1. Economic evaluation of MRC/BHF Heart Protection Study
Heart Protection Study Collaborative Group University of Oxford UK
The Heart Protection Study (HPS) involved the collaboration of a large number of doctors, nurses and patients around the UK.
The economic evaluation of HPS involved a further collaboration between the Clinical Trial Service Unit (CTSU) and the Health Economics Research Centre (HERC) at the University of Oxford.
The writing committee for the economic analyses, who undertook all data preparation, analysis and writing up, were (in alphabetical order): Jane Armitage, Andrew Briggs, Rory Collins, Alastair Gray, Borislava Mihaylova, Sarah Parish and Allen Young.

2. HPS: Eligibility criteria
Increased risk of CHD death due to prior disease:
Myocardial infarction or other CHD;
Occlusive disease of non-coronary arteries; or
Diabetes mellitus or treated hypertension
Age years
Total cholesterol  3.5 mmol/l ( 135mg/dl)
Statin or vitamins not considered clearly indicated or contraindicated by patient’s own doctors
The eligibility criteria for HPS are shown. Of note are that subjects recruited do not necessarily have high cholesterol levels by conventional standards and the patients’ own doctors did not consider statin treatment to be clearly indicated.
However, all subjects had an increased risk of coronary heart disease (CHD) death due to existing disease.
The original hypothesis was that reducing cholesterol would be beneficial in those at increased risk of CHD regardless of their initial cholesterol level.

3. Cost-effectiveness analysis of allocation to 40mg daily simvastatin
Based on within trial period only (mean 5 years)
Costs for UK National Health Service (2001)
Cost-effectiveness analyses undertaken:
per major vascular event* avoided
per vascular death avoided
overall and in subgroups at differing absolute risk
*first or subsequent heart attack, stroke or revascularisation following randomisation
The economic evaluation of HPS presented at the 2003 American Heart Association annual meeting is based on an analysis during only the scheduled trial treatment period (mean follow-up time of 5 years in both arms).
Only costs for the UK National Health Service were included and these were measured in pounds Sterling at 2001 prices.
The focus of the economic analysis was cost-effectiveness, with the outcomes of interest being the study treatment’s effect on major vascular events (MVEs: defined as first or subsequent occurrence of heart attack, stroke or revascularisation) and on vascular deaths.
These event rates and the health service costs are presented both separately and as cost-effectiveness ratios in terms of cost per MVE avoided and cost per vascular death avoided.
Although results are presented for the overall comparison of the two trial arms, it will be argued that the key results of interest are the cost-effectiveness estimates by subgroups at different absolute risks of CVD.

4. Major vascular events (MVE) and vascular deaths per 1000 patients
Simvastatin
allocated
Placebo
Avoided per
1000 (SE)
All MVE
270
359
89 (10)
Vascular death 76 91
15 (4)
This table shows the MVE and vascular death rates by treatment allocation presented as the number of events per 1000 patients over the five year mean follow-up of HPS.
The difference between the event rates in the two arms gives the numbers of events per 1000 patients that were avoided by allocation to simvastatin.
HPS showed that 89 MVEs were avoided per 1000 patients allocated 5 years of simvastatin and, for vascular deaths, 15 deaths per 1000 patients were avoided.
It should be noted that these intention-to-treat comparisons are based on an average compliance during the study of approximately two-thirds, which produced an average difference in LDL cholesterol of 1 mmol/l (or 40mg/dl) between the two treatment groups.

5. HPS: Methods of calculating costs
Costs included
Study simvastatin (40mg daily at £1/day) and any non-study statin
Hospitalisations for all major and other vascular events
Costs excluded
Non-statin drugs
Hospitalisations for non-vascular events (no significant differences between groups)
Primary and social care costs (no data available in HPS)
The cost of a daily 40mg dose of simvastatin, the study treatment in HPS, was approximately £1 per day in Since patients’ doctors were free to prescribe non-study statin if it became clearly indicated, the cost of non-study statin was also included in the costing.
A major cost of treating cardiovascular disease is, however, the cost of hospitalisation episodes. All hospitalisations were assigned a specialty code and days in hospital were costed by specialty to give a total cost for each hospitalisation episode. All hospitalisation costs related to major or other vascular events were included in the analysis.
Non-statin drugs and hospitalisations for non-vascular events were not included in the final analysis since there was no evidence of any material difference between the treatment groups of HPS.
Primary and social care costs that might ordinarily be included in an economic analysis were not collected in HPS, and no attempt was made to estimate these costs from external sources.

6. Mean costs per patient for statin use and hospitalisation for any vascular event
Cost category
Simvastatin
(n = 10,269)
Placebo
(n = 10,267)
Difference (SE)
Statin use
£1,712
£215
£1,497 (8)
Vascular events
£1,819
£2,319
-£500 (78)
Total
£3,530
£2,534
£996 (79)
This table shows the estimated costs of statin use and hospitalisation for vascular events during HPS split by treatment allocation.
The cost of statin drugs was approximately £1500 greater per patient among those allocated simvastatin over the 5-year mean follow-up of the trial.
Offset against this increased cost is the estimated cost-saving of £500 per patient reflecting the reduced number of hospitalisations for vascular events in the simvastatin group.
This gives an estimated net cost increase of approximately £1000 associated with allocation to simvastatin.

7. Overall cost-effectiveness within trial (95% CI)
Cost per MVE avoided:
£11,000 (£8-16,000)
Cost per vascular death avoided:
£66,000 (£42-135,000)
Combining the estimated reduction in MVEs and vascular deaths associated with allocation to simvastatin and the estimated net increase in cost generates cost-effectiveness ratios of £11,000 per MVE avoided and £66,000 per vascular death avoided.
However, these results represent an average across all patients in the trial. Of particular interest is the variability in the cost-effectiveness estimates for patients at differing absolute risks of CVD.

8. Assessing subgroup effects reliably
Analyses in different subgroups indicate:
Similar relative reduction in vascular events
Similar relative reduction in costs of vascular events
Similar absolute difference in statin treatment cost
Hence, cost-effectiveness for subgroups estimated by applying overall treatment effects to placebo event rates and costs observed in each subgroup
In estimating cost-effectiveness by patient subgroup it is important to avoid the well-known pitfalls of increased random error within subgroups.
The univariate subgroup analyses reported in the HPS clinical publications indicate similar relative reductions in vascular events in different subgroups.
Further analyses using cost showed similar relative reductions in the cost of vascular events.
By contrast, similar absolute differences in statin cost between treatment groups were found for different univariate subgroups.
Hence, the approach to cost-effectiveness was to apply these overall effects estimated from the whole study to the placebo event rates and costs observed in each subgroup.

9. 5-yr risk vascular death
Cox model on baseline characteristics used to create 5 multivariate risk groups
Risk Group
Number
5-yr risk MVE
5-yr risk MCE
5-yr risk vascular death 1 2 3 4 5
4107
4108
12%
18%
23%
28%
42% 4% 7%
10%
13%
22% 3% 5%
A standard multivariate Cox proportional hazards model was used to identify subgroups at differing absolute risk. Explanatory variables in the model included age, sex, treatment allocation, prior disease, LDL, HDL, blood pressure, treated hypertension, smoking and creatinine.
Using predictions from the Cox model, the 5 year risk of MVE was estimated and patients were stratified into five equal-sized groups based on this prediction. The average 5-year risk of MVE for each group is shown in the table, together with the 5-year risk of major coronary event (MCE) and vascular death.
Note that this yields an almost four-fold difference in the risk of MVE between the low and high risk groups. By contrast, the risk of MVE in the lower and upper third of LDL cholesterol at baseline was 22% versus 27%, which is a much narrower range of risk (and similar to the difference between the third and fourth multivariate risk groups).

10. Similar relative reduction in first MVE by prior disease and age
SIMVASTATIN
PLACEBO
Rate ratio & 95% CI
STATIN better
PLACEBO better
Baseline
feature
(10269)
(10267)
Prior disease
Any CHD
21.8%
27.5%
No prior CHD
CVD
18.7%
23.6%
PVD
24.7%
30.5%
Diabetes
13.8%
18.6%
Age (years)
<65
16.9%
22.1% ≥
65 <70
20.9%
27.2% 70
28.7%
ALL PATIENTS
19.8%
25.2%
24% SE 3
reduction
(2P< )
0.4
0.6
0.8
1.0
1.2
1.4
Similar RELATIVE reductions in risk of first MVE associated with simvastatin allocation were seen across subgroups by prior disease and by age.
Absolute event rates by treatment allocation are also shown.

11. Similar relative reduction in first MVE by LDL & HDL
Baseline
SIMVASTATIN
PLACEBO
Rate ratio & 95% CI
feature
(10269)
(10267)
STATIN better
PLACEBO better
LDL Cholesterol (mmol/l)
<3.0
17.6%
22.2% ≥
3.0 <3.5
19.0%
25.7% ≥
3.5
22.0%
27.2%
HDL Cholesterol (mmol/l)
<0.9
22.6%
29.9% ≥
0.9 <1.1
20.0%
25.1%
Similar RELATIVE reductions in risk of first MVE associated with simvastatin allocation were also seen across subgroups of baseline LDL and HDL-cholesterol. Absolute event rates by treatment allocation are also shown.
Of note is that the range of absolute risk of MVE across the LDL groups is only from 22% to 27% (compared with 12% to 42% for the multivariate risk group range). ≥
1.1
17.0%
20.9%
24% SE 3
ALL PATIENTS
19.8%
25.2%
reduction
(2P< )
0.4
0.6
0.8
1.0
1.2
1.4

12. Similar relative reduction in first MVE
by risk group
Risk
SIMVASTATIN
PLACEBO
Rate ratio & 95% CI
group
(10269)
(10267)
STATIN better
PLACEBO better
1 (low risk)
8.3%
11.8% 2
13.9%
18.2% 3
18.4%
24.9% 4
24.5%
29.6%
5 (high risk)
33.8%
41.4%
24% SE 3
ANY OF ABOVE
Similar RELATIVE reductions in the risk of first MVE associated with simvastatin allocation were also seen across the 5 multivariate risk groups.
19.8%
25.2%
reduction
(2P< )
0.4
0.6
0.8
1.0
1.2
1.4

13. Similar relative reduction in costs (£) of all vascular events by prior disease and age
Baseline
STATIN
PLACEBO
feature
Prior disease
Any CHD
2158
2675
No prior CHD  2
CVD
1281
1641 =
0.8 3
PVD
1866
2563
Diabetes
1076
1445
Age
< 65
1572
2066
For costs, the same pattern is seen with similar relative reductions associated with simvastatin allocation in the cost of all vascular events across subgroups of prior disease and age. Estimated absolute costs by treatment allocation are also shown.  2 =
0.8
1958
2369 2
> 70
2115
2710
ALL PATIENTS
1819
2319
0.78 ( )
0.4
0.6
0.8
1.0
1.2
1.4

14. Similar relative reduction in costs (£)
of all vascular events by risk group
Risk
STATIN
PLACEBO
group
1 (low risk)
784
1219 2
1364
1746  2 3
1773
2120 =
4.0 4 4
2168
2610
5 (high risk)
3000
3903
For costs, the same pattern is seen with similar RELATIVE reductions associated with simvastatin allocation in the cost of vascular events across 5 multivariate risk subgroups.
Also shown are the estimated absolute costs by treatment allocation. Note the increasing per patient hospital costs for vascular events with increasing absolute risk.
ALL PATIENTS
1819
2319
0.78 ( )
0.4
0.6
0.8
1.0
1.2
1.4

15. Similar absolute difference in statin costs
by risk group
Difference
£1555
£1536
£1534
£1485
£1371
(SE)
(16)
(17)
(19)
500
1000
1500
2000
Statin costs (£)
By contrast, similar ABSOLUTE statin drug costs and similar absolute statin cost differences are seen across the 5 risk groups.
Of note is that the use of non-study statin in the placebo group varied only very slightly across these risk groups. By contrast, the next slide shows that there was greater differential non-study statin use in the placebo group across age and prior disease subgroups. S P S P S P S P S P 1 2 3 4 5
Risk group

16. Similar absolute difference in statin costs by disease and age
£1494
£1503
£1514
£1488
(SE)
(10)
(12)
(11)
(16)
(14)
500
1000
1500
2000
Statin costs (£)
This slide shows the similar absolute statin drug costs and absolute cost difference by prior disease and age subgroups. S P S P S P S P S P
Any CHD
No CHD
< 65
≥ 65 < 70
≥ 70
Prior disease
Age (years)

17. Vascular event cost-savings by risk group
100%
% Current simvastatin price
50%
25%
This slide shows the estimated cost offsets for each of the 5 risk groups plotted as bars. The absolute cost offsets can be read from the left hand axis, and the percentage of the cost of statin treatment (at 2001 UK statin prices) offset from the right hand axis.
The horizontal red line indicates the absolute statin cost across the five risk groups at 2001 UK statin prices.
12%
5-yr risk MVE
42%

18. Vascular event cost-savings by risk group
100%
% Current simvastatin price
50%
25%
Since simvastatin has recently come off patent in the UK, and several other European countries, prices are likely to fall.
If the generic simvastatin price were to be 50% of the 2001 price, then the slide shows how it would become cost saving to treat the highest risk patients, and a greater proportion of the statin cost would be offset in the other risk groups. The white part of the bar shows the level of cost-saving.
12%
5-yr risk MVE
42%

19. Vascular event cost-savings by risk group
100%
% Current simvastatin price
50%
25%
At a generic price of 25% of the 2001 price of simvastatin, the four highest risk groups become cost-saving and most of the cost would be offset in the lowest risk group.
12%
5-yr risk MVE
42%

20. Cost per MVE avoided by risk group
Overall: £11,000
The cost per MVE avoided is plotted for each of the 5 risk groups along with the associated 95% confidence interval. Of note, is how different the between-risk-group cost-effectiveness ratios are compared to the overall estimate, with about a ten-fold difference in cost-effectiveness ratio between the lowest and highest risk groups.
12%
5-yr risk MVE
42%

21. Cost per vascular death avoided by risk group
Overall: £66,000
A similar pattern is seen for the cost per vascular death avoided, where there is more than a ten-fold difference between the highest and lowest risk groups.
12%
5-yr risk MVE
42%

22. On-going health economic analyses
Extrapolation to effects beyond trial period
Assessment of cost per QALY
Adapt analyses to other countries
The analyses presented here represent the within trial analysis based on the estimated UK costs. This within trial analysis is likely to underestimate the overall benefits of treatment over a life-time, and work is ongoing to project the long-term costs and potential savings.
Cost-effectiveness ratios based on MVEs are difficult to interpret (as MVEs include heart attack, stroke and revascularisation as well as death) and these outcomes can have very different impacts on length and quality of life. By contrast, focusing on vascular death underestimates the potential benefit of reducing non-fatal events. Work is ongoing to include quality of life measures in order to estimate cost per Quality Adjusted Life Years (QALYs).
Although HPS was conducted exclusively in the UK, it is the largest trial of cholesterol-lowering therapy, and its applicability extends beyond the UK. Work is ongoing to adapt the analyses for other countries.

23. CONCLUSIONS: Economic evaluation of HPS
Simvastatin allocation reduced vascular hospitalisation costs by 22% regardless of patient characteristics
Cost-effectiveness is chiefly determined by an individual’s overall risk of vascular events (rather than by single risk factors, such as LDL)
Statin treatment is cost-effective for a wide range of high-risk individuals (and will become increasingly so as statin prices fall)
Simvastatin allocation reduced vascular hospitalisation costs by 22%, regardless of patient characteristics. For statin cost it is the absolute cost that does not vary across patient characteristics.
Consequently, cost-effectiveness is chiefly determined by an individual’s overall risk of vascular events rather than by single risk factors (such as LDL). Therefore, the overall cost-effectiveness estimates for the trial, which by definition average across patients of differing absolute risk, are not useful summary measures of cost- effectiveness.
Statin treatment seems likely to be cost-effective for most, if not all, of the types of patients recruited in HPS - and may indeed be cost-saving over just 5 years for higher risk patients as statin prices fall.