PubMed Health. A service of the National Library of Medicine, National Institutes of Health.

Sharma P, Boyers D, Boachie C, et al. Elucigene FH20 and LIPOchip for the Diagnosis of Familial Hypercholesterolaemia: A Systematic Review and Economic Evaluation. Southampton (UK): NIHR Journals Library; 2012 Mar. (Health Technology Assessment, No. 16.17.)

3Assessment design and results: cost-effectiveness

Review of cost-effectiveness studies

Search strategy

Two separate searches were conducted for studies considering the cost-effectiveness of any of the intervention tests (Elucigene FH20 or LIPOchip) for proband testing or for the cascade testing of relatives. Studies were sourced from searching a range of electronic databases and websites. This was supplemented with a quality-of-life search. Contact with experts in the field and the scrutiny of bibliographies of retrieved papers were also used to identify any additional studies. Highly sensitive electronic searches were conducted to identify reports of published studies on the cost-effectiveness of tests for FH in index cases and for cascade testing of relatives. The search focused on identifying RCTs and comparative studies and the results were restricted to articles written in English. The search strategy included searches of all relevant journals since inception.

The databases searched were MEDLINE (1948 to Week 1 2011), MEDLINE In-Process & Other Non-Indexed Citations (10 January 2011), EMBASE (1980 to 2011 Week 1), BIOSIS (1956 to 10 January 2011), Science Citation Index (1970 to 10 January 2011), Conference Proceedings Citation Index – Science (1990 to 10 January 2011), Centre for Reviews and Dissemination databases including Database of Abstracts of Reviews of Effects, NHS Economic Evaluation Database (NHS EED) and Health Technology Assessment database. Searches were also carried out of the Cost-Effectiveness Analysis Registry. A supplementary quality-of-life search was also undertaken, including MEDLINE (1948 to Week 1 2011), MEDLINE In-Process & Other Non-Indexed Citations (10 January 2011), EMBASE (1980 to 2011 Week 1) and IDEAS Economics and Finance Research (February 2011). Full details of the search strategies used and websites consulted are documented in Appendix 3. In addition, reference lists of all included studies were scanned to identify additional potentially relevant studies

Methods (inclusion and exclusion criteria)

Studies were deemed to be relevant for the cost-effectiveness review if they included a measure of cost-effectiveness of the intervention tests (Elucigene FH20 – or alternative earlier versions or LIPOchip version 8–version 10) relative to any of the included clinical diagnostic criteria (Simon Broome, MedPed or Dutch criteria). The population and setting for the studies retrieved for further investigation were as described in Chapter 2. In terms of outcomes, the preferred type of analysis was cost-effectiveness measured as cost–utility analysis [cost per quality-adjusted life-year (QALY) gained]. However, because of a lack of data, we also considered other measures of cost-effectiveness, including cost per case detected or cost per diagnostic accuracy measurement. Study type inclusion and exclusion criteria were limited as we did not want to exclude any potentially relevant studies at this stage, the principal requirement being that studies were for a population of index cases or relatives of index cases with a clinical diagnosis of FH. Titles and abstracts of all reports identified by the search strategy were screened. Full-text copies of all studies deemed to be potentially relevant were obtained and assessed for inclusion. Disagreements were resolved by consensus or arbitration by a local clinical advisor. A data extraction form was developed, with data extracted by one health economist. A second health economist checked the data extraction and any disagreements were resolved by consensus among the review team. Additional further studies that did not meet our specific inclusion criteria but were none the less informative for development and population of the economic model were also retained. As these additional included studies did not form a vital part of the assessment, they have not been systematically critically appraised in depth but are included and narratively described in the following sections.

Results of the cost-effectiveness searches

A total of 258 papers were initially identified through the database searches, with a further 11 potentially relevant titles identified through the diagnostic accuracy search. However, on reading the titles and abstracts, only nine were judged potentially relevant to the cost-effectiveness review, with the remaining 260 not meeting the inclusion criteria of health economic analysis (cost-effectiveness or cost–utility) of a genetic test. We requested full-text articles of these nine papers that reported the cost-effectiveness of genetic testing and cascade testing techniques. These papers were further assessed by reading the full text of each retrieved paper and reapplying the inclusion and exclusion criteria. At this stage, only one study reported the cost-effectiveness of any of the comparators for this assessment. Of the remaining eight papers, three did not include any measure of cost-effectiveness and only briefly referred to cost implications, thus leaving a total of five relevant studies. Four of the five studies retrieved have been summarised in the previous systematic review undertaken as part of NICE clinical guidance CG71.1 Data are extracted and published in appendix D of the clinical guidance document. The remaining study, which was not previously summarised as part of CG71,1 is discussed below. In relation to additional searches for utility of diagnostic information, effect of mutation type on treatment choice and the efficacy of statins in children, potentially relevant full-text papers were retrieved and read in full, and have been considered in the economic modelling process and/or discussion where appropriate.

Discussion of included studies evaluating the cost-effectiveness of Elucigene FH20 and/or LIPOchip

One study57 was identified that met our inclusion criteria and evaluated the cost-effectiveness of one of the intervention tests. This study assessed the cost-effectiveness of LIPOchip in identifying and testing first-degree relatives of index cases identified with FH in a Spanish population. The analysis also included subsequent treatment with statins of test-positive individuals. Screening and treatment management was compared with a strategy of no screening and the perspective of the analysis was that of the national health system (payer). Cost-effectiveness outcome was measured as incremental cost per life-year gained. Clinical diagnosis of at-risk individuals was based on a uniform protocol for clinical diagnosis and genetic testing of index cases was carried out using the LIPOchip platform, which included the following diagnostic steps:

  1. LIPOchip DNA array
  2. multiplex quantitative PCR used to identify significant gene rearrangements (applied if DNA array was negative)
  3. complete sequencing of the LDLR gene (applied if the previous two steps were negative).

Among confirmed cases, the DNA array had a specificity and sensitivity of 99.7% and 99.9%, respectively, for all 118 mutations tested. Once index patients were identified, first-degree relatives were tested using steps 1 and 2 above only. Effectiveness among relatives was based on relative risks adjusted for age and sex59 and applied to national mortality rates. Once identified and treated, it was assumed that mortality risk reduced relative to untreated patients. The total cost of detecting a positive case was €1447 based on the assumption that, to detect one positive case, 3.4 relatives would need to be tested. This was combined with treatment costs based on simvastatin 40 mg and costs of acute myocardial infarctions (MIs) avoided based on risks calculated from Wonderling and colleagues.59 The cost-effectiveness was thus estimated based on cost per life-year gained as €3243 in the base-case analysis. Sensitivity analyses conducted varied the incremental cost-effectiveness ratio (ICER) from €1073 to €7235 per life-year gained. Probabilistic analyses indicated a 95% probability of cost-effectiveness at a societal willingness to pay for a life-year gained of > €7400 and a probability of 45% at a willingness to pay of €3450. The results suggest that genetic screening of first-degree relatives with LIPOchip in Spain is a cost-effective use of resources. The main limitation to this study in terms of this assessment is that there is no active comparator – it is assumed that no screening would take place in routine care. However, the study is useful and informative regarding the potential of LIPOchip. No studies were available reporting on cost-effectiveness for any of the other intervention tests.

Discussion of supplementary cost-effectiveness evidence

The remaining supplementary papers detailing cost-effectiveness of cascade testing among relatives using targeted cascade testing and other methods are briefly summarised and discussed below. Full data extraction pertaining to these reports is available from the NICE website as appendix D to the NICE clinical guideline document CG71.1 None of these studies evaluated the cost-effectiveness of any of the tests specifically in index patients; however, all indicated cost-effectiveness of cascade testing for FH among relatives of known FH index patients. The five included studies are discussed briefly below.

Marang-van de Mheen and colleagues60 compared five screening options in the Dutch population compared with no screening: treating (1) all patients with cholesterol level > the 95th percentile for the general Dutch population; (2) individuals fulfilling treatment criteria based on Dutch Institute of Health Care Improvement guidelines on hypercholesterolaemia; (3) (1) above but only those untreated at screening; (4) (2) above but only those untreated at screening; and (5) all FH-positive patients. The Framingham equation61 was used to estimate risk, survival and costs and the economic outcome measure is cost per life-year gained. This is explicitly not recommended as part of CG711 for calculating risk in the Simon Broome population. The most cost-effective option is option (2) with an associated ICER of €24,376 per life-year gained. Discounting was not conducted and there are questions relating to generalisability to a NHS perspective.

Marks and colleagues62 completed a cost-effectiveness analysis of screening for FH patients aged 16–24 years from the perspective of the NHS. Strategies evaluated were universal screening, opportunistic screening (unrelated reasons), opportunistic screening (patients with premature MI) and full screening of all first-degree relatives diagnosed with FH. The main comparison for the analysis was no screening. The primary outcome measure was cost per life-year gained and the study showed that tracing family members (first-degree relatives) systematically was the most cost-effective strategy with an ICER of £3097 per life-year gained.

Marks and colleagues63 conducted additional work over a 10-year period estimating the cost-effectiveness of (1) family tracing of index cases and (2) systematically screening all 16-year-olds. Primary economic outcomes were cost per case detected and cost per death averted. The main comparison for the analysis was no screening and no incremental analyses were conducted between groups. Costs per case identified were £3505 (family tracing) and £13,141 (universal screening). Costs per death averted were £3187 and £1.6M for the family tracing and universal options respectively. Therefore, the authors conclude that a more targeted screening programme identifying relatives of index cases is more cost-effective.

Wonderling59 used data from the Dutch screening programme from year 2000 in a sample of 18- to 60-year-olds to estimate the cost-effectiveness of screening compared with no screening. Treatment was administered using statins and it was estimated that screening would prevent 26 MIs per 100 patients receiving statin therapy. Primary outcome measures for the economic analysis were cost per case detected and cost per life-year gained, which were $7500 and $8800 respectively. Results were sensitive to the price of statins and a worst-case scenario estimated that the ICER could increase to $38,300 per life-year gained.

The additional included study was an older version of the currently included Marks study.64 Therefore, the up-to-date data have been reported. Other studies, including those by Leren,65 Humphries and colleagues66 and Hadfield and colleagues67 all suggest that genetic screening is a cost-effective use of NHS resources and should be implemented across the UK.

The main background for the economic modelling of these candidate tests comes from NICE clinical guideline CG71,1 in which an economic model was developed to compare DNA testing with LDL-C testing. The results showed that DNA testing was cost-effective with an associated ICER of £2676 per QALY gained. This model has been updated and integrated to account for the testing of Elucigene FH20 and LIPOchip and other plausible scenarios for the identification of FH and is described in more detail in the following sections.

We did not identify any other health economic models for the identification of FH that would be informative to the development of this assessment.

Summary

NICE clinical guidance (CG71)1 concluded that genetic testing of relatives of index cases with FH is cost-effective. There was, however, no available evidence detailing the cost-effectiveness of genetic testing of index patients specifically using any of the candidate tests in this review (i.e. Elucigene FH20 or LIPOchip). One study evaluated the cost-effectiveness of an intervention test for cascade testing of relatives.57 This is cascade testing based on LIPOchip; however, there are less costly methods of cascade testing of relatives (targeted sequencing) and so this analysis may be of limited use for informing the economic evaluation for this appraisal. A number of supplementary studies discussed provide strong evidence that cascade testing of relatives of index cases with FH is cost-effective. Based on this evidence together with the results of CG71,1 we have developed an economic model to assess the cost-effectiveness of Elucigene FH20, LIPOchip and comparators (including CGA and LDL-C) for the identification and treatment of index cases with FH and the identification of relatives by cascade testing.

Methods for economic analysis

The care pathway for this economic evaluation has been defined by NICE clinical guidance (CG71)1 and is as summarised in Chapter 1 (see Care pathways). In brief, the key points set out in this guideline that have implications for the economic evaluation recommend:

  • DNA testing to confirm clinical diagnosis of FH based on Simon Broome criteria in index (proband) patients suspected of having FH. A clinical diagnosis will include two LDL-C concentration measurements.
  • DNA testing for identified mutations in first-, second- and possibly third-degree family relatives.
  • Patients identified with FH should be offered a high-intensity statin therapy option.

A number of diagnostic pathways were specified as part of the NICE scope and review group protocol for analysis and are used to develop the economic modelling for this assessment; they are presented in Table 18.

TABLE 18. Diagnostic strategies for identifying a genetic mutation (or LDL-C level) in index cases.

TABLE 18

Diagnostic strategies for identifying a genetic mutation (or LDL-C level) in index cases.

Using these care pathways we developed an economic model to estimate the cost-effectiveness of several diagnostic strategies for the confirmation of clinical diagnosis of FH among index cases and the subsequent identification and treatment of FH-positive first-, second- and third-degree biological relatives of the index case.

Model structure

The model structure was developed based on clinical advice in line with the NICE scoping document and assessment group protocol. As diagnostic strategies in themselves do not lead to quality-of-life implications directly, the model follows a linked evidence approach in which intermediate outcomes (diagnostic accuracy) are linked to treatment outcomes and hence QALY gains. By a linked evidence approach we mean that, based on diagnostic test result, a patient will be either positive or negative. Positive-testing patients receive a high-intensity treatment and negative-testing index cases receive a low-intensity treatment as they will still be at risk of cardiovascular events based on high LDL-C levels. The treatment received by each group (true-positive, true-negative, false-positive, false-negative) will determine their cardiovascular events avoided and hence their QALYs gained from that treatment decision. The outcomes on the index diagnostic test also determine whether or not the relatives will receive targeted sequencing in combination with LDL-C or LDL-C alone as the cascade test of choice. Therefore, we can say that the diagnostic test outcome of the index case is ‘linked’ to treatment choice and overall health outcomes over a lifetime horizon.

A decision tree model has been developed to identify the most cost-effective method of identification of index cases and subsequent testing and identification of at-risk relatives. Diagnostic accuracy outcomes are linked to treatment outcomes and hence QALY gains using a previously developed economic Markov model used for clinical guidance (NICE CG711).

One of the most important advantages of genetic testing is the identification of family members for cascade testing. The test used to cascade test relatives of index cases will depend on the test used to identify the index case. Three tests (targeted gene sequencing, LIPOchip and Elucigene FH20) are substantially cheaper than CGA and may be used for cascade testing. For the majority of genetically confirmed index cases, targeted sequencing for the culprit mutation is the most commonly applied genetic cascade testing method (Dr Zosia Miedzybrodzka, University of Aberdeen, 2011, personal communication). For relatives of index cases identified using the Elucigene FH20 or LIPOchip tests, the cheaper test designed to detect the identified mutation (LIPOchip, Elucigene FH20 or targeted sequencing) may be used to cascade test relatives. LIPOchip or Elucigene FH20 may also be used to cascade known mutations picked up on other tests that would also be detected by the candidate tests.68 This scenario would apply only if LIPOchip or Elucigene FH20 were cheaper than targeted sequencing. For index cases identified based on Simon Broome criteria and not a genetic test, then LDL-C concentration measurement is the most common method used to cascade test relatives. The model structure for relatives assumes that, once a patient has a confirmed diagnosis, his/her close relatives will be identified and cascade testing will begin, first testing all first-degree relatives. For the base-case analysis it is assumed that each index case will have on average five first-degree relatives and each first-degree relative will have on average a further two first-degree relatives (second-degree relatives of each index case) who will require testing. For the purposes of this assessment we assume that once a first-degree relative tests positive, the process moves on to second-degree relatives and similarly on to third-degree relatives if appropriate. If a first-degree relative tests negative for FH, then the cascade testing process stops irrespective of the test used for cascading.

A copy of the model decision tree is illustrated in Figure 9, detailing the identification strategies for index cases in the model. Each circle represents a chance node at which probabilities of positive and negative test results are assigned. Index cases receive cost and QALY payoffs at each terminal node (triangle), at which point relatives are identified for cascade testing as described above.

FIGURE 9. Economic decision tree model for index cases.

FIGURE 9

Economic decision tree model for index cases. DFH, definite familial hypercholesterolaemia; PFH, possible familial hypercholesterolaemia.

Identification of probabilities for the decision model

The probabilities used to populate this model were estimated using standard conventions of Bayes' theorem. Basically, once we know the sensitivity and specificity of a test as well as the a priori probability of disease in the target population, we can calculate positive, negative, true-positive, true-negative and thus false-positive and false-negative values for the model. The formulae used for the calculation of each branch of the tree for single test strategies (e.g. CGA alone) are described in Table 19.

TABLE 19. Calculation of probabilities for decision tree.

TABLE 19

Calculation of probabilities for decision tree.

When tests are connected in series as add-ons to each other (i.e. the second test detects the same mutations as the first test plus additional FH-causing mutations), the theory is essentially the same but will be represented by the associated values of the second test. Taking Elucigene FH20 followed by CGA as an example, the positive rate will be [(proportion testing positive on Elucigene FH20 + proportion testing positive on CGA) – proportion testing positive on Elucigene FH20]. The proportions testing positive on Elucigene FH20 cancel each other out as they are incorporated in CGA and CGA detects all the mutations detected by Elucigene FH20 and more; therefore, the proportion testing positive on this example strategy is simply the value of the most comprehensive test in the strategy (i.e. CGA). A similar argument applies to Elucigene FH20 followed by LIPOchip.

For strategies in which MLPA is used as an add-on test to Elucigene FH20 or LIPOchip, the calculations are slightly different. As MLPA detects additional cases not detected using Elucigene FH20 or LIPOchip (we assume here that the detection of deletions and duplications on LIPOchip is inadequate and MLPA will still be needed to give a more robust estimate), the effect of the two tests in series is not as before. Therefore, for the calculation of true-negatives on Elucigene FH20 followed by MLPA, the effect will be multiplicative and can be calculated as [(1 – prevalence) × (specificity of Elucigene FH20) × (specificity of MLPA)]. The MLPA test has not been considered separately from CGA because, by definition, CGA will already include MLPA as part of the process.

Sensitivity and specificity values used in the calculations of the model are presented in Table 20 for information. More detailed information on sensitivity and specificity for all included studies is presented in Chapter 2. Studies chosen to inform the economic modelling fulfilled two main criteria: (1) they were based on patients with a Simon Broome definite FH or possible FH clinical diagnosis of FH (preferably in a UK population) where possible and (2) when tests were conducted in a number of different countries (outwith the UK) in a study, we have chosen the cohort with the largest sample size (unless some explicit reason existed why this would not be appropriate). These were assumed to offer the most robust estimates in the absence of UK data. When studies did not report clinical diagnosis based on the Simon Broome criteria or when evidence was of poor quality and limited usability, we obtained parameter values from Dutch and MedPed criteria instead. For reasons discussed in the statistical analysis, it has not been possible to pool estimates of sensitivity and specificity for a combination of definite FH and possible FH diagnoses across studies in a robust way because of study heterogeneity (see Chapter 2, Assessment of test performance). Therefore, single studies have been chosen based on the best available evidence and the most recent version of each test analysed. The impact of these choices on our base-case conclusions will be explored through the use of lowest and highest estimates available from all of the included studies, based on all clinical criteria (MedPed and Dutch criteria included), in sensitivity analysis.

TABLE 20. Sensitivity and specificity of tests used to populate the economic model.

TABLE 20

Sensitivity and specificity of tests used to populate the economic model.

It is important to note that there is likely to be some correlation between those patients detected on MLPA and those detected using LIPOchip. Clinical expert opinion (Dr Zosia Miedzybrodzka, University of Aberdeen, personal communication) suggests that the LIPOchip test may be inadequate to detect deletions and duplications and in practice MLPA may be required to give a more accurate diagnosis.

LIPOchip can be used within the model in two separate ways. First, the strategy ‘LIPOchip’ refers to the test purchased by a laboratory in the UK from the manufacturer and processed at the UK laboratory. Additionally, the manufacturer offers a service whereby blood samples can be sent to the manufacturer's plant in Spain for analysis using a two-stage process, first testing with LIPOchip and then sequencing of the LDLR gene for those testing negative. This is referred to as LIPOchip platform (Spain). Because of its second stage, at an additional cost of €100, this test has a higher sensitivity. It is, however, not CGA as the process does not include MLPA. Therefore, the sensitivity is less than that of CGA. Clinical expert opinion in the UK suggests that, to be able to fully detect all deletions and duplications of the gene, the MLPA test would be required as LIPOchip's own method of detecting these cases may be inadequate. Additional data presented at the spring meeting of the CMGS70 suggest that (using data from Bristol's NHS Hospital Genetics Laboratory) LIPOchip version 10 may be inadequate to detect copy number changes compared with MLPA, with only two cases out of a sample of seven correctly identified using LIPOchip.

In addition, there is much debate about the true prevalence of detectable FH-causing mutations among patients testing positive (definite FH or possible FH) based on the Simon Broome criteria. There is also great variation in this number between laboratories and this is likely to be because of issues of ethnicity as some tests will have different detection rates based on different ethnic groups (see Chapter 2, Assessment of test performance for additional information). For the purposes of our base-case analysis, we have assumed that 36.5% of clinically diagnosed patients (Simon Broome definite FH or possible FH) will have an identifiable mutation.37 Data from four regional Scottish genetics services (Aberdeen, Dundee, Edinburgh and Glasgow; Dr Zosia Miedzybrodzka, University of Aberdeen, 2010, personal communication) suggest that, between 2007 and 2010, this value was approximately 35% for the whole of Scotland based on data classifiable as definite FH or possible FH. This has been confirmed in personal communication with Dr Zosia Miedzybrodzka, who estimates that, for every three patients tested in Aberdeen using CGA, on average only one will have a detectable FH-causing mutation. NICE CG711 estimates, using data extracted from the UK FH Cascade Audit Project (FHCAP),70 that 80% of patients clinically diagnosed with definite FH will have a detectable FH-causing mutation and 30% of those diagnosed as possible FH will have a detectable mutation. Given that the FH audit 201018 identifies 36% as definite FH and 58% as possible FH (the remainder being homozygous or not stated), this would suggest that 46.2% of patients clinically diagnosed as definite FH or possible FH would have an identifiable genetic mutation using CGA. Other studies quote varying estimates of these values and so maximum and minimum values will be explored in the sensitivity analysis. It is estimated that 50% of first-degree relatives of an index case will have an inherited mutation. This evidence for first-degree relatives has been applied to second- and third-degree relatives in the model. The reason for this is that the process of cascade testing is an iterative approach. Second-degree relatives will not be tested using targeted sequencing unless a first-degree relative has an identified mutation. Therefore, it is assumed that the second-degree relative is in fact the first-degree relative of an individual with an identified FH-causing mutation and so will also have a 50% probability of having inherited that mutation.

Markov model

The Markov model for this assessment has been adapted from the model used for the estimation of treatment effect used to inform NICE CG71.1 The model was developed by the Royal College of Physicians Guideline Development Group and is updated in this assessment. This model calculated the lifelong treatment costs and outcomes of high-intensity statin therapy for the management of FH and low-intensity statin therapy for the management of others at risk of CHD because of elevated lipid levels. In addition to those who were classed as well, a total of eight further health states were modelled [unstable angina, MI, peripheral arterial disease (PAD), stroke, heart failure, revascularisation, cardiovascular death and other death]. Baseline risks were sourced from NICE technology appraisal 9471 and relative risks were sourced from the Simon Broome register. Utility weights were sourced from the literature and validated by the health economist working on this assessment. Utility of the general population was taken from the Health Survey for England 1996,72 which is the most up-to-date data source for the UK general population, and was adjusted for age and sex differentials. Beneficial health outcomes were used to estimate QALYs based on reduced risks of cardiovascular incidents. These treatment effects were sourced from a meta-analysis of two RCTs, the Incremental Decrease in Clinical Endpoints Through Aggressive Lipid Lowering (IDEAL) and the Treating to New Targets (TNT) trials conducted as part of the NICE CG71 assessment.1 Data from Versmissen and colleagues8 were checked against and found to be consistent with the assumptions and data used for CG71,1 in so far as they show the efficacy of statins in improving the clinical causes of cardiovascular disease and by extension the reduction in serious cardiovascular events such as MI. However, they do not describe the exact causal relationship between the improved clinical outcome and reduced events.The data from Versmissen and colleagues8 are consistent with those of the CG711 assessment in that they suggest efficacy of statins and by extension the reduction in serious cardiovascular events such as MI. Costs and outcome data have been updated to current values using the latest available literature in the field or inflated to current prices (2010/2011) if no updated literature was available. Further details of the model structure are available from the NICE website (appendix E to the clinical guideline document1). The perspective of this economic evaluation is that of the UK NHS and all costs and resource use are applied in accordance with NICE guidelines on the methods of technology appraisal. NICE recommends that, where possible, the desired economic outcome is cost per QALY gained. Treatment costs and QALYs gained are extrapolated to the patient's lifetime horizon and discounted at a rate of 3.5% per annum in line with standard NICE methods. It was not deemed necessary to discount diagnostic costs for each individual as the time taken for diagnosis is < 1 year. Sensitivity analyses explore the impact of varying the discount rate for both costs and QALYs between 0% and 6%. All other follow-up clinical costs that are expected to occur annually once a diagnosis of FH has been made are discounted as described.

Relevant patient populations

The relevant patient population for the base-case analyses is adults with heterozygous FH, focusing on index patients with a clinical diagnosis of FH based on the Simon Broome criteria (either definite or possible FH). Sensitivity and specificity of the tests included for the economic modelling both implicitly account for patients with either definite or possible FH. Data showing separate sensitivity and specificity rates for definite FH and possible FH were not available for all tests under consideration, thus making accurate subgroup analysis difficult. The data that were available are detailed in Tables 914 (index cases) and Table 15 (testing of relatives). Children with a clinical diagnosis are considered as a separate age subgroup in line with current CG71 recommendations.1 Patients with an identified mutation causing FH are informed of their diagnosis and first-, second- and third-degree biological relatives are identified. Sensitivity analysis explores a situation in which only first- and second-degree biological relatives are cascade tested.

Treatment options to be evaluated

Treatment options to be evaluated are based on NICE CG71,1 which recommends that a patient with FH should be offered a high-intensity statin therapy for the aggressive lowering of lipid levels by a recommended 50%. Index cases who have elevated lipids on the basis of the Simon Broome criteria (i.e. the majority of patients) will benefit from statin therapy as they are at a ≥ 20% 10-year risk of cardiovascular disease events.71 We assume that 10% of relatives testing negative on targeted sequencing will also require some cholesterol-lowering therapy. This is an author assumption based on clinical expert opinion and previous NICE guidance and is varied between 0% and 50% in the sensitivity analyses. This refers to the estimated percentage of relatives without an identified mutation who will require treatment on the basis of high cholesterol levels. Such cases receive a low-intensity treatment in the model. As relatives are not clinically diagnosed with FH based on the Simon Broome criteria, it would be inappropriate to treat all patients, as only a percentage will have elevated lipids. The impact of varying this assumption is explored in sensitivity analysis. There is, however, much debate among clinicians over how to treat FH and patients at an increased risk of cardiovascular disease as a result of elevated lipids, with some choosing a ‘start low’ treatment option (starting all patients on a low-intensity statin such as simvastatin 40 mg) and others giving everyone a high-intensity statin (e.g. atorvastatin 80 mg or rosuvastatin). For the base-case analysis, we have assumed a multitreatment regimen for FH patients based on and adapted from the FH clinical audit 2010.18 Patients with a Simon Broome-positive diagnosis but who have no genetic confirmation of FH will receive low-intensity statin therapy to reduce their elevated lipid levels. Such cases (especially those relatives who are false-positive) may also respond adequately to exercise and diet therapy, the effects on quality of life of which are beyond the scope of this assessment. Cole and colleagues73 have conducted a detailed systematic review of the literature to explore the evidence in relation to the effects of dietary and lifestyle interventions in chronic heart disease risk reduction. Also, NICE guidance on dietary interventions in CHD provides additional information in the UK. Personal communications from Dr Anthony Wierzbicki (2011, Guy's and St Thomas' Hospitals NHS Trust) and Dr William Simpson (2011, NHS Grampian) are used in sensitivity analyses to explore the sensitivity of the model to treatment choice in practice.

Resource use estimation

Clinical resource use

For the purposes of this evaluation, we have assumed that all index cases will have received a clinical diagnosis of FH based on the Simon Broome criteria. Resource use and costs associated with this diagnosis are common to all tests being evaluated and so are not included. This is standard economic evaluation practice to include only resource-use estimations which differ between tests under consideration. However, the resource use associated with tests after the initial diagnosis is important and has been considered in the analysis. It is assumed that, once the proband has a genetic test or LDL-C confirmation of FH, he or she will attend a lipid clinic to discuss treatment and lifestyle management of the condition. It is at this point that family pedigree will be identified and contact with relatives will be initiated. It is assumed that initially only first-degree relatives will be contacted as there would be no point in contacting second-degree relatives until a diagnosis was confirmed in first-degree relatives using a genetic screen. Table 21 details resource use and cost estimation for this process based on clinical expert opinion and Hadfield and colleagues.70

TABLE 21. Resource use of health-care professionals for both index and cascade testing of patients after diagnosis.

TABLE 21

Resource use of health-care professionals for both index and cascade testing of patients after diagnosis.

Index cases or relatives diagnosed with FH are offered an annual follow-up appointment with a lipids specialist at an outpatient clinic. In the absence of a specific unit cost tariff for a lipids specialist, this service is assumed similar to a cardiologist appointment (Dr William Simpson, University of Aberdeen, 2011, personal communication) and is costed at £222 per outpatient consultation.

Diagnostic resource use

A new national activity unit has been developed for molecular genetics and cytogenetic tests in the UK. This is based on a weighted report and uses for molecular genetics an amplicon as the base unit. All molecular genetic tests are then assigned a relative number of units that slot into bands with some efficiency built in as the number of amplicons increases. This new methodology for measuring activity for molecular genetic tests was developed by collaboration between the CMGS and the UKGTN. The objective was to devise a transparent and consensus system for measuring molecular test activity that could be implemented by all laboratories. Tests are weighted by complexity so that, for example, simply booking in a sample has the lowest weight and sequencing a gene of over 100 exons, for example RYR2, the highest. All realisable costs of each laboratory are collated and a total cost of the service is then calculated including salaries, consumables, overheads, etc. Each laboratory can derive its own unit cost, based on dividing budget by activity, and thus in effect derive a cost per test. For example, a £1.2M service producing 30,000 MOLUs will have a unit cost of £40.00. This system of pricing has been modelled by most of the laboratories in the UK and has been accepted by the professional bodies and UKGTN as a suitable approach to establishing a national tariff for genetic tests. Details of the national MOLU bands are included in Appendix 11 for information. The MOLU system is not a perfect system of estimating costs, however, and the limitations are outlined in Chapters 1 and 5.

For the base-case analysis, transportation costs of samples (preferably blood samples) for DNA testing and blood samples for LDL-C testing are included. Based on clinical expert advice (Dr Gail Norbury, Guy's Hospital, London, 2011, personal communication), an increasing number of genetics samples are tested by processing saliva samples. Saliva-based samples are less costly to transport as they are more stable and require only first-class postage; however, the kits to extract the DNA are substantially more expensive. These resource use differences, however, will be included in the MOLU consumables mentioned above based on 1 MOLU for DNA extraction. The majority of tests are carried out in the UK; however, LIPOchip may be processed by the manufacturer on site in Spain. The additional resource and transportation costs associated with sending a blood sample overseas via air are considered for the LIPOchip platform processed in Spain. This was assumed to take a cost of 1 MOLU, commonly applied in genetic testing to cost transferring samples to laboratories overseas. Therefore, a cost of £30 has been applied in the base case. Additionally, there may be extra costs associated with resampling an estimated 3% of samples (Progenika, 2011, personal communication). These costs are also incorporated.

Unit cost estimation

Clinical costs

Costs of clinician time for treating patients, identifying a family pedigree, counselling relatives on the importance of their condition and contacting relatives themselves are estimated using Payment by Results (PBR) national tariffs where available (e.g. for a first appointment with a lipid specialist). For all other resource use, including clinical nurse specialist (to identify pedigree and counsel patients), GP time to confirm second LDL-C test and administrator time to contact relatives, costs are estimated using Personal Social Services Research Unit (PSSRU) unit costs of health and social care.75 Costs are based on the median of the appropriate agenda for change pay scale and include overheads, training costs, insurance, annual leave, etc.

Diagnostic costs

Costs of genetic testing strategies vary greatly among laboratories, especially based on their area of expertise and also in relation to their size – the greater the laboratory size, the greater the throughput of samples tested and thus the lower the costs based on economies of scale through mass genetic testing. Laboratories that can keep their budget constant or can reduce it but increase the number of MOLUs produced will have lower unit costs. The incentive then is to reduce the total budget while maintaining or increasing output. This system is simplistic and transparent and is the method adopted by most laboratories in the UK in setting their genetic testing tariffs (Dr Zosia Miedzybrodzka, University of Aberdeen, and Dr Gail Norbury, Guy's Hospital, London, 2011, personal communication). For the purposes of the base-case analysis, it is assumed that the MOLU cost is £30 per MOLU (Dr Kevin Kelly, University of Aberdeen, 2011, personal communication). The cost of each MOLU will be varied in sensitivity analysis provided by Dr Gail Norbury (£33 per MOLU). Unit cost estimation is adjusted within the model for strategies that have more than one test in order to account for the cost differentials associated with earlier positive test identification. The cost of DNA extraction is also incorporated into the analysis and receives a unit of 1 MOLU. Details of MOLU units applied and the associated costs for each test strategy are presented in Table 22. The cost of testing a hypothetical cohort of 1000 index cases with combination strategies is dependent on the numbers testing positive on the first test in that strategy. For example, in a strategy such as Elucigene FH20 followed by CGA for negatives, an index case who tests positive on Elucigene FH20 will not receive the second more comprehensive test.

TABLE 22. Costs applied to each testing strategy in the model.

TABLE 22

Costs applied to each testing strategy in the model.

In addition to the tests outlined above, the LIPOchip platform (Spain) as a genetic testing platform is a potential alternative to CGA. The test, which involves using the LIPOchip followed by sequencing of test-negative cases, is offered by the manufacturer (Progenika) at a cost of €250 for a LIPOchip test and €350 for the whole process. The associated costs are incorporated into the analysis using an exchange rate of €1 = £0.89. The LIPOchip platform processed in Spain is explained in Chapter 1. Briefly, this is a two-stage process whereby, if the sample is positive on LIPOchip, no further testing takes place. If the sample is negative on LIPOchip then the sample is sequenced for an additional €100. Therefore, assuming that the sensitivity of LIPOchip is the same regardless of where it is processed and using similar methodology to that in Table 22, we estimate the total cost of the strategy (before transportation of samples costs) as (1000 × 250 × 0.89) + (713 × 100 × 0.89) = £285,957.

The cost of targeted sequencing may also be estimated using the MOLU system. Targeted sequencing (including DNA extraction) is allocated a MOLU of 3. At a cost of £30 per MOLU, this would amount to £90 per targeted sequencing test. Based on the MOLU system, targeted sequencing is cheaper than Elucigene FH20 and is therefore the strategy of choice for cascading relatives.

Low-density lipoprotein cholesterol concentration measurements will be taken for all members of the study, regardless of testing strategy. Additional measures will, however, be carried out to confirm the diagnosis. Therefore, an additional two LDL-C tests will be required (at least one of which will be a fasting blood sample) to confirm the Simon Broome diagnosis if this is the method of diagnosis being adopted. It is assumed that, in order to get an extra blood test taken for the additional LDL-C measurement, an additional visit to a GP will be required. It is not expected that transportation costs of samples sent to laboratories for analysis will differ significantly between LDL-C and genetic tests as both require the transportation of potentially hazardous blood specimens.

Treatment costs

As discussed in Treatment options to be evaluated and as recommended by CG71,1 treatment will be of either high or low intensity, predominantly with statins. Should a patient be intolerant to statins, treatment may be administered using ezetimibe as per the NICE CG711 guideline. There is, however, some debate as to the relative effectiveness of ezetimibe monotherapy; therefore, only a small proportion of patients are likely to receive this treatment in practice (Dr William Simpson, NHS Grampian, personal communication). Also based on personal communication (Dr Anthony Wierzbicki), ezetimibe as monotherapy is ineffective and patients who have an inadequate response to statins may need to be treated with ezetimibe plus bile acid sequestrants. A number of FH patients will receive polypharmacy incorporating treatment with statins and ezetimibe. Table 23 details the unit costs per year of treatment with all of the potential drugs included in the modelling process with costs sourced from the British National Formulary (BNF).76 To reflect differential treatment practice among clinicians, various combinations of these drugs (based on clinical expert opinions) are explored in the model. The most common combination therapies are included in Table 23.

TABLE 23. Unit costs of drug treatments used in the economic modelling.

TABLE 23

Unit costs of drug treatments used in the economic modelling.

For the base-case analysis, we used data from the FH audit 2010,18 the most up-to-date data source on FH treatment in practice. We also use data from clinical experts (Dr Anthony Wierzbicki, Guy's and St Thomas' Hospitals NHS Trust, 2011, personal communication, and Dr William Simpson, NHS Grampian, personal communication) to conduct sensitivity analysis surrounding the proportions of patients on each treatment as part of either a high- or a low-intensity statin therapy. The cost impact of atorvastatin, which is due to come off patent during the course of this assessment, will have implications for treatment costs in the model. This will be explored in sensitivity analyses.

Costs of cardiovascular events avoided as a result of treatment

Table 24 details the costs of cardiovascular events avoided. For the base-case analysis, these costs have been calculated using weighted averages of all Health Resources Group (HRG) codes pertaining to each cardiovascular event avoided. Elective and non-elective tariffs from PBR data for 2010–1174 are used and weighted for the numbers of elective and non-elective cases sourced from the Hospital Episodes Statistics online database (www.hesonline.nhs.uk/Ease/servlet/ContentServer?siteID=1937&categoryID=192).

TABLE 24. Costs of cardiovascular events.

TABLE 24

Costs of cardiovascular events.

Data sourced from current NICE guidelines1 such as for subsequent MI are not available as part of PBR nor do any national tariff prices exist for these events. Therefore, values have been sourced from CG711 and inflated to current price levels for use in the model. Costs of cardiovascular death or other deaths have been assumed to be equal to £0 as it is not envisaged that this would have cost implications for the NHS. However, such deaths avoided would have great impact on the results from a societal perspective.

List of assumptions

A number of assumptions have been made throughout the modelling exercise and for the base-case model; the impact of each will be explored in relevant sensitivity analyses. Table 25 summarises the main assumptions made throughout the health economic modelling process.

TABLE 25. List of major assumptions, justification and method for dealing with associated uncertainty.

TABLE 25

List of major assumptions, justification and method for dealing with associated uncertainty.

Data analysis

Base-case analysis

For the base-case analysis, we analyse an index patient of age 50 years, with an assumed average first-degree relative age of 50 years. The decision model is run on the basis of a hypothetical cohort of 1000 patients with a clinical diagnosis of FH based on the Simon Broome criteria (including both definite FH and possible FH). Cost and QALY values are estimated as described in the preceding sections and applied to the number of people passing through each branch of the decision tree illustrated in Figure 9. On the basis of test accuracy, a proportion of all 1000 index patients are positive (true-positive or false-positive) or negative (true-negative or false-negative).These patients are assigned the relevant cost and QALY values as described and total costs and QALYs are generated for the full cohort.

Test strategies are ranked in ascending order of cost. Those strategies that are more costly and less effective are excluded on the basis of simple dominance. Additional tests that are dominated by a combination or two or more alternative strategies are excluded by extended dominance. ICERs are calculated as incremental costs divided by incremental QALYs between non-dominated strategies. This is the most common method of presenting ICERs and relates the options sequentially ranked by costs. For the purposes of this assessment, the most relevant comparators are:

  1. CGA, recommended indirectly by NICE guidance CG71.1
  2. LDL-C concentration measurement only. The reason for this is that, in practice, LDL-C is the main method of identification presently adopted in the UK (although genetic testing is more common in Scotland, Wales and Northern Ireland than in England).

Therefore, ICERs are presented as cost per QALY compared with the two suggested reference standards for this evaluation (LDL-C and CGA).

This process is applied to two distinct research questions. First, we investigate the cost-effectiveness of each of the 12 strategies for index cases alone. However, of greater importance and thus the primary focus of the analysis is to present cost-effectiveness estimates for the complete process of index case confirmation of clinical diagnosis but also for the identification and testing of relatives (i.e. the whole cascade testing process).

Subgroup and additional scenario analysis

The cost and QALY results for different age groups are explored in this section for the full cascading project only (i.e. index and relative cases). Results for index cases alone are presented in Appendix 12.

These subgroup analyses include a range of age profiles and also include the incorporation of any available evidence relating to the efficacy of statins in the treatment of children. To this end, we have completed a structured search of the literature, which has identified four systematic reviews of the efficacy of statins in children, the most recent of which is a Cochrane review of high quality that is used to inform the discussion and the model.77 The data suggest that statins are efficacious in children in reducing cholesterol and have non-significantly different adverse events to placebo. Therefore, statins are likely to be safe in children with FH although long-term follow-up of this patient group is required. As data relating directly to CHD are lacking, treatment effect relative to CHD is assumed to be similar to that of a young adult (equivalent to a 30-year old index case in the economic model).

A number of age-specific subgroups were considered (probands aged 15, 30, 50, 65, 75 and 85 years). These age subgroups are similar to those used in previous economic modelling for FH1 and represent a good distribution of the ages of the population who may present for testing. Table 26 details the calculated number of relatives for each index case and their average age used in the model.

TABLE 26. Details of index case age and associated number and age of relatives.

TABLE 26

Details of index case age and associated number and age of relatives.

As discussed in Model structure, there may be alternative estimates of cost-effectiveness based on whether the index case is identified as definite FH or possible FH as their clinical diagnosis. It should be noted, however, that because of a lack of sensitivity data for each test separated into definite FH and possible FH subgroups, it was not possible to conduct robust analyses of FH cases split by clinical diagnosis subgroup. We have, however, conducted threshold analyses which show the combination of mutation prevalence and test sensitivity that would be required for the candidate test to be considered cost-effective as a pre-screen to CGA. The probabilistic sensitivity analysis accounts for the combined variation in all of the input parameters.

Sensitivity analyses

As many assumptions are made throughout the modelling process and selective data are chosen to inform the parameters, it is possible that the results generated will be sensitive to some of the judgement calls, assumptions and decisions made in the analysis. Therefore, we carry out a range of sensitivity analyses to determine the sensitivity of the base-case results to changes in our assumptions. A range of univariant deterministic analyses are presented in Appendix 14, the main results of which are reported and discussed in Analysis of uncertainty, including probabilistic sensitivity analysis. In addition, a probabilistic sensitivity analysis is also presented to explore uncertainty in the model.

Areas of uncertainty that are explored include:

  1. Prevalence rates of FH-causing mutations among clinically diagnosed index cases and at-risk relatives.
  2. Treatment differences for those with genetically confirmed FH and those without a genetic confirmation. The implication of forthcoming price reductions of atorvastatin is also explored.
  3. Uncertainty surrounding the proportion of probands and relatives with a given test result receiving treatment (e.g. the proportion of those with a false-negative or true-negative test result receiving statin therapy).
  4. The costs of diagnostic strategies, especially issues of uncertainty surrounding the MOLU pricing system and the likely cost of a 1-unit MOLU output.
  5. Key assumptions relating to the model structure, including cascade testing only of first- and second-degree relatives, discount rates applied to costs and effects, the impact of not cascade testing negative index cases and the proportion of index and relative cases agreeing to participate in the identification and testing process.
  6. Uncertainty associated with assumptions listed in Table 25 including structural assumptions regarding management of negative-testing index and relative cases.

Probabilistic sensitivity analysis

Deterministic one-way sensitivity analyses and point estimates of ICERs do not adequately provide information on the true impact of uncertainty surrounding the model parameters. Because of imperfect information on both the resource use and effectiveness of each treatment strategy, costs and QALYs are highly likely to be subject to at least some degree of uncertainty. Therefore, we conducted additional probabilistic sensitivity analysis using Monte Carlo simulation (5000 repetitions). Distributions were fitted to each of the parameters based on published studies (where available), CG71 data1 and a number of assumptions where no data were available. For example, where insufficient data existed in published sources to fit distributions to parameters, standard errors were assumed in order to calculate alpha and beta values. This may slightly under- or overestimate the variation in some of the parameters; however, it is not likely to impact greatly on resultant cost-effectiveness acceptability curves (CEACs). For sensitivity of test strategies (Elucigene FH20 and LIPOchip) the analysis was bounded by the highest and lowest reported mean values in all of the studies identified from the systematic review of the literature. Full details of probabilistic sensitivity analysis parameters are presented in Appendix 16.

The net benefit framework was used to estimate net monetary benefits for each simulation as described in Briggs.78 The defining characteristic of this approach is that all strategies add to a probability of cost-effectiveness equal to 1. This uncertainty is illustrated in the form of CEACs for each of the non-dominated strategies of testing. CEACs for the base-case analysis are presented in the text, with supplementary analyses following the same approach for each age subgroup in the model presented in Appendix 15 for completeness. The analysis is presented for non-dominated test strategies only. The comparison for the calculation of incremental costs and QALYs for this analysis is LDL-C as this is current practice in the NHS. As the remit of this report is primarily to assess the cost-effectiveness for index cases and relatives, we have not conducted probabilistic sensitivity analysis for index cases alone. In addition, CEACs are presented for 5%, 10%, 20% and 50% mutation prevalence rates in order to reflect the uncertainty surrounding mutation detection rates in various subgroups of the population, primarily varying based on ethnic background.

Results of economic analysis

Results presented for the base-case analysis are subject to the assumptions listed in Table 25.

Summary of test results for a hypothetical cohort of 1000 familial hypercholesterolaemia patients

Table 27 details the flow of a hypothetical cohort of 1000 patients through the model based on those testing false-positive, true-positive, false-negative and true-negative. The values for sensitivity and specificity are combined values for all definite FH or possible FH patients based on the Simon Broome criteria. We have used estimates of sensitivity and specificity derived from the clinical effectiveness review and applied these to the model as discussed in Methods for economic analysis.

TABLE 27. Test results for a hypothetical cohort of 1000 patients by testing strategy for index cases and cascading of both test-positive and test-negative relatives.

TABLE 27

Test results for a hypothetical cohort of 1000 patients by testing strategy for index cases and cascading of both test-positive and test-negative relatives.

Assumptions relating to the incidence of FH among the tested population are discussed in Model structure; however, for the base-case model we assume a mutation detection rate of 36.5%37 among people who are either possible or definite FH using the Simon Broome clinical diagnosis (i.e. approximately one in three reporting for testing will test positive based on CGA). Sensitivity analysis explores the variation in this estimate.

As we have assumed that each genetic test is associated with specificity equal to 1, there are no false-positives for the base-case analysis. However, data suggest that LDL-C among index cases has a specificity of 0.29,44 indicating that a substantial number of positive test results will in fact be false-positives.

Mean cost and mean treatment effects associated with each diagnostic strategy

Index (proband) familial hypercholesterolaemia patients

Table 28 presents total costs and total QALYs for each treatment strategy for index cases alone ranked according to cost with dominance or otherwise indicated. Patients without FH will have a slightly longer survival prognosis and will thus receive slightly greater QALY gains than those with FH. Such patients are clinically diagnosed as having FH, have high lipid levels and are at an increased risk of CHD and so will have a positive response to cholesterol-lowering therapy.

TABLE 28. Total costs, total QALYs and sequentially presented ICERs for the identification of index cases.

TABLE 28

Total costs, total QALYs and sequentially presented ICERs for the identification of index cases.

The Elucigene FH20 diagnostic strategy alone generates the lowest costs for identifying index patients for two reasons: first, it is the cheapest genetic diagnostic test available and, second, it detects the lowest number of true-positive index cases. Therefore, it confirms the clinical diagnosis in the fewest index cases with FH and for that reason is associated with the lowest QALYs of all of the tests included. LDL-C identifies the largest proportion of positives (not necessarily true-positives for FH – although all index cases are technically true-positives based on their clinical diagnosis) and has the highest QALY gain as it detects the greatest number of patients at increased risk of CHD. Of the non-dominated sequences, LIPOchip platform (Spain) and CGA are both associated with ICERs between £20,000 and £35,000 per QALY gained.

Index cases and relatives

However, where genetic testing has the greatest advantage over LDL-C is in the identification of relatives for cascade testing. Cascade testing using LDL-C alone is less likely to be cost-effective in this population because of the large number of false-positives (including relatives incorrectly identified as having FH) who may be treated using high-intensity statin therapy when low-intensity therapy would have sufficed to reduce their cholesterol levels. Assumptions regarding false-positive relatives are detailed in Table 25 and can be tested in sensitivity analysis within the model. Table 29 presents total costs and QALYs for index and relative cases combined (i.e. a whole integrated strategy for the identification and management of index cases and relatives with FH). Cascade testing of relatives of an index case with an identified mutation is by targeted sequencing. This is because targeted sequencing is less costly than both of the other candidate tests. Therefore, as all tests would detect the identified mutation they are supposed to in the relatives, targeted sequencing is the most cost-effective way to do this in relatives of a mutation-positive index case.

TABLE 29. Total cost and QALY implications for index and relative cases.

TABLE 29

Total cost and QALY implications for index and relative cases.

In the analysis presented in Table 29, LDL-C is the least effective of all tests. Elucigene FH20 is the least costly genetic testing strategy and is also the most cost-effective of all non-dominated genetic testing strategies relative to LDL-C, being less costly, more effective and thus dominant. CGA is the most effective non-dominated strategy in terms of QALYs gained, with an associated ICER of £2135 per QALY gained relative to the next most effective non-dominated strategy (LIPOchip platform, Spain). Combination genetic tests are dominated by single genetic test strategies. For example, Elucigene FH20 followed by LIPOchip is dominated by LIPOchip alone-meaning that the extra cost of pretesting with Elucigene FH20 does not add any additional QALYs over and above LIPOchip. The reason for this is that LIPOchip will detect the same mutations and cost more when added to Elucigene FH20. A similar argument can be made for tests used for pre-screening prior to CGA. The case for test strategies including MLPA as a component is slightly different in that MLPA detects deletions and duplications of the gene and so detects approximately an extra 5% of mutations that would not otherwise be detected by Elucigene FH20. MLPA is incorporated and included in the CGA process and has not been considered separately here. In relation to LIPOchip, there is some uncertainty in relation to the detection of deletions and duplications of the gene. Therefore, we have taken a pragmatic approach and included LIPOchip alone and LIPOchip followed by MLPA. This will allow the reader to decide, based on further investigation of LIPOchip, whether or not MLPA would be required to obtain a definitive diagnosis among positive test results (i.e. a specificity of 1) as assumed in the model.

Incremental analysis for reference case and other scenarios

Index cases

Tables 30 and 31 evaluate the non-dominated sequences compared with the relevant comparators for index cases. The scope and protocol for this assessment define two important comparators: (1) the comparator recommended as part of the NICE clinical guidelines – full genetic DNA testing (or CGA as defined in our protocol) and (2) LDL-C, which is currently the most commonly used method as part of the Simon Broome criteria to identify FH in practice in the UK. Currently, DNA testing is available only in 15% of UK primary care trusts (UK FH audit project 201018) and therefore LDL-C is deemed an appropriate comparator based on current clinical practice in the UK (NICE diagnostic advisory group, 2011, personal communication).

TABLE 30. Comparison of non-dominated sequences vs LDL-C (index cases only).

TABLE 30

Comparison of non-dominated sequences vs LDL-C (index cases only).

TABLE 31. Comparison of non-dominated sequences vs CGA (index cases only).

TABLE 31

Comparison of non-dominated sequences vs CGA (index cases only).

Of the non-dominated sequences, LDL-C is the most costly and most effective test overall (see Table 30). Elucigene FH20 is the least costly but also the least effective test in terms of QALYs. In fact, all of the non-dominated testing strategies are cheaper overall and generate fewer QALYs than LDL-C. Although diagnosis costs for LDL-C are lower than the alternatives presented, treatment costs are much higher. This is because as all index patients will technically have FH based on their clinical diagnosis on the Simon Broome criteria, they will benefit from statin therapy. Additionally, even if they were not true FH, they would still be at an increased risk of coronary artery disease based on their cholesterol levels and so would benefit from treatment. LDL-C is therefore also associated with the greatest number of QALYs gained for index cases. This is because, should a negative diagnosis be based on a genetic test, patients who test false-negative may be inappropriately treated and would thus gain fewer QALYs than if they were prescribed high-intensity treatment for their FH based on LDL-C levels.

Table 31 presents similar information for index cases alone when the comparator of interest is CGA.

When the comparison of interest for index cases alone is CGA, all other non-dominated genetic tests are less costly and less effective than CGA. The question for a decision-maker in this scenario would thus be whether or not the cost savings are worth the associated QALY loss. ICERs are not reported in informing such a question as there is lack of evidence regarding how much society is willing to accept in compensation (in the form of cost savings) for a QALY loss.

Figure 10 presents the cost-effectiveness plane comparing all tests for index cases. This confirms the results alluded to in the tables of results above.

FIGURE 10. Cost-effectiveness plane (index cases).

FIGURE 10

Cost-effectiveness plane (index cases).

There are two important things to note from this illustration. First, LDL-C is the most costly strategy (driven by high-intensity statin treatment costs). As all patients are at risk of cardiovascular disease, however, QALY gain is highest driven by the extra-intensive treatment based on false-positive diagnoses of FH by LDL-C. These patients benefit from the increased statin therapy as they are at increased risk of cardiovascular disease based on their cholesterol levels. Second, the graph illustrates the dominance of single test strategies over similar strategies preceded by less-sensitive screening tests such as Elucigene FH20. As all strategies ending in CGA generate the same QALY gains, dominance is due to greater costs amongst multiple test strategies.

Confirmation of clinical diagnosis in index cases and cascade testing of relatives

Tables 32 and 33 evaluate the non-dominated sequences compared with the relevant comparators for the full process of index case confirmation of the clinical diagnosis and cascade testing of relatives. The comparators are LDL-C and CGA as in the preceding section.

TABLE 32. Comparison of non-dominated sequences vs LDL-C (identification of index cases and cascade testing of relatives).

TABLE 32

Comparison of non-dominated sequences vs LDL-C (identification of index cases and cascade testing of relatives).

TABLE 33. Comparison of non-dominated sequences vs CGA (identification of index cases and cascade testing of relatives).

TABLE 33

Comparison of non-dominated sequences vs CGA (identification of index cases and cascade testing of relatives).

Multiple testing strategies are dominated by single testing strategies generating the same sensitivity and test-positive rate overall. All non-dominated genetic tests are highly cost-effective compared with LDL-C in the identification of index cases with FH and cascade testing of relatives (assuming that society's willingness to pay for a QALY gain is £20,000). The Elucigene FH20 single test strategy is the most cost-effective, being less costly and more effective and thus dominant over LDL-C (see Table 32). However, should a decision-maker wish to have a DNA test with a definitive genetic diagnosis, (i.e. CGA) then this is more expensive but generates the most QALYs gained compared with LDL-C. Relative to LDL-C (current practice), CGA could be considered a cost-effective testing strategy with an associated ICER of only £1030 per QALY gained. This is also well below a willingness-to-pay value of £20,000 per QALY gained.

When cost and QALY pairs are compared with CGA (current NICE recommendations) for the whole process of identification of index cases and cascade testing of relatives, all non-dominated tests are less costly and less effective than CGA. Table 33 presents this comparison.

Again, as discussed previously, the reporting of ICERs for this scenario does not inform the question of what reduction in QALYs a decision-maker is willing to accept in order to achieve a predefined cost saving. All non-dominated testing strategies are less costly and also less effective than CGA.

Figure 11 presents the cost-effectiveness plane comparing all tests for index cases and cascade testing of first-, second- and third-degree biological relatives. This confirms the results alluded to in the tables of results above.

FIGURE 11. Cost-effectiveness plane for index cases and cascade testing of relatives.

FIGURE 11

Cost-effectiveness plane for index cases and cascade testing of relatives.

In this scenario (including cascade testing of relatives in the analysis), LDL-C used as a method of identification of relatives is less costly than all other tests (with the exception of Elucigene FH20) but does not generate the same QALY gains as any of the genetics-based tests. LDL-C is an inexpensive test to carry out (relative to other more costly genetic options); however, LDL-C alone will falsely diagnose many index cases as having FH and hence many relatives will be cascade tested unnecessarily for fewer QALYs gained. LDL-C is thus dominated by the lower-cost Elucigene FH20 test.

Differential results for subgroups

The impact of varying the age of the index case and associated average age of relatives is explored in this section. As with the base-case analysis, results are presented sequentially and also incrementally relative to LDL-C and CGA. Analyses for index cases only are presented in Appendix 12.

For index cases alone, the results are quite difficult to interpret and there appears to be much variability in the ICERs depending on age (see results tables in Appendix 12). As in all other analyses, all pre-screen tests are dominated by more effective tests that generate cost savings due to treatment effects. For all index case ages, non-dominated test strategies appear to be less costly and less effective than LDL-C, with the exception of an 85-year-old index case, for which genetic tests are dominant over LDL-C. These results should, however, be interpreted with caution. The wide variability in the presented ICERs is due to small or indeed negligible QALY differences between strategies. This is because, for index cases alone, most if not all patients will be at risk of cardiovascular events and all will have a clinical diagnosis of FH.

Genetic testing has the advantage in the identification and treatment of relatives through the cascade testing process for all age subgroups and this is evident from associated tables (for index cases and relatives combined) reported in Appendix 13 and discussed in the following paragraph.

The results presented suggest that, as in the base case, all pre-screening strategies are dominated by single test strategies detecting the same number of people, regardless of age. The reason for this is that costs associated with savings on test-positive cases are offset by submitting a whole cohort of negative patients through two or maybe three tests. As only a proportion will have a genetic mutation, these additional costs outweigh cost savings from those tested positive on pre-screens such as Elucigene FH20 or LIPOchip. This confirms the base-case results presented in Tables 29, 32 and 33. As reported for the base case in Table 29, relative to LDL-C, Elucigene FH20 is the most cost-effective option for all age groups analysed, with all ICERs under £1400 per QALY gained. This probably represents a highly cost-effective use of NHS resources. The next most cost-effective testing options after Elucigene FH20 are LIPOchip (platform processed in Spain), for which the costs per QALY gained are between £714 and £2513 irrespective of age group analysed, and CGA, with ICERs only slightly higher than those of LIPOchip (platform processed in Spain). Therefore, as in the base case, there are a number of tests that could be deemed cost-effective, all with very low ICERs relative to LDL-C. As discussed, should we wish to achieve a definitive diagnosis and generate the greatest QALY gain then CGA is a cost-effective means to achieve such an objective.

Summarising these results together, all of the age group analyses are consistent with the conclusions of the base-case analysis for an index case aged 50 years. Therefore, one may conclude that the conclusions of the model for index and relative cases are not sensitive to the age of the index case or associated relatives. CEACs based on probabilistic sensitivity analysis of the age subgroup results show some uncertainty at low threshold values of willingness to pay but, at threshold values > £5000 per QALY gained, CGA is the most likely cost-effective testing strategy, increasing to almost 100% as the threshold value increases towards a threshold ceiling ratio of £20,000 per QALY gained. CEACs reporting these results are presented for illustration in Appendix 15.

Because of a lack of good-quality data differentiating the sensitivities of the tests for definite and possible FH, we have conducted a threshold analysis indicating the prevalence and sensitivity that would be required for the candidate tests to be cost-effective as a pre-screen for CGA. Additional sensitivity analysis around the maximum and minimum values of all reported studies is presented in the following section. At the current estimate of sensitivity of Elucigene FH20, there would need to be an underlying prevalence of mutations of 61% at current prices of CGA. Should the price of CGA drop in the future as a result of next-generation sequencing, the required prevalence of underlying mutations would need to be 93%. This is based on an assumed price reduction of 40% in the cost of DNA sequencing in the future. The results for LIPOchip are less favourable at current levels of sensitivity as the lower cost of Elucigene FH20 followed by CGA would dominate LIPOchip followed by CGA at high prevalence rates, irrespective of whether or not we apply a cost reduction of 40% to DNA sequencing as part of CGA.

From an alternative perspective, one may be interested in the sensitivity of Elucigene FH20 and/or LIPOchip that would be required to generate cost savings as a pre-screen to CGA at current levels of mutation prevalence. Elucigene FH20 would be required to have a sensitivity of at least 73% to be a cost-saving pre-screen to CGA for a mutation prevalence rate of 36.5% as used in the base-case economic model. LIPOchip would not be cost-effective as a pre-screen to CGA for any plausible sensitivity values at this mutation prevalence level. Plausible values are defined as those sensitivities below the sensitivity of CGA. The reason for this is that, because of the relatively low prevalence of mutations, even at a sensitivity of 90%, only 33% of cases would be positive, with the remaining 67% requiring CGA to confirm the presence or otherwise of a FH-causing genetic mutation.

Therefore, if the goal is to gain an unequivocal diagnosis, for low mutation prevalence rates, pre-screening with Elucigene FH20 or LIPOchip prior to CGA is not cost-effective. At high prevalence rates, > 61%, Elucigene FH20 may offer a cost-effective option; however, this is less likely once the costs of next-generation sequencing fall.

Analysis of uncertainty, including probabilistic sensitivity analysis

One-way deterministic sensitivity analyses

A range of one-way deterministic analyses are presented to investigate the sensitivity of the model to uncertainty in some of the key parameters and in relation to model structural assumptions as outlined in Table 25. All deterministic sensitivity analyses were carried out on the base case of an age 50 years index case.

A full range of sensitivity analyses in relation to treatment effect have been carried out previously in the NICE clinical guidance CG71.1 The model was found to be insensitive to a range of sensitivity analyses including assumptions surrounding nurse and consultant time with patients, costs of cholesterol testing, costs of letters to relatives for cascading, cascading from alternative numbers of relatives (first and second degree), relative risks of non-cardiovascular disease deaths and treatment effect used in the model. As data from the CG711 assessment have been updated and used for the purposes of this report, it is highly unlikely that these sensitivity analyses will have any impact on the sequences of ICERs for this analysis. We have additionally explored the impact of including a cost of £80 (standard A&E tariff) for those patients who die in the model. This is to reflect any additional costs that may be involved over the £0 assumed in the base-case analysis. The results are not sensitive to this assumed value.

Therefore, the focus of sensitivity analyses for this assessment centres on parameters and assumptions that we hypothesise may have an impact on the sequence of ICERs or on the overall cost-effectiveness conclusions. Many parameters alter the cost-effectiveness of identifying index cases alone; however, as the remit for this report is primarily the detection and treatment of relatives with FH, we focus mainly on analyses that affect the overall outcome (i.e. the confirmation of index cases and the cascade testing of at-risk relatives). Full analyses for both groups are included in Appendix 14 for information. In the appendix, results for the index case analysis are presented first, followed by results for index cases and relatives together. The order of tables follows the sequence of results presented below.

The following discussion refers to index cases and relatives together.

Prevalence of familial hypercholesterolaemia-causing mutations among index cases and relatives

Prevalence of FH-causing mutations among index cases is varied between 28%79 and 52%.41 For both low and high estimates of mutation prevalence, the order of the ICERs remains unchanged compared with the base case. Elucigene FH20 remains the most cost-effective strategy relative to LDL-C (associated ICERs = dominant and £395 per QALY gained for low and high estimates respectively). The next most cost-effective options after Elucigene FH20 are LIPOchip (platform processed in Spain) and CGA for both low and high mutation prevalence rates with all ICERs < £1300 per QALY gained. See Differential results for subgroups for a threshold analysis estimating the prevalence required for Elucigene FH20 or LIPOchip to be deemed a cost-effective pre-screen to CGA.

Prevalence of FH-causing mutations among relatives of index cases is an uncertain parameter that is generally held to be approximately 50%, based on the logic that one out of every two offspring will inherit a genetic mutation. Sensitivity analysis varied this assumption by ±20% to between 40% and 60% of first-degree relatives inheriting the culprit gene (author assumption). A low estimate suggests that Elucigene FH20 is dominant, being less costly and generating more QALYs than LDL-C. After that, as in the base-case analysis, LIPOchip platform (processed in Spain) and CGA remain the next most cost-effective testing strategies. A higher estimate of mutation prevalence among relatives of 60% suggests the same three non-dominated test strategies as in the base case, all with ICERs of < £1200 per QALY gained. Therefore, as similar tests are recommended as being cost-effective for all prevalence values considered, the base-case conclusions remain insensitive to any assumptions surrounding prevalence rates in either index cases or relatives with all ICERs for non-dominated strategies < £1300 per QALY gained relative to LDL-C.

Familial hypercholesterolaemia treatment

Analyses reducing the cost of atorvastatin did not change the base-case conclusions, with no difference in the sequence of the presented ICERs. Elucigene FH20 remains the most cost-effective option relative to LDL-C; LIPOchip (platform processed in Spain) is the next most cost-effective option followed by CGA, as was reported in the base-case analysis. ICERs for all three non-dominated tests are insensitive to changes in the cost of treatment used in the model (all reported ICERs are < £1100 per QALY gained relative to LDL-C).

Data in relation to the base-case model sourced treatment proportions for FH from the FH audit 201018 and assumed generic simvastatin treatment for those without confirmed FH. This assumption was tested using treatment proportions provided by Dr Anthony Wierzbicki (personal communication, Guy's and St Thomas' Hospitals NHS Trust, 2011). This assumed that both genetically confirmed FH and genetically non-confirmed FH patients would receive a range of treatments. This included polypharmacy for some patients including treatment with ezetimibe as well as statins. The order and magnitude of the ICERs relative to LDL-C remain similar to that in the base-case analysis.

The conclusions drawn are therefore not sensitive to changes in treatment pattern or to costs of treatment administered to patients.

The impact of the decision to treat negative-testing relatives or index cases

The base-case analysis assumes that 10% of negative-testing relatives will require treatment. However, it may be that 0% or at least no more than in the general population will require treatment. Therefore, sensitivity analysis investigates a scenario in which none of these relatives would receive statin therapy. In this scenario, the magnitude and order of the ICERs are very similar to those in the base-case analysis, with Elucigene FH20 remaining the most cost-effective strategy, dominating LDL-C. LIPOchip (platform processed in Spain) and CGA are the next most cost-effective options (ICERs of £902 and £1062 per QALY gained, respectively, relative to LDL-C). Hypothetically increasing this proportion to 50% does not lead to any significant change in the order or magnitude of the ICERs presented.

In an unlikely situation that negative index cases do not receive treatment or clinical follow-up, Elucigene FH20 is the only non-dominated genetic testing strategy and is less costly but less effective than LDL-C, the reason being that index cases testing negative for a FH-causing genetic mutation are still at significant risk of cardiovascular events and so not treating based on genetic mutation alone would lead to large numbers of at-risk individuals being missed, hence the reason LDL-C would be the most cost-effective strategy. It is important to note, however, that the above-mentioned analysis is for illustration only and is not necessarily a reflection of the true care pathway.

Costs of diagnostic strategies

Increasing or decreasing the MOLU costs associated with each test by ±£10 (varying cost per MOLU from £20 to £40) does not impact on the overall test order, with only minimal changes in the relevant ICERs. This is because the model is determined primarily around lifelong costs and health outcomes associated with treatment for FH or otherwise.

Sensitivity of key assumptions (model structure)

The assumption that cascade testing takes place of first-, second- and third-degree biological relatives of the index case is tested by assuming that the process stops after the second-degree relative regardless of test result. All genetic tests are even more cost-effective in this scenario. Elucigene FH20 is less costly and generates greater QALYs than LDL-C and is thus dominant. LIPOchip (Spain) and CGA are both associated with ICERs of < £800 per QALY gained.

The base-case analysis assumes that all index patients with a clinical diagnosis will have their family pedigree investigated, with cascade testing using targeted sequencing for relatives of genetically confirmed index cases. However, those that do not receive a genetic test or are test-negative will still be cascade tested using LDL-C. This is because, although a genetic mutation may not be detected, it is possible that such individuals have mutations or genes that have not yet been identified as causing FH. However, as a sensitivity analysis, we have explored the impact on the results of not cascade testing from such genetically test-negative index cases. In this scenario, all non-dominated genetic tests are actually less costly and less effective than LDL-C testing. Although the results are sensitive to this aspect of the model, clinical advice suggests that this would be highly unlikely in practice as cascade testing from negative index cases is a very important part of the cascade process. The results are not sensitive to assumptions regarding the proportion of index and/or relatives agreeing to have their family history investigated or agreeing for cascade testing to take place.

We varied the discount rate between 0% and 6% for costs and benefits as is standard practice in economic modelling to test our model to assumptions regarding uncertainty surrounding the value of future costs and health gains accrued over a lifetime horizon. For a discount of both 0% and 6% the order of the ICERs relative to LDL-C remained the same as in the base-case analysis. The magnitude of these ICERs showed no significant changes either. The results for the base-case analysis present estimates of cost-effectiveness based on current costs of CGA. However, the cost of genetic DNA sequencing will fall in the coming months and years with the development of next-generation (non-Sanger based) sequencing techniques. Therefore, we have explored the impact on the results of reducing the cost of sequencing by an estimated 40% (Dr Gail Norbury, Guy's Hospital, London, 2011, personal communication). In this scenario, LIPOchip (platform processed in Spain) becomes extendedly dominated. Elucigene FH20 is dominant and CGA is associated with an ICER of £995 per QALY gained relative to LDL-C.

Sensitivity relating to diagnostic test accuracy

For each of the main tests we have investigated the cost-effectiveness based on studies reporting the highest and lowest sensitivity values for Elucigene FH20 and for LIPOchip. This gives a greater picture of the uncertainty across studies and the impact on associated cost-effectiveness results. It also reflects the sensitivity of our analyses to different population groups, some of whom may have greater sensitivity on Elucigene FH20, with others doing better with LIPOchip.

In relation to Elucigene FH20, the upper limit of the sensitivity analysis (0.5238) increases the ICER associated with Elucigene FH20 relative to LDL-C. This suggests higher proportionate increases in costs relative to proportionate increases in QALYs, thereby increasing the ICER between the two tests. Lower estimates (0.28636) work in the opposite direction and lead to Elucigene FH20 being dominant over LDL-C. Such findings are somewhat counterintuitive, with there usually being a positive relationship between higher test sensitivity and improvements in cost-effectiveness. The situation here, however, is more complex because of the clinical benefit (and QALY gain) of LDL-C at minimal cost as well as the addition of cascade testing. Higher sensitivity tests lead to a greater number of positive relatives being given a targeted sequencing test (which is more expensive). Although this test detects more true FH cases and generates greater QALY gain, this is offset somewhat by the advantages of LDL-C (individuals will gain improvements in QALYs regardless of whether or not they have FH, through statin-based therapy for their high cholesterol) that form part of the comparator testing. A similar situation arises with LIPOchip strategies relative to LDL-C. However, in all of these analyses, the rank ordering of Elucigene FH20, LIPOchip and CGA in terms of effectiveness and cost-effectiveness remains the same. As the sensitivity of Elucigene FH20 and LIPOchip increases, their associated ICERs approach that of CGA.

The sensitivity and specificity of LDL-C among relatives are taken from Starr and colleagues49 and varied according to the upper and lower bounds of the 95% CIs. In both scenarios, Elucigene is the most cost-effective option relative to LDL-C. ICERs tend to be slightly lower for genetic tests using the higher bound of the CI for sensitivity and slightly higher for the lower bound. These differences are, however, small in magnitude and the counterintuitive effect of test sensitivity in relation to the ICER can be explained as discussed above. Similar analysis of the specificity of LDL-C among relatives does not alter the sequences of the ICERs or the conclusions drawn from the relevant comparisons. Again, all non-dominated sequences are highly cost-effective relative to LDL-C.

In conclusion, based on the above analyses, the results show some sensitivity to changes in some parameters and structure for the confirmation of index cases alone, but are more robust to variations in key parameters when index cases and relatives are analysed together. In all scenarios presented, Elucigene FH20, LIPOchip (Spain) and CGA are cost-effective uses of NHS resources relative to LDL-C. There is some uncertainty surrounding the direction of movement of the ICER as a result of changes in the sensitivity of the tests that may seem counterintuitive. The results of the one-way sensitivity analyses should therefore be interpreted with caution and, for a more accurate measure of overall model uncertainty, probabilistic sensitivity analysis is likely to offer a better estimate.

Probabilistic sensitivity analysis

Probabilistic sensitivity analysis is carried out for the base case as described in the methods section. Figure 12 illustrates the results in the form of a CEAC.

FIGURE 12. Cost-effectiveness acceptability curves for the base-case analysis.

FIGURE 12

Cost-effectiveness acceptability curves for the base-case analysis.

This figure shows that, at low threshold values of willingness to pay for a QALY gain relative to LDL-C (< £2500), Elucigene FH20 has the highest probability of being cost-effective, but this reduces as the willingness to pay for an additional QALY increases. CGA is the most cost-effective option at threshold values > £2500 and is associated with a > 90% probability of being cost-effective at all threshold values > £3500 per QALY gained (Figure 12 is scaled down to aid discussion of low threshold values). The other non-dominated strategy, LIPOchip platform processed in Spain, is never associated with a probability of cost-effectiveness > 20%. Probabilistic analysis also generates similar results and conclusions for each age subgroup in the analysis (see Appendix 15). At threshold values of willingness to pay for a QALY gain approaching £20,000, CGA is always the most cost-effective option. This is an important point and confirms the generalisability of the base-case probabilistic results to other age groups.

In addition to deterministic analysis surrounding the mutation detection rate among clinically diagnosed FH patients, we considered some extra analysis based on input from Dr Anthony Wierzbicki (Guy's and St Thomas' Hospitals NHS Trust, 2011, personal communication), who states that, among his patient group, the majority of patients are possible FH and he estimates that the proportion likely to be detected by CGA is approximately 20–25% or may even fall to 5% in some population groups. With this in mind, we have conducted probabilistic sensitivity analysis for a range of potential mutation detection rates for CGA. The relevant CEACs are presented in Appendix 17 and show that the results are somewhat sensitive to this value in the model, especially at low threshold values and for lower rates of prevalence. For the lowest prevalence rate considered (5%), there is quite a bit of uncertainty at threshold values < £6000 per QALY gained. At very low values of willingness to pay (< £3000 per QALY), Elucigene FH20 is the strategy most likely to be cost-effective. LIPOchip platform processed in Spain is less likely to be cost effective except at very specific values of willingness to pay between £3000 and £4000 per QALY gained and at a low prevalence rate of 5%. However, this test is never associated with a probability of cost-effectiveness of > 50% regardless of prevalence rate or willingness-to-pay threshold. For higher estimates of prevalence (i.e. 10–50%), the results mirror those of the base-case analysis. However, of greater importance is that, for all prevalence rates of FH considered, CGA is the most cost-effective strategy at threshold values of > £5000 per QALY gained, increasing to 70% at the conventional value of willingness to pay of £20,000 per QALY gained for a prevalence of 5%. This probability increases to almost 100% for all other prevalence rates considered (i.e. 10%, 20% and 50%). Therefore, although there is some uncertainty surrounding the results based on varying mutation detection rates in clinically diagnosed index cases, probabilistic analysis shows CGA to be the most likely cost-effective use of NHS resources. The conclusion of the cost-effectiveness of CGA confirms the results of the previous NICE guidance in that the most comprehensive test for FH is cost-effective. NICE CG711 estimated that CGA was cost-effective with an associated ICER of £2676 per QALY gained versus LDL-C. Our results generate similar conclusions with a lower estimate of the ICER of £1030 per QALY gained relative to LDL-C. This is likely to be because of the cost reductions in CGA and in treatment over time.

Summary

Base-case results from deterministic analyses show that Elucigene FH20 is the most cost-effective diagnostic test, being less costly and more effective and thus dominant over LDL-C. However, this test strategy is less effective than recommended alternatives such as CGA (the most comprehensive diagnostic test for FH). Other non-dominated test strategies, LIPOchip platform processed in Spain and CGA, are also highly cost-effective. The latter strategy generates the greatest QALY gain but at additional cost. The sequences of the ICERs remain robust to the majority of deterministic sensitivity analyses; however, some plausible variations change the magnitude of these ICERs slightly. It is likely that CGA will become more cost-effective going forward because of the emergence of next-generation sequencing techniques, reducing the time and cost required to conduct large gene sequencing operations. More important, though, is the fact that all three non-dominated test strategies are cost-effective at all conceivable threshold values of willingness to pay for a QALY gain. In all cases it is more cost-effective to cascade test relatives using targeted sequencing instead of either Elucigene FH20 or LIPOchip. This is because of the relative diagnostic cost savings for the same high level of accuracy in a targeted group of relatives.

Probabilistic sensitivity analysis more clearly shows the relative cost-effectiveness of the three test strategies mentioned above. At usual threshold values of willingness to pay for a QALY gain of £20,000, CGA is the most cost-effective test strategy. Although the probabilistic sensitivity analysis shows some uncertainties surrounding alternative mutation detection rates among clinically diagnosed index cases, CGA still remains the most likely option to be cost-effective. The probabilistic results are not sensitive to the age of the index case or associated average age of relatives.

Amongst the test strategies identified as being cost-effective, there are other factors that may need further consideration before arriving at a judgement on which strategy to recommend. For example, there may be practical and resource issues associated with full-scale implementation of CGA if this is recommended as a test strategy for all. If so, then judgement is required on whether it is ethical to implement cascading based on an index test result that is not as accurate as alternative more effective and cost-effective strategies such as CGA. In addition, cost-effectiveness will also depend on how clinicians view the outcome of tests such as Elucigene FH20, which detect only approximately 44% of cases with a FH-causing mutation; for example, there is the potential for missing cases, especially at-risk relatives who may not show high LDL-C levels when tested but who may have a FH-causing mutation. These patients may forgo potentially life-saving treatment if index cases are managed only on the basis of their clinical diagnosis as opposed to their genetic test. This issue does not arise for CGA for which an unequivocal diagnosis is reported in so far as this method detects all known FH-causing mutations.

© 2012, Crown Copyright.

Included under terms of UK Non-commercial Government License.

Cover of Elucigene FH20 and LIPOchip for the Diagnosis of Familial Hypercholesterolaemia: A Systematic Review and Economic Evaluation
Elucigene FH20 and LIPOchip for the Diagnosis of Familial Hypercholesterolaemia: A Systematic Review and Economic Evaluation.
Health Technology Assessment, No. 16.17.
Sharma P, Boyers D, Boachie C, et al.
Southampton (UK): NIHR Journals Library; 2012 Mar.

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...