756 people with multiple sclerosis in Tayside during 1993-5. Of these, 178 had secondary progressive disease, 46 of whom were already wheelchair dependent. There were 132 people in the population cohort, and 17 people in the neurology subset.

Wheelchair dependence

Twenty four (18%) of the population cohort became wheelchair dependent within 36 months of follow up. We estimated that the number needed to treat with interferon beta-1b for 30 months to postpone wheelchair dependence by nine months was 18 (95% confidence interval 5 to 26). Four (24%) of the neurology subset became wheelchair dependent. The number needed to treat to postpone wheelchair dependence by nine months in the neurology subset was 14 (number needed to treat (harm) 9 to ∞ to number needed to treat (benefit) 3).

Relapse rates

We identified 32 admissions for treatment of a relapse in the 36 month follow up period, giving a rate of 0.08 (95% confidence interval 0.06 to 0.11) per patient-year. A total of 56 discrete corticosteroid episodes were identified in 21 people. Three people had concomitant bronchodilators prescribed, so their steroids were assumed to be for asthma. One oral steroid prescription preceded an admission to hospital for relapse by 15 days; all other episodes of community and hospital treated relapses were mutually exclusive. The rate of community treated relapses was 0.15 (0.11 to 0.18) per patient-year. In the 30 months to the end of December 1995, the neurology subset experienced seven hospital admissions for relapse and four community corticosteroid prescriptions and therefore an estimated 11 untreated relapses.

Quality adjusted life years gained

We sent a postal survey to 569 people with multiple sclerosis, of whom 402 (71%) returned valid responses. Of these 402, 84 had secondary progressive disease. Using these 84 responses and a discount rate of 6% for benefits, treatment of 18 members of the population cohort and 14 members of the neurology subset for 30 months would result in a gain of 0.397 and 0.357 QALYs respectively.

Cost utility ratios

Table 3 shows the main values used to calculate the cost utility ratios. The net cost per QALY gained from interferon beta-1b in secondary progressive multiple sclerosis was £1 024 393 (95% confidence interval £276 191 to £1 484 824). For the neurology subset, the cost per QALY gained was £883 209 (£179 209 to ∞).

Sensitivity and threshold analyses

A 190% increase in the cost of each additional nine months of wheelchair dependence from £2840 to £5153 (which takes account of a societal perspective) reduced the cost per QALY gained by only 0.2%. A 25% increase in the number of QALYs gained in secondary progressive multiple sclerosis (to 0.496 discounted QALYs) reduced the cost per QALY gained in the population cohort by 20% (to £819 514). If the unit cost of interferon beta-1b was reduced by half to £4800 per patient a year, the cost per QALY gained fell by 49.4% to £506 407. If we assumed that 31% of hospital admissions and community treated relapses were prevented (in the trial 31% of all relapses were prevented, but only 12% of hospital admissions and 21% of steroid use was prevented4) the cost per QALY gained was £832 399 (£221 831 to £1 208 133) in the population cohort and £628 797 (£122 079 to ∞) in the neurology subset. None of the proposed thresholds could be breached without extreme changes to the model's variables (table 4).

Discussion

The recent trial showing benefit from interferon beta in secondary progressive multiple sclerosis4 has led to a call for the treatment to be made available immediately to all patients with that form of the disease.25 The Multiple Sclerosis Society of Great Britain and Northern Ireland issued minimum standards of care for people with multiple sclerosis26 and is planning a national campaign to raise the profile of the disease in the United Kingdom.27 Despite the need for improvements in the care of people with multiple sclerosis, our analysis shows that treatment with interferon beta-1b has a significant opportunity cost and resources could probable be used better elsewhere.

Robustness of cost utility model

Studies that take a health service perspective are often criticised for ignoring social costs of disease. This is particularly relevant to multiple sclerosis, as healthcare expenditure is typically 15% to 20% of the total cost of the disease. 14 15 28 However, our cost utility ratio was unaffected by changes in the cost of care, and arguments that interferon beta-1b might have an acceptable cost utility ratio if social costs were considered29 now seem implausible. Other countries have estimated similar costs for nine additional months of severe disability (Belgium: 9200 ecu (£6345) health and social service perspective30; United States $2911 (£1712) healthcare perspective31; and Canada $11 427 (£4395) societal perspective28) This model should be valid outside the United Kingdom unless the cost of an additional nine months of wheelchair dependence and six relapses exceeds £400 000.

The high cost utility ratio was driven by the cost of interferon beta-1b and the modest clinical effect. The cost of interferon beta-1b will probably fall over time because of competition from other products,32 but price reductions are unlikely to be sufficient to make its use cost effective (table 4).

Comparisons with other healthcare interventions

Treatment of secondary progressive multiple sclerosis with interferon beta-1b is one of the least efficient interventions for improving health in a population. We identified 104 citations on Medline using the terms “cost” and “quality adjusted life year” from 1995 to October 1998. The median cost per QALY from the 128 cost utility ratios cited was £15 625 (interquartile range £9167 to £74 969) and only one other study exceeded £1 million per QALY gained.33 Although comparisons of the cost per QALY gained may conceal important methodological differences,34 Richards was probably correct to conclude that most competing demands for healthcare funds are likely to produce greater health gains than interferon beta for multiple sclerosis.5

Alternative uses for funds for interferon beta-1b

Our cost utility ratios are of a similar size to estimates in studies of relapsing-remitting multiple sclerosis— £328 000 to £809 000 per QALY in Britain35 and $300 000 in Canada.36 Although targeting treatment at patients with more active disease seemed more efficient than treating all patients with secondary progressive disease, the opportunity cost was still substantial.

The health related quality of life of people with multiple sclerosis is much lower than that of the general adult population,37 matched controls with other chronic diseases,38 and controls with another neurological disorder (epilepsy).39 Therefore, interventions which improve the quality of life of people with multiple sclerosis more efficiently than interferon beta-1b need to be identified. It is probably appropriate to allocate more resources to people with secondary progressive multiple sclerosis, but access to interferon beta-1b should be restricted Realistically, far more benefit would be obtained from directing funds into improved supportive care.