Cost effectiveness of lowering cholesterol concentration with statins in patients with and without pre-existing coronary heart disease: life table method applied to health authority population

Introduction

The results of the Scandinavian simvastatin survival study provide robust evidence for the effectiveness of lowering cholesterol concentration with simvastatin in patients aged 35-70 years who have a history of coronary heart disease and a serum cholesterol concentration >5.4 mmol/l.1 More recently, the west of Scotland coronary prevention study has provided convincing evidence for the benefit and safety of pravastatin in middle aged men without coronary heart disease and a moderately raised cholesterol concentration (initial value >6.4 mmol/l).2

Given this evidence for the effectiveness of statins, can the NHS afford to treat all who might benefit? A typical health commission with a population of 500 000 will have around 50 000 men aged 45 to 64, of whom some 20 000 will have a cholesterol concentration over 6.4 mmol/l and therefore would probably benefit from treatment. The current cost of statins at typical daily doses is around £540 a year,3 and to treat all who might benefit would cost £10.8m per annum. In addition, up to 2000 adults in this age group with cholesterol concentrations between 5.5 and 6.4 mmol/l and pre-existing coronary heart disease would also benefit from treatment.

Cost effectiveness studies of lowering cholesterol with statins have used estimates of the likely changes in serum lipid concentrations to predict the benefits of intervention on morbidity and mortality.4 5 6 7 Various estimates of the cost effectiveness of secondary prevention based on the results of the Scandinavian trial have been reported, suggesting an average cost of £85 000 to £136 000 per life saved8 9 or £23 110 to £32 440 per life year saved.10

The Scandinavian trialists reported that the relative risk reduction with treatment is the same whatever the initial cholesterol concentration.11 Because serum cholesterol concentration is an independent risk factor for fatal coronary events,12 13 14 15 16 17 18 19 the absolute risk reduction will depend on the patient's cholesterol value before treatment. Also, the relative risk reduction is likely to be the same for patients in different absolute risk groups because of other independent risk factors such as age, sex, and the nature of the patient's pre-existing coronary heart disease. This assumption is supported by the fact that the relative risk reductions reported in the Scottish trial were similar to those in the Scandinavian trial, although the study populations were dissimilar. The various combinations of risk factors will therefore result in widely differing risks of fatal cardiovascular disease, and so the cost effectiveness of lowering cholesterol concentration will also vary widely.

We estimated the cost effectiveness of statins in lowering serum cholesterol concentration in people at varying risk of fatal cardiovascular disease on the basis of published estimates of the benefits of intervention. We also explored the cost and cost effectiveness implications for a typical purchasing authority of changing the criteria at which treatment of hypercholesterolaemia might be recommended.

Methods

The benefits of lowering cholesterol concentration have been assumed to be a reduction in the numbers of deaths from cardiovascular disease and in the numbers of non-fatal acute coronary events, coronary angiographies, and revascularisation procedures.

We used a life table method to estimate the life years gained by lowering serum cholesterol concentration in the population of this health commission. The method is illustrated in table 1, which is the life table for an imaginary cohort of 1000 men with pre-existing coronary heart disease aged 45-54 (average age at start 50 years). The method for deriving the numbers in each column in table 1 are as follows:

Columns 3 and 4 are the age specific death rates from all causes and from cardiovascular disease in one year age bands for the general population, estimated from a log-linear regression of published death rates in five year age bands.20 Cardiovascular disease includes codes 401-405, 410-414, 430-438 of the ninth revision of the International Classification of Diseases (ICD-9).

Column 5 shows the death rates from cardiovascular disease for men with pre-existing coronary heart disease, estimated with the formula r/ (pr + (1 - p), where r is the mortality ratio of those with pre-existing coronary heart disease to those without and p is the proportion of the population with pre-existing coronary heart disease. We assumed that r = 3.4.14

Column 6 shows the death rate from all causes for those with pre-existing coronary heart disease, or column 3 minus column 4 plus column 5, given that the death rate from non-cardiovascular causes in those with pre-existing coronary heart disease is the same as in the general population.

Column 7 shows the mortality from cardiovascular disease in men with pre-existing coronary heart disease receiving treatment with a statin; the rate is 0.58 x column 5, given that that treatment with statin reduces cardiovascular mortality by 42%.1

Column 8 shows the death rate from all causes in men with pre-existing coronary heart disease, or column 3 minus column 4 plus column 7, given that the reduction in all cause mortality after treatment is entirely due to a reduction in deaths from cardiovascular disease.

Column 9 shows the expected number of deaths without statin treatment, or the number surviving at start of year x column 6.

Column 10 shows the expected number of deaths with treatment, or the number surviving at start of year x column 8.

Column 11 shows the number surviving at end of year without treatment, or number surviving at start of year x (1 — column 6).

Column 12 shows the number surviving at end of year with treatment, or the number surviving at start of year x (1 - column 8).

Total shows the life years survived by the cohort, or the sum of the number surviving at the end of each year + 0.5 x total number of deaths.

A similar life table can be constructed for men and women in the two age bands without previous coronary heart disease, with angina, and with previous myocardial infarction, and within these categories for subjects with cholesterol values in different fifths. The assumptions about the relative risk of fatal cardiovascular disease for these different categories of patient are given in table 2.

Estimating the numbers of non-fatal myocardial infarctions, coronary angiographies, and revascularisation procedures prevented in each risk group is difficult because the age-sex specific rates of these events and their relative risks for different patient categories are not known. However, the results of the Scandinavian and Scottish trials can be used to estimate the numbers of events prevented. There were 0.28 non-fatal myocardial infarctions and 0.38 revascularisation procedures prevented for every life year saved in the Scandinavian trial,10 with broadly similar results in the Scottish trial, which also reported that for every life year saved 0.4 coronary angiographies were avoided. The cost of a non-fatal myocardial infarction and revascularisation procedure are about £430022 and £5000.23 These published figures are similar to local contract prices for these procedures.

The costs of the treatment have been limited to drug costs. Given that in the Scandinavian trial two thirds of the patients required 20 mg daily and one third 40 mg daily, the average cost at current prices is £540 a year.3 The Department of Health traditionally discounts future costs and savings at 5-6%,24 though the validity of this procedure has recently been questioned.25 In the model we used a discount rate of 5% and a range of values from 0-10% in the sensitivity analysis.

The results of the model depend on assumptions made about the values of various input parameters. We carried out a sensitivity analysis with a range of potential values for each input parameter. The model was recalculated 1000 times using the Latin hypercube sampling technique to ensure that each input parameter was sampled across the complete range of its putative distribution.26 The values used for each input parameter and the sampling distribution used are given in table 3.

Results

In Cambridge and Huntingdon the values for the cholesterol fifths are roughly <5.5 mmol/l, 5.5-6.0 mmol/l, 6.1-6.5 mmol/l, 6.6-7.2 mmol/l, and >7.2 mmol/l. Thus those in the lowest fifth would have cholesterol concentrations below those in the Scandinavian trial, and only those in the top two fifths would qualify for treatment in the Scottish trial. Of 95 800 resident men and women aged 45-64 years, 3795 had pre-existing coronary heart disease and a cholesterol concentration >5.4 mmol/l and some 18 100 men without coronary heart disease had a cholesterol concentration >6.4 mmol/l.

Table 4 shows the benefits of treatment for 10 years for patients with pre-existing coronary heart disease, stratified by age, sex, type of coronary heart disease, and cholesterol fifth. Table 5 shows the costs of treatment for the same groups. Table 6 shows the benefits and costs of treating men without a history of coronary heart disease. The values given in these tables are the results when middle values of the input parameters were entered in the model. In those with pre-existing coronary heart disease, 35 619 treatment years would save 442 life years and prevent 137 deaths, 128 non-fatal myocardial infarctions, 164 revascularisation procedures, and 177 coronary angiographies at a net cost of £14.1m—an average cost of £103 000 per life and £32 000 per life year saved. The cohort of men without coronary heart disease would undergo 174 364 treatment years to save 512 life years and prevent 90 deaths, 153 non-fatal myocardial infarctions, 207 revascularisation procedures, and 218 coronary angiographies at a net cost of £74.6m—an average cost of £0.8m per life saved and £147 000 per life year saved.

These averages hide large variations in cost effectiveness for different groups of patients at different degrees of risk. Treatment of hypercholesterolaemia is most cost effective in men aged 55-64 who have previously had a myocardial infarction and whose cholesterol concentration is above 7.2 mmol/l. Treatment costs £6000 per life year saved compared with £361 000 per life year saved in women aged 45-54 with angina and a cholesterol concentration of 5.5-6.0 mmol/l.

Table 7 shows the results of the sensitivity analysis for several key results of the model. For patients with pre-existing coronary heart disease the estimated average cost effectiveness ranged from £15 000 to £70 000. Rank correlation of the inputs and outputs in the sensitivity analysis showed that the effectiveness of treatment was the parameter having the greatest influence on the result (r=0.76). Correlation coefficients for all other input variables were <0.5. For men without coronary heart disease, average cost effectiveness ranged from £70 000 to £424 000, with the effectiveness of treatment again being the parameter having the greatest influence (r=0.95). The sensitivity analysis also shows that even if the “best case” scenario for the lowest risk group is compared with the “worst case” scenario for the highest risk group, treatment is still seven times more costly per life year saved in the lowest risk group.

In deciding which patients should be treated, while taking into account cost effectiveness data, the effect of changing criteria for intervention on the total cost of treatment, average cost effectiveness, and marginal cost effectiveness should be considered. The following are five groups of patients at decreasing risk of coronary heart disease, all of whom fulfil the criteria for intervention of the Scandinavian or Scottish trial:

• Group 1—men aged 45-64 and women aged 55-64 with previous myocardial infarction and a cholesterol concentration >5.4 mmol/l

• Group 2—men aged 45-64 and women aged 55-64 with angina and a cholesterol concentration >5.4 mmol/l

• Group 3—men aged 55-64 with no history of coronary heart disease and cholesterol >6.5 mmol/l

• Group 4—women aged 45-54 with angina or previous myocardial infarction and a cholesterol concentration >5.4 mmol/l

• Group 5—men aged 45-54 with no history of coronary heart disease and a cholesterol concentration >6.5 mmol/l.

The cost implications for Cambridge and Huntingdon Health Commission of expanding the indications for cholesterol lowering treatment to include each successive risk group is shown in table 8. If only those patients in the highest risk group are treated, 1371 patients would require treatment at a net cost of £4.5m (over 10 years) and a benefit of 282 life years saved at £16 000 per life year. If the next highest risk group is also treated an additional 1688 patients will require treating at an additional cost of £11.1m and an additional benefit of 138 life years saved at a marginal cost of £47 000 per life year gained. The average cost effectiveness of treating both groups is £26 000 per life year saved. For treating the fifth risk group the marginal cost per life years saved is £230 000 compared with an average cost of £92 000.

Discussion

Lowering cholesterol has been shown to be safe and effective for patients with and without pre-existing coronary heart disease, but the cost of treating all those in whom treatment is likely to be effective is not sustainable within current NHS resources. In addition, other equally effective secondary prevention interventions, such as antiplatelet treatment27 and dietary intervention,28 29 are cheaper. We have estimated the cost effectiveness of lowering cholesterol concentration in various at risk patient groups to identify groups in which intervention will be most efficient.

Although the validity of several assumptions was explored in a sensitivity analysis, no attempt was made to account for other assumptions. In particular, although the benefits of lowering cholesterol concentration were derived from large trials, our model assumes that they occur immediately treatment is started, whereas not much effect was apparent in the Scandinavian or Scottish trial for six months to two years.1 2 Taking this into account would result in the intervention becoming less cost effective. The population on which this model is based is affluent compared with the rest of the United Kingdom. In an area with higher mortality from coronary heart disease the absolute benefits would be greater and thus the intervention would become more cost effective. Other than drug costs, costs of managing hypercholesterolaemia in the population have not been included in the model. This is reasonable in patients with pre-existing coronary heart disease, since the drug will simply be added to standard treatment and will incur few additional costs associated with doctor time. In patients without coronary heart disease there would be additional costs associated with measuring cholesterol concentrations in a large healthy population group, and in those subsequently treated there will be costs from increased use of doctor time.

The model does not account for several direct and indirect benefits that might accrue from lowering cholesterol concentration. For example, treatment may reduce morbidity from angina or congestive heart failure. This would reduce direct costs through reduction in need for other treatment and a reduction in patient consultations. Measuring these benefits is, however, difficult without published information about such outcomes, and the benefits are likely to be small in comparison to the total cost of drugs. Indirect societal costs, such as preventing illness related job losses, are also relevant but are, again, difficult to account for. We also did not assess the differing utility of preventing death or morbidity in the different risk groups as this type of analysis was beyond the scope of this study.