NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Collaborating Centre for Primary Care (UK). Lipid Modification: Cardiovascular Risk Assessment and the Modification of Blood Lipids for the Primary and Secondary Prevention of Cardiovascular Disease [Internet]. London: Royal College of General Practitioners (UK); 2008 May. (NICE Clinical Guidelines, No. 67.)

  • This publication is provided for historical reference only and the information may be out of date.

This publication is provided for historical reference only and the information may be out of date.

Cover of Lipid Modification

Lipid Modification: Cardiovascular Risk Assessment and the Modification of Blood Lipids for the Primary and Secondary Prevention of Cardiovascular Disease [Internet].

Show details

Appendix CHealth Economic Modelling

3.1. Results for the cost-effectiveness of Identifying people at high risk of developing cardiovascular disease

Introduction

Current guidelines recommend statin treatment at a ten-year CVD risk of 20%.(National Institute for Health and Clinical Excellence, 2006) Therefore determining an individual’s ten-year CVD risk is necessary in order to determine both their eligibility for treatment and their likelihood of benefiting from treatment. CVD risk assessment is therefore central to the problem-identifying patients for CVD prevention in primary care.

Calculating cardiovascular risk requires knowledge of a patient’s age, sex, diabetic status, smoking status, total cholesterol, HDL cholesterol and whether or not they have existing cardiovascular disease. However, because it requires both staff time and blood tests, there is a cost to obtaining risk factor information through patient assessment.

Since risk factor assessment has a cost, it would be helpful to pre-select and prioritise those patients most likely to benefit from further assessment. How could we do this? There are a number of potential strategies. One strategy is to see patients at random: in effect this means assessing patients in random order as they present to the practice for other reasons. A second strategy is to prioritise for assessment all patients who are on antihypertensive treatment. This follows the recommendations of the National Service Framework for Coronary Heart Disease. A third strategy is to prioritise patients by their age. A fourth strategy is to undertake a prior estimate of CVD risk on all patients and to prioritise patients by their estimated CVD risk.

This paper describes these four strategies in more detail and then determines the costs and consequences of identifying patients using each strategy.

Methods

The methods section is divided into sections, firstly the methods for the markov model which generated the costs and QALYs for statin treatment. Secondly the modelling of the identification strategies.

Methods 1: Modelling statin treatment benefit

People at risk of developing CVD were modelled. Different CVD risks were modelled for different age groups. Baseline annual CVD risks modelled ranged from 0.5% to 4.75% for people aged between 30 and 74 years old.

The choice of comparators

The analysis compared lower dose of statins versus placebo to estimate the QALY gain per person and the cost per person on statins. The model used the cost of simvstatin 40mg which the GDG recomemded for use in primary prevention. In sensitivity analysis a combination of drugs (50% atorvastatin 10mg, 25% on simvastatin 40mg and Pravastatin 40mg) was used. This was taken from the placebo trials reported in the statins TA 94 (National Institute for Health and Clinical Excellence, 2006).

Outcomes

The treatment effect was measured in terms of prevention of CVD events: non-fatal unstable angina, MI, Stroke, and deaths from these causes and all causes (total mortality).

Health outcomes for the cost-effectiveness analysis are summarised in the form of Quality Adjusted Life Years (QALYs), where one QALY represents one year of healthy life.

The model did not explicitly include cost impacts of withdrawals, non-concordance or transfers between treatments. The impact of such changes on effectiveness is implicitly included through the use of intention-to-treat trial data. These assumptions are conservative, as they will tend to increase the estimated cost and hence the incremental cost-effectiveness ratios for statins, which are expected to have higher withdrawal rates due to adverse events.

Model structure and assumptions

A Markov model was developed to evaluate the incremental costs and effects of lifetime treatment with statins compared with placebo for primary prevention of cardiovascular disease from a UK NHS perspective.

Figure 1 in appendix C 1 shows a schematic representation of the patients’ pathways. All patients start in the well state. During each annual cycle of the model, a proportion of patients enter one of the qualifying event health states (MI, unstable angina, Stroke, or death, while the remainder stay in the event free state. Patients can experience more than one non-fatal event in subsequent periods of the model. However, for members of the hypothetical cohort who experience more than one event, the model only applies the additional costs and adverse health consequences of the most recent event: the model does not have any ‘memory’ of previous states. This is a simplifying assumption, which will tend to underestimate the additional benefits of more effective CVD preventive interventions.

The model was run first assuming that the cohort was receiving no statins. The model was then re-run assuming people were receiving statins with transition probabilities adjusted to reflect the expected reduction in CVD events applying the relative risks observed in the meta-analysis. Health care costs and QALYs were then estimated for each option by weighting the time spent in the various states by mean costs and ‘utilities’ (health-related quality of life) of the health states. The cost and utility data used in the model was sought from literature.

The time horizon modelled is lifetime, with an assumed upper age of 100, by which time most of the cohort has died. Due to the nature of Markov models, some proportion of the cohort remains alive, no matter how high the mortality rates assumed.

Baseline risks

Most of the baseline data was taken from the placebo arm of the meta-analysis done for the guideline and from the statins TA 94 (National Institute for Health and Clinical Excellence, 2006) by Ward et al. Non-CVD related mortality by age and sex was taken from the life tables for England and Wales prepared by the Government Actuaries Department (Government Actuaries Department, 2006) and from data on the proportion of deaths due to CVD-related causes from the Office for National Statistics (Office for National Statistics, 2006). Tables 1 and 2 in appendix C 1 shows the annual distribution of CVD events and annual transition probabilities from the well state. Table 3 shows transitions probabilities once an individual has had a primary event.

Treatment effects

The relative treatment effects of statins were taken from a meta-analysis of primary prevention trials from the Statins TA94 (National Institute for Health and Clinical Excellence, 2006). Table 4 in appendix C 1 shows the mean relative risks and 95% confidence interval. The treatment effect was applied to the annual transition probabilities without treatment in table 2 and 3.

Cost data

The NICE reference case specifies that costs should be measured from an NHS and personal social services perspective. These should include the direct cost of drug treatment and also potential savings from avoided treatments due to reduced incidence of CVD disease. Costs were calculated using cost weights for each of the health states of the model, multiplied by the time spent in each state. Costs are at 2006/2007 prices. As per current NICE guidance, an annual discount rate of 3.5% was used for both costs and health benefits. (National Institute for Health & Clinical Excellence., 2006). Costs of health states were taken from Statins TA (National Institute for Health and Clinical Excellence, 2006) table 12 in appendix C 1. Drug costs were taken from the prices quoted by the Prescription Pricing Authority (NHS Prescription Pricing Authority, 2008) while GP costs were taken from PSSRU (PSSRU, 2005), biochemical tests costs were taken from NHS reference costs 2005/06 (Department of Health Reference Costs, 2007), see table 13 in appendix C 1.

Quality of life (Utility)

In the NICE reference case, the value of health outcomes – including beneficial and harmful impacts of treatment on mortality and morbidity – is estimated using the Quality Adjusted Life Year (QALY) approach. This requires estimates of survival and quality of life associated with each health state included in the model. Utility data was obtained from published literature, mainly from the Statins TA. (National Institute for Health and Clinical Excellence, 2006). We also used age specific utility data from the health survey of England 1996, which shows that utility decreases by age. Table 14 and 15 in appendic C 1, shows heath state utilities and utility by age respectively

Methods 2 Modelling the identification strategies

The costs and effects of using different strategies for identifying patients were determined by modelling the consequences of each strategy in the same population. The population used was derived from the Health Survey for England 2003 (Health Survey for England 1998, 2002) All patients with complete risk factor information were selected. Individuals aged less than 30 and more than 74 were excluded. Individuals with a history of CVD were excluded and individuals with diabetes mellitus were excluded. This left a population similar to those considered for primary prevention of CVD. A separate model was then constructed to estimate the lifetime QALYs and costs for people on statin treatment as described avove.

Identification strategies

The four patient identification strategies under consideration were:

  • Random assessment.
  • Patients who are over 50, then patients who are over 40.
  • Prioritise patients by their age.
  • Prioritise patients by a prior estimate of CVD risk.
Random assessment

Since patients present to the practice more or less in random order, it means that they would be assessed in random order. This was simulated by allocating each individual in the model population a random number and assessing patients in descending order of this number.

Prioritise patients aged over 50 followed by patients who are over 40

Under this strategy individuals over 50 would be assessed first, followed by individuals over 40, followed by all others. There would be no basis on which to prioritise one individual over 50 over another, therefore individuals in the same age group would be seen in random order.

Prioritise patients by their age

Under this strategy older individuals would be assessed first and younger individuals afterwards. Among individuals of the same age there would be no basis on which to prioritise one individual over another, therefore individuals of the same age would be seen in random order.

Prioritise patients by a prior estimate of CVD risk

Under this strategy, a prior estimate of ten-year CVD risk is calculated for every individual. This prior estimate is calculated from all available risk factor information held on the practice electronic medical records. Where risk factor information is missing (for example, no blood pressure or no cholesterol level) a default value is used instead.

Default risk factor values

Default risk factor values were derived from the Health Survey for England of 1998. The population aged 16 and over was divided into 256 subgroups based on their age (in eight ten-year age bands), gender, whether they were taking antihypertensive treatment, whether they had a diagnosis of cardiovascular disease, diabetic status and smoking status. For categorical risk factors – smoking status, diabetic status, and cardiovascular disease status – the default is that the risk factor is absent. This is because in all subgroups it is more common that the risk factor status is absent than present.

For continuous risk factors (blood pressure and cholesterol levels) average blood pressures (systolic and diastolic) and cholesterol levels (total and HDL) were calculated for each of the 256 subgroups. Subgroups containing fewer than ten individuals were combined. In the first place, smokers and non-smokers were combined and common mean blood pressures and cholesterol ratios calculated for both. If this did not produce ten measurements, diabetics and non-diabetics were combined; then those with and without CVD. In this way a list of default blood pressures was calculated for every possible age, sex and risk factor category. These values are available on-line (Prevention of Cardiovascular Disease, 2005).

Calculating CVD risk and determining treatment eligibility

Risk factor data on all individuals in the model population (individuals aged 30 to 74, without CVD or diabetes) were exported into an Excel spreadsheet. A series of ten-year CVD risks was calculated for each individual in the model population.

CVD risks
True CVD risk

The first CVD risk calculated was based on all risk factor information as recorded in the Health Survey for England. These risk factors were taken to be true measured blood pressure and for each individual, the CVD risk calculated from these risk factor values was taken to be their true CVD risk. This risk is used to determine the total burden of CVD events in the population. It is also used to determine whether patients have been correctly or incorrectly classified as treatment eligible in the sensitivity analysis.

Clinically determined CVD risk

In the Survey only a single blood pressure or cholesterol reading is available. However, in clinical practice, CVD risk is estimated from one total cholesterol and HDL cholesterol measurement and the average of blood pressure measurements taken at two clinic visits. Because of biological variation in these measurements are not identical to the true average cholesterol or blood pressure. In order to simulate these readings, a “clinically determined” cardiovascular risk was calculated for each individual, based on the mean of two estimated blood pressures and estimated total cholesterol and HDL cholesterol levels.

Clinically measured blood pressure incorporates the effects of biological variation on blood pressure and cholesterol measurement. Two estimated blood pressures were generated for each individual in the population using a previously described methodology. (Marshall, T., 2004) This method adjusts true blood pressure (the survey blood pressure) by an error term. [Measured BP = True BP · (1 + Error term)]. A series of normally distributed error terms were generated in Excel as random numbers with a mean of zero and a standard deviation equal to the coefficient of variation of between-visit, measured blood pressure. This between-visit coefficient of variation was derived from meta-analysis (Wright, J. M. and Musini, V. J., 2000). The result of this is that in addition to their true blood pressure (survey blood pressure) each individual had two further estimated blood pressures that represent blood pressures obtained at two separate clinic visits.

Two estimated cholesterol levels were also generated for each individual in an analogous way. These were derived from the measured cholesterol levels with an incorporated error term based on the coefficient of variation derived from published studies: 7.2% for total cholesterol and 7.5% for HDL cholesterol (Nazir, D. J., Roberts, R. S., Hill, S. A. et al, 1999). From the estimated blood pressures and cholesterol levels a ten-year CVD risk was calculated. This is the individual’s clinically determined ten-year CVD risk. This CVD risk is used to determine whether the individual is clinically eligible for treatment.

Prior estimates of CVD risk

The more risk factor data are available in electronic medical records, the more accurate is the prior estimate of CVD risk. Practices could undertake a prior estimate of ten-year CVD risk in any of a variety of different ways. For this analysis one scenario was modelled, the situation where the electronic medical records include a basic minimum of details on patients’ risk factors such as age, gender, cardiovascular disease status, diabetic status, antihypertensive drug treatment status plus smoking status and a measured blood pressure.

Determining treatment eligibility

Treatment eligibility criteria are derived from recent UK guidelines: these recommend antihypertensive treatment for those whose blood pressures exceed 160/100 mm Hg, or with blood pressures exceeding 140/90 mm Hg and ten-year CVD risk over 20% (Williams, B., Poulter, N. R., Brown, M. J. et al, 2004). They also recommend aspirin for those at over 20% ten-year CVD risk who are aged over 50 and statins for those at over 20% ten-year CVD risk, or with familial hyperlypidemia (defined here as total cholesterol ≥9.0 mmol/l) (Wood, D., Wray, R., Poulter, N. et al, 2005).

True treatment eligibility status is determined from complete risk factor data, using the true blood pressure and cholesterol levels recorded in the Health Survey for England. Clinically determined treatment eligibility status is determined using the mean of two estimated blood pressures and estimated cholesterol levels.

Treatment eligibility criteria are written as logical functions in Excel which determines whether a patients is eligible for antihypertensive or statin treatment (1 = eligible, 0 = not eligible). Treatment eligibility for aspirin is not included because since they have a ten-year CVD risk greater than 20% all patients eligible for aspirin are also eligible for at least one other preventive treatment.

Logical function 1: Eligibility for antihypertensive treatment

=IF(OR(“SystBP”>=160,”DiastBP”>=100),1, IF(AND(OR(“SystBP”>=140,”DiastBP”>=90), OR(“10 year CVD Risk”>0.2,”CVD History”=TRUE,” Diabetes”=TRUE)),1,0))

Logical function 2: Eligibility for statin treatment

=IF (OR (“TC”>=9, “10 year CVD Risk”>=0.2,”CVD History”=TRUE,” Diabetes”=TRUE), 1, 0)

Effectiveness

The effectiveness of each patient identification strategy was measured in two ways. The first outcome is the number of patients correctly identified as eligible for treatment. The second outcome is the total number of QALYs gained predicted to occur in patients identified as eligible for statin treatment over a lifetime taken from the markov model. The model also gives the total number of CVD events prevented that gives an indication of the total burden of CVD that could be influenced by prevention in identified patients. A method that can identify a greater number of treatment eligible patients earlier and a greater burden of more CVD in treatment eligible patients will generate more QALYs.

Costs

Costs are considered from the NHS perspective. Obtaining risk factor information on individual patients has a cost. Cost data was obtained from literature see table 13 in appendix C 1. Using 2006/7 costs, practice nurse time costs £32 per hour of patient contact and general practitioner time £118 per hour of patient contact. In the base case analysis, all patient assessment is carried out by a practice nurse. A full patient assessment will require at least one 20 minute visit to measure blood pressure and cholesterol and inquire about smoking history; and two 10 minute visit to check blood pressure again. Biochemistry/lipid profile costs are taken (£3.56 (GDG expert opinion): The total cost of each strategy is the total number identified by each strategy multiplied by the assessment cost plus the CVD cost attributable to patients eligible for the statin treatment including the drug costs.

Cost-effectiveness

Under any identification strategy, with increasing resources the number of treatment eligible patients identified and the total burden of CVD prevented increases. In order to compare strategies we can represent cost-effectiveness of each identification strategy graphically by decile: the total cost of the strategy on the horizontal axis and the total effectiveness of the strategy on the vertical axis.

False positives/low risk patients misclassified as needing treatment

Under any strategy, some individuals whose ten-year CVD risk is less than 20% may be misclassified as eligible for treatment. These individuals will derive some benefit from treatment, but it will be less than the benefit of treating higher risk patients. The main predictor of benefit is the patient’s CVD risk. It follows, that there is a CVD risk level at which the inconvenience of treatment to the patient outweighs the benefits of treatment. This level is likely to vary from one individual to another and it is difficult to predict a CVD risk level at which few patients would choose treatment. (Trewby, P. N., Reddy, A. V., Trewby, C. S. et al, 2002), (Steel, N., 2000), (McAlister, F. A., Laupacis, A., Teo, K. K. et al, 1997), (Marshall, T., Bryan, S., Gill, P. et al, 2006). However only a minority of clinicians would recommend treatment at less than 10% ten year CVD risk. (Greenfield, S., Bryan, S., Gill, P. et al, 2005) It is assumed that an individual at less than 10% ten year CVD risk who is not otherwise eligible for treatment – because their true blood pressure exceeds 160/100 mm Hg or their true total cholesterol exceeds 9.0 mmol/l – has been misclassified as needing treatment. The analysis compares the proportion of false positives that would result under each strategy costs and QALY losses.

Modelling the strategies

Details of all individuals in the model population are entered into an Excel spreadsheet. All individuals have three sets of recorded characteristics. The first of these are obtained from the Health Survey for England: age, gender, CVD risk factor status and antihypertensive drug treatment status. The second characteristics are derived from data obtained in the Health Survey for England: ten-year CVD risk; estimated blood pressures and cholesterol levels and estimated ten-year CVD risk; and treatment eligibility status derived from estimated risk factors and ten-year CVD risk. The third set of characteristics includes a random number and a series of prior estimates of ten-year CVD risk. These characteristics are used to determine the order in which individuals are invited for assessment.

Assessment costs are calculated for each individual. In order to simulate each strategy, individuals are ranked in an order indicated by one of the third set of characteristics. To simulate random assessment, individuals are ranked using the random number (i.e. in random order). To simulate assessment in order of a prior estimate of CVD risk patients are ranked in descending order of their prior estimate of ten-year CVD risk.

The total cost of assessing one patient under any given strategy is the total cost of assessing the first patient (plus any additional costs e.g. to extract data and undertake the ranking exercise). The total cost of assessing ten patients under any given strategy is the total cost of assessing the first ten patients (plus any additional costs) and so on.

Sensitivity analysis

A number of different scenarios are explored in a sensitivity analysis. These include changes to the costs of assessing patients: allowing for higher assessment costs in clinical practice than in the base case model; allowing for the fact that random assessment uses GP time rather than practice nurse time.

Results

If all 4264 patients were assessed, the model estimates that 652 individuals will be diagnosed as clinically eligible for treatment. Untreated, we would expect these individuals to suffer from 81 CVD events over the next ten years. We would expect the 652 individuals diagnosed as clinically eligible for treatment to include 14 (2% of the total) individuals at low risk of CVD (less than 10% ten-year CVD risk) who had been misclassified as eligible for treatment. The screening process will identify 1% of the population aged between 35–44 years as eligible while the majority 87% of the patients will be aged over 65. See Table 1.

Table 1. Results showing percentage of people eligible for treatment by CVD risk bands for all age groups.

Table 1

Results showing percentage of people eligible for treatment by CVD risk bands for all age groups.

We estimated the QALY gain from a markov model per person on statin treatment broken down by age and CVD risk band. The model estimated that within each CVD risk band, the younger patients gain more QALYs than the elderly patients, which is expected given that QALYs are calculated using life expectancy and quality of life estimates. Also for each age band the higher the CVD risk, the greater the QALY benefit from statin treatment. We then calculated the QALY gain by each screening strategy and decile. Table 2 below shows the lifetime QALY gain by age and risk band and table 3 shows the QALY gain by decile and strategy for all age groups and risk bands. 95% of the QALY gains from prior CVD assessment are realized in decile 3, i.e. when 30% of the practice population is screened. In contrast random screening will have 28% while age has 76% when 30% of the total practice is screened.

Table 2. Results showing lifetime QALY gain from statin treatment by age and risk band from the Markov model.

Table 2

Results showing lifetime QALY gain from statin treatment by age and risk band from the Markov model.

Table 3. Results showing number of QALY gain in each decile by strategy.

Table 3

Results showing number of QALY gain in each decile by strategy.

Table 4. Results showing percentage of QALY gain by decile in each strategy.

Table 4

Results showing percentage of QALY gain by decile in each strategy.

When the second decile is considered i.e. screening 20% of the relevant population, 88% of the high risk patients will be identified by prior CVD method while the random method will identify a mere 19%, age alone 50% and age>50 then>40 identifies 38%. This means prior CVD will ensure that those at high risk are identified earlier and given the necessary treatment on time.

Table 5. Results showing QALY loss due to misclassification (people with less than 5% 10 year risk).

Table 5

Results showing QALY loss due to misclassification (people with less than 5% 10 year risk).

Total costs

Table 6 includes lifetime costs on statin treatment from the markov model. The costs include costs of drugs and costs of CVD events. Table 7 includes all the costs by strategy which includes the treatment costs from the markov model, plus assessment costs and invitation costs for non random strategies.

Table 6. Results showing lifetime treatment cost by age and risk band from the Markov model.

Table 6

Results showing lifetime treatment cost by age and risk band from the Markov model.

Table 7. Results showing total cost by decile and by strategy.

Table 7

Results showing total cost by decile and by strategy.

Table 8. Results showing total costs of misclassification (people with less than 5% 10 year risk).

Table 8

Results showing total costs of misclassification (people with less than 5% 10 year risk).

Cost-effectiveness

The cost-effectiveness results showed that using prior CVD information is the most cost-effective method of identifying those at risk of developing heart disease. The use of prior CVD information will result in an estimated ICER of about £2,109/QALY when compared with no screening. Age alone is dominated by prior CVD method. This means prior CVD will result in more quality adjusted life years and less cost compared with Age if the first 10% of the population is considered. When all the relevant 12 strategies are compared, the analysis suggests that it’s cost-effective to screen 20% of the relevant population. The ICER is about £7,604/QALY when prior CVD is compared with the next best non-dominated option (10% prior CVD). The ICER for 30% prior CVD compared with the next best non-dominated option (20% prior CVD) is about £37,644 per QALY..

Table 9. Cost-effectiveness results of identification strategies in the prevention of CVD for decile 1–3.

Table 9

Cost-effectiveness results of identification strategies in the prevention of CVD for decile 1–3.

Figure 1. Cost-effectiveness plane for the different screening strategies decile 1 to 3.

Figure 1

Cost-effectiveness plane for the different screening strategies decile 1 to 3.

Sensitivity analysis

The base case analysis assumes that patient assessment takes 20 minutes of practice nurse time. Assuming that the consultation will take upto 45 minutes increases the nurse cost from £11 to £24. The ICER for the relevant 20% of the population doubles to about £15,900 per QALY. This suggests that model results are not sensitive to duration of consultation, but there is a tendency as expected that the longer the durartion of consultation the less favourable the ICER. The results show that prior CVD assessment is still the optimal option.

The base model assumed that the assessment was done by a practice nurse who cost about £32/hour. We assumed that the assessment was done by the GP whose hourly cost is £118/hour. This was assumed to have no impact on outcomes. The consultation was assumed to remain the same (20 minutes), this results in all assessment costs being higher from £52 per person to £138 per person. The most cost effective strategy will no longer be screening 20% of the relevant population since the ICER is well above the £20,000/QALY threshold (£25,500/QALY), but rather screening 10% of the relevant population. The ICER will be £11,381 per QALY when prior CVD is compared with no screening. Thus the model is sensitive to assumptions about the helth care personnel responsible for screening. The model suggests that using practice nurses will result value for money than using doctors..

The base model assumed the cost of simvastatin 40mg (£18.12). When we assumed a combination of drugs were used as identified in the primary prevention trials. 50% used atorvastatin 10mg, 25% simvastatin and Pravastatin 40mg (£146.63), the conclusions remained the same. The ICERs however were borderline cost effective for screening 20% of the relevant population £21,500/QALY.

When treatment effect was varied, using the upper and lower confidence interval of the treatment effect, the model conclusions did not change, screening 20% of the relevant population using prior CVD estimates remained cost-effective all the time.

Discussion

Our model estimates that 652 individuals will be diagnosed as clinically eligible for treatment if all 4262 patients were assessed. Untreated, we would expect these individuals to suffer from 81 CVD events over the next ten years. We would expect the 652 individuals diagnosed as clinically eligible for treatment to include 14 (2% of the total) individuals at low risk of CVD (less than 10% ten-year CVD risk) who had been misclassified as eligible for treatment. The screening process will identify 1% of the population aged between 35–44 years as eligible while the majority 87% of the patients will be aged over 65. If we consider the first three deciles, 95% of eligible patients will be on treatment using prior CVD assessment, while only 82% and 28% will be on treatment using age and random assessment.

The cost-effectiveness results showed that using prior CVD information is the most cost-effective method of identifying those at risk of developing heart disease. Screening 20% of the relevant population using prior CVD information is the optimal strategy. The estimated ICER of about £7,604/QALY when prior CVD is compared with the next best non-dominated option (10% prior CVD).

Our model had some limitations. The model assumed that patients present to the practice more or less in random order especially for random assessment. This is likely to be biased, underestimating the effectiveness and cost-effectiveness of random assessment, as it does not reflect the increased likelihood or frequency of consultation in patients with a higher CVD risk. In practice, patients with symptoms of undiagnosed CVD are more likely to consult. However, we feel the overall results are unlikely to change because of this especially given that prior CVD information has the ability to rank patients in order of risk, which in-turn will influence the effectiveness and cost-effectiveness.

We used the Framingham equation in the estimation of the CVD. The problem with the Framingham CVD algorithm is that it tends predominantly to identify older people (because age is a strong risk factor), at the expense of younger people whose absolute risk may be low but whose relative risk is high compared to age-matched peers. Another problem is that the Framingham equation does not include risk of non-CHD, non-cerebrovascular disease outcomes in its predictions unlike the JBS2 algorithm. Because of this, for a given risk profile Framingham tends to give slightly higher risk estimates.

Because age is such an important determinant of CVD risk, CVD risk assessment has to be an ongoing exercise rather than a ‘one-off’ exercise. On an individual level, patients who are below the at-risk threshold of >20% this year may not be in 2–3 years time, as age is such a strong risk factor and blood pressure, if untreated, may also increase.

We would expect the efficiency of patient identification in a real clinical setting to be somewhat lower than the efficiency modelled here, because not every patient attends for assessment and not every assessed patient is treated. If patients do not attend for assessment, they incur the cost of the invitations which were incorporated in the base case model, without any potential benefits. This reduces the overall cost-effectiveness of the non-random strategies. Similarly, if patients are identified as eligible for treatment, but choose not to start treatment, they incur the costs of invitation and assessment, without any benefits of treatment. However the relative efficiency of one strategy compared to another will be unchanged.

Conclusions

Primary prevention of CVD should make use of strategies to prioritise patients likely to be at highest risk and to invite patients in descending order of CVD risk estimated from available data in the GP database. UK general practices have enough data to use this systematic way. Random assessment of patients in order to identify those eligible for CVD prevention is probably the least efficient strategy. Nurses rather than GPs should undertake random assessment of CVD risk in unselected patients.

A model to estimate the cost-effectiveness of higher versus lower intensity statins in the treatment of coronary heart disease

Introduction

Recent trials have demonstrated that high intensity statins provide cardiovascular benefits beyond the conventional lower intensity statins in people with coronary heart disease (CAD) (de Lemos, J. A., Blazing, M. A., Wiviott, S. D. et al, 2004), (LaRosa, J. C., Grundy, S. M., Waters, D. D. et al, 2005), (Pedersen, T. R., Faergeman, O., Kastelein, J. J. et al, 2005), (Cannon, C. P., Braunwald, E., McCabe, C. H. et al, 2004). However, only one study has tried to quantify the incremental cost and benefits of high intensity compared with low intensity statins.(Chan, P. S., Nallamothu, B. K., Gurm, H. S. et al, 2007) This study, set in the USA concluded that high intensity statins were cost-effective in patients with acute coronary syndrome. In patients with stable CAD the cost-effectiveness was highly sensitive to treatment efficacy and drug costs. The GDG asked for a model to estimate the cost-effectiveness of high intensity compared with low intensity statins in patients with coronary disease taking the perspective of the UK National Health Service.

Modelling Assumptions
Population and sub-groups

There is no trial evidence for the effectiveness of higher intensity statins versus lower intensity statins in primary prevention. This model considers the use of higher versus lower intensity statins in secondary prevention of cardiovascular disease. The cost-effectiveness of higher versus lower intensity statins is likely to vary between subgroups of patients. Thus, in this model, separate analyses have been run for patients with stable coronary artery disease (CAD) and those with acute coronary syndrome (ACS) respectively.

Treatment comparators

The analyses that follow compare treatment with higher intensity statins compared to treatment with lower intensity statins. Lower intensity refers to less aggressive lipids lowering doses such as (Simvastatin 20mg, Pravastatin 40mg, Atorvastatin 10mg). The higher intensity statins used in the trials were Atorvastatin 80mg and Simvastatin 80mg). Both of the trials that considered patients with CAD used atorvastatin 80mg as the higher intensity statin, whilst of the two trials that considered ACS patients, one used aorvastatin 80mg and the other used simvastatin 80mg.

Model structure

A Markov model was developed to evaluate the incremental costs and effects of lifetime treatment with higher intensity statins compared with lower intensity statins for the secondary prevention of CHD from a UK NHS perspective.

In a Markov model there are a finite number of health states. It is assumed that at any point in time, all patients must be in one and only one of the states. The model then replicates how a hypothetical cohort of people moves between the states. Figure 2 in appendix C 1 shows a schematic representation of the patients’ pathways. All patients start in the well state post a primary event. During each annual cycle of the model, patients can enter any one of the assumed event health states (MI, unstable angina, Stroke, revascularisations, transient ischemic attack (TIA), heart failure (HF), peripheral artery disease (PAD) or death) while the remainder stay in the event free state. The rate at which people move through the model is driven by transition probabilities, which describe the likelihood of moving between states from one 12 month cycle of the model to the next. These transition probabilities are age adjusted.

Patients can experience more than one non-fatal event in subsequent periods of the model. However, for members of the hypothetical cohort who experience more than one event, the model only applies the additional costs and adverse health consequences of the most recent event. The model does not have any ‘memory’ of previous states. The model ensures that patients cannot transition from a more severe health state to a less severe one.

The transition probabilities for patients on higher intensity statins are adjusted to reflect the expected risk reduction in CVD events compared to the risk for patients on lower intensity statins by applying the relative risks observed in the trial data. Health care costs and QALYs were then estimated for each health state using unit costs and health utility values. Assumed cost and utility values used in the model and their sources are shown in appendix C 1. Cost and QALY outcomes are calculated by the model until a patient dies or until they reach the age of 100, which ever of these two events occurs first.

Baseline risks

The baseline incidence of CVD events assumed in this model are taken from the clinical trials for the two populations of interest. A beta distribution was fitted for the baseline incidence of CVD events. For patients with ACS, risk data was taken from the A to Z (de Lemos, J. A., Blazing, M. A., Wiviott, S. D. et al, 2004), and the PROVE-IT trials.(Cannon, C. P., Braunwald, E., McCabe, C. H. et al, 2004) (table 5 in appendix C 1). For patients with stable CAD, risk data was taken from the TNT(LaRosa, J. C., Grundy, S. M., Waters, D. D. et al, 2005) and the IDEAL(Pedersen, T. R., Faergeman, O., Kastelein, J. J. et al, 2005) trials (table 6 in appendix C 1). In the base case the model assumes that patients with cardiovascular disease have a two-fold risk of dying from non-CVD causes compared with the general population.(Packham, C., Gray, D., Silcocks, P. et al, 2000) All cause mortality rates were taken from the life tables for England and Wales prepared by the Government Actuaries Department(Government Actuaries Department, 2006) and the Office for National Statistics(Office for National Statistics, 2006). PAD data were taken from a study by Murabito J et al (Murabito, J. M., D’Agostino, R. B., Silbershatz, H. et al, 1997) Revascularisation data were taken from Johansen H et al (Johansen H, Nair C Taylor G., 1998) Heart failure data were taken from a study by Cowie MR (Cowie MR, Wood DA Coats AJS Thompson SG Poole-Wilson PA Suresh V and Sutton GC, 1999) in Hillingdon UK. Transition probabilities from other health states are shown in table 9 in appendix C 1.

Treatment effects

The relative risks of CVD events following treatment are estimated from the identified trials. Data were meta-analysed for the higher intensity versus the lower intensity statins using a random effects model. To evaluate the effects of high intensity statins, the predicted incidence of CVD events was reduced by applying the reported relative risks and their confidence interval modelled as a log-normal distribution. For patients with ACS, we also took treatment effect from the individual trials in sensitivity analysis to assess the impact of atorvastatin and simvastatin separately. (Appendix C1 table 10 and 11 and figure 3 and figure 4).

Cost Assumptions

The NICE reference case specifies that costs should be measured from an NHS and personal social services perspective. These should include the direct cost of drug treatment and also potential savings from avoided treatments due to reduced incidence of cardiovascular disease. Costs were calculated using annual cost for each of the health states of the model. As per current NICE guidance, an annual discount rate of 3.5% was used for both costs and health benefits.(National Institute for Health & Clinical Excellence., 2006)

Costs of health states were taken from the Statins TA 94 (National Institute for Health and Clinical Excellence, 2006). In the base case model for CAD it was assumed that patients would be on a single 80mg dose of atorvastatin (£28.21 per month) while for ACS a weighted price of simvastatin and atorvastatin was assumed (weighted by the numbers of patients in each trial). . Drug prices used are as of the PPA Drug Tariff February 27th 2008(NHS Prescription Pricing Authority, 2008). GP costs were taken from PSSRU (PSSRU, 2005), and biochemical test costs were taken from NHS reference costs 2006/07.(Department of Health Reference Costs, 2007) All costs used are tablualted in appendix C1 (tables 12 and 13). We fitted a gamma distribution to costs reflecting published ranges.

The model assumes that people on high intensity statins will incur one additional visit per year over and above the visits for patients on low intensity statins for check ups (monitoring). The model assumes no adverse events for patients on higher intensity statins and therefore no adverse event costs.

Health and Utility Outcomes

Treatment effect of statins is measured in the model in terms of prevention of CVD events: non-fatal unstable angina, MI, HF, Stroke, revascularisations, TIA, peripheral artery disease, and deaths from CVD and all other causes. Quality adjusted life years are calculated in the model as each of the CVD health states is allocated a health utility value. Health utility data was obtained from published literature, primarily from the Statins TA 94.(National Institute for Health and Clinical Excellence, 2006) Health state utilities were adjusted to reflect the fact that health related quality of life in the general population decreases with age (i.e. multiply the disease utility weight by age utility weight). Age utility weights were taken from the Department of Health, Health Survey for England, 1996(Department of Health, 1998). We fitted a beta distribution on quality of life estimates and the standard errors were obtained from published studies. (Table 14 and 15 in appendix C1).

Cost-effectiveness

The results of cost-effectiveness analysis are usually presented as Incremental Cost-Effectiveness Ratios (ICERs), which measure the additional cost of using high intensity statins per additional QALY gained compared with low intensity statins. The incremental cost per QALY is given by:

(cost of high intensity statins-costof low intensity statins)(QALYof high intensity statins-QALY of lowintensity statins)
Sensitivity Analysis

A range of sensitivity analyses has been undertaken on the basecase results. Firstly univariate sensitivity analyses have been undertaken to assess the impact of a range of input parameter, including costs, utilities and relative risks of treatment. Probabilistic sensitivity analyses have been undertaken to determine the impact of the imprecision of input values on decision uncertainty. We present cost-effectiveness acceptability curves (CEAC) based on probabilistic analyses of model parameters which are sampled randomly. The CEAC shows the probability that each strategy is cost effective at different levels of society’s willingness to pay for an additional QALY.

Results

The base case results are presented below, and cost-effectiveness is assessed against a threshold of £20,000/QALY. We have separately modelled high intensity stains for both ACS and the stable CAD pupulations.

1. Results for patients with acute coronary syndrome (ACS)

Table 10 indicates the modelled number of CVD events for a hypothetical population of 1,000 ACS patients treated with either high intensity or low intensity statins. The table indicates that fewer cardiovascular events occur in the population treated high intensity statins. This translates to a gain of 0.32 discounted QALYs per patient when compared with low intensity statins. The additional cost of achieving this gain in QALYs is £1,418.

Table 10. Lifetime modelled events for a cohort of 1,000 ACS patients treated with either low or high intensity statins.

Table 10

Lifetime modelled events for a cohort of 1,000 ACS patients treated with either low or high intensity statins.

Cost-effectiveness results

The model estimates the lifeltime incremental cost per QALY of using high intensity statins (both simvastatin and atorvastatin 80mg) compared with low intensity statins (both simvastatin and pravastatin) is about £4,700. indicating that high intensity statins are cost-effective in ACS patients. The probability that high intensity statins are cost effective when compared to lower intensity statins is 94%. The probabilistic ICER is almost similar to the deterministic ICER as shown in table 11 below

Table 11. Base case deterministic results for the cost-effectiveness of a higher intensity statins compared with a lower intensity statins in a 65 year old patient with acute coronary syndrome.

Table 11

Base case deterministic results for the cost-effectiveness of a higher intensity statins compared with a lower intensity statins in a 65 year old patient with acute coronary syndrome.

2. Results for patients with stable coronary artery disease (CAD)

Table 12 indicates the modelled number of lifetime events for a hypothetical 1,000 stable CAD patients treated with either high or low intensity statins. The table indicates that fewer cardiovascular events occur in the population treated high intensity statins. This translates to a gain of 0.08 discounted QALYs per patient when compared with low intensity statins. The additional cost of achieving this gain in QALY is £2,389.

Table 12. Lifetime modelled events for a cohort of 1,000 CAD patients treated with either low or high intensity statins.

Table 12

Lifetime modelled events for a cohort of 1,000 CAD patients treated with either low or high intensity statins.

Cost-effectiveness results

The model estimates the lifeltime incremental cost per QALY of using high intensity statins (atorvastatin 80mg) compared with low intensity statins (simvastatin 40mg) is about £27,800 indicating that high intensity statins are not cost-effective in patients with stable CAD. The probability that high intensity statins are cost effective when compared to lower intensity statins is 42%. The probabilistic ICER is almost similar to the deterministic ICER as shown in table 13 below

Table 13. Base case deterministic results for the cost-effectiveness of a higher intensity statin compared with a lower intensity statin in a 65 year old patient with stable coronary artery disease.

Table 13

Base case deterministic results for the cost-effectiveness of a higher intensity statin compared with a lower intensity statin in a 65 year old patient with stable coronary artery disease.

Figure 2. Increamental cost-effectiveness plane for a 65 year old patient with CAD.

Figure 2

Increamental cost-effectiveness plane for a 65 year old patient with CAD. Scatter Plot showing the mean differences in costs and QALYs (based on high intensity –low intensitity statins) illustrates the uncertainity surrounding the estimates of (more...)

Figure 3. Cost effectiveness acceptability curve for higher intensity statin compared with a lower intensity statin in a 65 year old patient with stable coronary artery disease.

Figure 3

Cost effectiveness acceptability curve for higher intensity statin compared with a lower intensity statin in a 65 year old patient with stable coronary artery disease. The CEAC shows the probability that high intensity statins are cost-effective when (more...)

Figure 4. Increamental cost-effectiveness Plane for high intensity statins compared to low intensity statins for a 65 year old patient with ACS.

Figure 4

Increamental cost-effectiveness Plane for high intensity statins compared to low intensity statins for a 65 year old patient with ACS.

Figure 5. Cost effectiveness acceptability curve for higher intensity statin compared with a lower intensity statin in a 65 year old patient with stable coronary artery disease.

Figure 5

Cost effectiveness acceptability curve for higher intensity statin compared with a lower intensity statin in a 65 year old patient with stable coronary artery disease. Our model shows that using a threshold of £20,000/QALY, high intensity is 94% (more...)

Deterministic sensitivity analysis

A range of univariate sensitivity analyses was undertaken to assess the impact of different input parameters on the base case ICERs. The results are interpreted using a cost-effectiveness threshold of £20 000 per QALY.

Using data from A-Z trial and PROVE-IT

ACS trials used both simvastatin 80mg and atorvastatin 80mg as high doses. We analysed simvastatin and atorvastatin separately. When data from A-Z was used alone, thus comparing simvastatin 80mg with simvastatin 20mg. Simvastatin 80mg resulted in more QALYs 0.41. The estimated ICER was £2,108/QALY, indicating that simvasatin 80mg is very cost-effective.

Table 14. Base case results for the cost-effectiveness of a higher intensity statins compared with a lower intensity statins in a 65 year old patient with acute coronary syndrome using data from A-Z trial.

Table 14

Base case results for the cost-effectiveness of a higher intensity statins compared with a lower intensity statins in a 65 year old patient with acute coronary syndrome using data from A-Z trial.

When treatment effect from PROVE-IT was used, thus comparing atorvastain 80mg with pravastatin 40mg, high intensity resulted in 0.20 additional QALYs. The estimated ICER was £10,714/QALY. This analysis indicates that atorvastatin is cost-effective when compared with low intensity statins in ACS patients. Thus all higher intensity statins are cost-effective in this patient group. However simvastatin 80mg is cheaper and more cost-effective than atorvastatin.

Table 15. Base case results for the cost-effectiveness of a higher intensity statins compared with a lower intensity statins in a 65 year old patient with acute coronary syndrome using data from PROVE-IT trial.

Table 15

Base case results for the cost-effectiveness of a higher intensity statins compared with a lower intensity statins in a 65 year old patient with acute coronary syndrome using data from PROVE-IT trial.

Efficacy of treatment (using lower and upper confidence intervals)

The efficacy of higher intensity statins compared to lower intensity statins was assessed using the upper and lower 95% confidence RR intervals from the meta-analysis. The ICERs were not senstive for all outcomes except for the more extreme ranges of CVD mortality risks. Taking the upper confidence interval of cardiovascular mortality, the higher intensity statins result in more CVD deaths than lower intensity statins. In this case higher intensity statins will become borderline cost-effective for the ACS sub-group with an estimated ICER of about £20,590/QALY. For the stable CAD sub-group high intensity statins becomes dominated by lower intensity statins. When lower limits of the 95% CI high intensity statins become cost-effective for patients with stable CAD suggesting that the model results are sensitive to this assumption for patients with stable CAD.

Table 16. Sensitivity analysis, efficacy of treatment (using lower and upper confidence intervals).

Table 16

Sensitivity analysis, efficacy of treatment (using lower and upper confidence intervals).

Assuming the cost of simvastatin 80mg

The trials for ACS used atorvastatin 80mg as the high dose. In our base model we used the price of Artovastatin 80mg. We used the price of simvastatin 80mg, assuming that simvastatin 80mg and atorvastatin are of the same efficacy. High intensity statins resulted in more QALYs (0.08) and was less costly (£38 cheaper) when compared to low intensity statins. Thus high intensity statins when we used generic simvastatin are highly cost-effective since they dominate low intensity statins.

Results by age

The model results suggest that higher intensity statins are cost-effective for all ages after acute coronary syndrome with ICERs ranging between about £5,400/QALY for those aged 45 years to about £3,800/QALY for those aged 75 years. However for patients with stable CAD, higher intensity statins are not cost-effective using the £20,000/QALY threshold, suggesting the model results are not sensitive to age.

Table 17. Results by Age for patients post acute coronary syndrome (ACS) and stable coronary disease (CAD) patients.

Table 17

Results by Age for patients post acute coronary syndrome (ACS) and stable coronary disease (CAD) patients.

The use of simvastatin 80mg costs

We also tested the assumption about cost of statins in aptients with stable CAD. The base model used the cost of atorvastatin 80mg. We used the cost simvastatin 80mg in sensitivity analysis. High intensity statins will dominate lower intensity stains in patients with stable CAD. This result suggests that assuming same effectiveness between simvastatin 80mg and atorvastatin 80mg, patients needing high intensity statin can benefit by being offered the cheaper high intensity statin. This model is thus sensitive to assumptions about costs of statins.

The base case ICERS are insensitive to the costs of cardiovascular events. For both lower and upper costs over the ranges reported and in some cases 50% less and 100% more (see table 12 in appendic C1) the ICERs ranged between about £26,800 and £29,000/QALY for a 65 year old patient with stable CAD. Only small changes were seen for the ICERs of patients with ACS.

We also tested the assumptions about the health state utilities. The health state utilities used in the model were obtained from the literature. The literature provided 95% confidence interval for some health state utilites and for some they did not. Where the 95% confidence interval was not provided or ranges we tested the assumption that the mean health state utilities were 0.2 less or more than the ones we got from the literature (GDG expert opinion). (see table 14 in appendic C1) The modelled ICERs remained insensitive to these assumed changes in utility values. For patients with ACS the modelled ICERs were insensitive to changes in health state utility values.

The base case model assumed there was one extra annual visit to the GP attributable to being on higher intensity statins, over and above the normal visits on people taking lower intensity statins. For each of these visits a bloods test is assumed. We tested the assumption that there would be no extra visits as suggested in the BNF, and alternatively that there will be two extra visits to the GP with blood tests as suggested by the GDG. The estimated ICER increased slightly when we assumed two visits (£30,832/QALY) and improved slightly when we assumed no extra visits (£25,890/QALY) but still retained the same conclusions suggesting the model is not sensitive to this assumption.

In the base model we assumed that people with cardiovascular disease have a two-fold increase in risk of dying from other causes compared with the general population. This was a conservative assumption based on information presented in Peckham et al(Packham, C., Gray, D., Silcocks, P. et al, 2000). Robinson et al(Robinson, M., Palmer, S., Sculpher, M. et al, 2005), however, demonstrated that this non CVD mortality could be more than four fold. In sensitivity analysis we assumed that there was no difference and that there was a four fold risk of dying from other causes for people with coronary heart diseases. The model results are not sensitive to changes in this assumption as the base case conclusions are retained in all cases in both sub-groups of patients.

Discussion & Conclusions

A model developed for this guideline demonstrates that high intensity statins compared with low intensity statins in patients with ACS are cost-effective. The estimated ICER is £4,700/QALY with a 94% probability that high intensity statins are cost-effective at £20,000/QALY threshold.. For patients with stable CAD, the estimated ICER is £27,800/QALY suggesting that high intensity statins are not cost-effective for patients with stable CAD with a 42% probability that high intensity statins are cost-effective at £20,000/QALY threshold. The model results for ACS are stable in sensitivity analysis except for an extreme assumption about treatment effect on cardiovascular mortality.

Our model results concur with one published economic analysis done in the USA by Chan 2007.(Chan, P. S., Nallamothu, B. K., Gurm, H. S. et al, 2007) Chan concluded that high intensity statins are highly cost-effective in patients with ACS and that the cost-effectiveness in patients with stable CAD is highly depended on statin efficacy and costs.

One of the main limitations in our analysis is the lack of credible long-term safety and utility data for higher intensity statins in trials. The trials reported that there was no significant difference between higher and lower intensity with regards to major adverse events. This has been challenged by Ravnskov et al (Uffe Ravnskov, Paul J Rosch Morley C Sutter and Mark C Houston, 2006) who argue, “Clinical experience has taught us that a dose increase of any drug will inevitably increase both the number and the seriousness of side effects”. There is also lack of utility data to estimate the loss of quality of life as a result of high intensity statins. The base case model assumed no difference in quality of life due to side effects for patients on higher intensity statins. Further research in this area is likely to be worthwhile.

The lack of data on treatment effects for patients the under 55s and the over 75s makes it harder to be so clear about the relative cost-effectiveness of statins in these age-groups. In sensitivity analysis, we showed that age had only a relatively affect on the modelled ICERs.

All models are simplifications of the real world. A simplification in our model model arises because of the nature of Markov models. These assume that the probability of an individual moving to any given health state in one time period depends only their current health state (there is no longer ‘memory’ in the model). Thus the probability of heart failure for a patient whose last event was an MI is assumed to be the same irrespective of how many other cardiovascular events they have previously had.

Similarly, a patient’s health outcome and health care costs incurred are assumed to depend only on their current health state. These assumptions are unlikely to be strictly true, and will tend to underestimate overall costs and overestimate health outcomes for the cohort. Thus, interventions that prevent more CVD events will tend to appear rather less cost-effective than they may be in reality. So the model is conservative in this respect.

In conclusion, compared with low intensity statins, high intensity statins in patients with ACS are cost-effective There is some sensitivity for the basecase ICER to assumptions about treatment effect on cardiovascular mortality. In patients with stable CAD, Atorvastatin 80mg is not cost-effective using a £20,000/QALY threshold. However, generic simvastatin 80mg is cost-effective using our model. These results are relatively robust in sensitivity analyses.

3.2. Modelling the cost-effectiveness of a strategy titrating to a higher intensity statin using titration threshold targets in patients with cardiovascular disease

Introduction

In light of lack of published evidence on the cost-effectiveness of treatment using statins to pre-specified cholesterol level targets, a model was built in to estimate the cost per QALY of titrating using pre-specified targets to a maximum dose, compared with a fixed dose treatment strategy using simvastatin 40mg. The population to be modelled is a hypothetical cohort of 1,000 adults with hyperlipidemia who have had a myocardial infarction but who are free from diabetes. The model takes a UK NHS costing and health care perspective.

Model Assumptions

Treatment Strategies

In addition to the base line comparator fixed dose treatment strategy (simvastatin 40mg) we modelled four titration strategies using targets of 5 or 4 mmol/l total cholesterol, and using both 1 and 2-step titration strategies. In the one-step treatment strategy, the model assumes that patients not reaching target on simvastatin 40mg are then treated with the higher intensity simvastatin 80mg with no further measurement against target, and no further dose increase to follow. In the two-step model, patients not reaching target on simvastatin 80mg are assumed to be treated with atorvastatin 80mg with no further measurement against target, and no further dose increase to follow. Each increase in dose is assumed to be preceded by a GP consultation and blood test.

Patient Population

The population modelled is a hypothetical cohort of patients with a previous myocardial infarction. The population is defined with an initial distribution of total cholesterol levels corresponding to results from The Health Improvement Network (THIN) database (table 18) below. The average age of these patients is 61 years and the average initial total cholesterol level is 1.1mmol/l.

Table 18. Distribution of initial total cholesterol for secondary prevention population.

Table 18

Distribution of initial total cholesterol for secondary prevention population.

The results of the STELLAR trial (Jones, P. H., Hunninghake, D. B., Ferdinand, K. C. et al, 2004) were then used to estimate reductions in total cholesterol from statin treatment, and thereby to estimate the proportion of patients reaching target on each of the included drugs. These results are summarised in tables 19 and 20.

Table 19. The estimated reduction in total cholesterol obtained by simvastatin and atorvastatin from the STELLAR trial.

Table 19

The estimated reduction in total cholesterol obtained by simvastatin and atorvastatin from the STELLAR trial.

Table 20. Cumulative proportion of modelled cohort estimated to reach target on each of the modelled drugs.

Table 20

Cumulative proportion of modelled cohort estimated to reach target on each of the modelled drugs.

Baseline annual transitional probabilities

Having determined the proportion of patients on each of the included drugs (as determined by the assumed target total cholesterol), a markov model was built to estimate the impact of statin treatment on cardiovascular disease (CVD) events (defined as MI, stroke, PAD, TIA, heart failure, revascularisation, angina, cardiovascular death, and death from other causes). The markov model is a lifetime model which uses transitional probabilities (annual cycles) to estimate the number of CVD events from the initiation of statin treatment until death, or until the patient reaches an age of 100, whichever is the earlier of these two events. Using health state utility values assigned to each of the above health states, the model then calculates quality adjusted life years for each of the modelled treatment strategies.

Baseline annual transitional probabilities of CVD events following a previous MI are estimated from data reported in the TNT (LaRosa, J. C., Grundy, S. M., Waters, D. D. et al, 2005) and IDEAL (Pedersen, T. R., Faergeman, O., Kastelein, J. J. et al, 2005) clinical trials. These transitional probabilities have then be apportioned across patient age bands using data reported in the literature. (Kaplan RC, Heckbert SR Furberg CD Psaty BM., 2002),(Michiel L. Bots, MD PhD, Eline C. van der Wilk, MSc, Peter J. Koudstaal, MD PhD et al, 1997), (Wouter T. Meijer, Arno W. Hoes Dominique Rutgers Michiel L. Bots Albert and Hofman and Diederick E. Grobbee, 1998) (Office for National Statistics, 2000), 7784}, (Department of Health (20042004), These transitional probabilities are presented in Table 8 of Appendix C1.

The probability of CVD events following non-MI CVD events have been taken from the literature as given in Table 9 of Appendix C1. CVD events are apportioned across age bands using the same assumptions used to estimate the post MI transitional probabilities. Transitional probabilities following stroke, TIA, and PAD have been apportioned across age bands using data presented by Hardie. et al (Hardie K, Hankey GJ Jamrozik K Broadhurst RJ Anderson C., 2004)

Non CVD mortality is modelled by using the age adjusted ‘all cause mortality’ rates (Government Actuarial Department.(Government Actuaries Department, 2006) ) and adjusting for CVD mortality. The model makes the conservative assumption that the non-CVD mortality rate in the stable CAD population is twice that of the general population (Packham, C., Gray, D., Silcocks, P. et al, 2000).

Treatment Effects

Having defined the baseline transitional probabilities, the statin treatment effects were estimated using equations derived from a meta-analysis by Law et al.(Law, M. R., Wald, N. J., Rudnicka, A. R. et al, 2003)

The equations were applied in a 2-stage procedure. Firstly, the cholesterol lowering effects using both simvastatin and atorvastatin were measured using the following equations:

Reduction in Total Cholesterol by drug and dosage is given by:

= -1.123 + 0.238TC + 0.384 × Logn (dose of simvastatin)

= -2.205+0.419TC+0.475LN (dose of Atorvastatin)

Then the relative risks of CVD/CVA events were estimated using the following equations respectively:

Average relative risk (RR) of CHD is given by:

RR of CHD per 1.2 mmol/l reduction in total cholesterol = -0.745 × Logn(Age) + 3.47, so

RR of CHD = (-0.745 × Logn(Age) + 3.47)(Reduction in TC / 1.2)

Where Age = average age of patient cohort in years

Average relative risk of cerebrovascular disease/PAD is given by:

RR of PAD/cerebrovascular disease per 1.2 mmol/l reduction in total cholesterol = 0.94, so

RR = 0.94(Reduction in TC / 1.2)

The resulting relative risk estimates from statin treatment effect are presented in table 21 by drug and dose. Only CHD, and not cerebrovascular disease/PAD risk are age dependant as specified by the Law and Wald equations.

Table 20. Proportion of patients modelled to be on each of the three included drugs under four treatment strategies.

Table 20

Proportion of patients modelled to be on each of the three included drugs under four treatment strategies.

Table 21. Treatment effect (relative risk of cardiovascular events) by age, baseline cholesterol level, and dose of statin.

Table 21

Treatment effect (relative risk of cardiovascular events) by age, baseline cholesterol level, and dose of statin.

Modelled Costs

Statin drug costs are taken from prices quoted on December 4th 2007 by the Prescription Pricing Authority (NHS Prescription Pricing Authority, 2008).

Costs of treatment for CVD events are taken from published literature (appendix C1 table 12.

Each up titration in the target treatment arm of the model is assumed to be preceded by a standard ( approx 10 minute) GP consultation and a blood test (Assumed total cost per up titration of £26). Unit costs of GP visits and blood test are taken from literature (Appendix C1: table 13). In line with current NICE guidance(National Institute for Health & Clinical Excellence., 2006), an annual discount rate of 3.5% has been applied to future costs in the markov model

Quality of life (Utility)

In order for the model to estimate QALYs, each of the modelled CVD health states has been assigned an assumed health related quality of life utility score using previously published values (Appendix C1, Table 14). Future QALY values are discounted at 3.5% per annum as recommended by NICE. (National Institute for Health & Clinical Excellence., 2006)

Cost-effectiveness Analysis

The results of cost-effectiveness analysis are summarised using an incremental cost-effectiveness ratio (ICER). Specifically, the incremental cost per QALY.

Sensitivity Analysis

In addition to the deterministic base cases, which use the mean values of the included model variables, the sensitivity of the base case ICER results to a range of univariate deterministic sensitivity analyses have been tested.

Base Case Results

The base case results are presented for the hypothetical modelled cohort patient population with a mean age of 61 years and a mean total cholesterol level of 6.1mmol/l. Analyses are presented for both the 2-step and single step titration strategies and using both total cholesterol targets of 5mmol/l and 4mmol/l.

Table 22 indicates that with a target of 5mmol/l total cholesterol, the majority of patients (69%) are modelled to reach target on simvastatin 40mg. This is true of both the fixed and the titration population groups in the model. With a target of 4mmol/l, only 31% of patients will reach target on simvastatin 40mg. In the 2 step titration model an additional 15% of patients reach target on simvastatin 80mg, if the target is 5mmol/l, and an additional 6% reach target using a 4mmol/l target.

Table 22. Base case incremental cost-effectiveness results.

Table 22

Base case incremental cost-effectiveness results.

Cost-effectiveness

Table 23 presents an example of the models output values and indicates the number of events for the hypothetical 1,000 patient cohorts having assumed a 2-step titration and a target total cholesterol of 5mmol/l. The table indicates that fewer CVD events occur in the population treated using the titration strategy.

Table 23. Lifetime event outputs modelled for a cohort of 1,000 patients using a 2-stage titration treatment strategy with a target of 5mmol/l total cholesterol compared with a fixed low dose treatment strategy.

Table 23

Lifetime event outputs modelled for a cohort of 1,000 patients using a 2-stage titration treatment strategy with a target of 5mmol/l total cholesterol compared with a fixed low dose treatment strategy.

The discounted costs and QALYs for the fixed dose and both of the 1 and 2-step titration strategies using both targets of 5mmol/l and 4mmol/l are presented in table 24 in order of accending cost.

The fixed dose treatment strategy is the strategy which is least costly, but also generates the smallest number of QALYs. As expected, the 2-step titration strategies are more costly than the 1-step titration strategies and having a target of 4mmol/l is more expensive than a target of 5mmol/l.

The model indicates that the 1-step target 4 treatment strategy has extended dominance over the 1-step target 5 strategy and has an ICER of £14,089 compared to the fixed dose strategy. The 2-step titration to 5mmol strategy is also dominated by the 1-step 4mmol strategy (that is, it costs more and produces less QALYs) and so both 5mmol target strategies are excluded due to dominance. The incremental cost-effectiveness ratio of the 2-step target 4mmol/l compared to the 1-step target 4mmol/l strategy is £67,395 and is therefore not cost-effective using the usual UK thresholds.

In summary, only the 1-step target 4mmol/l treatment strategy is cost-effective using a threshold of £20,000 per QALY. Only 37% of the modelled population achieve target using this strategy, and the model indicates that it would not be cost-effective to push a higher proportion of patients to target using drugs such as atorvastatin 80mg at its current price. .

Univariate Sensitivity Analyses

Sensitivity analysis: costs of cardiovascular events

Increasing the costs of treatments for cardiovascular events will improve the cost-effectiveness of interventions for CVD diseases all else being equal. Using the upper range of the the assumed base case costs of CVD treatments (Appendix C1 table 12) only marginally lowers the incremental cost per QALY. Conversely, using the lower range of the assumed costs of these interventions in the 1-step target 4 model only increases the ICER to at most £15,100/QALY. Thus, the base case model results are insensitive to the CVD event cost assumptions.

Sensitivity analysis: Relative Risk of non-CVD mortality

The base model assumed that people for secondary prevention have a two-fold increase in the risk of dying from any cause compared to the general population. This assumption was tested by assuming that the risk of non CVD mortality is the same as that of the general population, and also that the difference is four fold. The above sensitivity analysis gives a range of £11,426/QALY to £18,564/QALY for the ICER for the 1-step target 4mmol/l treatment strategy. Using a threshold cost per QALY of £20,000, the cost-effectiveness conclusions of the base case analyses are not changed by this sensitivity analysis. For the ICER for the 1-step model using a target of 4mmol/l to reach £20,000, the risk of dying from non CVD causes needs to be over 4 times greater than that of the general population.

Sensitivity analysis: health state utilities

The health state utilities used in the model were obtained from previously published models. These values have been varied up and down by 10% in sensitivity analyses. Lowering the utility values of all health states other than the starting state, and holding all other variables constant, the ICERs will improve. For the 2-step model with a target of 4mmol/l total cholesterol, the ICER remains well above the upper £30,000/QALY threshold likely to be acceptable to NICE. Improving the utility values of CVD health states by 10% but holding the initial state utility and all other variables constant in the 1-step model with a target of 4mmol/l increases the ICER to £18,200/QALY. This is still under the £20,000 per QALY threshold. As such, although the modelled ICERs are relatively sensitive to changes in health state utility values, our sensitivity analyses indicates that the base case conclusion regarding cost-effectiveness are not affected by 10% changes in health state utility values.

Sensitivity Analyses: Starting Age

The sensitivity of the ICERs was also tested against changes in the assumed starting age of the patient cohort. As age of the starting cohort varies from 45 years to 75 years, then the ICER for the 1-step target 4 titration model varies from £13,273 to £17,417. The ICERs are thus relatively stable to changes in patient age with a trend to slightly higher ICERs for older patient groups. The conclusions of the base case analyses are however unchanged by this sensitivity analysis.

Sensitivity analysis, discounting cost and health benefits

NICE recommends that’s both future costs and future benefits are discounted at a rate of 3.5% per annum in order to allow for societal time preference. We tested the sensitivity of the base case ICERs to the discounting assumption using rates of 0% and 6%. Using these assumptions, the ICERs for the 1-step target 4 model vary from £11,186 to £16,701/QALY. So, the higher the discount rate, the higher the ICER, although the base case cost-effectiveness conclusions are not affected by this sensitivity analysis.

In Summary, the sensitivity anlyses have indicated that the base case ICERs are relatively stable to changes in input variable values. There is some evidence that it may not be cost-effective to treat to target in patients who are at increased risk of dying from non-CVD causes.

Discussion and Conclusion

Economic models are by definition a simplification of the real world. There is lack of long-term clinical trials comparing titration strategies with fixed lower intensity statin treatment strategies. As such, our targets model is predicated on the assumption that reductions in CVD events, resulting from reductions in total cholesterol levels from statin treatment are adequately represented by the Law and Wald equations. These equations are themselves predicated on the Framingham risk equations. The equations reflect the fact that higher intensity statins lead to greater reductions in cholesterol. Relative risk reductions are greater for patients with a higher starting cholesterol level and for younger patients. Our base case model assumes a hypothetical cohort of patients with an average starting total cholesterol of 6.1mmol/l and and average age of 61 years.

The guideline group acknowledged that the results of the Law meta-analysis overestimate reduction in cholesterol and CVD events in comparison to the longer-term trial results described by the Cholesterol Trialists Collaboration, and may yield over-optimistic estimates of treatment effects. The external validity of our model should be tested if and when long term outcome data becomes available from trials comparing a fixed dose treatment strategy with a target driven strategy.

There is also lack of good long-term safety and utility data for statin use. Although a number of safety studies and a meta-analysis on statin use were identified, the GDG felt the recruitment in these trials made it difficult to demonstrate any significant difference in side effects, since only those who could tolerate statins were included in the trials. As a result the trials reported that there was no significance difference between higher intensity and lower intensity statins with regards to major side effects, though there is a trend of greater ‘minor’ adverse events with increasing dose. There is also lack of health related quality of life utility data, with which to estimate quality of life reductions resulting from adverse events associated with higher intensity statin treatment. Consequently, and in line with previously published cost-effectiveness analyses in hyperlipidemia(National Institute for Health and Clinical Excellence, 2006), our model assumes no adverse events from treatment with higher intensity statins.

Another limitation of the model arises because of the nature of Markov models. These assume that the probability of an individual moving to any given health state in one time period depends only on their current health state (there is no ‘memory’ in the model). Thus the probability of HF for a patient whose last CVD event was an MI is assumed to be the same irrespective of how many CVD events they have previously had. Similarly, a patient’s health outcome and health care costs incurred are assumed to depend only on their current health state. These assumptions are unlikely to be strictly true, and will tend to underestimate overall costs and overestimate health outcomes for the cohort. Thus, interventions that prevent more CVD events will tend to appear rather less cost-effective than they may be in reality. So the model is conservative in this respect.

Our model indicates that both modelled treatment strategies using a target of 5mmol/l are dominated by the 1-step titration strategy using a target of 4mmol/l (that is, initiating patients on simvastatin 40mg and up-titrating to simvastatin 80mg those patients who do not reach target on 40mg of simvastatin. This treatment strategy is the only cost-effective treatment strategy modelled although only 37% of patients are likely to achieve the target. Our model indicates that it is not cost-effective to try to get more patients to target by adding atorvastatin 80mg because the ICER then increases to over £60,000 per QALY.

The ICERs, and therefore the conclusions of the base case analyses, were robust to sensitivity analyses. However, our base case assumed a 2-fold risk of non-CVD mortality in the modelled population. The relative risk of dying from non-CVD death in a cohort of patients with coronary heart disease compared to the general population has been estimated to be more than four fold. (Packham, C., Gray, D., Silcocks, P. et al, 2000) If this is the case, then our model indicates that the cost-effectiveness of using cholesterol targets may be more borderline than indicated by our base case.

3.3. Appendix C 1, Data tables and Figures

Table 1. Distribution of primary CVD events taken from the Statins TA94 (National Institute for Health and Clinical Excellence, 2006)

Table 2. CVD event rates per person per annum without treatment annual transition probabilities from the well state

Table 3. CVD event rates per person per annum without treatment annual transition probabilities from other health states

Table 4. Treatment effect of statins versus placebo taken from the Statins TA 94 (National Institute for Health and Clinical Excellence, 2006)

Table 5. Baseline absolute risks of CVD events for patients with ACS taken from PROVE IT (Cannon, C. P., Braunwald, E., McCabe, C. H. et al, 2004) & A to Z (de Lemos, J. A., Blazing, M. A., Wiviott, S. D. et al, 2004)

Table 6. Baseline absolute risks of CVD events for patients with stable CAD taken from IDEAL (Pedersen, T. R., Faergeman, O., Kastelein, J. J. et al, 2005) & TNT (LaRosa, J. C., Grundy, S. M., Waters, D. D. et al, 2005)

Table 7. Baseline annual transition probabilities acute coronary syndrome patients following MI

Table 8. Baseline annual transition probabilities, stable coronary artery disease patients following MI

Table 9. B40 aseline transitional probabilities from other health states following MI

Table 10. Relative risks of CVD events from high intensity compared to low intensity statin treatment from meta-analysis of PROVE IT (Cannon, C. P., Braunwald, E., McCabe, C. H. et al, 2004) & A to Z (de Lemos, J. A., Blazing, M. A., Wiviott, S. D. et al, 2004) for patients with acute coronary syndrome

Table 11. Relative risks of CVD events from high intensity compared to low intensity statin treatment from meta-analysis of IDEAL (Pedersen, T. R., Faergeman, O., Kastelein, J. J. et al, 2005) & TNT (LaRosa, J. C., Grundy, S. M., Waters, D. D. et al, 2005) for patients with stable coronary artery disease

Table 12. Costs of CVD events

Table 13. Cost of drugs and GP visits

Table 14. Health state utilities

Table 15. Age-related utility from Health Survey for England 1996

Figure 1. Model structure for cost-effectiveness of lower intensity statins versus placebo in primary prevention of cardiovascular disease (statin benefit used for the identification model)

Figure 2. Model structure for cost-effectiveness of lower intensity statins versus higher intensity in the secondary prevention of cardiovascular disease (used for the high low dose and treat to target models)

Figure 3. Treatment effect used in the model for patients with acute coronary syndrome, meta-analysis of PROVE-IT (Cannon, C. P., Braunwald, E., McCabe, C. H. et al, 2004) and A to Z (de Lemos, J. A., Blazing, M. A., Wiviott, S. D. et al, 2004)

Figure 4. Treatment effect used in the model for patients with stable coronary artery disease, meta-analysis of IDEAL (Usher-Smith, J. A., 2007) and TNT (LaRosa, J. C., Grundy, S. M., Waters, D. D. et al, 2005)

Reference List

1.
Prevention of Cardiovascular Disease. Department of Public Health and Epidemiology, University of Birmingham Website. 2005. [Accessed 2005 Oct. 17]. Available from: URL: http://pcpoh​.bham.ac​.uk/publichealth/cardiovascular/index​.htm.
2.
Health Survey for England 1998. UK Data Archive Website. 2002. [Accessed 2005 May 25]. Available from: URL: http://www​.data-archive​.ac.uk/findingData/snDescription​.asp?sn=4150.
3.
Cannon CP, Braunwald E, McCabe CH, Rader DJ, et al. Intensive versus moderate lipid lowering with statins after acute coronary syndromes. N Engl J Med. 2004;350(15):1495–1504. [PubMed: 15007110]
4.
Caro J, Migliaccio-Walle K, Ishak KJ, Proskorovsky I. The morbidity and mortality following a diagnosis of peripheral arterial disease: long-term follow-up of a large database. BMC Cardiovasc Disord. 2005;5:14. [PMC free article: PMC1183197] [PubMed: 15972099]
5.
Chan PS, Nallamothu BK, Gurm HS, Hayward RA, et al. Incremental Benefit and Cost-Effectiveness of High-Dose Statin Therapy in High-Risk Patients With Coronary Artery Disease. Circulation. 2007;115:2398–2409. [PubMed: 17452609]
6.
Cowie MR, WDCATSP-WPSVaSG Incidence and aetiology of heart failure. A population-based study. European Heart Journal. 1999;20:421–428. [PubMed: 10213345]
7.
Davies A, Hutton J, O’Donnell J, Kingslake S. Cost-effectiveness of rosuvastatin, atorvastatin, simvastatin, pravastatin and fluvastatin for the primary prevention of CHD in the UK. British Journal of Cardiology. 2006;13(3):196–202.
8.
de Lemos JA, Blazing MA, Wiviott SD, Lewis EF, et al. Early intensive vs a delayed conservative simvastatin strategy in patients with acute coronary syndromes: Phase Z of the A to Z trial. JAMA. 2004;292(11):1307–1316. [PubMed: 15337732]
9.
Department of Health. Health Survey for England 1996, The Stationary Office. 1998.
10.
Department of Health (2004) Health Survey for England 2003, Prevalence of angina, adults aged between 55 and 74*, latest available year, UK studies compared. The Stationery Office; London: 2004.
11.
12.
Government Actuaries Department. Interim Life Tables. 2006. GAD website Available from: URL: http://www​.gad.gov.uk​/life_tables/interim_life_tables.htm.
13.
Greenfield S, Bryan S, Gill P, Gutridge K, et al. Factors influencing clinicians’ decisions to prescribe medication to prevent coronary heart disease. J Clin Pharm Ther. 2005;30(1):77–84. [PubMed: 15659007]
14.
Hardie K, HGJKBRAC Ten-year risk of first recurrent stroke and disability after first-ever stroke in the Perth Community Stroke Study. Stroke Stroke. 2004;35:731–735. [PubMed: 14764929]
15.
Hartwell D, Colquitt J, Loveman E, Clegg AJ, et al. Clinical effectiveness and cost-effectiveness of immediate angioplasty for acute myocardial infarction: systematic review and economic evaluation. Health Technol Assess. 2005;9(17):1-iv. [PubMed: 15876363]
16.
Harvard CEA Registry. Cost-effectiveness analysis (CEA) Registry. Tufts-New England Medical Centre website. 1997. Available from: URL: http://www​.tufts-nemc​.org/cearegistry/index.html.
17.
Henderson RA, Pocock SJ, Clayton TC, Knight R, et al. Seven-year outcome in the RITA-2 trial: coronary angioplasty versus medical therapy. J Am Coll Cardiol. 2003;42(7):1161–1170. [PubMed: 14522473]
18.
Johansen H, NCTG Health Reports Winter. 3. Vol. 10. Statistics Canada; 1998. Variation in angiography and bypass surgery; pp. 63–76. [PubMed: 9926348]
19.
Jones PH, Hunninghake DB, Ferdinand KC, Stein EA, et al. Effects of rosuvastatin versus atorvastatin, simvastatin, and pravastatin on non-high-density lipoprotein cholesterol, apolipoproteins, and lipid ratios in patients with hypercholesterolemia: additional results from the STELLAR trial. Clin Ther. 2004;26(9):1388–1399. [PubMed: 15531001]
20.
Kaplan RC, HSFCPBM Predictors of subsequent coronary events, stroke, and death among survivors of first hospitalized myocardial infarction. J Clin Epidemiol. 2002;55(7):654–664. [PubMed: 12160913]
21.
Karnon J, Brennan A, Pandor A, Fowkes G, et al. Modelling the long term cost effectiveness of clopidogrel for the secondary prevention of occlusive vascular events in the UK. Curr Med Res Opin. 2005;21(1):101–112. [PubMed: 15881481]
22.
LaRosa JC, Grundy SM, Waters DD, Shear C, et al. Intensive lipid lowering with atorvastatin in patients with stable coronary disease. N Engl J Med. 2005;352(14):1425–1435. [PubMed: 15755765]
23.
Lavender MCNHDKR. Computer simulation to estimate the effectiveness of carotid endarterectomy. J Health Serv Res Policy. 1998;3(1):6–11. [PubMed: 10180392]
24.
Law MR, Wald NJ, Rudnicka AR, Law MR, et al. Quantifying effect of statins on low density lipoprotein cholesterol, ischaemic heart disease, and stroke: systematic review and meta-analysis. BMJ. 2003;326(7404):1423. [PMC free article: PMC162260] [PubMed: 12829554]
25.
Marshall T. When measurements are misleading: modelling the effects of blood pressure misclassification in the English population. BMJ. 2004;328(7445):933. [PMC free article: PMC390212] [PubMed: 15087340]
26.
Marshall T, Bryan S, Gill P, Greenfield S, et al. Predictors of patients’ preferences for treatments to prevent heart disease. Heart. 2006;92(11):1651–1655. [PMC free article: PMC1861237] [PubMed: 16709699]
27.
Mason JMFNGJMMJMNJP. Specialist nurse-led clinics to improve control of hypertension and hyperlipidemia in diabetes: economic analysis of the SPLINT trial. Diabetes Care. 2005;28(1):40–46. [PubMed: 15616231]
28.
McAlister FA, Laupacis A, Teo KK, Hamilton PG, et al. A survey of clinician attitudes and management practices in hypertension. J Hum Hypertens. 1997;11(7):413–419. [PubMed: 9283056]
29.
Mehta SR, Yusuf S, Peters RJ, Bertrand ME, et al. Effects of pretreatment with clopidogrel and aspirin followed by long-term therapy in patients undergoing percutaneous coronary intervention: the PCI-CURE study. Lancet. 2001;358(9281):527–533. [PubMed: 11520521]
30.
Michiel L, Bots MP, Eline C, van der Wilk M, Peter J, Koudstaal MP, Albert Hofman MPDEGMP. Transient Neurological Attacks in the General Population. Prevalence, Risk Factors, and Clinical Relevance. Stroke. 1997;28:768–773. [PubMed: 9099194]
31.
Murabito JM, D’Agostino RB, Silbershatz H, Wilson WF. Intermittent claudication. A risk profile from The Framingham Heart Study. Circulation. 1997;96(1):44–49. [PubMed: 9236415]
32.
National Institute for Health & Clinical Excellence. The guidelines manual. NICE technical manual. 2006.
33.
National Institute for Health and Clinical Excellence. Statins for the prevention of cardiovascular events. London: National Institute for Health and Clinical Excellence; 2006.
34.
National Institute for Health and Clinical Excellence. Hypertension: Management of hypertension in adults in primary care: partial update. London: National Institute for Health and Clinical Excellence; 2006.
35.
Nazir DJ, Roberts RS, Hill SA, McQueen MJ. Monthly intra-individual variation in lipids over a 1-year period in 22 normal subjects. Clin Biochem. 1999;32(5):381–389. [PubMed: 10480454]
36.
NHS Prescription Pricing Authority. NHS Electronic Drug Tariff. 2008. NHS Business Services Authority website Available from: URL: http://www​.ppa.org.uk​/edt/July_2006/mindex.htm.
37.
Office for National Statistics. Key Health Statistics from General Practice, Prevalence of heart failure, adults aged between 45 and 84, latest available year, UK studies compared. The Stationery Office; London: 2000.
38.
Office for National Statistics. Statbase. 2006. National Statistics website Available from: URL: http://www​.statistics​.gov.uk/statbase/expodata​/spreadsheets/d8986.xls.
39.
Packham C, Gray D, Silcocks P, Hampton J. Identifying the risk of death following hospital discharge in patients admitted with a suspected acute myocardial infarction in whom the diagnosis is not confirmed. Eur Heart J. 2000;21:213–219. [PubMed: 10639303]
40.
Pedersen TR, Faergeman O, Kastelein JJ, Olsson AG, et al. High-dose atorvastatin vs usual-dose simvastatin for secondary prevention after myocardial infarction: the IDEAL study: a randomized controlled trial. JAMA. 2005;294(19):2437–2445. [PubMed: 16287954]
41.
PSSRU. Personal Social Services Research Unit PSSRU. 2005. www​.pssru.ac.uk.
42.
Robinson M, Palmer S, Sculpher M, Philips Z, et al. Cost-effectiveness of alternative strategies for the initial medical management of non-ST elevation acute coronary syndrome: systematic review and decision-analytical modelling. Health Technol Assess. 2005;9(27):1–72. [PubMed: 16022802]
43.
SOLVD Investigators. Effect of enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med. 1991;325(5):293–302. [PubMed: 2057034]
44.
Steel N. Thresholds for taking antihypertensive drugs in different professional and lay groups: questionnaire survey. BMJ. 2000;320(7247):1446–1447. [PMC free article: PMC27388] [PubMed: 10827049]
45.
The Clopidogrel in Unstable Angina to Prevent Recurrent Events Trial Investigators. Effects of Clopidogrel in Addition to Aspirin in Patients with Acute Coronary Syndromes without ST-Segment Elevation. NEJM. 2001;345(7):494–502. [PubMed: 11519503]
46.
Trewby PN, Reddy AV, Trewby CS, Ashton VJ, et al. Are preventive drugs preventive enough? A study of patients’ expectation of benefit from preventive drugs. Clin Med. 2002;2(6):527–533. [PMC free article: PMC4953904] [PubMed: 12528966]
47.
Uffe Ravnskov PJRMCSaMCH. Should we lower cholesterol as much aspossible. BMJ. 2006;332:1330–1332. [PMC free article: PMC1473073] [PubMed: 16740566]
48.
Usher-Smith JA. Evaluation of the cost savings and clinical outcomes of switching patients from atorvastatin to simvastatin and losartan to candesartan in a primary care setting (Structured abstract) Int J Clin Pract. 2007;61(1):15–23. [PubMed: 17229176]
49.
Williams B, Poulter NR, Brown MJ, Davis M, et al. Guidelines for management of hypertension: report of the fourth working party of the British Hypertension Society, 2004-BHS IV. J Hum Hypertens. 2004;18(3):139–185. [PubMed: 14973512]
50.
Wood D, Wray R, Poulter N, Williams B, et al. JBS 2: Joint British Societies’ guidelines on prevention of cardiovascular disease in clinical practice. Heart. 2005;91(supp 5):v1-52. [PMC free article: PMC1876394] [PubMed: 16365341]
51.
Wouter T, Meijer AWHDRMLBA, Hofman, Diederick E, Grobbee Peripheral Arterial Disease in the Elderly: The Rotterdam Study. Arterioscler Thromb Vasc Biol. 1998;18:185–192. [PubMed: 9484982]
52.
Blood pressure variability: lessons learned from a systematic review. 8th International Cochrane Colloquium; October 2000; Cape Town. 2000.
53.
Yock CABDBGAMHMAODK. Cost-effectiveness of bypass surgery versus stenting in patients with multivessel coronary artery disease. Am J Med. 2003;11(5):382–389. [PubMed: 14553874]
54.
Yusuf SM. The HOPE (Heart Outcomes Prevention Evaluation) Study: The design of a large, simple randomized trial of an angiotensin converting enzyme inhibitor (ramipril) and vitamin E in patients at high risk of cardiovascular events. Can J Cardiol. 1996;12(2):127–137. [PubMed: 8605634]
Copyright © 2008, Royal College of General Practitioners.

All rights reserved. No part of this publication may be reproduced in any form (including photocopying or storing it in any medium by electronic means and whether or not transiently or incidentally to some other use of this publication) without the written permission of the copyright owner. Applications for the copyright owner’s written permission to reproduce anypart of this publication should be addressed to the publisher.

Bookshelf ID: NBK55502

Views

  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (749K)
  • Disable Glossary Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...