Alternative cascade-testing protocols for identifying and managing patients with familial hypercholesterolaemia: systematic reviews, qualitative study and cost-effectiveness analysis

Introduction

The typical pathway for FH identification involves physicians, often the primary care provider, referring individuals with suspected FH to a specialist who confirms the diagnosis, often using genetic testing. The specialist will then arrange testing of relatives of confirmed FH cases, usually by the patient contacting the relatives themselves (indirect cascade testing). The exception could be testing children of affected parents, which may be done directly. In fact, the family are often traced to two or three generations. 8,22 Initially, this usually starts with the affected individuals’ children. 23 Internationally, most cascade testing starts with adult index patients and cascading testing to other relatives including children (‘forward cascade testing’). ‘Reverse’ cascade testing is also under consideration: starting identification from affected children. 24 However, despite being recognised as a cost-effective strategy,25 there are still many patients not being diagnosed, with one of the reasons being the relatively low yield, which could be related, partly, to using the indirect approach. The alternative approach to the indirect approach is direct cascade testing, whereby the genetic specialist contacts the relatives directly.

In 2019, a systematic review found that the proportion of cascade-tested relatives was higher with the direct approach;26 however, as this review did not synthesise the studies quantitatively, the magnitude of the differences between the approaches remains unclear. Therefore, we have performed a systematic review and a meta-analysis to quantify the yield of different approaches (direct, indirect, combination) for cascade testing for FH.

Materials and methods

The protocol for the systematic review was registered in PROSPERO (CRD42019125775). Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines27 were adhered to throughout the conduct and reporting of the systematic review.

The systematic review encompasses relevant study designs, including controlled trials and epidemiological studies, which assessed the effectiveness of cascade testing for FH among relatives. Eligible participants were first- and second-degree relatives of index cases with confirmed FH, determined using a clinical diagnosis [i.e. Simon Broome,5 DLCN,3 make early diagnosis to prevent early death (MEDPED)28 or another criterion appropriate to the population being studied]; LDL-C levels, using age-specific cut-off points; or genetic diagnosis of mutation-positive cases. The protocol for cascade testing was the intervention of interest, which could be conducted via (1) a direct method of contact (whereby the relatives of the index case are contacted directly by the clinic, usually using personalised letters or telephone calls, once consent has been sought from the index case), (2) an indirect method of contact (whereby the index case acts as an intermediary by passing on personalised letters or information to their relatives) or (3) a choice of indirect or direct methods (or a combination of direct and indirect methods). The primary outcome measure was the proportion of relatives of the index cases tested out of those contacted, henceforth referred to as yield. Secondary outcome measures included the proportion of relatives of the index cases with confirmed FH out of those tested, the proportion of relatives of the index cases contacted out of those eligible, the proportion of relatives of the index cases who responded out of those contacted and the proportion of index cases who participated in cascade testing out of those genetically or clinically confirmed with FH.

Comprehensive literature searches of three databases [MEDLINE, from 1946 to May 2020; EMBASE, from 1980 to May 2020; and Cochrane Central Register of Controlled Trials (CENTRAL), from 1966 to May 2020] were performed using a highly sensitive search strategy based on keywords and medical subject heading (MeSH) terms relating to the population (e.g. proband$, index patient$, relative$, family$, patient$) and intervention of interest (e.g. cascade, mass screening, contact tracing) (see Appendix 1 and Table 23, for the MEDLINE search strategy), and other publications were identified through contact with topic experts.

In addition, grey literature was identified from the following sources: the British Cardiovascular Society Annual Conference, the HEART UK Annual Scientific Conference, the European Human Genetics Conference and the EAS Congress, from dates of inception to March 2020, and through hand-searching the Atherosclerosis journal and the HEART UK (www.heartuk.org.uk/) and US Family Heart Foundation (https://thefhfoundation.org/) websites. No language restrictions were applied, and translations were sought when necessary.

Results

The searches identified a total of 3742 studies. Following title and abstract screening, 217 studies were assessed for full-text screening (see Figure 1). At the full-text screening, 193 studies were excluded, related to ineligible study design (77 studies), ineligible or duplicate population (35 studies), ineligible or unclear intervention (62 studies) or ineligible outcome reporting (19 studies); therefore, 24 studies were included in the systematic review and meta-analysis32–55 (see Table 1).

FIGURE 1.

The PRISMA flow chart for systematic review of effectiveness of cascade-testing strategies.

Of the 24 included studies, 16 were conducted in Europe (England,35,36,41,44,45,51 Wales,39 Belgium,53,56 Denmark,33,52 Latvia,43 the Netherlands,50 Norway,55 Spain,40 Malta49 and Estonia32), one in Australia,34 three in the Americas (the USA38,47 and Brazil42), one in New Zealand,46 two in Asia (India48 and Hong Kong37) and one in South Africa. 54 All studies used an observational design to assess the outcome measures. The average number of confirmed index cases enrolled in the studies was 242, with sample sizes ranging from 2 to approximately 1300 participants.

The direct method of contact was used in 12 studies,33,35,38,40,42,45,46,48,50,52–54 a further 7 studies used an indirect method,32,36,37,43,44,47,55 and the remaining 5 used a combination of direct and indirect methods. 34,39,41,49,51 Contact could be made via a range of approaches, including postal invitation, telephone, in person or a combination of approaches. Forward cascade testing was used in the majority of included studies (23 studies), with the remaining study using reverse cascade testing. 52 Fourteen of the included studies reported the extent of cascade: the majority (eight studies32,33,37,42,45,48,50,53) cascaded to second-degree relatives, with only five studies cascading to first-degree relatives35,38,43,49,54 and one study cascading to third-degree relatives. 34

The majority of included studies (14 studies32,34,36–40,42,46–48,50,51,55) confirmed FH diagnosis for the index cases using genetic testing; nine studies confirmed FH diagnosis for the index cases using clinical assessment based on Simon Broome (four studies35,41,44,45), DLCN (three studies43,49,53) or study-specific criteria (serum cholesterol of ≥ 8 mmol/l, LDL-C of ≥ 6 mmol/l and family history of hypercholesterolaemia;33 apolipoprotein B: apolipoprotein A-1 ratio > 97th centile or apolipoprotein B > 99th centile, LDL-C > 95th centile and no secondary causes for raised cholesterol52); and one study stated that diagnosis was based on either genetic or clinical criteria, but did not provide additional details. 54 For the relatives, genetic confirmation of FH was used in the majority of studies (15 studies32,34,36–40,42,46–48,50,51,55,57). A further eight studies used clinical assessment based on either Simon Broome,35,41 DLCN,43,49 MEDPED,44 a combination of DLCN and MEDPED,53 or study-specific criteria (serum cholesterol of ≥ 7 mmol/l;33 LDL-C > 95th centile52). The final study used genetic testing or clinical assessment based on Simon Broome criteria depending on which arm of the trial the index case had been randomised to. 45

For the 16 studies using genetic testing for confirmation of FH, testing of only the low-density lipoprotein receptor (LDLR) gene was performed in three studies,46,48,50 testing of the LDLR and apolipoprotein B-100 (APOB) genes was performed in two studies,45 testing of the LDLR, APOB and proprotein convertase subtilisin/kexin type 9 (PCSK9) genes was performed in eight studies,32,34,37,38,40,42,47,51 and testing of the LDLR, APOB, PCSK9, and low-density lipoprotein receptor adaptor protein 1 (LDLRAP1) genes was performed in one study. 54 It was unclear what genes were tested in the remaining two studies. 36,39

The median number of new relatives with FH per known index case was 0.98 (range 0.15–8.6), with the largest medians seen in the studies using the indirect (median 1.39, range 0.22–8.60) and direct (median 1.27, range 0.15–3.86) testing strategies, compared with the combination approach (median 0.72, range 0.26–1.88); however, this is a crude analysis that does not consider the relative contribution of each study.

Primary outcome measure

Proportion of relatives of index cases tested for familial hypercholesterolaemia of those contacted

Four studies32,41,44,45 provided data to estimate the primary outcome. On average, 39% of relatives were tested for FH out of those contacted (95% CI 31% to 47%, four studies); however, the estimates varied significantly by the cascade approach used (p-value for subgroup differences, p = 0.01) (see Figure 2). The largest yield was seen in the study conducted in England that used a combination approach (40%, 95% CI 37% to 42%, one study); however, similar, but slightly lower, yields were seen for the direct and indirect strategies [direct: 33%, 95% CI 28% to 39% (one study, conducted in England); indirect: 34%, 95% CI 30% to 37% (two studies, conducted in England and Estonia)], although the results from the last two studies varied considerably (57%32 and 20%44).

FIGURE 2.

Proportion of relatives tested out of those contacted for FH cascade testing, by cascade approach.

Secondary outcome measures

The proportion of relatives contacted for familial hypercholesterolaemia testing out of those eligible

Only three studies reported data to estimate the proportion of relatives contacted for FH testing out of those eligible. 41,44,45 For the studies that reported this outcome, on average, 95% of relatives were contacted out of those eligible (95% CI 59% to 100%, three studies). Using either a direct or an indirect approach resulted in all the relatives who were eligible for testing being contacted (direct: 100%, 95% CI 99% to 100%, one study; indirect: 100%, 95% CI 99% to 100%, one study) (see Figure 3). However, in the single study that used a combination of direct and indirect methods, a significantly lower proportion of relatives were contacted out of those eligible (65%, 95% CI 63% to 67%; p-value for subgroup differences, p < 0.001). However, in this study,41 only 26% of the index cases had a diagnosis of definite FH, with the remaining having a possible diagnosis of FH.

FIGURE 3.

Proportion of relatives contacted out of those eligible for FH cascade testing, by cascade approach.

The proportion of relatives who responded out of those contacted for familial hypercholesterolaemia testing

Three studies reported data on the proportion of relatives who responded to cascade screening out of those contacted;41,44,45 on average, 43% of relatives responded out of those contacted (95% CI 28% to 58%, three studies)

Using a combination of direct and indirect approaches yielded a significantly greater proportion of relatives responding out of those contacted (54%, 95% CI 51% to 56%, one study) than using either an indirect approach (31%, 95% CI 27% to 35%, one study) or a direct approach alone (45%, 95% CI 39% to 51%, one study) (p-value for subgroup difference, p < 0.001).

The proportion of relatives with confirmed familial hypercholesterolaemia of those tested

Twenty-one of the included studies reported data on the proportion of relatives confirmed as having FH out of the number of relatives tested (see Figure 4). On average, 47% of relatives were confirmed to have FH out of those tested (95% CI 42% to 52%, 21 studies). Contact strategies were found to produce similar pooled results (direct: 50%, 95% CI 41% to 58%, I² = 97%, 10 studies; indirect: 45%, 95% CI 38% to 53%, I² = 82%, 7 studies; combination: 43%, 95% CI 29% to 58%, I² = 98, 4 studies; p-value for subgroup differences, p = 0.67).

FIGURE 4.

Proportion of relatives confirmed as having FH out of those tested, by cascade approach.

The proportion of index cases that participated in cascade testing out of those confirmed as having familial hypercholesterolaemia

Seventeen studies reported data on the proportion of index cases who participated in FH cascade testing out of those confirmed as having FH. On average, 89% of index cases participated in cascade testing out of those confirmed with a diagnosis of FH (95% CI 73% to 99%, 17 studies); however, the estimates varied significantly by the cascade approach used (p-value for subgroup differences, p < 0.001). The yield was highest using a direct approach (95%, 95% CI 83% to 100%, I² = 98%, nine studies), a slightly lower yield was seen using an indirect approach (78%, 95% CI 46% to 99%, I² = 99%, six studies), and the lowest yield was seen using a combination of direct and indirect approaches (60%, 95% CI 56% to 63%, two studies).

Conclusion

The review provides tentative support for the combination approach to cascade testing, whereby the index case determines which method is used to contact relatives. However, further evidence to support the combination approach requires experimental studies to compare the cascade approaches and/or interrogation of routine data sets and FH registers held on the cascade testing and the modality of contact with relatives.

Materials and methods

The protocol for the systematic review adhered to the PRISMA guidelines27 throughout, for both the conduct and the reporting of the review. Eligible participants were adults from the general population and adults with confirmed FH, determined using a clinical diagnosis (e.g. Simon Broome,5 DLCN,3 MEDPED28 or another criterion appropriate to the population being studied); LDL-C levels, with age-specific cut off points; or genetic diagnosis of mutation-positive cases. The eligible interventions and comparators were any licensed LLTs, including statins, ezetimibe, statins with ezetimibe, PCSK9 inhibitors, PCSK9 inhibitors with statins, PCSK9 inhibitors with statins and ezetimibe, or comparisons of different doses of the same treatment. Therefore, placebo or other non-active interventions were not eligible for inclusion as the comparator. The primary outcome measure was LDL-C levels, which could be reported as mean changes from baseline, mean percentage changes, or absolute level at follow-up. The study design included in this overview of reviews was systematic reviews of randomised controlled trials; however, we also included systematic reviews of observational cohort studies when there was a dearth of literature on adults with confirmed FH. We excluded studies that focused solely on index cases with homozygous FH.

Comprehensive literature searches of four databases [MEDLINE, from 1994 to June 2018; EMBASE, from 1994 to June 2018; Cumulative Index to Nursing and Allied Health Literature (CINAHL), from 1994 to June 2018; and CENTRAL, from 1994 to June 2018] were performed using a highly sensitive search strategy based on keywords and MeSH terms relating to the intervention of interest, outcome measures and study design (see Appendix 1 and Table 25, for the MEDLINE search strategy).

Results

The searches identified 2747 hits from MEDLINE, 1736 hits from EMBASE, 991 hits from CINAHL and 260 hits from CENTRAL, totalling 5734 hits. Following deduplication, 4829 hits were screened by title and abstract. Of the 214 papers identified for full-text screening, 14 papers59–72 were assessed for methodological quality (see Figure 5). The methodological quality of the 14 papers ranged from 1–11; thus, as none of the papers met the threshold of a score of at least 12, no papers were included in the overview of reviews.

FIGURE 5.

The PRISMA flow chart for the review of the effectiveness of cholesterol-lowering therapies.

Conclusion

The search did not identify any high-scoring systematic reviews assessing the effectiveness of head-to-head comparisons of cholesterol-lowering therapies for LDL-C and CVD events in the general adult population or among adults with confirmed FH.

Background

Cascade testing consists of systematically testing individuals who are at risk of a hereditary disease. 73 Cascade testing has long been proposed to diagnose relatives of patients genetically or clinically confirmed to have FH (the initial patient is known as the index case) because, as the inheritance of heterozygote FH is autosomal dominant, the probability of having heterozygote FH is 50% for first-degree relatives, and 25% for second-degree relatives. 74–76 Cascade testing can be conducted with a genetic test for FH mutation or based on the relative’s clinical and biochemical characteristics (e.g. LDL-C level given their age and sex), or a combination of the two (e.g. using LDL-C level to triage relatives before the genetic test). An example of biochemical criteria is using LDL-C cut-off points specific to age and sex, which is recommended by Starr et al. 77 Similarly, the MEDPED criteria use cut-off points that are age-specific for both total cholesterol (TC) and LDL-C. 28

The 2008 NICE clinical guidelines (CGs) recommended cascade testing with a genetic test for relatives of index cases with FH who had an identified mutation and age- and sex-specific LDL-C criteria to diagnose relatives of probands with FH. 8 The 2017 update, however, recommends using only the genetic test to conduct cascade testing. This systematic review aimed to assess the diagnostic accuracy of clinical and biochemical characteristics, including LDL-C and TC, and clinical signs, for the diagnosis of FH among relatives of index case with confirmed FH to inform the parameters for the economic model.

Review method

The systematic review was conducted and reported according to the PRISMA guidelines. 78 The study protocol was published on PROSPERO (CRD42018117445).

Results

After deduplication, 15,191 titles and abstracts were screened, 280 of which were screened at the full-text stage. Of these studies, 117 reported an ineligible outcome, 9 reported ineligible index cases, 6 reported ineligible reference standards, 76 reported an ineligible population, and 63 were not full-text articles or insufficient information was presented (see Figure 6).

FIGURE 6.

The PRISMA flow diagram of the review of the diagnostic accuracy of biochemical and clinical criteria among relatives.

Nine studies met the inclusion criteria for this systematic review. 37,41,50,55,82–86 The studies were completed in the following countries: Australia,87 Denmark,41 Iceland,85 Japan,83 the Netherlands,41,82 Norway,41,50,55 South Africa86 and Vietnam. 84 Two studies used the same cohort of individuals; however, one assessed the diagnostic potential of TC50 and the other used LDL-C levels (Norwegian cohort). 41 None of the included studies used other index tests.

The nine studies included a total of 40,079 relatives, of whom 14,310 were diagnosed with FH. The number of index cases was 964 across five studies;37,55,84–86 this could not be determined for the remaining four studies. 41,50,82,83 Three studies enrolled only adults (mean age 40 years);55,82,83 one enrolled only children (mean age 13 years);37 two studies enrolled both adults and children,41,84 and three studies did not report the ages of relatives. 50,85,86 The extent of the cascade testing, in terms of the degree of relative, was unclear in the majority of studies (n = 5); however, only first-degree relatives were considered in two studies,37,41 and first- and second-degree relatives were considered in the remaining two studies. 50,83 Eight of the nine studies also collated triglyceride levels. Triglyceride levels among FH individuals were high in one study. 83

The index test used in four studies was LDL-C. 37,41,82,84 A further four studies used TC. 50,55,85 One study assessed both LDL-C and TC as the index test. 83 Starr et al. 77 compared the accuracy of their own age- and sex-specific cut-off points with that of the MEDPED criteria.

The five studies that used LDL-C as a diagnostic criterion reported varying cut-off values. 37,41,82–84 For example, Huijgen et al. 82 reported four statistical models with differing levels of genetic severity in a cohort from the Netherlands. Each analysis used a 90th percentile cut-off point, which is derived from the sample. Starr et al. 77 reported analyses for differing age and sex groups for three different countries (Denmark, the Netherlands and Norway), and presented the results of the MEDPED criteria for comparison. In each cohort, cut-off points varied from 3.01 mmol/l for males and 3.32 mmol/l for females (aged 15–24 years) to 4.31 mmol/l for males (aged 45–54 years) and 4.36 mmol/l for females (aged ≥ 55 years). The Netherlands and Norway cohorts included patients aged 0–14 years, which had cut-off points of 3.11 mmol/l and 3.37 mmol/l for males and females, respectively. 77 Similarly, Truong et al. 84 used different cut-off points for Vietnamese children (3.0 mmol/l) and adults (4.1 mmol/l). The same cut-off point was used for adults of Japanese origin. 83 The final study was conducted only with children, with a cut-off point of 3.5 mmol/l. 37

Five studies used TC as diagnostic criterion. 50,55,83,85,86 There were variations in cut-off values; however, one study, using a Norwegian cohort, did not report the cut-off values used. 55 In a Japanese cohort, a cut-off point of 5.8 mmol/l was assessed. 83 In a South African cohort, a cut-off point of 4.8 mmol/l (80th percentile of sample) was used. 86 The other two studies analysed the 90th and 95th percentiles of the samples, derived from Netherlands and Iceland cohorts, respectively. 50,85

Discussion

The aim of this review was to review the current evidence of the diagnostic accuracy of clinical and biochemical characteristics in identifying relatives with FH, compared with genetic identification. We found nine studies, all of which examined LDL-C and/or TC levels to diagnose FH among relatives of index cases, with one study assessing the diagnostic accuracy of both. 83 The studies varied in terms of geographical location, relatives’ ages, cut-off points (for LDL-C and TC levels) for diagnosis and quality. The small number of studies, coupled with substantial heterogeneity, means that clinical utility of these biochemical tests could not be determined.

Conclusion

Our systematic review on clinical and biochemical characteristics to diagnose FH among relatives of known index cases found nine studies on LDL-C and/or TC. However, the evidence on diagnostic accuracy and appropriate cut-off points for diagnosis was poor, owing to the paucity of studies and high heterogeneity, which we could not investigate quantitatively.

Context

Although the association of FH with premature CHD is well known, the risk of atherosclerotic disease in other vascular regions among patients with FH is less clear, and evidence from previous studies is conflicting. 6,93–95 The work reported in this chapter sought to determine the CVD risk profile of patients with FH in the general population using longitudinal data from patients’ primary care electronic health records. We assessed the incidence and risks of CHD, stroke/transient ischaemic attacks (TIAs) and peripheral vascular disease (PVD) among patients with clinical diagnoses of FH identified in primary care.

Research objective

The objective was to determine the CVD risk profile of patients with FH in the general population.

Methodology

We conducted a retrospective matched cohort study using data from the CPRD. For each patient with a clinical diagnosis of FH, we randomly identified three patients without FH and individually matched on age, sex and general practice. Those without FH had cholesterol testing done within 6 months of the start date of their matched FH case, no documented diagnosis of FH and no pre-existing CVD. Patients who had disease (Read) codes suggesting pre-existing CVD (CHD, stroke, TIA or PVD) prior to study entry were excluded.

Results

There were 3,936,934 patients in the CPRD with records of either a TC or LDL-C measurement between 1 January 1999 and 22 July 2016. Of these, 14,097 patients had clinical FH and no prior history of CVD at baseline. This comprised 5152 patients with documented diagnosis of FH in the electronic health record, and 8945 patients who had no documented FH diagnosis but had the clinical phenotype of FH based on the Simon Broome or DLCN diagnostic criteria for definite FH. Of these identified patients, 53.3% were females, the mean age at the start of follow-up was 42 years and the mean BMI was 27.3 kg/m². FH patients were matched with 42,506 non-FH patients. As individuals with and those without FH were matched on age and sex, the distribution of these characteristics were similar across both groups. The median follow-up time for both FH and non-FH patients was 13.8 years (IQR 8.4–17.7 years), and the average follow-up times for patients with and patients without FH were 174,950 and 588,470 person-years, respectively.

The baseline demographic and clinical characteristics of patients with and patients without FH are shown in Table 2. As expected, a significantly higher proportion of FH patients had a family history of premature CHD than non-FH patients (5.1% vs. 2.7%, respectively; p < 0.001), and more FH than non-FH patients were on lipid-lowering medication at the start of the study (19.1% vs. 4.7%, respectively; p < 0.001). Ethnicity records were missing for 85% of patients; for those for whom ethnicity was recorded, 80% were white. BMI records were available for 40% of patients, so multiple imputation with chained equations was used to estimate missing BMI values.

For patients on statins or other cholesterol-lowering medication with known potency, the corrected levels of TC were estimated from observed levels, based on estimated percentage reduction in LDL-C with statins of different potencies. 99 As expected, the mean cholesterol concentration was significantly higher among those with FH [9.30 mmol/l (SD 0.02 mmol/l)] than among those without FH [5.98 mmol/l (SD 0.01 mmol/l)]. The median triglyceride concentrations in the FH and non-FH cohorts were 2.10 mmol/l (IQR 1.34–3.52 mmol/l) and 1.30 mmol/l (IQR 0.90–1.92 mmol/l), respectively. See Appendix 2 and Figure 18.

Conclusion

Individuals with FH have been shown to have greatly increased risk of a range of cardiovascular outcomes, including not only CHD, but also stroke, TIA and PVD. This has important clinical implications and emphasises the need for improved case identification of clinically recognisable FH in the general population for targeted preventative intervention. The findings also suggest incorporating a broader range of cardiovascular outcomes for economic modelling.

Context

Historically, the Simon Broome Register has been linked to ONS data for ascertainment of annual death records, which have provided outcomes related to annual mortality rates (up to 2016). However, much of the participant journey through their life course is missing, including outcomes related to disease morbidity, treatments, procedures and operations, and health resource use. These outcomes have not been previously assessed.

Research objective

The objective was to evaluate the long-term cardiovascular outcomes of individuals with FH.

Methodology

A total of 3553 people in the Simon Broome Register were recruited from participating lipid clinics between 1 January 1980 and 20 December 2010. Of these, 2988 (84%) had linked HES Admitted Patient Care records; these comprised the final study cohort. Individuals without linked HES records had comparable baseline demographic characteristics to those with linked data, but a higher proportion of them had a record of previous history of CVD.

Results

The characteristics of the study population, at the time of registration on Simon Broome Register are shown in Table 4. Of the cohort, 1418 (47.5%) were male. Compared with men, women were 5 years older at registration and 4.8 years older at time of commencing LLT. Although women had a slightly lower BMI than men, their mean untreated TC concentration was significantly higher, but median triglyceride concentration was significantly lower. Consumption of alcohol was significantly higher among men than women; significantly fewer women reported ever smoking, whereas the prevalence of current smoking was similar among men and women. Fewer women than men reported a prior history of CVD, with significant difference in myocardial infarction, CHD and previous revascularisation, and women had their first myocardial infarction 8 years later than men. Although significantly more women than men had a history of hypertension, the prevalence of type 2 diabetes was similar among men and women. Overall, women on the Simon Broome Register had a better CVD risk factor profile, but a later age at FH diagnosis and commencement of LLT.

Conclusion

This study finding of significantly higher risk of CVD in all age groups of patients in this registry-based cohort, compared with the general population, emphasises the importance of early diagnosis and treatment of FH. Our study provides confirmatory evidence of higher excess CVD morbidity in younger age groups of patients with FH and, importantly, provides novel insight into sex differences in the diagnosis and management of FH, as well as substantial sex disparities in the excess CVD burden associated with FH. We noted in these patients, cared for in a specialist setting, that excess CVD morbidity is markedly higher among women than men aged 30–50 years and aged > 50 years; this highlights the need for optimisation of lipid lowering and risk factor management for all FH patients, with particular attention to women with FH.

Background and objectives

A number of important papers have characterised long-term outcomes for FH patients in the UK. However, these studies do not provide an ideal basis for economic modelling for a number of reasons. First, these studies have not always included non-fatal CVD events,6,91,94,98 which are important to include in economic modelling as they have significant cost and quality-of-life implications. Second, these studies have not separated out treated and untreated FH patients,101 which makes it challenging to make inferences about the effect of treatment on outcomes, as required when modelling the benefits of diagnosis and treatment. Finally, these studies have not extrapolated risk beyond the censoring time. These approaches are important as they allow prediction of event risk over time horizons beyond the observed data, and can provide risk predictions according to patients’ characteristics. This work seeks to address these limitations, and provide more robust long-term outcome modelling to support cost-effectiveness analyses in FH.

With this work, we wanted to understand the lifetime consequences of diagnosing and treating individuals with FH in terms of both morbidity and mortality. Contemporary UK primary and secondary care data on long-term outcomes in the FH population reflect a group of individuals who are generally under active treatment with LLTs. Therefore, in this section of the economic evaluation chapter, the focus was on appropriately characterising outcomes among these individuals. The objectives of this section were to address two key questions: (1) what is the impact of diagnosis and treatment on LDL-C levels? and (2) what is the long-term risk of major CVD events and mortality for individuals diagnosed with, and treated for, FH?

In Cost and health benefits of diagnosis of people with familial hypercholesterolaemia in the long term, and Cost-effectiveness of alternative cascade-testing protocols for familial hypercholesterolaemia, these findings are used within a long-term cost-effectiveness model to simulate the counterfactual, that is what would have been the long-term outcomes of these individuals had they not been diagnosed with, and consequently treated for, FH. This, in turn, allows the estimation of the benefits of diagnosis and treatment in terms of long-term CVD risks, health-related quality of life, health resource consumption and costs, and mortality.

In Methods, we introduce the utilised data source and the methods used within to achieve our objectives. Results will present the results of the implemented analysis and, finally, Summary of results and study limitations discusses key findings and important limitations.

Methods

Results

Treatment practices and low-density lipoprotein-cholesterol response

After applying the Friedewald equation, LDL-C missingness was substantial (overall, 36.4%; ‘untreated’, 34.6%; ‘newly treated’, 29.3%; and ‘treatment retainers’, 49.3%), and was addressed with multiple imputation; see Report Supplementary Material 2 for further details. The results of the analysis on the change from baseline to 24 months in LDL-C levels for the different FH patient subgroups are shown in Table 7. These effects relate to pre and post FH diagnosis analysis of imputed LDL-C levels for the ‘untreated’ and the ‘newly treated’ subgroups and to pre and post first prescription analysis of imputed LDL-C levels for the ‘treatment retainers’ subgroup; see Report Supplementary Material 2 for further details.

TABLE 7 - Absolute and relative reductions from baseline in LDL-C levels, by treatment subgroup

Accounts for variability within and between imputed data sets.

Time point and reduction	LDL-C level (95% CI)
	Untreated (N = 744)	Newly treated (N = 1291)	Treatment retainers (N = 844)
Baseline LDL-C level (mmol/l)
Pre imputation	4.32 (2.10 to 7.06)	5.39 (3.00 to 8.00)	5.58 (3.20 to 8.80)
Post imputationa	4.27 (2.07 to 7.10)	5.44 (2.90 to 8.50)	5.62 (3.00 to 9.38)
Post-baseline LDL-C level (mmol/l)
Pre imputation	4.15 (2.10 to 7.09)	3.46 (1.68 to 6.91)	3.49 (1.60 to 6.76)
Post imputationa	4.11 (2.01 to 7.70)	3.49 (1.60 to 6.95)	3.54 (1.60 to 7.20)
Post-imputation reduction from baseline in LDL-C level
Absolute reduction (mmol/l)	0.26 (0.17 to 0.34)	1.94 (1.84 to 2.04)	2.08 (1.95 to 2.20)
Relative reduction (%)	0.01 (–0.1 to 0.03)	32.6 (30.9 to 34.3)	34.4 (32.5 to 36.3)

Results of the before-and-after analysis show that a small reduction in LDL-C was observed for the ‘untreated’ group, potentially a consequence of lifestyle changes following FH diagnosis. In the ‘newly treated’ and the ‘treatment retainers’ subgroups, similar absolute and relative reductions in LDL-C levels were found to have occurred. These changes were clinically meaningful, with the effect of LLTs implying an average absolute reduction in LDL-C of approximately 2 mmol/l and a relative reduction in LDL-C levels of > 30%, on average, although substantially lower than the 50% reduction target108,121,122 and the absolute target of 2.6 mmol/l for primary CVD prevention123 advocated for FH patients. Figure 7 shows the distribution of proportional LDL-C reduction across the ‘untreated’, ‘newly treated’ and ‘treatment retainers’ subgroups. Indeed, approximately only 30% of the subset of individuals receiving treatment achieve the ≥ 50% LDL-C reduction target. Similar findings were obtained for males and females.

FIGURE 7.

Distribution of proportional LDL-C reductions from baseline to 24 months, by treatment subgroup.

Risk of major cardiovascular first events and mortality among treated individuals

Table 8 shows the number of first events of each type occurring in the treated CPRD cohort after FH diagnosis. The probability of a post-FH diagnosis first major CVD event occurring during the 20 years of follow-up of the CPRD cohort is depicted in Figure 8. The median follow-up time was 4.8 years, with Kaplan–Meier survival estimates of 0.93 (95% CI 0.92 to 0.95), 0.90 (95% CI 0.88 to 0.92) and 0.88 (95% CI 0.84 to 0.92), for 5, 10 and 15 years, respectively. Approximately half of the sample was censored at 5 years, which, together with a low event rate, implied an undefined median survival; see number of FH patients at risk at different time points in Figure 8. A decreasing hazard of the event occurring throughout time is observed.

TABLE 8 - Occurrence of key clinical events as first major CVD events post FH diagnosis in the CPRD treated cohort (primary diagnosis only)

Treated individuals’ subset includes the ‘newly treated’ individuals (n = 1291) and the ‘treatment retainers’ (n = 844).

Nature of CVD	CPRD treated cohorta (N = 2135)
CVD first events included in model, n (%)
ACS	91 (4.3)
TIA/stroke	35 (1.6)
CVD death	15 (0.7)

FIGURE 8.

Kaplan–Meier of time to first major CVD event post FH diagnosis in the CPRD cohort, number of individuals at risk and predicted parametric curves. GenGamma, generalised gamma.

An assessment of the suitability of the different survival models fitted was carried out. The Royston–Parmar model with five knots was the model that visually provided the best fit, and had the lowest Akaike information criterion; see Report Supplementary Material 2 for further details. Nonetheless, in terms of internal validity, and when weighing model complexity and model fit, the generalised gamma model was considered the optimal choice as it provided a good fit during the first 12 years of data. The external validity assessment through the Perak et al. 96 data indicated that none of the models fitted particularly well. The external data suggested an increasing hazard over time, and this was the case for all age groups. The available models estimated either decreasing hazards over time or, in the case of the exponential model, a stable hazard over time (see Report Supplementary Material 2 for further details). We therefore included the exponential model as a sensitivity analysis. We return to the appropriateness of the models for long-term extrapolation in the discussion.

The exponential and generalised gamma models’ parameters, pooled using Rubin’s rules for multiple imputed data sets, are provided in Table 9. Both models provided similar results. Being older, male, having a history of CVD and a higher LDL-C level are all factors associated with increased risk of first major CVD events. We highlight that, controlling for age, male FH treated individuals are, on average, two to three times more likely to have a first CVD event than females. Similarly, FH patients with a history of CVD are, on average, 4 to 13 times more likely to have a first major CVD event than individuals with no history of CVD. The heterogeneity in event risk is evident when these risk factors are combined. The average event risk for the generalised gamma model at 5 and 10 years is 0.07 (95% CI 0.05 to 0.10) and 0.11 (95% CI 0.07 to 0.16), respectively. As an example, for a 28-year-old female with a baseline LDL-C level of 5.5 mmol/l and no history of CVD, the predicted event risk at 5 and 10 years is 0.01 (95% CI 0.01 to 0.02) and 0.02 (95% CI 0.01 to 0.03), respectively, whereas, for a male, it is 0.02 (95% CI 0.01 to 0.04) and 0.04 (95% CI 0.03 to 0.07) at 5 and 10 years, respectively. For a 49-year-old female with a baseline LDL-C level of 5.5 mmol/l and a history of CVD, the predicted event risk at 5 and 10 years is 0.16 (95% CI 0.09 to 0.28) and 0.24 (95% CI 0.14 to 0.39), respectively; for a male, it is 0.33 (95% CI 0.19 to 0.49) and 0.48 (95% CI 0.30 to 0.71) at 5 and 10 years, respectively.

TABLE 9 - Risk factors for first major CVD event for the CPRD cohort for the exponential and the generalised gamma survival models

CPRD (N = 2135, events, n = 141)	Exponential survival model			Generalised Gamma survival model
	HR	95% CI	p-value	HR	95% CI	p-value
Age at baseline (years)	1.05	1.04 to 1.07	< 0.001	1.08	1.05 to 1.10	< 0.001
Sex (male)	2.19	1.53 to 3.14	< 0.001	3.35	1.86 to 6.03	< 0.001
History of CVD	4.36	2.77 to 6.88	< 0.001	13.13	4.14 to 41.65	< 0.001
PT-LDL-C (mmol/l)	1.33	1.20 to 1.47	< 0.001	1.56	1.29 to 1.88	< 0.001
Mu	–	–	–	13.19	10.54 to 15.84	–
Sigma	–	–	–	0.79	0.19 to 1.40	–
Q	–	–	–	0.50	–0.25 to 1.24	–

Summary of results and study limitations

The analysis shows that individuals with a recorded diagnosis of FH in primary care are often not treated with LLT within 2 years of their diagnosis, and of those who do receive LLT, < 30% achieve the recommended reductions in LDL-C. We also noted that, at 10 years of follow-up, the average risk of a first major non-fatal CVD event or CVD-related death was 11% for the treated cohort as a whole, with higher estimates among men than women.

Our analysis used the CPRD cohort, which reflects the treatment and management of FH patients as recorded within primary care in the UK. The use of this data set has its caveats, as it might not have correctly coded individuals with FH, the cohort might not be representative of the UK FH population as a whole or of patients identified via cascade testing who are expected to have genetically confirmed FH rather than a diagnosis based on clinical phenotype. Furthermore, this cohort is unlikely to characterise the current clinical practices in tertiary specialised care, such as lipid clinics, where FH patients are closely followed and managed9 (although some patients in the CPRD may also be managed in lipid clinics). Finally, the CPRD cohort presented a relatively long follow-up of up to 20 years, although event data were sparse from around year 10.

Background and motivation

Cascade testing offers the opportunity to expand FH diagnoses by systematically testing family members (relatives) of affected individuals (index cases). It has been shown to markedly enhance FH identification and to be cost-effective in the UK health-care setting relative to an absence of cascade testing. 110,124,125

To evaluate the benefits and costs of alternative ways to design a cascade service, it is important to understand how different features of the cascade service affect the yield of FH cases. This chapter reports the analysis of a large sample of index cases and relatives considered for genetic testing within the Welsh and Wessex FH services (see Chapter 3, PASS Wales and Wessex Cascade Service Data, for a description of the services). All available records within PASS for each service were made available for this analysis. Although designed to support the operational needs of FH cascade services, PASS provides a rich source of information on the characteristics of the service and the index cases and relatives engaging with the service. The specific research questions this analysis seeks to inform within a UK context are as follows.

• What are the characteristics and diagnoses of tested index cases?

• How well do the clinical criteria used to select index cases for genetic testing perform?

• How many relatives complete cascade testing, how many relatives could potentially be reached by cascade services and what service or patient factors might explain participation by relatives?

• What are the characteristics of relatives who complete the cascade and for whom cascade testing offers the opportunity for improved management and outcomes?

Results from this chapter inform the cost-effectiveness model comparing cascade-testing protocols, presented in Cost-effectiveness of alternative cascade-testing protocols for familial hypercholesterolaemia.

Methods

Results

Index and relative database exclusions

In the Welsh index extraction, 99 cases were deemed out of area, the majority of which were cases reported from health boards in close proximity to the Welsh border, leaving a primary analysis set with a sample of 2618 Welsh index cases. From the Wessex index extraction, only six cases were deemed out of area, leaving a sample of 1116 within-area index cases (see Figure 9).

FIGURE 9.

Index and relative selection process.

The number of relative entries without a family number totalled 6 in Wales and 101 in Wessex, leaving 8571 and 6273 unique and potentially linkable relative cases in Wales and Wessex, respectively. Of these cases, 3815 Welsh and 2143 Wessex relatives were linked to a FH-positive within-area index and formed analysis set 1. Analysis set 2 comprised 2020 Welsh and 1002 Wessex relatives deemed to have been contacted and within-area from analysis set 1. From analysis set 2, 1205 Welsh and 501 Wessex relatives successfully completed genetic cascading and informed analysis set 3.

Data availability

Age at diagnosis, sex, area and LDL-C level data showed high levels of completion for within-area index cases. Unadjusted WDLCN scores were available for 97% of Welsh indexes, age-adjusted WDLCN scores were available for 38% and inferred age-adjusted WDLCN scores were available for 88%. The Wessex-modified Simon Broome score was available for 91% of Wessex indexes.

Area data were available for 77.1% of Welsh and 66.0% of Wessex relatives in analysis set 1. In analysis set 2, sex, relative degree to index and method of contact showed very high levels of completion. Relatives’ age and sex data were complete for cascaded relatives, whereas other variables were available for only a smaller subsample of cascaded relatives (LDL-C level: Wales 37.0% and Wessex 55.9%; CVD history: Wales 54.4% and Wessex 41.7%; LLT history: Wales 51.6% and Wessex 47.1%).

The analysis presented uses a complete-case approach.

Findings from analysis of index cases

Index cases recorded in Welsh and Wessex services were predominantly female (Wales, 60.3%; Wessex, 64.7%), had elevated mean observed LDL-C levels (Wales, 5.85 mmol/l; Wessex, 5.88 mmol/l), and had an average age of 54.9 years in Wales and 51.3 years in Wessex. Index cases were mostly aged > 40 years in both services (Wales, 88.9%; Wessex, 81.5%); adolescent index cases (aged < 18 years) were rare (Wales, 0.5%; Wessex, 1.7%).

Overall, 552 (21.1%) genetically confirmed FH cases and 103 (3.8%) VUSs have been identified from 2618 index cases in the Welsh service. From 215 initial variants reported as VUSs in Wales, 84 (39.1%) cases were reclassified to FH and 28 (13.0%) as not pathogenic. In the Wessex service, 323 (28.9%) monogenic and 44 (3.9%) VUSs were identified from 1116 indexes. Four reclassifications (8%) were made to FH and two (4%) to not pathogenic (from 50 initial VUS diagnoses). LDLR mutations were the most common in both services (Wales, 86.8%; Wessex, 77.7%) followed by APOB (Wales, 12.0%; Wessex, 19.0%). PCSK9 and apolipoprotein E (APOE) variants were rare (see Table 10).

TABLE 10 - Index patient characteristics

APOE, apolipoprotein E.

Age at diagnosis.

The highest observed/unadjusted LDL-C level on record prior to testing.

Diagnosis as of 15 April 2020 criteria.

Characteristics and diagnosis	Wales (N = 2618)	Wessex (N = 1116)
Characteristics
Age (years),a mean (SD)	54.9 (12.3)	51.3 (13.3)
0–17, n (%)	12 (0.45)	19 (1.70)
18–39, n (%)	279 (10.7)	187 (16.8)
40–59, n (%)	1378 (52.7)	595 (53.3)
≥ 60, n (%)	947 (36.2)	315 (28.2)
Female (%)	60.3	64.7
LDL-C level (mmol/l),b mean (SD)	5.85 (1.9)	5.88 (1.6)
Diagnosis c
Monogenic FH, N (%)	552 (21.1)	323 (28.9)
LDLR mutation, n (%)	479 (86.8)	251 (77.7)
APOB mutation, n (%)	66 (12.0)	63 (19.5)
PCSK9 or APOE mutation, n (%)	7 (1.3)	9 (2.8)
VUS, n (%)	103 (3.8)	44 (3.9)
Negative, n (%)	1963 (75.0)	749 (67.1)

Of the 2531 individuals referred to the Welsh service with an unadjusted WDLCN score, 1988 (78.5%) scored ≥ 6. From the 1005 indexes with age-adjusted WDLCN scores, and 2295 indexes with inferred scores, 772 (76.8%) and 1710 (74.5%), respectively, scored ≥ 6. Of the Welsh index cases who scored ≥ 6 using unadjusted, age-adjusted and inferred age-adjusted WDLCN scores, 505 (25.4%), 306 (39.6%) and 491 (28.7%), respectively, had FH diagnoses. This compared with 32 (5.9%) FH diagnoses among index cases with unadjusted WDLCN of < 6 and 18 (3.2%) FH diagnoses among index cases with inferred age-adjusted WDLCN scores of < 6 (numbers suppressed for the age-adjusted WDLCN score).

In Wessex, the Simon Broome criteria classified 136 (13.4%) genetic referrals as unlikely FH, 853 (84.0%) as probable FH and 26 (2.6%) as definite FH. FH was diagnosed in 8.8%, 29.2% and 57.7% of unlikely, probable and definite FH classifications, respectively.

Figure 10 shows the sensitivity and specificity of unadjusted and age-adjusted WDLCN scores for identifying FH and non-FH cases across alternative referral thresholds. The maximum absolute difference in sensitivity occurred at a threshold score of ≥ 8 (9.2%). The unadjusted score had a higher specificity at thresholds up to 13, and converged with age-adjusted specificities at higher thresholds. Increasing the referral threshold score from ≥ 6 (current) to ≥ 8 for the unadjusted (age-adjusted) score would mean that 56.1% (42.9%) of individuals who do not have FH would be correctly excluded from genetic testing; however, it would mean that 29% (20%) of individuals with FH would be incorrectly excluded from testing and subsequent cascading. The diagnostic accuracy of inferred age-adjusted scores was broadly comparable to observed age-adjusted scores (further details in Report Supplementary Material 3).

FIGURE 10.

Observed data on the accuracy of the Welsh scoring criteria in detecting FH with and without age-adjustments among Welsh service indexes eligible for testing. tnr, true negative rate; tpr, true positive rate.

Relative cases

On average, for each FH index identified in the Welsh and Wessex services, 7.35 and 7.01 relatives were available within PASS (see Table 11). Of the relatives with observed area data (Wales, 69%; Wessex, 66%), approximately 25% of Welsh and 29% of Wessex cases were out of area. If we conservatively assume that cases with missing area data are within area, this brings these proportions down to 16% and 18% for Wales and Wessex, respectively. The number of within-area relatives potentially accessible to each service was 6.17 in Wales and 5.69 in Wessex per FH-positive index case. The Welsh and Wessex services contacted approximately 64.3% and 58.0%, respectively, of relatives not known to be residing out of area. Relatives were not contacted for a variety of reasons, such as index non-participation and second-degree relative whose first-degree relative was found not to have FH. The average number of relatives with evidence of being contacted by the service was 3.97 in the Welsh service and 3.30 in the Wessex service. Of the relatives contacted by the service, 59.7% in Wales and 50.0% in Wessex went on to complete the cascade, corresponding to 2.41 Welsh relatives and 1.66 Wessex relatives per FH index case. In Wales, the final number of FH relatives identified was 666, representing 1.35 cases detected per FH case. The comparable figures for Wessex were 285 relatives identified, representing 0.95 cases per FH case.

TABLE 11 - Mean number of relatives per FH index case

The number of linkable relatives per FH index with a family number in the PASS extractions.

Those who were missing area data were assumed to be within area.

Relatives may not have been contacted for a variety of reasons (e.g. index not returning necessary forms, index having no interest in testing within family or is unreachable, relative unreachable).

Relatives may not have been cascaded owing to incompletion on their part, or they did not require testing in the first place (e.g. determined not to be at risk, already tested by another service).

Relatives per FH index case	Relatives, n (%)
	Wales	Wessex
Relatives registered in PASS per FH indexa	7.35 (6.60)	7.01 (5.64)
Within-area relatives in PASS per FH indexb	6.17 (6.31)	5.69 (5.78)
Contacted within-area relatives per FH indexc	3.97 (4.95)	3.30 (2.43)
Relatives completing cascade per FH indexd	2.41 (3.60)	1.66 (2.41)
FH relatives identified per FH index	1.35 (2.13)	0.95 (1.46)

Table 12 displays the estimated probabilities for completing the cascade in each cascade service by relative sex, degree and contact profiles; the results of the logistic models are presented in Report Supplementary Material 3.

TABLE 12 - The estimated probabilities of a relative completing the cascade, by service, method of contact, degree to index and sex

Degree of relative and sex	Wales (%)		Wessex (%)
	Direct contact (95% CI)	Indirect contact (95% CI)	Indirect contact (95% CI)
First
Female	71.1 (67.0 to 75.1)	53.7 (48.4 to 59.1)	43.2 (38.0 to 48.4)
Male	61.3 (56.7 to 66.0)	42.8 (37.5 to 48.2)	30.5 (25.5 to 35.5)
Second and more distant
Female	61.3 (56.2 to 66.4)	42.9 (37.2 to 48.6)	47.2 (40.3 to 54.1)
Male	50.6 (45.1 to 56.1)	32.6 (27.3 to 37.9)	34.0 (27.6 to 40.4)

In Wales, sex, degree and the direct method of contact (relative to indirect) were all statistically significant predictors of cascade completion (p < 0.01). Females were 53% more likely to complete the cascade than males, first-degree relatives were 55% more likely than second-degree or further relatives to complete the cascade and relatives contacted directly were 111% more likely to complete the cascade than those contacted indirectly [odds ratio comparing direct with indirect testing 2.11 (95% CI 1.66 to 2.69)]. Paediatric cases were 165% more likely to complete the cascade than adults cascaded indirectly.

Female first-degree relatives were estimated to have a probability of completing the cascade of 71% when directly contacted and 57% when indirectly contacted. In comparison, male first-degree relatives had 62% and 43% probabilities of completion when directly and indirectly contacted, respectively. Female second-degree or further relatives were estimated to have a probability of completing the cascade of 61% when directly contacted and 43% when indirectly contacted, each 10% above the same probabilities for males. No evidence was found to suggest that direct contact was more effective by sex or degree [interaction terms between direct contact and sex, and direct contact and degree, were not statistically significant predictors of cascade completion (p > 0.24)].

In Wessex, sex was a statistically significant predictor of cascade completion (p < 0.01), with contacted females 74% more likely to complete genetic testing than males. First-degree relatives were 15% less likely to complete the cascade than second-degree relatives, although this result was not statistically significant (p = 0.275). Female first-degree relatives were estimated to have a 43% (and males a 31%) probability of completing the cascade when indirectly contacted; female second-degree relatives were estimated to have a 47% (and males a 34%) probability of completing the cascade when indirectly contacted.

Of the relatives who completed cascading, 58.2% in Wales and 60.1% in Wessex were female, the average age was 40.1 years in Wales and 30.9 years in Wessex, and the mean observed LDL-C level was 4.92 mmol/l in Wales and 4.33 mmol/l in Wessex. Table 13 presents the CVD and LLT histories of cascaded relatives prior to genetic testing in each service by FH status and age group (see Report Supplementary Material 3 for further details). In both services, higher rates of CVD history were associated with older age groups and FH status. The probability that individuals had received LLT prior to the cascade was more common among relatives with CVD history and had a positive gradient with respect to age at testing. Relatives without a CVD history were more likely to be treated if they tested positive for FH. Numbers were too small to assess the influence of FH status on the probability of being on LLT among those with a history of CVD.

TABLE 13 - The CVD history and prior LLT rates of cascaded relatives, by service, age, FH status and CVD history

Age band (years)	Wales (%)						Wessex (%)
	FH positive			Non-FH and non-VUS			FH positive			Non-FH and non-VUS
	Individuals with CVD history	Individuals with prior LLT		Individuals with CVD history	Individuals with prior LLT		Individuals with CVD history	Individuals with prior LLT		Individuals with CVD history	Individuals with prior LLT
		CVD history	No CVD history		CVD history	No CVD history		CVD history	No CVD history		CVD history	No CVD history
Pooled	7.2	85.7	38.5	2.3	50.0	7.9	10.0	100.0	18.9	5.1	100.0	1.8
0–9	0.0	–	6.7	0	–	0.0	–	–	–	–	–	–
10–17	0.0	–	15.5	0	–	0.0	0.0	–	0	0.0	–	0.0
18–39	1.9	75.0	36.3	0	–	2.5	0.0	–	8.0	0.0	–	0.0
40–59	13.3	91.7	62.4	3.7	0	13.5	6.0	–	27.3	5.3	–	0.0
≥ 60	30.9	83.3	86.2	9.4	100.0	31.0	34.6	100.0	50.0	15.0	100.0	10.0

Summary of results and study limitations

This analysis used data from two of the largest FH cascade services in the UK and includes detailed patient-level records on 3839 indexes and 14,951 linked relatives. Analysis of index cases indicated high yields of individuals with FH within both services, with 21% of index cases in the Welsh service and 29% in the Wessex service testing positive for FH mutations. Among index cases meeting the full criteria for genetic testing, rates of FH were even higher, at 25–40% (depending on scoring criteria used) for Wales and 30% for Wessex.

In line with previous studies,10,126–128 we found a strong relationship between the stringency of thresholds used to select index cases for genetic testing and the likelihood that an individual carries a FH mutation. Increasing or decreasing the threshold used to inform which indexes are eligible for genetic testing will have implications for the number of individuals correctly identified as having FH and the number of individuals requiring genetic testing. The cost-effectiveness of alternative thresholds is evaluated in Cost-effectiveness of alternative cascade-testing protocols for familial hypercholesterolaemia.

Detailed analysis of the linked relatives shows that 25–29% of individuals who could potentially be contacted for cascade testing live outside the geographic area covered by the cascade service. The lack of a nationally co-ordinated service may prevent access to appropriate diagnosis and management for these individuals, and lead to duplicate genetic testing. In Cost-effectiveness of alternative cascade-testing protocols for familial hypercholesterolaemia, this information is used to assess the potential for a co-ordinated national service to improve the cost-effectiveness of cascade testing.

As shown by the systematic review in Chapter 2, Review 1: effectiveness of cascade-testing protocols among relatives for familial hypercholesterolaemia, very limited data are available to inform the preferred approach to contacting relatives. The available studies are not comparative in nature, and attempting to make inferences across studies is hampered by heterogeneity. The analysis of the Welsh data represents the first available within-service comparison of direct and indirect contact. This analysis shows that direct contact by the service, rather than contact indirectly via family members, is associated with a significantly higher probability that an individual will complete the cascade process. The cost-effectiveness of direct and indirect contact methods is evaluated in Cost-effectiveness of alternative cascade-testing protocols for familial hypercholesterolaemia. We also show that the probability of completing the cascade is significantly lower among men and among more distant relatives (second-degree or further), although the latter effect was observed in Wales only. Policy measures to enhance engagement among these groups are therefore warranted. Results were not comparable between services given fundamental differences in demography and relevant contextual factors.

We also identified a significant group of relatives who are within area, but not contacted by the service. Detailed information on why these individuals were not contacted was not available from PASS and remains an important priority for future research and data collection.

The strengths of this study lie in the large available sample size, the direct derivation of the data from two services, and the availability of FH specialist nurse input to support the development of additional variables (e.g. by recoding free-text fields within PASS) and provide a detailed understanding of the data.

The limitations of the study relate to its observational nature, the large proportion of missing data and generalisability for some variables. The assessment of unmodified Simon Broome scores in Wessex reduced the representativeness of our findings to the Wessex service. For the comparison of direct and indirect contact methods, we were able to adjust for the degree of relative and relative sex, but other confounding factors that are difficult to measure are also likely to be present. For example, age has been identified as a predictor of cascade success in previous work;41 however, our analysis failed to include age as a predictor because of missingness. As the choice of contact method was made on a case-by-case basis, the indirect method may have been used more frequently when direct contact was challenging because of a lack of information and engagement with the family. This would exert a downwards bias in our estimates of completion for indirectly contacted relatives. However, it is also likely that some indirect contacts that did not lead to further interaction with the service went unrecorded. This would exert an upwards bias in our estimates of completion for indirectly contacted relatives.

All analyses were conducted on a complete-case basis; this may have resulted in biases if the complete cases are not representative of the patients diagnosed by the services. This bias is likely to be particularly acute for the analysis of the characteristics of those who completed the cascade. For this group, CVD history and LLT history were missing for a substantive proportion of patients. This is an important consideration as these results may have significant influence on the results of the cost-effectiveness analysis.

Introduction

There is widespread consensus that people with FH should be diagnosed and treated early, given that treatment is effective, safe, cost-effective and inexpensive. 8,25,123,129–131 However, the magnitude of health losses and costs due to underdiagnosis, and the magnitude of the benefits from diagnosis and treatment, have had relatively little attention. The cost-effectiveness of diagnosis and treatment has been investigated mainly in the context of cascade testing. 11–13,25,110,130–134 Most studies predicted CVD risk based on data from the general population, adjusted upwards to reflect the increased risk for people with FH, and did not consider the impact of prognostic risk factors on CVD risk or the increased effect on CVD risk of exposure to LDL-C over time (known as ‘LDL-C burden’). The exception is the study by Ademi et al.,131 a cost-effectiveness analysis that compared screening and treatment at 10 years with no screening in the Australian setting, which used historical data from people with FH and accounted for LDL-C burden. However, in Ademi et al.,131 CVD risk depended only on age and results were presented for the entire individual population without exploring differences by prognostic factors.

Our decision model of the benefits of diagnosis and treatment is unique in that it uses data from UK people with FH (CPRD cohort), considers LDL-C burden17 and accounts for the impact of prognostic factors on lifetime risk of CVD events. A model was required because the CPRD cohort comprised diagnosed people who were on LLT and included few younger people, its follow-up did not span to individuals’ expected lifetime and there was no information on health-related quality of life over time.

The objective was to estimate the health benefits and health-care costs of diagnosis and treatment, compared with no diagnosis (and no treatment), to inform the cost-effectiveness analysis of alternative cascade protocols, which is reported in the next section. Our model allows us to capture the differential outcomes of policies that diagnose different individual groups, considering CVD risk factors (e.g. sex, LDL-C level) and age at diagnosis.

Methods

We took the perspective of the UK NHS and used 2019 prices, discounting future costs and health benefits at 3.5% per annum. 135 We validated the model using the Assessment of the Validation Status of Health-Economic decision models (AdViSHE) and the TECHnical VERification (TECH-VER) checklists132,136 (see Report Supplementary Material 4). We built the model in Microsoft Office Excel^® 2016 (Microsoft Corporation, Redmond, WA, USA).

Model structure

We designed the model based on previous cost-effectiveness models in FH,11–13,25,110,111,130–134 our understanding of the disease and of the impact of diagnosis and treatment, discussions with our stakeholder group, relevant clinical literature and the availability of data to parameterise key aspects of the model. Our conceptual model assumes that people with FH are at risk of CVD events, both fatal and non-fatal. The risk of a CVD event depends on an individual’s characteristics, defined earlier and included in the analysis of the time to the first major CVD event (see Economic analysis of the Clinical Practice Research Datalink database).

Figure 11 shows the model structure. The model is a cohort Markov model. People with FH enter the model in the ‘well’ state, either diagnosed (hence undergoing management of FH) or undiagnosed (hence untreated), depending on the option under evaluation. The age at model entry represents the age at diagnosis if they were diagnosed; the counterfactual assesses what would happen if they were not diagnosed at this same age. In the ‘well’ state, people are at risk of having the first major CVD event since diagnosis (i.e. model entry) and of non-CVD death.

FIGURE 11.

Model structure. Acute coronary syndrome includes unstable angina, myocardial infarction and ACS unspecified. TIA/stroke includes TIAs and ischaemic stroke. The first CVD event represents the first CVD event since model entry, which is the age at diagnosis if the model is evaluating the option in which patients are diagnosed, or the same age but under the counterfactual scenario that patients were not diagnosed.

The risk of first major CVD event depends on the age at diagnosis, sex, PT-LDL-C level and prior CVD history, as discussed in Economic analysis of the Clinical Practice Research Datalink database. This first CVD event can be fatal, non-fatal ACS (which includes unstable angina, myocardial infarction and unspecified ACS) or non-fatal ischaemic stroke/TIA. We distinguished between non-fatal ACS and non-fatal ischaemic stroke/TIA given their different lifetime implications to the risk of death, costs and health-related quality of life. The risk of all-cause death following a first non-fatal CVD event depends on the nature of this first major CVD event, either ACS or ischaemic stroke/TIA. The model cycle length is annual, with half-cycle correction. The model structure in Report Supplementary Material 4 justifies the model structure in more detail.

Model inputs

Table 14 summarises the model inputs, with details provided in the model inputs section of Report Supplementary Material 4.

TABLE 14 - Model inputs for the base-case scenario

Parameter	Value	Source	Details in (if applicable)
Risk of events
Reduction in LDL-C due to FH diagnosis (weighted average of reduction in ‘newly treated’ and ‘treatment retainers’)	33.4%	Analysis of CPRD cohort	Table 7 and Report Supplementary Material 4, Table 1
Risk of first CVD event	Generalised gamma risk equation	Analysis of CPRD cohort	Report Supplementary Material 4, Table 2–4, and Table 9
Distribution of people by type of first CVD event
Death	11%	Analysis of CPRD cohort	Report Supplementary Material 4, Table 5
Non-fatal ACS	65%
Non-fatal TIA/stroke	25%
Risk adjustment to CVD risk (HRs multiplied by hazard rate depending on time from model entry)
People aged < 40 years		Calculated from Perak et al.,96 figure 2	Report Supplementary Material 4, Table 6
10–20 years vs. 0–10 years	4.13
≥ 20 years vs. 0–10 years	6.44
People aged ≥ 40 years
10–20 years vs. 0–10 years	1.48
20 + years vs. 0–10 years	2.36
Effect of reducing LDL-C by 1 mmol/l on risk of CVD events	Calculated according to EAS equation	EAS consensus statement,17 table 2	N/A
Effect of reducing LDL-C by 1 mmol/l on the risk of non-CVD death	0.96 (95% CI 0.92 to 1.01)	Cholesterol Treatment Trialists’ Collaboration,18 webfigure 4A	Report Supplementary Material 4, Table 7
Risk of all-cause death following the first non-fatal CVD event	Constant probability over time by age at event, sex and type of event	Lewsey et al.137 Analysis of CPRD cohort General population mortality statistics138	Report Supplementary Material 4, Table 8
Risk of non-CVD death	Depends on age and sex	General population mortality statistics138	Report Supplementary Material 4, Table 9
Effect of FH diagnosis on costs
Costs of treatment, per annum
Cost of LLT, per annum	£21	Analysis of CPRD cohort Unit costs from the 2019 drug tariff139	See Report Supplementary Material 4, Table 10
Adverse effects of LLT: additional costs due to earlier cases of diabetes
Primary prevention, per annum	£3	NICE CG181,108 inflated to 2019 prices140	N/A
Secondary prevention, per annum	£6
Costs of monitoring, per annum
Adults in primary care in the year of LLT initiation	£137	Resource use based on NICE CG181,108 NICE CG71,8 HEART UK consensus statement141 and feedback from the stakeholder group Unit costs obtained from national costs140,142 and NICE CG181,108 inflated to 2019 prices	See Report Supplementary Material 4, Table 1–14
Adults in primary care in the subsequent years following LLT initiation	£35
Adults in secondary care in the year of LLT initiation	£556
Adults in secondary care in the subsequent years following LLT initiation	£170
Children in secondary care in the year post diagnosis who are not on LLT	£272
Children in secondary care in the subsequent years post diagnosis who are not on LLT	£112
Children and adolescents in secondary care in the year post diagnosis who were started on LLT	£723
Children and adolescents in secondary care in the subsequent years post diagnosis who were started on LLT	£336
Other inputs
Cost of health states, per annum
Cost of non-fatal ACS year 1	£8195	Walker et al.,143 inflated to 2019 prices140	See Report Supplementary Material 4, Table 15
Cost of non-fatal ACS year 2 and beyond	£2137
Cost of non-fatal ischaemic stroke/TIA year 1	£9244
Cost of non-fatal ischaemic stroke/TIA year 2 and beyond	£1990
Cost of CVD death	£2300
Cost of non-CVD death	£1929
Health-related quality of life
Age and sex adjustments	Regression model	Ara et al.144	N/A
Post first non-fatal ACS year 1	0.76	NICE CG181108	See Report Supplementary Material 4, Table 16
Post first non-fatal ACS year 2 and beyond	0.88
Post first non-fatal ischaemic stroke/TIA	0.72
N/A, not applicable.

Risk of first major cardiovascular disease event

To inform the risk of the first CVD event among patients who were diagnosed and treated, we took the risk estimated from the CPRD cohort, given that they were all diagnosed and treated, and the observed distribution by type of event (see Economic analysis of the Clinical Practice Research Datalink database). These risk equations predicted that the CVD risk was relatively stable over time, which we thought was unrealistic and did not reflect another study with a longer follow-up and a large sample size. 96 Perak et al. 96 used data from six large US epidemiological cohorts to identify people with a PT-LDL-C level indicating FH over a follow-up of up to 30 years. Using the curves of adjusted 30-year survival free from CHD presented in Perak et al.,96 which we digitised, we estimated a HR associated with longer follow-up by comparing the cumulative hazard rate for 0–10 years with those for 10–20 years and ≥ 20 years. We multiplied this HR by the estimated hazard rate (according to the risk equations estimated in Economic analysis of the Clinical Practice Research Datalink database after 10 years from model entry or being at risk of CVD, whichever was earlier).

To estimate the risk of the first major CVD event among undiagnosed (and untreated) people, we assumed that diagnosis affected the risk of the first CVD event only via LDL-C, and that the effect of LDL-C on CVD risk increases over time (known as ‘LDL-C burden’17,145). The proportional reduction in LDL-C was the same for all patients and corresponded to the average reduction observed among the treated patients in the CPRD cohort (estimated in Economic analysis of the Clinical Practice Research Datalink database), given that the reduction was similar between those treated before and those treated after a formal diagnosis (34.5% vs. 32.6%, respectively). The absolute reduction depended on the subgroup’s PT-LDL-C level. This assumed that (1) the PT-LDL-C level represents the LDL-C level that people would have had if they had not been diagnosed and managed; and (2) the proportional reduction in LDL-C is independent from the pre-treatment levels,99 and it is generalisable to all FH people irrespective of age, sex and PT-LDL-C level.

To account for the effect of LDL-C burden, we estimated the effect of absolute reductions in LDL-C on CVD risk with the equation proposed by the 2017 EAS consensus statement [risk reduction per 1-mmol/l reduction in PT-LDL-C = exp(−0.249 + (number of years of treatment − 5) × (−0.0152)], which relates the number of years in treatment to reduction in atherosclerotic CVD risk. 17 We assumed that the number of years of treatment corresponded to the number of years since diagnosis (i.e. model entry).

Figure 12 shows how the reduction in CVD risk depends on PT-LDL-C level and duration of treatment (i.e. time since model entry and diagnosis) according to the EAS equation. 17 For a 1-mmol/l reduction in PT-LDL-C, the risk reduction ranges from 17% after the first year of treatment to 61% after 50 years of treatment. At 5 years, the risk reduction is 22%, in line with the meta-analysis of trials of statin treatment. 18 Longer treatment durations result in greater risk reductions. For example, 54% risk reduction is achieved after 40 years, which is similar to the risk reduction estimated from Mendelian randomisation studies. 146–149 Greater risk reductions are achieved for larger reductions in PT-LDL-C. For example, a 2-mmol/l reduction results in risk reductions of 31–85%, depending on the time since model entry (or diagnosis).

FIGURE 12.

Relationship between treatment duration and reduction in CVD risk.

As only 22% of the CPRD cohort were aged < 40 years at the time of FH diagnosis, there is uncertainty in the extent to which the risk equations generalise to younger people. Following feedback from the stakeholder group and given that the youngest age at CVD event recorded in PASS was 25 years, we assumed that people are at risk of a CVD event from 25 years of age.

Results

Full results are available in Report Supplementary Material 4.

Gains in health outcomes among diagnosed people

Figure 13 shows the gain in event-free life expectancy due to diagnosis among males (see Figure 13a) and females (see Figure 13b). The gains are greater if diagnosis occurs at a younger age (e.g. approximately a gain of 20.86 years for males aged 0–9 years with PT-LDL-C = 5.5 mmol/l, compared with 5.34 years for males aged 40–59 years with the same PT-LDL-C level). People with higher PT-LDL-C have greater gains (e.g. gains among males aged 40–59 years range between 0.18 and 10.02 years, depending on their PT-LDL-C level). Gains are also more pronounced among people with prior CVD history. The gains among females follow the same pattern. The differences in gains in event-free life expectancy due to diagnosis are driven by differences in the risk of CVD events, the competing risk of non-CVD mortality, the absolute reduction in LDL-C given the same proportional reduction from diagnosis and the increased effect of lowering LDL-C over time. Report Supplementary Material 4, Figure 5, shows the gain in life expectancy. Gains in life expectancy are less pronounced, but still substantial. For example, among males, the gain is 0.08–10.35 years.

FIGURE 13.

Gain in event-free life expectancy due to diagnosis (not discounted, in years). (a) Males; and (b) females.

Figure 14 shows the gain in quality-adjusted life expectancy predicted by the model. This is the ‘QALY gain’, which, after discounting to present values, enters the calculation of cost-effectiveness. Similarly to the gain in event-free life expectancy and overall life expectancy, the differences in QALY gains reflect the different CVD risk and risk reduction given the PT-LDL-C level. The QALY gain ranges between 0.04 and 11.20 QALYs, depending on the subgroup. Once discounted to present values, the QALY gain is lower, at 0.01–2.11 discounted QALYs (see Report Supplementary Material 4, Figure 6).

FIGURE 14.

Gain in quality-adjusted life expectancy from diagnosis (not discounted, in QALYs). (a) Males; and (b) females.

Impact of diagnosis on costs

Figure 15 shows the impact of diagnosis on NHS costs over people’s expected lifetimes (see Report Supplementary Material 4, Figure 7, for discounted results). In subgroups with lower PT-LDL-C levels (< 1.5–3.5 mmol/l, depending on age and prior CVD history), diagnosis increases costs. This is because the risk of CVD events among these people is lower, and the risk reduction due to diagnosis is also lower, as it depends on their PT-LDL-C level. Therefore, the additional costs of LLT and monitoring are not fully offset by the savings in the costs of the avoided CVD events. The cost increase is more pronounced in younger subgroups with lower PT-LDL-C levels because they incur monitoring costs from diagnosis, are treated from 10 years of age, but benefit from CVD events only from 25 years of age. With greater levels of PT-LDL-C, the diagnosis results in larger savings in younger subgroups (than older subgroups), as their CVD risk is higher and diagnosis prevents a larger number of CVD events. The savings are greater in subgroups with prior CVD history because they are at higher risk of CVD events. The additional costs of FH diagnosis are mostly due to monitoring, given the small acquisition cost of statins.

FIGURE 15.

Impact of FH diagnosis on costs (undiscounted). (a) Males; and (b) females.

Impact of diagnosis on net health to the NHS

Figure 16 shows the net health gain of diagnosis at a cost-effectiveness threshold of £15,000 per QALY, discounted to present values (see Report Supplementary Material 4, Figure 8 for results according to the cost-effectiveness threshold of £20,000 per QALY). The net health gain from diagnosis ranged between –0.27 and 2.51 QALYs at the £15,000 per QALY threshold. Diagnosis represents a net health gain for the NHS for most males (see Figure 16a) unless their PT-LDL-C level is < 1.5 mmol/l if aged < 39 years or < 0.5 mmol/l if aged ≥ 40 years or had prior CVD history. The results are similar for females (see Figure 16b), although the net health gains are generally not as large, and diagnosis is not a net health gain for females aged 0–39 years with PT-LDL-C levels of ≤ 2.5 mmol/l. This is because females are at lower risk of CVD events.

FIGURE 16.

Net health gain of FH diagnosis and treatment (compared with no diagnosis and no treatment) results, by PT-LDL-C level, in QALYs per individual, at a cost-effectiveness threshold of £15,000 per QALY. (a) Males; and (b) females. Results are probabilistic. Net health gain = (QALYs if individual group was diagnosed – QALYs if individual group was not diagnosed) – (costs if individual group was diagnosed – costs if individual group was not diagnosed)/ cost-effectiveness threshold at £15,000 per QALY.

The net health gain increases with higher PT-LDL-C levels. The pattern of net health gain by age is not as clear, because of the effect of discounting and the age at which we assumed people start being at risk of CVD events (25 years of age in the base-case scenario). As a result, people diagnosed younger start incurring costs, but derive benefits only later in life, both of which are discounted to present values at 3.5% per annum.

Uncertainty analysis

The probability that diagnosis is a net health gain to the NHS (i.e. cost-effective) follows a similar pattern (see Report Supplementary Material 4, Figure 9). The probability is close to 1 in most subgroups for which diagnosis is a net health gain, and close to 0 in subgroups for which the diagnosis is a net health loss (e.g. subgroups with PT-LDL-C levels of < 1.5 mmol/l and no prior CVD history).

Table 15 summarises the results of the scenario analyses (details are in Report Supplementary Material 4, Table 28). Diagnosis was a net health gain in the base-case scenario and in all scenarios in most of the subgroups (cells in white). In the subgroups with PT-LDL-C levels ≥ 2.5 mmol/l, diagnosis was a net health gain in the base-case scenario and in most of the additional scenarios (cells in light grey). In the subgroups with no prior CVD and low PT-LDL-C levels (0.5–1.5 mmol/l), diagnosis was not a net health gain in any (cells in black) or most of the scenarios (cells in dark grey).

TABLE 15 - Results of the scenario analyses

Note

The cells show the number of times (base-case and additional scenarios) that diagnosis was a net health gain, for the cost-effectiveness threshold of £15,000 per QALY:

Black – if diagnosis was not a net health gain in either the base-case scenario or additional scenarios.

Dark grey – if diagnosis was a net health gain in 1–10 of all the scenarios (i.e. base case and additional).

Medium grey – if diagnosis was a net health gain in 11–20 of all the scenarios.

Light grey – if diagnosis was a net health gain in 21–29 of all the scenarios.

White – if diagnosis was a net health gain in all 30 of all the scenarios.

Sex and PT-LDL-C (mmol/l)	No prior CVD, age (years)					With prior CVD, age (years)
	0–9	10–17	18–39	40–59	≥ 60	18–39	40–59	≥ 60
Males
0.5	0	0	0	0	1	0	3	13
1.5	0	1	3	20	27	28	28	28
2.5	20	25	25	29	29	30	30	30
3.5	26	27	30	30	30	30	30	30
4.5	28	30	30	30	30	30	30	30
5.5	30	30	30	30	30	30	30	30
6.5	30	30	30	30	30	30	30	30
7.5	30	30	30	30	30	30	30	30
8.5	30	30	30	30	30	30	30	30
Females
0.5	0	0	0	0	0	0	1	1
1.5	0	0	0	1	14	9	28	28
2.5	1	1	3	24	29	30	30	30
3.5	2	26	24	30	30	30	30	30
4.5	26	27	30	30	30	30	30	30
5.5	28	30	30	30	30	30	30	30
6.5	29	30	30	30	30	30	30	30
7.5	30	30	30	30	30	30	30	30
8.5	30	30	30	30	30	30	30	30

The model was most sensitive to scenarios with alternative assumptions about the long-term CVD risk and the effect of LDL-C burden over time, and to alternative assumptions about the benefits and costs of diagnosis. For example, not considering the effect of LDL-C burden on CVD risk (instead assuming that 1-mmol/l reduction in LDL-C reduces CVD risk by 21%, as per the effect observed in trials of statins18) results in diagnosis not being a net health gain in 14 additional subgroups, as the (counterfactual) CVD risk among undiagnosed/untreated people is lower than in the base-case scenario. If the risk of the first CVD event does not increase over time, as suggested by our analysis of the Perak et al. 96 data, diagnosis is not a net health gain in an additional 12 subgroups. Relating to the effect of diagnosis on health outcomes, the scenario with the greatest impact was if diagnosis reduces LDL-C to the EAS targets of 3.5 mmol/l in children and adolescents, 1.8 mmol/l in adults in primary prevention and 1.4 mmol/l in adults in secondary prevention. 123 Under this assumption, diagnosis is not a net health gain in an additional eight subgroups, namely subgroups with PT-LDL-C levels lower than the target, as they are assumed to experience no further reductions, but incur the costs of monitoring. The scenario on the cost implications of diagnosis with the greatest impact was the assumption that 50% of adults are monitored in primary care (vs. 75% in the base-case scenario).

The expected value of perfect information is small, at a maximum of 23.51 and 25.12 QALYs per 1000 people at the £15,000 per QALY and £20,000 per QALY thresholds, respectively (see Report Supplementary Material 4, Figure 10). The small expected value of perfect information suggests that the impact of parameter uncertainty on decision uncertainty is small.

Introduction

Cascade screening/testing is recommended by NICE8 and international guidelines,123,161 and has been shown to be cost-effective. 11–13,25,110,130–134 The cascade-testing process can differ in various dimensions, such as method of contacting relatives (directly by the FH service or indirectly via the index case), type of cascade contact pattern (contacting second-degree relatives sequentially only when their first-degree relative was confirmed to have FH or contacting first- and second-degree relatives simultaneously) and testing strategies (e.g. using LDL-C level as a screening test prior to the genetic test); we term the various ways in which cascade testing can be implemented as ‘cascade protocols’. Most cost-effectiveness studies compared cascade screening plus genetic testing with no cascade,12,110,130,131,133 with three studies including testing strategies relying on lipid levels,13,25,134 and no studies, as far as we are aware, comparing protocols with different methods of contact, contact pattern and testing strategies.

This cost-effectiveness analysis compares a wide range of alternative cascade protocols, comprising alternative cut-off points to select index cases to genetic testing and/or the cascade, contacting relatives directly or indirectly, contacting relatives sequentially or simultaneously, and using lipid and genetic tests in combination. The aim was to estimate the short-term and long-term health outcomes, costs and diagnosis yields, and to identify the cost-effective cascade protocols.

We developed a new decision model to simulate cascade protocols, so that we could compare them without requiring their implementation in practice, and to predict long-term health outcomes and costs, which are observable only in a study with very long follow-up. We informed the model with the characteristics of index cases and relatives estimated in the analysis of PASS data from the Welsh and Wessex FH services (the data are described in Chapter 3, PASS Wales and Wessex Cascade Service Data, and the analysis is described in Service data analysis); the long-term health outcomes and costs of diagnosed and undiagnosed FH patients estimated by our cost-effectiveness model, described in Cost and health benefits of diagnosis of people with familial hypercholesterolaemia in the long term; and other sources of data and input from the stakeholder group.

Methods

The cost-effectiveness analysis took the perspective of the UK NHS at a 2019 price base. We discounted future costs and health benefits to their present value at 3.5% per annum. 135

Conceptual model

Figure 17 shows our conceptual model. We interpreted the cascade as a three-stage process: (1) selecting index cases to be cascaded, (2) contacting relatives and (3) testing relatives.

FIGURE 17.

Conceptual model. Relatives are correctly diagnosed if their index is selected to the cascade, the relatives are tested, and the test is correct (full arrows). Relatives are not diagnosed if their index is not selected to the cascade, or if relatives are not tested. FH relatives who are correctly diagnosed are treated with LLT, and hence have lower CVD risk than FH relatives who are not diagnosed. Our base-case scenario conservatively assumes that FH relatives who are on LLT at the time of the cascade do not have additional health benefits or costs from FH diagnosis, and hence do not benefit from diagnosis. This is tested in a scenario analysis, in which we assume that diagnosis reduces LDL-C level by 50% irrespective of prior treatment. Non-FH relatives who are correctly diagnosed have no change in management. We assumed that non-FH relatives who are incorrectly diagnosed incur greater costs (because of the monitoring), but experience no health effects. Even though some non-FH relatives are at high CVD risk (because of other risk factors, e.g. age) and may benefit from LLT, we excluded this effect in the base-case scenario because the management of these patients should be conducted independently from cascade testing. MFH, monogenic familial hypercholesterolaemia.

Results

Context

Three regions were chosen to reflect differing current approaches in the UK, including varied methods of contacting relatives for FH cascade testing, different service structures, and varied numbers and types of health-care professionals and appointments involved in cascade-testing pathways. Region A has a mixed urban, semi-rural and rural, predominantly white, population. Region B includes a major city, with a non-white population of 14%, and surrounding semi-rural population. Region C includes two large, predominantly semi-rural counties, one with two major cities that include a non-white population of 15%. These regions are summarised in Box 1 by cascade-testing approach, based on information from service leads and participating professionals, and operating service specifications.

Sampling and data generation

A purposeful sample of individuals and relatives was chosen to reflect a sociodemographic range of index patients and relatives identified by cascade testing, across the three regions, including differing FH assessment and testing experiences, including contact within or outside an index patient’s area and with different health providers. A purposeful sample of health professionals was similarly sought to include a spectrum of roles and experiences in relation to differing types of FH index and cascade-testing provision in the three regions.

Semistructured one-to-one interviews were conducted by telephone (from October 2018 to December 2019) using a broad topic guide to explore the views and experiences of all respondents, and lasted for an average of 45 minutes. Prior to interview, all participants provided written informed consent.

Data analysis

Interviews were audio-recorded, anonymised, transcribed verbatim and checked for accuracy, and data were organised in NVivo version 11 (QSR International, Warrington, UK). For each set of patient/relative and health professional interviews, transcripts were read and reread to gain familiarity with the data and underwent line-by-line coding following a developed coding framework; similar codes were grouped together to construct themes.

Themes for each set of interviews were revised iteratively until final themes were agreed on. Two researchers, with differing disciplinary backgrounds in health psychology (field researcher/interviewer, LC) and academic primary care (health researcher and practising GP, JK), independently developed the coding and themes. These were developed in collaboration with the study PPIE lead (WR). Data analysis was ongoing alongside data collection until thematic saturation had been reached. Framework analysis170 was then used to chart the data and compare themes across regions in relation to the perceptions, views and concerns of patients and relatives, and health-care professionals involved in different cascade-testing approaches.

In a further stage, member checking171 was used to help enhance validity. All participants who agreed to being approached later in the study were subsequently invited to review and comment on summaries of preliminary findings (from patient/relative or health professional interviews according to whether they were a patient/relative or health professional, respectively) to check for accuracy and resonance with their experience, and to enable opportunity to review, develop or confirm findings.

Communication about familial hypercholesterolaemia and testing in families: influence on acceptability of indirect and direct approaches

Index patients felt able to have conversations about FH testing with family members if they had close relationships. They described using communication they felt would help, such as focusing on the practicalities of blood testing, or using a jocular or informal tone, rather than attempting to convey detail:

I think it is probably better the more simplified it is...., so if people are just expecting to go for a blood test for screening...., I think people might have maybe less resistance to something which seems relatively harmless, than if they get overwhelmed with the big terms.

Index, male, 27, region A – direct

We kept it tongue in cheek and light-hearted and it was just a case of ‘There is this gene that we have inherited, our suspicions are that it goes back to my grandmother’.

Index, female, 40, region C – hybrid

Across all regions, health professionals found use of a solely indirect contact approach less successful. They were concerned that index patients might not convey letters from their service or explain the importance of getting tested to their relatives. Professionals also felt that they could not be confident what information about FH or testing had been communicated. In particular, success of the approach was beholden to relationships within the family. These factors could all increase opportunity for relatives to disengage from a cascade-testing offer:

If we take the pure indirect method, we have got no way of being really sure that the correct information has been passed on to relatives, so obviously there is a point of weakness there... and of course in some families... they refuse to speak to relatives and then we don’t really get to see those people because the family communication has broken down.

FH nurse, direct contact approach

An indirect approach had worked well for some patients with good family relationships, and for those who chose to informally support a direct approach by communicating with relatives to forewarn them about future correspondence they would receive from the specialist service:

We are a close family; I am close with my brother, so it was very easy for me to speak to him about it and saying ‘look I have had this done, I have got this diagnosis, you are the only one that really it could affect next’.

Index, female, 46, region B – indirect

I was sort of tipping the relatives off that somebody would be getting in touch with them from the hospital, so they knew that I had had this incident [diagnosed with FH] and that they were looking at screening everyone [in the family].

Index, male, 52, region A – direct

However, an indirect approach could often leave index patients feeling pressured, or unable to do this in the context of a negative relationship. They felt that they lacked the necessary information and wherewithal themselves, and that this was better and more appropriately undertaken independent of the family:

My [daughter] is a nightmare and it would be good for her to speak to somebody outside of the family, because us trying to get through to her the seriousness of the condition, she just says ‘oh’, you know, ‘you’re always on at me’.

Relative, female, 42, region C – hybrid

... He [brother] doesn’t want to know, and I don’t have enough to do with him really to thrash that one out with him.

Index, female, 56, region B – indirect

... people maybe don’t want you to contact them. I have fallen out with my sister, so it would be quite hard for me to contact my sister about this. I would rather go for someone else contacting her.

Index, male, 46, region B – indirect

In contrast, index patients spoke of several advantages of a direct approach. Even when close family relationships existed, some were concerned about communicating information to family members appropriately. Some also felt frustrated at being unable to persuade reluctant relatives to engage with testing, and would have opted for health professionals’ direct contact if this had been offered. Contact by service health professionals was seen as carrying more helpful credibility and authority:

... I understand people might not be too grateful that they have been summoned to hospital for something that they might have never heard of... but at the same time I think it might be quite difficult for family members to explain what it is.

Index, male, 27, region A – direct

Receiving a letter from somebody in the medical profession who is an expert in it carries more weight than just me saying you need to get this test done.

Relative, F, 77, region B – indirect

I think if a hospital was to contact you and say you really need to come for this test, I think it would be much more effective than your mum....

Relative, male, 52, region C – hybrid

Corresponding with patients’ perspectives, health professionals found that using a direct approach better ensured that contact with the offer of testing was made to relatives and that information about FH and testing could be conveyed appropriately and accurately. Some also highlighted their concerns about a potential ethics difficulty arising with an indirect approach, that returning to the index patient to determine if or why their relatives had not engaged may breach patient confidentiality. This meant that a solely indirect approach could create a dead end for engaging relatives. Using a direct contact method meant that services were better able to follow up non-response from relatives by telephone or by sending another letter:

I think the one frustrating thing about indirect cascade testing is, if a family doesn’t engage, where can you go from there really? You know what are the ethics in contacting the index again and saying ‘Did they get the letters?’ because you can’t really let the index know that people haven’t come in. [With] the direct method you can send letters again.

FH nurse, hybrid approach

Health professionals perceived, or had found, that a more flexible combined or ‘hybrid’ approach worked more effectively. This was informed by exploring the nature of relationships within families with the index patient, and their views on what would best work for them and their relatives. This entailed negotiating using direct or indirect or both methods for each relative. This might, for example, involve index patients wishing to inform a relative and pass information to them (indirect), but also with direct follow-up contact from the service. Conversely, an index patient might prefer their relative be contacted directly by the service only because of the nature of their relationship or because of insights about how their relative may prefer to be contacted or to hear about FH:

[I]t is just having that freedom for that hybrid approach and getting consent at that point to say ‘right, which relatives can we just go straight to? Which ones do you think need a bit more nurturing?’ Direct isn’t [always] better than indirect, it is a combination of the two that works best for each individual family... that’s another privilege of having [FH nurse-led model] within this service, that [they] have an hour with each patient,... while you draw the pedigree, you’re really finding out what the dynamic is in the family.

FH service lead, hybrid approach

For similar reasons, most index patients and relatives had experienced or perceived this ‘blended’ combination approach to be more acceptable:

... a blended approach, definitely, because I think if you do receive that letter, that can be, in itself, quite worrying, but purely relying on the individual [index], it is entirely dependent on that individual’s ability to convey the message and the seriousness of it.

Index, male, 43, region C – hybrid

Use of a FH nurse-led model appeared well suited to this approach. This was perceived to offer particular strengths in establishing continuity of care for index patients and their relatives, providing a single point of contact for them to maintain engagement with the process, and facilitating a therapeutic relationship and understanding of families’ needs. This was greatly valued by families who had experienced this approach:

... it definitely works better, with the nurse doing the whole thing, I think it is better for the patient and the family because they have got one consistent contact who then knows the dynamics of the family better and also the referral process.

FH nurse, indirect contact approach

The [FH] nurse that does all of the genetic testing and things and it is perfect, she is really, really nice and you don’t feel rushed, your family get to ask all of your questions and she came to my daughter’s first appointment too.

Index, female, 40 region C – hybrid

Communication and being informed about familial hypercholesterolaemia

Having positive interactions with health-care professionals who communicated effectively about FH reassured patients about FH testing and what a positive result would mean for their lives. People who had been able to discuss FH testing with their FH nurse, genetic counsellor or consultant recalled a good understanding of FH and the principles of genetic inheritance for FH:

They explained it to me as trying to find the gene in the human genome is like trying to find a spelling mistake in a novel, but once you know what page and what word is spelt wrong, then it is much easier for them to then see if my boys had the same thing.

Index, male, 43, region C – hybrid

Some health professionals felt that they or other colleagues lacked adequate time during busy appointments to ensure that messages had been understood by the patients, or to discuss family issues appropriately, compromising uptake of testing. This required well-developed communication skills and confidence to address any deeper issues that might act as a barrier to engagement with cascade testing, such as guilt or loss related to FH diagnosis, or sensitivity about inheritance or family issues:

... I think genetics is perceived as being a little bit mysterious, and in a busy medical appointment it is quite difficult to explain that.... I have got plenty of people who don’t turn up to their genotyping appointment and that is possibly my fault because I haven’t had the opportunity to explain in a way that they recognise the value of bothering to get tested.

Consultant lipidologist, indirect contact approach

... we have had issues with paternity come up, you know there are all sorts of potential issues,.... It is that whole perspective of the individual versus the family, but also having the confidence to explore because, if there are issues around bereavement or guilt and blame,....

FH nurse, hybrid approach

Familial hypercholesterolaemia nurses were considered to be well placed to take on this counselling role with support and training, while bringing specialist clinical knowledge and potentially greater time during appointments than medical colleagues.

A strong theme prioritised by health professionals was that communication about testing be focused on FH having effective treatment to prevent major cardiac events, and rather less foregrounded as genetic testing for a genetic disease. Introducing this too early could be too disconcerting and lead to patient disengagement:

When we come from a genetic test perspective, the language is far too alarming and we almost put them off the idea of having a test at all... thinking about this as a cardiac condition which is preventable, I think, is more helpful.

Consultant lipidologist, indirect contact approach

Engaging and communicating with a range of people with different health literacy was seen as an issue in this context. Professionals perceived that those from more socially advantaged and educated backgrounds may access FH services to a greater extent and be better able to assimilate the information presented to them:

... [focusing on genetics] does mean that you lend yourself as a service to people who are better resourced [in understanding], so people who perhaps have got higher levels of education, whose first language is English, etc., and it makes it more challenging for people who aren’t of that background, I think, to engage.

Consultant lipidologist, indirect contact approach

... we see a lot of middle-class people; we do see other people, but we see a lot of people from... the higher your social class and your intellect, the better you tend to be at accessing services.

FH nurse, hybrid approach

An example of a family’s confusion about FH testing emphasised the importance of how FH and its inheritance is communicated to wider family members:

... we traced it [FH] back on my mum’s side but they didn’t offer my dad a test... he is absolutely convinced that he will be [have FH] as well because of his family medical history.

Index, female, 40, region C – hybrid

Some patients and relatives had been signposted towards more formal medical resources, such as journal articles and NICE guidelines, and found this off-putting and overwhelming. They appreciated provision of trusted information they were better able to understand, such as from the charity HEART UK, and also accessed wider NHS and external resources online or in written forms from other sources.

Participants felt that wider awareness raising and public understanding about FH was needed if its detection and successful cascade testing was to occur, including at school and through public media:

I mean, there is no one that goes on Holby City with FH is there?... make [people] look at themselves... we realised my grandma’s dad had a massive heart attack and died, grandma had a triple heart bypass... and my mum run ultra-marathons... yet there is still problems, why is that?... So it’s up to individual families to question it, isn’t it?

Relative, female, 23, region C

Access to and organisation of familial hypercholesterolaemia testing pathway

Improving service pathways and patient-centred services

Individual contexts for engaging with familial hypercholesterolaemia testing

Seeking an explanation for family history of cardiac events or premature death, and the desire or strongly felt obligation to protect younger generations of the family from the same, were the most common motivations for engaging with FH testing. Most people had been positive about and unsurprised to receive the offer of FH testing and subsequent diagnosis or its exclusion:

Everybody [in family] has been receptive to it [FH testing] and keen to know if potentially they have it because my mother’s stroke was so sudden and unexpected....

Index, female, 40, region C

Index patients explained why their relatives had not taken up the offer of cascade testing. Some, particularly if only indirectly contacted, were deterred by long waiting times for first appointments, or if several successive appointments with different professionals had been involved. Others noted their relatives’ nature, concerns about impact on lifestyle if diagnosed, having to take statins or similar medication, uncertainties about future employment or obtaining life insurance:

I was fine about it [testing], because the consultant had explained... it is manageable, I was OK about it. My brother wasn’t, to be honest, he went into a bit of a meltdown and started panicking about life, etc., but he is a different character to me.

Index, female, 42, region C

There has been all of this talk as well about the statins being not good... conflicting evidence all of the time... conflicting news reports... I think that doesn’t help people or puts people off.

Relative, male, 52, region A

I can remember having the conversation with my eldest son. He was worried that if he had got [tested], it was going to affect his insurance... that is why he thought he best not know.

Index, female, 68, region A

Parents with younger children

Respondents who were parents of younger children, and who had experience of being tested for FH, spoke about their conflicting feelings when considering testing their children. Considerable uncertainly and anxiety existed for some in not being able to test their children of pre-school age, with a concern that they be able to plan for the future as early as possible:

I understood that they wouldn’t medicate him... but I just wanted to know what I was in for. Every time I had an appointment with the consultant, I said ‘Can he come and get tested now?’, ‘No, he is only 5’. When he was 8, I said ‘I really want to know, he is not that far off being 10’.

Index, female, 46, region B

For older children, the decision could be more straightforward, with respondents feeling empowered by knowledge that could help their children’s future health, and for their child to be similarly informed. Others were more ambivalent and wondered if they had rushed their decision to establish if their child had FH. For example, they reflected on whether or not they should have given more thought to how this could affect their children in later life, particularly around potentially limiting career options and the ability to get life insurance:

... it never even occurred to me not to get them tested. Especially for my son... having such a high chance of heart attacks before the age of 35.

Relative, female, 42, region C

I have a new appreciation, I think, for why you may not want to get your children tested; I went gung hoe and all guns blazing. Now I think maybe I shouldn’t have got him tested, not actually got him a diagnosis, just had his cholesterol treated, if I really had thought it through properly. What have I set him up for in adult life?... his future and job prospects, certain careers, and a mortgage... have I scuppered his chances of getting life insurance?

Index, female, 46, region B

One mother described how she was unable to access psychological support for her teenage daughter following her cascade testing and FH diagnosis, which had amplified anxiety against a background of an existing condition:

My daughter has had a really bad year, she now has quite severe anxiety... You know the problem is her processing the fact that for the rest of her life she has a hole in her heart, plus a faulty gene... it has generated another concern and we feel like we have got nothing and nowhere with it.

Index, female, 40, region C

Overall, however, respondents were positive about their experience following cascade testing of their children, and again underlined the value of having continuity of care with a FH professional who had become known and trusted:

We took the children to the clinic... for blood tests, and I then had a phone call from [my FH nurse] to give me their results.... When my daughter sees the child consultant, [FH nurse] is always there as well.

Index, female, 42, region C

He [young son] has got a positive [FH] diagnosis, but... the main thing for me actually was having my consultant for both of us... he knows my history and you build up a rapport.... I am just so glad he is involved in [son’s] treatment as well.

Index, female, 46, region B

Summary of results and study limitations

This qualitative study has provided insights into the acceptability of different cascade-testing approaches in practice, and what happens in service pathways that may help or hinder engagement with testing.

For index patients and relatives, family relationships shaped experience of, and preference for, an indirect or a direct cascade approach or a combination of both. Experience of quality of communication about FH, the accessibility and organisation of pathways, and continuity of care further determined acceptability of approach. For index patients, making indirect contact worked less well, creating pressure about engaging relatives unless their relationship was conducive, or concern about communicating appropriate FH information. Direct contact of relatives by a specialist service was preferred as offering an independent, more credible and authoritative approach.

Professionals’ experiences echoed the perspectives of index patients and relatives, reporting more success using a direct approach. Using a solely indirect approach by index patients could work well for some, but was commonly problematic because of the unreliability of invitations being passed on, uncertainty about what may be communicated and ethics issues in following up relatives’ non-response. Deploying a direct approach improved confidence that these issues would be avoided.

Combining approaches by using a flexible ‘hybrid’ model was regarded as effective by both patients and professionals. This involved FH professionals developing an understanding of what was appropriate for individual families in discussion with index patients, to inform whether to use an indirect or a direct approach, or a combination of both, for each relevant family member. A FH specialist nurse-led model providing adequate time for enhanced communication and continuity of care for families from commencement and throughout the cascade-testing pathway was preferred and strongly supported.

To further improve acceptability of cascade testing, professionals emphasised that communication with families should focus on FH as effectively treatable to prevent major cardiac disease, rather than as a forbidding genetic condition. From both professionals’ and patients’/relatives’ perspectives, barriers to uptake of cascade testing could be reduced by removing the conventional requirement for GP referral, enabling self-referral to FH services when possible, and limiting the number of appointments and different health professionals in testing pathways. Having continuity of care with a single FH professional, such as a consultant or a FH nurse, for ongoing family support greatly helped. This was particularly valued by parents with younger children who had been tested.

We acknowledge some limitations, which are challenges for future research to address in this field. Index patient and relative participants were white and English-speaking; this lack of ethnic diversity reflects that seen in FH services, and thus who we were able to access and engage for interview and sampling.

Although some relevant experience was captured through the accounts of those participating, we did not secure direct testimony from relatives who declined FH cascade testing or from second-degree relatives. We were unable to identify participants in these groups willing to be interviewed. We note that more women than men were interviewed, perhaps reflecting women’s commonly greater willingness to participate in research or as parents of index patients. Finally, although genetic counsellors were not directly involved in two FH service settings, those in the third were in the midst of genetic services reconfiguration and did not respond to the invitation to participate.

These qualitative findings must be interpreted with regard to the selected samples as described, which may aid assessment of the relevance of these findings beyond this study’s context.

Yield of cases

Our systematic review of different cascade protocols included 24 non-comparative studies. Based on four studies, the combined approach to contacting relatives led to a slightly higher yield of relatives tested (40%, 95% CI 37% to 42%) than the direct (33%, 95% CI 28% to 39%) or indirect approach alone (34%, 95% CI 30% to 37%). However, only 26% of the index cases in the combined approach had a confirmed diagnosis. The systematic review to assess the diagnostic accuracy of clinical and biochemical criteria to diagnose relatives included nine studies; no study was found reporting relatives’ characteristics that could inform yield of cases, and hence the cost-effectiveness analysis.

Analyses of the PASS databases identified that 21–29% of index cases have genetically confirmed FH, and women, first-degree relatives and those contacted directly by health professionals were more likely to complete cascade testing (p < 0.01). The Welsh PASS database analysis also showed that direct contact by the service, rather than contact indirectly via family members, is associated with a significantly higher probability that an individual will complete the cascade process, although the probability of completing the cascade is significantly lower among men and among more distant relatives (second-degree or more distant). The database also identified that the yield of cases could be affected by locality, with 25–29% of individuals who could potentially be contacted for cascade testing living outside the geographic area covered by the cascade service.

Treatment patterns, and short- and long-term outcomes

The search for relevant systematic reviews of the effectiveness of cholesterol-lowering therapies on LDL-C levels and CVD among adults identified 14 relevant systematic reviews; none of these met the methodological quality on the AMSTAR critical appraisal tool to be included in the review of reviews. However, the CPRD–HES data set indicated that 26% of patients with coded diagnosis of FH in their primary care records remain untreated after 2 years, and < 30% achieve the reductions in LDL-C recommended by NICE FH guidelines (≥ 50% reduction).

Using the primary care CPRD and specialist FH register data sets, we were also able to explore the lifetime consequences in terms of morbidity and mortality among diagnosed and treated individuals with FH. The CPRD demonstrated substantially increased CVD risk among individuals with FH codes in primary care records, compared with controls (HR 9.14, 95% CI 8.55 to 9.76), and, over 20 years of follow-up, they were more likely to develop CVD if they had previous disease or a higher baseline cholesterol level. Furthermore, we estimate that, at 10 years of follow-up, the average risk of first major non-fatal CVD event or CVD-related death is 11%. History of CVD was identified as a key prognostic variable, with age, sex and (baseline) PT-LDL-C level also playing important roles. A specialist FH register confirmed excessive CVD morbidity among those with FH (SMbR 7.17, 95% CI 6.79 to 7.56), with particularly high morbidity among women and younger patients with FH.

However, identifying evidence of long-term CVD outcomes among children diagnosed with FH was limited. A preliminary scoping review identified two systematic reviews, one from 201620 and one from 2019. 19 In particular, Lozano et al. 20 had indicated no studies looking at the (long-term) relationship between LLT for children and CVD events in adulthood, or the association between intermediate outcomes in childhood (e.g. lipid concentrations, atherosclerosis) and CVD event. We concluded that, owing to the nature of these studies, such data would not be available.

Cost-effectiveness of alternative protocols

The cost-effectiveness analysis involved the development of two new cost-effectiveness models. One model predicted the long-term health outcomes of individuals with FH. The results informed the second model, which compared alternative cascade-testing protocols.

Considering long-term health outcomes and costs, the diagnosis of people with FH generally represents a positive net health gain to the NHS (i.e. it is cost-effective) for most people with a PT-LDL-C level of ≥ 2.5 mmol/l or who have prior CVD history. The net health gain from diagnosis and treatment ranged between –0.27 to 2.51 QALYs at the £15,000-per-QALY threshold. In general, the net health gain of diagnosis is greater among males, people with higher PT-LDL-C levels and people with prior CVD history at diagnosis. The assumptions with the largest impact were those about the risk of first major CVD event, in particular the effect of LDL-C burden on CVD risk and the increase in CVD risk over time.

The cascade-testing protocol with the highest net health gain, the base-case cost-effective protocol, involved diagnosing relatives according to their age, whether or not they were on LLT, their PT-LDL-C level and, when this information provided insufficient diagnostic certainty, conducting genetic testing (as presented in Table 20). Compared with not conducting cascade testing, per index family assessed for cascade, the cost-effective protocol diagnoses 52% (n = 0.31) of relatives with the disease, at a cascade cost of £536, and with an incremental cost-effectiveness ratio of £13,996 per QALY gained. The cost-effective protocol using the same testing strategy for all relatives regardless of age, the harmonised cost-effective protocol, achieves similar outcomes and may be preferable if additional nurse time (unaccounted for here) is required to implement a testing approach that differs according to the age of relatives.

Acceptability of cascade-testing approaches

For index patients and relatives, family relationships shaped experience of, and preference for, an indirect or a direct cascade approach, or a combination approach. Experience related to quality of communication about FH, the accessibility and organisation of pathways, and continuity of care further determined acceptability of approach. For index patients, making indirect contact worked less well, creating pressure about engaging relatives unless their relationship was conducive, or concern about communicating appropriate FH information. Direct contact of relatives by a specialist service was preferred as offering an independent, more credible and authoritative approach.

Professionals’ experiences echoed the perspectives of index patients and relatives, reporting more success using a direct approach. Using a solely indirect approach by index patients could work well for some, but was commonly problematic because of unreliability of invitations being passed on, uncertainty about what may be communicated and ethics issues in following up relatives’ non-response. Deploying a direct approach improved confidence that these issues would be avoided.

From both professionals’ and patients’/relatives’ perspectives, barriers to uptake of cascade testing could be reduced by removing the conventional requirement for GP referral, enabling self-referral to FH services when possible, and limiting the number of appointments and different health professionals in testing pathways. Having continuity of care with a single FH professional, such as a consultant or a FH nurse, for ongoing family support greatly helped.

Yield of cases and acceptability of different cascade-testing approaches

The acceptability of different contact strategies may partly explain their effectiveness. Qualitative interviews with FH patients have demonstrated mixed views about the approach to cascade testing: a 2011 Scottish study favoured indirect cascade testing,88 whereas, in a 2015 Australian study,100 index patients supported health professionals directly contacting relatives, perceiving health professionals to have greater credibility and authority. Furthermore, it has been suggested that indirect contact by index patients may lead to inadequate counselling and a sense of social pressure to be tested in solidarity with other family members. 166 With the autosomal dominant mode of inheritance, 50% of the first-degree relatives of index cases will be affected. This finding was confirmed in our systematic review, with similar proportions seen for each cascade strategy.

Considering other factors that affect yield, age has been identified as a predictor of cascade success in previous work;41 however, our analysis of the PASS data set failed to include age as a predictor as a result of missingness. Furthermore, in the Welsh FH cascade service, as the choice of contact method was made on a case-by-case basis, the indirect method may have been used more frequently in cases where direct contact was challenging because of a lack of information about and engagement with the family. This would exert a downwards bias in our estimates of completion for indirectly contacted relatives. However, it is also likely that some indirect contacts that did not lead to further interaction with the service went unrecorded. This would exert an upwards bias in our estimates of completion for indirectly contacted relatives.

The uncertainty related to mode of contact also arose in the qualitative interview study. The study confirmed that practising health professionals also experience direct contact as a more reliable, certain and successful method to use. This trend in favour of the greater effectiveness of direct approaches is consistent with a previous quantitative audit of FH nurse cascade testing,41 and a 2019 systematic review. 26 In this study, when family connections were strong, some index patients preferred to discuss testing with their relatives indirectly, prior to a direct approach from a health-care professional, to prepare them. As other work has found, communication about genetic screening may happen informally within the family,172 and index patients may have more confidence communicating about FH testing with their families if they are well supported with information and education by a health professional. 169 Nevertheless, using a solely indirect approach, we and others100 have found that some index patients become frustrated trying to persuade relatives to engage with testing; they are fearful of relationship breakdown, and thus would welcome direct specialist approaches to support them.

Treatment patterns, and short- and long term outcomes

The results of the analysis of the LDL-C response to treatment are in line with the FH 2010 audit,9 in which a mean reduction in LDL-C (in mmol/l) of 37% (median 33%) was found. Nevertheless, our estimate of approximately 30% of individuals achieving the ≥ 50% LDL-C reduction target falls short of what the same national audit concluded (44%),9 but is above figures reported elsewhere in 2019 (19.6%173).

As highlighted in Bianconi et al.,174 different traditional predictors of CVD risk have been investigated in the literature. 175,176 LDL-C has been unequivocally associated with a premature atherosclerosis onset and an increased risk of CVD,177–179 and our findings confirmed that. Age is considered one of the strongest independent predictors of CVD, and a debate is still ongoing about whether or not sex is an independent predictor of CVD in this population. Our CVD risk modelling indicates that an age dependency of risk exists in this population, potentially accelerated compared with the general population. The same was observed previously in the Simon Broome Register5 and in Perak et al. 96 A 2017 systematic review180 presented similar findings to ours, with most studies identifying that men had a ≈2.5-fold higher CVD risk than women, although the magnitude of the difference varied by study.

As reported in Bianconi et al.,174 a history of previous CVD events among FH patients may confer a higher CVD risk. Among treated FH patients, lower risk estimates have emerged in the literature for primary prevention groups than for secondary prevention groups, with, for instance, the standardised mortality ratio for CHD ranging from 1.03 among people with FH without previous CVD events (primary prevention) to 5.15 among FH patients with previous CVD events (secondary prevention). 181 Our findings reinforce history of CVD as one of the key predictors of CVD risk.

Cost-effectiveness of alternative protocols

Previous cost-effectiveness analyses of screening have estimated the lifetime costs and health outcomes of diagnosis and treatment. 11–13,25,109,137,138,147–149 As all concluded that cascade is cost-effective, compared with no cascade, we can infer that diagnosis and treatment were found to be cost-effective, which is in line with our results. More specific comparisons are difficult because the previous studies did not consider the same subgroups or compare the same range of cascade protocols as our study. For example, in terms of the selection of index cases to the genetic test, similar to Crosland et al.,130 we found that less restrictive clinical score cut-off points were a net health gain to the NHS. Crosland et al. 130 estimated that the cost of cascade testing relatives with the genetic test was £191 per relative (2016 prices), whereas our model estimated £179–192 per relative tested, depending on the type of cascade (sequential or simultaneous) and the method of contact (direct or indirect) at 2019 prices. In Nherera et al.,25 a testing strategy using only LDL-C to diagnose relatives had a sensitivity of 0.64 and specificity of 0.84 (age-averaged estimate; LDL-C cut-off level not reported), based on an analysis of data from the Dutch cascade programme. This is approximately in line with the results from our cost-effectiveness model, if using a LDL-C cut-off level of 4 mmol/l, with a sensitivity of 0.71 and a specificity of 0.80 if the same cut-off level is used for all relatives.

Yield of cases

(See Chapter 2, Review 1: effectiveness of cascade-testing protocols among relatives for familial hypercholesterolaemia, Review 4: the diagnostic accuracy of clinical and biochemical criteria and scoring systems for the diagnosis of familial hypercholesterolaemia among relatives, and Chapter 5, Service data analysis) In the systematic review of the effectiveness of cascade testing and a previous review,26 there were very limited data available to inform the preferred approach to contacting relatives. The available studies were not comparative in nature and attempting to make inferences across studies is hampered by heterogeneity. The most robust evidence for comparing the effectiveness of the strategies for cascade testing would have been from studies that made within-study comparisons; however, no studies using such designs were identified. Therefore, we had to rely on comparing strategies across studies. We have assumed that the differences in yield between the cascade strategies can be wholly ascribed to the contact method used; however, it is likely that differences in the setting, approaches, method used to diagnose FH (genetic testing or clinical criteria) and time may all have influenced the yield. Furthermore, only four of the 24 included studies provided information to estimate the proportion of relatives tested out of those contacted for cascade testing. Of the other 20 studies, 18 did not report the number of relatives contacted. With approximately half of the included studies reporting the extent of cascading to other relatives, we were unable to explore whether there were differences in yields by cascade strategy related to extent of cascading to other relatives. In the study assessing combined approach, only 26% of index cases had a definite diagnosis of FH, with the remaining having a probable diagnosis;41 therefore, this was likely to result in a reduction in the efficiency of the cascade programme, compared with restricting cascading to relatives for whom index cases had a definite diagnosis of FH. As a result, the true yield using the combination approach could be substantially greater.

Owing to the limited evidence, particularly lack of within-study comparison of modalities of contact, the data were not in a format that could be used to parameterise the economic model. However, the Welsh PASS data provide the first available within-service comparison of direct and indirect contact. This was enhanced by direct derivation of the data from two services, and the availability of FH specialist nurse input to support the development of additional variables (e.g. by recoding free-text fields within PASS) and provide a detailed understanding of the data. Furthermore, the Wales and Wessex PASS data were derived from two of the largest FH cascade services in the UK and include detailed patient-level records on 3839 indexes and 14,951 linked relatives. The limitation of using the PASS Welsh data to compare contact methods is that there may have been confounding factors that we were unable to adjust for, given the limited data (e.g. age, which has been identified as a predictor of cascade success in previous work41). We note that these databases did identify a significant group of relatives who are within area, but not contacted by the service. Detailed information on why these individuals were not contacted was not available from PASS and remains an important priority for future research and data collection. All analyses were conducted on a complete-case basis; this may have resulted in biases if the complete cases are not representative of the patients diagnosed by the services. This bias is likely to be particularly acute for the analysis of the characteristics of those who completed the cascade. For this group, CVD history and LLT history were missing for a substantive proportion of patients. This is an important consideration as these results may have significant influence on the results of the cost-effectiveness analysis.

Yield of new relatives with FH will be affected by number of index cases identified. The latter would require evaluation of robust methods of identifying FH cases. It was recognised by Health Technology Assessment commissioners and the team that this was outside the remit of this commissioned project. However, we were able to provide some indication of the impact of index case identification using the Welsh PASS data set. In line with previous studies,10,131–133 we found a strong relationship between the stringency of thresholds used to select index cases for genetic testing and the likelihood that an individual carries a FH mutation. Increasing or decreasing the threshold used to inform which index cases are eligible for genetic testing will have implications for the number of individuals correctly identified as having FH and the number of individuals requiring genetic testing. In terms of the selection of index cases for genetic testing, similar to Crosland et al.,130 we found that less restrictive clinical score cut-off points were a net health gain to the NHS.

Treatment patterns, and short- and long term outcomes

(See Chapter 4, and Chapter 5, Economic analysis of the Clinical Practice Research Datalink database.) We were able to take advantage of using long-term data from two large, contemporary populations of FH individuals from the UK (CPRD and Simon Broome data sets). These data sets also benefited from linkage to HES data. Our work allows the quantification of risk according to a patient’s age, sex, CVD history and PT-LDL-C level.

In the CPRD data set, we were able to evaluate the absolute long-term CVD risk in both a diagnosed and a treated cohort of individuals with FH (see Chapter 5, Economic analysis of the Clinical Practice Research Datalink database) and compare the CVD risk profile of individuals with FH with that of the general population (see Chapter 4, Risk of cardiovascular disease among those with familial hypercholesterolaemia, from primary care records). Unlike previous analysis of CVD risk with FH, as well as higher rates of CHD and stroke, the primary care data set demonstrated comparable increased risk of PVD. The findings also suggested incorporating a broader range of CVD risk outcomes for economic modelling. On a cautionary note, PVD is less well phenotyped in primary care records than CHD and stroke.

As previously highlighted, the CPRD primary care data set might not have correctly coded individuals with FH, and the cohort might not be representative of the UK FH population as a whole or of patients identified via cascade testing who are expected to have genetically confirmed FH, rather than a diagnosis based on clinical phenotype. Although the CPRD has a relatively long follow-up of up to 20 years, event data were sparse from around year 10. For validation and extrapolation, it seemed relevant to focus on studies with longer follow-up than this. From a number of studies assessed for this purpose, the Perak et al. 96 study was chosen as the most relevant, but there remained a number of limitations to using this study. We had to make a number of adjustments to make the estimates comparable to those in the CPRD, and this study focused on US patients with elevated cholesterol, but without a diagnosis of FH. Furthermore, the CPRD is unlikely to characterise the current clinical practices in tertiary specialised care, such as lipid clinics, where FH patients are closely followed up and managed9 (although some patients in the CPRD may also be managed in lipid clinics). However, the Simon Broome data set identified many patients who were followed up in lipid clinics.

Cost-effectiveness of alternative protocols

(See Chapter 5.) To our knowledge, this is the first cost-effectiveness analysis of cascade testing for FH that compared a wide range of protocols, comprising strategies to select index cases to the cascade, with and without genetic testing; contact methods (direct or indirect), cascade pattern (sequential or simultaneous) and testing strategies for relatives, combining genetic and cholesterol testing; age; and treatment status. We informed the analysis with routine data from FH services in England and Wales about the characteristics of index cases and relatives, and the probability that relatives come forward for testing. Furthermore, we estimated the long-term outcomes of relatives with and relatives without diagnosis with a new cost-effectiveness model, informed by routinely collected data from a cohort of FH patients in the UK. This new cost-effectiveness model predicts lifetime outcomes conditional on age at diagnosis, sex, PT-LDL-C level and prior CVD event history.

The cost-effectiveness model of long-term health outcomes and costs of individuals with FH was informed by risk equations and LDL-C response to treatment estimated from an analysis of the individual-level data of the CPRD cohort of individuals with FH. The focus of this analysis was only on first events post diagnosis, as the context of the long-term economic model is on improving diagnosis. The primary benefit of this is expected to be among individuals who have not yet experienced a CVD event. Once individuals have experienced a CVD event post FH diagnosis, the role of FH diagnosis is likely to be less important, as individuals with CVD history will generally be in receipt of high-intensity statins. The individual-level analysis focused on treated individuals as, being the majority of individuals in our cohort, we believed that they would be more generalisable to the broader FH population.

The weaknesses of the cost-effectiveness analysis mostly relate to the assumptions required to design and inform the two cost-effectiveness models. These relate to the available data on the characteristics and degree of kinship of relatives in the PASS databases, lifetime CVD risk, effect of treatment on LDL-C level, and the generalisability of the CPRD cohort to FH patients. Given the limited data on lipid levels routinely collected by the Welsh and Wessex services, we requested data from the Dutch cascade screening programme (although a comparison with the available Welsh and Wessex FH service data suggested that they were generalisable). In addition, the costs of monitoring people were based on the national guidelines and stakeholder input, rather than informed by a systematic process or empirical data. Our estimates of the resources involved in conducting the cascade were based on information from FH nurses, literature and feedback from the stakeholder group, because the PASS data did not include details of the staff time and grade. Importantly, the cost-effectiveness analysis relies on in silico comparisons of cascade protocols, many of which have not been used in clinical practice, and on the validity of the assumptions about the generalisability of available data, the effect of LDL-C burden and long-term CVD risk. Therefore, the generalisability of the cost-effectiveness results to clinical practice is uncertain.

Our base case assumes that longer duration of exposure to raised LDL-C increases CVD risk (known as ‘LDL-C burden’), based on the relationship proposed by the 2017 EAS consensus statement. 17 This assumption is supported by epidemiological studies,17 although there is uncertainty about the size of the effect and its generalisability to people with FH. Similarly, although the assumption that CVD risk increases with age is uncontroversial, the magnitude of this increase is uncertain. For these reasons, further research is warranted on the effect of LDL-C burden and age on CVD event risk in the long term, particularly among younger people, who were under-represented in the CPRD cohort. An uncertainty that we were unable to explore in the scenario analysis is whether or not the CVD risk observed in the CPRD cohort is representative of the CVD risk experienced by people with FH, given the uncertainty regarding the extent to which the CPRD cohort includes people who do not have FH (e.g. because of miscoding or misdiagnosis in primary care).

Acceptability of cascade-testing approaches

(See Chapter 6.) To our knowledge, this is the first study to explore the experiences and perceptions of both index patients and relatives, and health professionals, across different settings in the UK, chosen to reflect a range of current FH cascade-testing approaches. The purposeful study sample included people with a diversity of cascade-testing experiences and a range of professionals involved in the cascade-testing service. The analysis of the data was developed by two researchers of different disciplinary backgrounds, and a process of member checking with respondents themselves was also undertaken, confirming interpretation of their experiences.

We acknowledge some limitations, which are challenges for future research to address in this field. Index patient and relative participants were white, predominantly female and English-speaking. The lack of ethnic diversity reflects that seen in FH services, and thus who we were able to access and engage for interview and sampling. Moreover, we did not secure direct testimony from relatives who declined FH cascade testing, second-degree relatives or genetic counsellors.