Objective To estimate the probabilistic cost-effectiveness of cascade screening methods in familial hypercholesterolaemia (FH) from the UK NHS perspective.
Design Economic evaluation (cost utility analysis) comparing four cascade screening strategies for FH: Using low-density lipoprotein (LDL) cholesterol measurements to diagnose affected relatives (cholesterol method); cascading only in patients with a causative mutation identified and using DNA tests to diagnose relatives (DNA method); DNA testing combined with LDL-cholesterol testing in families with no mutation identified, only in patients with clinically defined ‘definite’ FH (DNA+DFH method); DNA testing combined with LDL-cholesterol testing in no-mutation families of both ‘definite’ and ‘probable’ FH patients (DNA+DFH+PFH). A probabilistic model was constructed to estimate the treatment benefit from statins, with all diagnosed individuals receiving high-intensity statin treatment.
Population A cohort of 1000 people suspected of having FH aged 50 years for index cases and 30 years for relatives, followed for a lifetime.
Main outcomes Costs, quality-adjusted life-years (QALY) and incremental cost-effectiveness ratios (ICER).
Results The DNA+DFH+PFH method was the most cost-effective cascade screening strategy. The ICER was estimated at £3666/QALY. Using this strategy, of the tested relatives 30.6% will be true positives, 6.3% false positives, 61.9% true negatives and 1.1% false negatives. Probabilistic sensitivity analysis showed that this approach is 100% cost-effective using the conventional benchmark for cost-effective treatments in the NHS of between £20 000 and £30 000 per QALY gained.
Conclusion Cascade testing of relatives of patients with DFH and PFH is cost-effective when using a combination of DNA testing for known family mutations and LDL-cholesterol levels in the remaining families. The approach is more cost-effective than current primary prevention screening strategies.
- cascade testing
- delivery of care
- DNA testing cost-effectiveness
- familial hypercholesterolaemia
- lipid lowering
- public health
Statistics from Altmetric.com
- cascade testing
- delivery of care
- DNA testing cost-effectiveness
- familial hypercholesterolaemia
- lipid lowering
- public health
What is already known on this topic
FH is a common genetic disorder in which patients have a very high risk of early-onset heart disease that can be effectively treated with high-intensity statins.
Currently in the UK, fewer than 15% of the predicted 110 000 patients are diagnosed and there are no commissioned services to identify people with this condition.
Clinical and cost-effective strategies for the identification of patients with this condition and their implementation are urgently required.
What this study adds
Cascade testing from index patients with both clinically defined definite and possible FH is highly cost effective when using a combination of DNA testing for the family mutation when it can be found and LDL-cholesterol levels when it cannot.
Cascade testing to identify relatives of patients with FH is also more cost effective than recently recommended primary prevention screening strategies.
The approach will become even more cost-effective as technological advances reduce the cost of DNA testing and increase its sensitivity and following the patent expiry of high-intensity statins.
Familial hypercholesterolaemia (FH) is characterised by hypercholesterolaemia, xanthomas and premature coronary heart disease (CHD) and affects approximately one in 500 people in western countries.1 2 It is a monogenic disorder caused by mutations in three genes: those coding for the receptor for low-density lipoprotein (LDL) particles (LDLR), for apolipoprotein B (APOB) and for an enzyme involved in the degradation of the receptor as it recycles, PCSK9.3 Treatment with statins is effective and reduces mortality.4 5 In the UK fewer than 15% of the predicted 110 000 affected people are diagnosed.6
FH is diagnosed in the UK using the Simon Broome criteria2 7 based on cholesterol levels (typically the 95th percentile of total serum cholesterol or LDL-cholesterol is a cut-off value), family history of hyperlipidaemia or early CHD and the presence of (or family history of) xanthomas. Individuals fulfilling these criteria, and those found to carry an FH-causing mutation are given the diagnosis of definite familial hypercholesterolaemia (DFH), while those showing only elevated cholesterol levels together with a family history of hyperlipidaemia or early CHD are given the diagnosis of possible familial hypercholesterolaemia (PFH).
There is considerable overlap in the distribution of LDL-cholesterol levels between individuals with and without FH. In children, in whom the overlap is least,8 9 using a simple cut-off results in a false-positive rate of 8–10% and a false-negative rate of 10–15%. In adults, the false negative is greater.10 Also, an individual's cholesterol levels may fluctuate, moving from below to above the cut-off value on repeat measurements. Therefore, some patients will be given a false-negative diagnosis (ie, told that they do not have FH when they do), whereas others will be given a false-positive diagnosis (told that they have FH when they do not).
When DNA testing is used to diagnose FH, a mutation can be identified in 60–90% of DFH patients,11 12 depending on the sensitivity of the methods and the population under consideration. By comparison, a mutation can be identified in only 20–30% of PFH patients.13 Once the underlying mutation has been identified, molecular genetic screening of first-degree relatives has a sensitivity and specificity close to 100%.
The UK National Institute for Health and Clinical Excellence (NICE) has recommended, based on deterministic economic evaluation of the alternative approaches to cascade screening,14 15 the use of ‘cascade screening’ of first-degree relatives of patients with FH using cholesterol measurement and DNA methods in combination. This paper presents the results of a probabilistic economic analysis to compare the costs and benefits of alternative screening strategies in terms of quality-adjusted life-years (QALY).
Model structure, assumptions and analytical methods
We constructed a decision tree in Excel, in which a hypothetical 1000 patients referred from general practice with a suspicion of heterozygous FH entered the model. Figure 1 shows a schematic presentation of the decision problem and a full breakdown of the decision pathways. The decision pathways for all the methods under consideration each have three disease states that depict the initial diagnosis that is definite FH, possible FH and no FH, as defined by the Simon Broome7 and the FHCAP study.16 Subsequent branches of the tree are dependent on the cascade screening methods under consideration, which are described in detail below. Once individuals are identified as true positives, false positives, true negatives and false negatives at the end of the decision tree, they enter into a Markov model according to the treatment protocol, also described below. Four cascade screening methods were compared.
The cholesterol method
This is the standard method of clinical diagnosis and identification of affected relatives using elevation of LDL-cholesterol levels. Only patients meeting the criteria of DFH or PFH were included for cascade testing.
The DNA method
The identification of an FH-causing mutation by molecular genetics methods, first in the index patient and then in first-degree relatives. Only patients with an identified FH-causing mutation were included for cascade testing.
The DNA+DFH method
Following DNA testing of the index cases cascade testing of relatives is undertaken in all mutation-positive index cases but additionally, in the relatives of DFH index cases in whom no mutation can be found (M-ve), cascade testing is undertaken using cholesterol (LDL-cholesterol) levels to identify affected relatives.
The DNA+DFH+PFH method
Following DNA testing of the index cases cascade testing of relatives is undertaken in all mutation-positive index cases but additionally, in the relatives of DFH and PFH index cases in whom no mutation can be found (M-ve), cascade testing is undertaken using cholesterol (LDL-cholesterol) levels to identify affected relatives.
Treatment protocol and estimated long-term benefits from statin treatment
All index cases and relatives with a diagnosis of FH (whether DFH or PFH) are assumed to be offered high-intensity statin therapy, in line with NICE guidelines on FH,14 while true and false negatives were assumed to be on low-intensity statins. For the relatives, a proportion (1.3%)17 of the subjects with either a true-negative or a false-negative diagnosis will require treatment with low-intensity statins, because the combination of their lipid and other cardiovascular risk factors brings their 10-year cardiovascular disease (CVD) risk to over 20%. False positives were given high-intensity statins but did not benefit from the statin, rather they incurred a disutility (reduction in quality of life) estimated to be approximately 3% and then varied in sensitivity analysis (expert opinion). We developed a Markov model using Microsoft Excel to estimate the treatment benefit from statins. The structure of the model has been described in detail previously,18 and used data from the Simon Broome Study.7 16 Death from other causes was assumed to be the same as that of the general population and was taken from the life tables of the England and Wales government actuary department (2006).19 The model assumes that the risk of CVD increases with age for both men and women.20 The risk of stroke and peripheral artery disease were assumed to be the same as seen in the general population, because data from the Simon Broome Register indicated that these risks are not significantly higher in the FH population.21 Treatment effects of statins were taken from a meta-analysis of the four trials that compared high-intensity statins with low-intensity statins after myocardial infarction—see supplementary table 1.22–25
Drug costs were taken from the British National Formulary26 (number 61, 2011) and are shown in supplementary table 5 (available online only). Costs of full fasting and non-fasting cholesterol measurements and costs of CVD events were taken from NHS reference costs.27 28 All costs were at 2010/11 prices and as per current NICE guidance; an annual discount rate of 3.5% was used for both costs and health benefits.
Outcomes and quality of life (utility)
Clinical outcomes modelled were myocardial infarction, stroke, heart failure, transient ischaemic attack; peripheral arterial disease, unstable angina, revascularisation, cardiovascular and total mortality. Utility weights for the various health states and age-adjusted utility were taken from our earlier study.18 Age-adjusted utility was solicited from the general population using time trade-off29 (see supplementary tables 2 and 3, available online only). The beneficial value of health outcomes was estimated using the QALY. We did not allow for any harmful effects of treatment with statins because significant side-effects are relatively uncommon,30 but assessed their impact in sensitivity analysis.
Probabilistic sensitivity analysis
As a result of imperfect information on the effectiveness of intervention and the resources consumed for treatment, both the costs and effects of health interventions are inevitably associated with some degree of uncertainty, and this introduces the possibility of error into decision-making.31 In our analysis we used Monte Carlo simulation to generate the sampling distribution of the joint mean cost and efficacy in order to quantify the uncertainty around the estimates of costs and effects. In addition, we performed one-way sensitivity analysis on variables that had uncertain estimates and yet were likely to influence overall conclusions. These included the cost of the cholesterol method, a reduction in the cost of statins, any potential loss in quality of life due to side-effects of high-dose statins and the costs of DNA testing.
The four cascade screening methods identified differing numbers of true and false negatives and positives among both cases and relatives (table 1). The DNA-only strategy required the least number of relatives to be tested, but did not identify as many true positives as the DNA+DFH+PFH strategy. This last strategy was also the strategy that required the largest number of relatives to be screened.
Costs of diagnosis and treatment
Table 2 shows the cost of diagnosis and treatment of people diagnosed with either monogenic or polygenic hypercholesterolaemia for each of the four strategies, using the treatment protocol outlined above in the Methods section, while table 3 shows the QALY gained in each strategy. The total costs of diagnosis for the index case included the total cost of clinical confirmation for index cases (lipid profile plus healthcare professional costs, estimated to be £240 per index case and £139 per relative) and the cost of DNA testing. The cost per relative included the costs of sending out letters. These costs were multiplied by the numbers of people tested under each strategy. DNA testing and cascading was not done in those 100 individuals identified in each strategy as true negatives.
The cost of treatment and QALY gain per individual was estimated from the Markov model for each strategy under consideration. The number of index and relative cases identified by each strategy was multiplied by the cost and QALY gain per individual. The total cost of each strategy was thus the sum of the diagnosis and treatment costs.
As shown in table 4, all DNA-based methods were cost-effective relative to the cholesterol-only method. However, cascade testing from DNA+DFH is ruled out by extended dominance. The principle of extended dominance is applied in incremental cost-effectiveness analyses to eliminate from consideration strategies whose costs and benefits are improved by a mixed strategy of two other alternatives.32 The combinations of DNA-only and DNA+DFH+PFH are thus both more cost-effective than DNA+DFH. After accounting for the options ruled out by extended dominance, the relevant incremental comparison is between the DNA method and the DNA+DFH+PFH method. The estimated base case incremental cost-effectiveness ratio (ICER) is £3666/QALY, as shown in table 4.
We assessed uncertainty around this ICER by Monte Carlo simulation using 2000 iterations. Figure 2 illustrates the probability that any one strategy is cost-effective, as a function of the willingness to pay. Given a maximum acceptable ceiling ratio of £20 000/QALY the probability that DNA+DFH+PFH is cost-effective compared with the DNA method is 100%. Therefore, given the data, there is a 100% chance that the additional cost of DNA+DFH+PFH, compared with the DNA method, is at or below £20 000 per QALY gained.
One-way sensitivity analysis showed that the model results were not sensitive to changes in assumptions about loss in quality of life due to side-effects of high-dose statins as the ICER remained below £20 000/QALY when the assumption was varied between 1% and 10%.
We also varied the cost of statins as we expect atorvastatin to be off patent in 2011. We thus reduced the cost of atorvastatin by 60% and the cost-effectiveness results became more favourable, with the ICER falling from the current estimate of £3666 to £3070/QALY. We also varied other variables like the proportions of index cases and relatives who agreed to testing, the age at identification for index cases and relatives, and the cost of cholesterol testing and DNA costs, and in all cases the ICER remained below £4000/QALY, suggesting the model is not sensitive to changes in these parameters. The base model assumed that there was no quality of life loss associated with side-effects of statins, and when we assumed a 5% loss in quality of life, the ICER increased only slightly to £4028, demonstrating that the model results are also not sensitive to this assumption.
Our economic analysis indicates that the most cost-effective cascade screening strategy for people suspected of having FH is DNA testing plus cascading from both mutation negative definite and possible FH individuals, with an estimated ICER of £3666/QALY when compared with the DNA-only method. Our results were stable in univariate sensitivity analysis. Probabilistic sensitivity analysis also showed that the DNA+DFH+PFH strategy is 100% cost-effective as it falls below the recommended £20 000/QALY threshold currently used in the UK for evaluating interventions. Altering assumptions about several key determinants of cost and effectiveness including the cost of statins (which will fall in the near future as some of the potent statins recommended for FH patients come off patent), the cost of DNA testing, the overall cost of the cholesterol measures, the proportions of index cases and relatives who agreed to testing, and the age at identification for index cases and relatives did not materially influence the ICER. There is no uncertainty that DNA+DFH +PFH is the most cost-effective option.
Strengths and weaknesses
To our knowledge this is the first probabilistic analysis of FH screening strategies for the UK.
Because patients with FH have very high LDL-cholesterol levels from birth, they will frequently require high-intensity lipid-lowering therapy sufficient to reduce LDL-cholesterol to recommended levels,14 15 and studies have shown that statin treatment reduces their premature mortality.5 In our model we have used a combination of statins, but the model was not sensitive to different combinations of statins. We have assumed that any individual identified with elevated LDL-cholesterol levels will be treated whether or not they carry the family mutation. Individuals who do not carry the mutation are likely to be treated with a lower dose of statins and the costs and benefits for this have been included in the model. The results of our model were not sensitive to the documented side-effects of statins33 34 because variations in screening methods influence the numbers of people allocated to high or low dose, but have no influence on the QALY gain from the statin treatment.
Because observational data from the Simon Broome Register cohort showed no significant increase in mortality in FH patients over 60 years of age,4 we have assumed that people over the age of 60 years will benefit from statins to the same degree as the general population. This does not support ceasing treatment at age 60 years in people diagnosed with FH. People with FH who have reached this age or beyond without treatment and without experiencing any cardiovascular event or symptoms appear to have a risk low enough not to warrant high-intensity treatment, that is, a survivor effect.
Our model did not consider cascade testing from children due to a lack of data on the effectiveness of statins in children. If children were included in the case-finding approach, this strategy is likely to become even more cost effective as the number of relatives per index case would increase. A false-negative diagnosis may deny both the patient who has FH and that person's relatives with FH the benefit of more intensive cholesterol-lowering therapy. By contrast, cascade screening from false-positive cases will not identify any true FH patients and will waste resources. It would be possible to reduce the numbers of false positives and negatives if better data were available on the range of LDL-cholesterol levels to be expected in the mutation-carrying relatives of patients with FH, and the extent to which this range overlaps with that in non-mutation-carrying relatives.
Markov models have inherent limitations. They 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 longer ‘memory’ in the model). Similarly, a patient's health outcome and healthcare 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 the costs and overestimate the health outcomes for CVD events. Therefore, interventions that prevent more CVD events will appear less cost-effective than they may be in reality.
Comparison with other studies
Our findings that DNA-based screening methods are more cost effective are consistent with other published studies. Marks et al35 undertook a cost-effectiveness analysis from the NHS perspective. This included universal screening, opportunistic screening in primary care, screening of people admitted to hospital with premature myocardial infarction, or tracing family members of affected patients. The authors concluded that screening family members was the most cost-effective strategy, with an estimated ICER of £3097 per life-year gained (LYG) using cholesterol measurement for diagnosis and £4914/LYG using DNA testing, while universal population screening using cholesterol measurement only was a much less cost-effective strategy with an estimated ICER of £13 029/LYG. Marks et al36 also considered the costs and deaths averted over 10 years from either a population strategy of screening 16-year-olds or tracing family members of affected patients. They concluded that family tracing was again the most efficient strategy, with the cost per death averted being £3187. A cost-effectiveness study of the FH genetic screening programme in The Netherlands resulted in a similar cost per LYG of US$8800.37 The result was sensitive to the price of statin treatment and the number of LYG.
Our results also compare favourably with strategies to identify individuals at a lower risk of CVD (ie, primary prevention), which were evaluated in a previous NICE guideline in which the recommended screening strategy (targeted screening) has an ICER of £7604/QALY.38
Implications and future research
There is a lack of UK data describing the range of LDL-cholesterol levels to be expected in the mutation-carrying relatives of patients with FH, and the extent to which this range overlaps with that in non-mutation-carrying relatives. Further research is required to characterise the distributions of LDL-cholesterol levels in mutation-carrying relatives of patients with FH and the extent of overlap with levels in other relatives to improve the performance of screening strategies. The cost-effectiveness of DNA screening is likely to improve in the future as the proportion of definite FH patients in whom a mutation can be identified increases because of improvements in techniques for mutation identification, and also because of the identification of new genes in which mutations cause FH. However, even now this economic analysis supports the identification and treatment of individuals with FH as a highly cost-effective strategy in the prevention of CVD.
National strategies to reduce the burden of CVD in the UK would be made more effective and more cost effective by incorporating the screening strategy that was recommended in national guidance from NICE for the identification and treatment of people with FH. To date we are not aware that there has been any local implementation of such a strategy in England.
Funding The National Collaborating Centre for Primary Care was commissioned and funded by the National Institute for Health and Clinical Excellence to develop the guideline for the identification and management of adults and children with familial hypercholesterolaemia. This paper reports work that was undertaken at the request of the Guideline Development Group. The full guideline can be accessed at http://www.nice.org.uk/Guidance/CG71. SEH would like to acknowledge grants PG2008/008 from the British Heart Foundation.
Competing interests LN, RM, MT and SEH were members of the Guideline Development Group for the guideline for the identification and management of adults and children with familial hypercholesterolaemia. DM has no conflicts to declare. Any opinions expressed in this paper are those of the authors and are not intended to represent those of any affiliated organisations.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.