Article Text

Download PDFPDF

Original article
Effectiveness of alternative strategies to define index case phenotypes to aid genetic diagnosis of familial hypercholesterolaemia
  1. Rosemary E J Clarke1,
  2. Soundrie T Padayachee2,
  3. Rebecca Preston2,
  4. Zofia McMahon1,
  5. Mitchell Gordon3,
  6. Colin Graham4,
  7. Martin A Crook1,
  8. Anthony S Wierzbicki5
  1. 1Department of Metabolic Medicine, Guy's and St Thomas Hospitals, St Thomas Hospital, London, UK
  2. 2Department of Radiology, Guy's and St Thomas Hospitals, St Thomas Hospital, London, UK
  3. 3Department of Blood Sciences, GSTS Pathology, Guy's and St Thomas Hospitals, St Thomas Hospital, London, UK
  4. 4Department of Genetics, Northern Ireland Regional Genetics Centre, City Hospital, Belfast, UK
  5. 5Department of Chemical Pathology, Guy's and St Thomas’ Hospital, London, UK
  1. Correspondence to Dr Anthony S Wierzbicki, Department of Chemical Pathology, St Thomas’ Hospital, Lambeth Palace Road, London SE1 7EH, UK; Anthony.Wierzbicki{at}kcl.ac.uk

Abstract

Objective To determine the utility of secondary stratification measures to improve the ascertainment of index cases of familial hypercholesterolaemia (FH).

Design A retrospective study of genotyped index patients with Simon Broome (SB) FH.

Setting University teaching hospital.

Patients 204 patients aged 55±14 years; 36% had tendon xanthoma (TX), 21% had coronary heart disease (CHD), low-density lipoprotein cholesterol (LDL-C) was 6.20±2.24 mmol/l and 55% had genetic FH.

Interventions The effects of different staging systems (SB vs Dutch criteria), presence of TX, use of LDL-C level, personal history of CHD and imaging evidence of atheroma by carotid intima-media thickness or coronary artery calcium score to identify genetic FH was explored.

Outcome measures Changes in C-statistic and net reclassification index (NRI).

Results SB criteria gave a C-statistic of 0.64 comprising C=0.65 in TX(+) and C=0.5 in TX(−) patients. Genetic FH was present in 75% of TX(+) compared with 44% in TX(−) patients. The Dutch criteria gave C=0.72. Addition of imaging criteria to prior CHD raised C=0.64 to C=0.65 in all patients with a NRI of 19% (p=0.06). In TX(−) patients imaging raised C=0.50 to C=0.65 with a NRI of 0.38 (p=0.001) and a weighted comparison index of 0.28, implying the detection of 14 more FH cases per thousand.

Conclusions Patients with tendon xanthoma (definite FH) should be genotyped. In patients with possible FH, the presence of a personal history of CHD or imaging evidence of increased atheroma offers the best method of identifying index patients likely to have monogenic FH.

Statistics from Altmetric.com

Introduction

Familial hypercholesterolaemia (FH) is an autosomal dominant genetic disorder characterised by a raised low-density lipoprotein cholesterol (LDL-C) concentration and an increased incidence of premature coronary heart disease (CHD).1 The incidence of CHD is increased 20–100-fold in untreated younger individuals and by two- to threefold in untreated older individuals.2 FH is associated with the presence of tendon xanthomata (TX) and premature corneal arcus. FH is easily treatable with statin and allied treatments and adequate treatment is associated with a 75% reduction in lifetime risk of death from cardiovascular disease (CVD).3 Given the autosomal dominant inheritance of the FH cascade family screening is event and cost-effective.4–7 Recommendations for identification, cascade screening and treatment of FH exist both in the Netherlands and in the UK.5 ,6 Both Dutch8 and UK guidelines5 ,6 recommend genetic diagnosis of FH, whereas the USA recommends LDL-C screening.9

Problems exist in the diagnosis of index cases with FH as hypercholesterolaemia and a family history of premature CHD (<60 years) may have other causes, including familial combined hyperlipidaemia,10 raised lipoprotein (a)11 or familial clustering of multigenic12 or environmental risk factors. Screening primary care databases can help to identify cases of FH but many individuals meeting the general criteria do not have this condition.13 The LDL-C distribution in FH significantly overlaps that of the general population.14 Diagnosis of FH in lipid clinics relies on pre-stratification of patients using clinical criteria such as the Simon Broome (SB) or Dutch Lipid Clinic Network criteria.5 ,6 ,15 The yield of genetic screening in FH varies from 50% to 75% in patients with physical signs (TX) to 10–20% in those lacking these signs.8 ,16 The prevalence of TX and rates of CHD are falling in the UK17 thus making the initial diagnosis more difficult and currently <25% of cases of FH in the UK have been identified.18 There is increasing interest in the use of additional biomarkers to re-stratify risk in population risk screening. This study investigated whether the use of radiological biomarkers (carotid intima-media thickness (cIMT) and coronary artery calcium score (CACS)) would improve diagnostic stratification (including sensitivity and specificity) in FH as both are related to total cholesterol-years of exposure.19

Methods

This retrospective study aimed to identify the utility of imaging biomarkers of established atherosclerosis in FH. Patients were identified as potential cases of FH on the basis of SB criteria from a University teaching hospital lipid clinic covering new presentations from 2005 to 2011.2 These define a person as having either definite or probable FH. Secondary causes of raised cholesterol must be excluded first. An alternative tool for the phenotypic diagnosis of FH is the Dutch Lipid Clinic Network criteria.15 Peripheral and cerebral vascular disease, personal medical history and premature (<45 years old) arcus cornealis are also included, and it is a risk scoring system as opposed to the dichotomous classification system in the SB criteria.

All unrelated proband patients had undergone fasting measurement of lipid profiles and lipoprotein (a) on presentation as well as detailed recording of other CVD risk factors. Three-generation pedigrees had been recorded (if possible) and data on age of onset of CHD in any family member were recorded as were total cholesterol levels if known. Xanthomas were sought on the dorsum of the hands, elbows, pretibial tuberosities and Achilles tendons and were considered to be present if tendons appeared diffusely enlarged or had focal nodularities.20 New patients underwent measurement of the atheroma burden if they did not have a personal history of major CHD events—admission with acute coronary syndrome, myocardial infarction, percutaneous coronary intervention or coronary artery bypass grafting. This was performed using semiquantitative ultrasonic angiology between 1995 and 2009 and since then by either measurement of cIMT or CACS as detailed below.

  1. Bilateral cIMT measurement of internal carotid rear wall thickness was performed using a 13 MHz transducer and edge-detection software on an Acuson system (Siemens, Frimley, UK). Results were reported as mean thickness and standard deviation related to centile for age and gender based on the European Carotid Atherosclerosis Study charts.21 A cIMT value >75th centile, presence of plaque or >20% stenosis was taken as a positive result indicating a high atheroma burden and increased risk in line with American Society of Echocardiography recommendations.21 A plaque was defined as a focal widening of the IMT relative to adjacent segments, appearing on ultrasound as a localised protrusion of the vessel wall into the lumen.21

  2. Measurement of CACS using a 256-slice CT scanner (Philipis iCT256; Best; The Netherlands). CAC distribution was reported as an Agatston score related to vessel-specific distribution and presented as centile for age and gender based on the St Francis Heart Study.22

Genotyping was performed after clinical consent for genetic testing at an accredited regional genetics centre by initial screening for 57 common mutations using an Illumina system and has been implemented since 2008. If no mutation was detected on initial screening then further comprehensive genetic analysis was performed by sequencing all 18 exons of the LDL receptor gene and its promoter, part of apoB-100 exon 26 and pro-protein convertase subtilisin kexin (PCSK)-9 exon 7. Such genetic analysis is predicted to identify 95% of UK mutations.23

Clinical characteristics and biochemical data were analysed in an open format using Systat 12 (Systat, Hounslow, London, UK). The power of screening strategies was investigated by receiver operator curve (ROC) analysis and areas under curves compared using C-statistics. Logistic regression analyses were also conducted to validate the ROC analyses and included clinician as a dependent (confounding) variable. A variety of scenarios for the whole population and SB/Dutch definite/possible subgroups were explored. In addition the clinical utility of re-stratification was investigated by calculation of reclassification rates and net reclassification rate.24 The effect of reclassification on detection rates was analysed by χ2 analysis. A p value <0.05 was considered significant. A weighted comparison (WC) analysis25 was performed for the same dataset assuming that the prevalence of FH in the population with raised initial LDL-C was 5% and that the relative weighting for diagnostic certainty was 4.26

Results

The clinical characteristics of the 204 patients are shown in table 1. The average age of the population was 55±14 years, 47% were male, LDL-C was 6.20±2.24 mmol/l and 21% had established CHD. The prevalence of other CVD risk factors were hypertension in 21%, current smoking in 13% (42% were ex-smokers) and diabetes in 1%. TX had been recorded as present in 36% at some time. Patients with TX(+); n=73) were older those without (TX(−); 59 vs 53 years; p=0.003) but more often had CHD (35% vs 12%; p<0.001), higher total cholesterol (9.75 vs 8.43 mmol/l; p<0.001) and LDL-C (7.45 vs 5.56 mmol/l; p≤0.001) but did not differ in other characteristics. A genetic diagnosis was obtained in 75% of TX(+) patients compared with 44% in TX(−) patients.

Table 1

Characteristics of patients with identified FH gene mutations (+) and patients not having monogenic FH gene mutations (−)

The effects of different screening strategies on this dataset were explored in the whole cohort and TX(−) patients (table 2). These included use of SB criteria, LDL-C criteria alone, calculation of Dutch risk scores and re-stratification measures. In this group SB criteria gave a ROC area under the curve (AUC) of 0.64 which comprised a ROC of 0.65 in TX(+) patients but 0.5 in TX(−) patients. Use of the numerical Dutch scoring system with a cut-off point of 5 gave an overall ROC of 0.72.

Table 2

Specificity (spec), sensitivity (sens), net reclassification indices (NRI) and weighted comparison (WC) reclassification for strategies to detect genetic familial hypercholesterolaemia

Many patients present with minimal data apart from lipid levels and diagnosis of premature arcus cornealis, and TX, in particular, shows interobserver variability. The simplest screening method is to use lipids alone. LDL-C criteria were derived from the Dutch cohort dataset and showed peak power for detection of FH with LDL-C>6.5 mmol/l with a number needed to diagnose of 2.7 in the whole population and 3.1 in TX(−) patients. This method was highly specific (82%) but not very sensitive (55%) with a ROC AUC of 0.68. Addition of data on family history of premature CHD or gross hypercholesterolaemia did not add any significant predictive capacity to the simpler measure and decreased the WC measure.

Many index patients with FH are identified on presentation to cardiology services with premature CHD. Screening using SB criteria allied with a personal history of premature CHD was specific (88%) but insensitive (23%) with a ROC AUC of 0.68 and a net reclassification index (NRI) of 11% overall and 30% in TX(−) patients (p=0.01). Further details of the characteristics of risk stratified TX(−) patients are given in table 3. As formal presentation with CHD does not exclude asymptomatic significant atherosclerosis a further analysis was conducted of adding either cIMT, identifiable plaque/stenosis or CACS in patients without demonstrable disease. All patients with prior CHD were assumed to have a positive cIMT or CACS. Additional cIMT data were available in 51 patients (31%) and CACS in 23 (14%) in the whole cohort and 39 patients for cIMT (24%) and 14 patients (9%) for CAC in the TX(−) group. Addition of imaging criteria to prior CHD raised the ROC AUC for all patients from 0.64 to 0.65 but with a NRI of 19% (p=0.06). In TX(−) patients imaging/atheroma criteria raised the ROC AUC to 0.65 from 0.50 with a NRI of 0.38 (p=0.001) with a 28% improvement in the WC index (WCI) implying the detection of 14 extra patients with FH per 1000.25 Patients with detectable atheroma included 31% with CHD, a cIMT of 0.67±0.15 vs 0.48±0.08 mm (p=0.02) and CACS 91 (centile 78–100) vs 19 (0–55) (p<0.001). The rate of diagnosis of FH in TX(−) patients improved from 33% to 44% (p<0.001).

Table 3

Characteristics of TX(−) patients identified as high risk or low risk on the basis of coronary heart disease and coronary calcium score or carotid IMT >75th centile

Discussion

The diagnosis of FH is becoming increasingly problematic. National screening strategies such as the National Health Service (vascular) health check27 are identifying individuals with raised LDL-C. A total cholesterol >7.5 mmol/l can be found in 2–15% of the population depending on age and only 1/10–1/75 of these would be predicted to have FH.5 Family histories are useful and reliable28 ,29 to help identify FH and other hereditary hyperlipidaemias. Both the SB and Dutch FH criteria for diagnosis of FH include the presence of family history of CHD and further substratify by TX. The development of TX is related to age, male gender, triglycerides and LDL-C.30

Given the expense of genetic testing there is an increasing need to develop algorithms to ensure that these tests are used effectively. Previous studies have found frequencies of FH mutations of 75–80% in TX(+) patients but only 10–20% in TX(−) patients but the rates of TX are declining. Patients with FH have a raised cholesterol level from infancy and hence their ‘cholesterol-years burden’ is higher, effectively giving them an increased ‘vascular age’. This can be visualised by many modalities including duplex ultrasound of the carotid artery and CACS on CT of the chest.

This retrospective study reviewed the clinical signs, family history, biochemistry and imaging data gathered on a cohort of patients genetically screened for FH and referred over 20 years to a secondary/tertiary care centre. It found that for this population of FH index cases SB criteria gave an AUC on the ROC curve of 0.65 and Dutch criteria an AUC of 0.72 in line with other studies.5 ,6 Similar to other studies TX was present in 73 (36%) patients, and 75% of TX(+) patients were diagnosed with genetically proven FH.8 ,16 This translates into a number needed to diagnose (NND) of 1.3. However, in 130 TX(−) patients the AUC on the ROC curve was only 0.50, indicating that the SB criteria were inadequate to diagnose FH in this group and indeed, on testing, only 44% were found to have genetic FH—a NND of 2.3. Other studies have shown a lower prevalence of genetic FH in SB ‘possible’ patients of 28%, suggesting that the NND is likely to be higher in a less selected population.16

The effectiveness of a series of other strategies to identify likely probands with FH was investigated. Using LDL-C alone, peak effectiveness was found with LDL-C>6.5 mmol/l which reclassified 63% of cases and gave a NND of 3.1. Other LDL-C cut-off points either decreased sensitivity or gave higher NNDs—up to 16. Many patients with FH present with premature CHD but this was found in only 11% of TX(−) patients. Imaging detected either plaque, stenosis, a raised cIMT or CACS in a further 25% of these patients. Addition of imaging to premature CHD raised the AUC on the ROC curve from 0.50 to 0.65 with a net reclassification index of 38% (p=0.001). The WCI, including a fourfold certainty measure derived from patient confidence in the diagnosis from the GRAFT trial26 and assuming a prevalence of 5% of FH in this hypercholesterolaemic population, suggested that this strategy improved diagnosis (WCI=0.28) and would theoretically detect a further 14 cases of FH per thousand. The actual performance in the cohort showed an increase in the detection rate for genetic FH from 33% to 44% (p<0.001)—that is, 11% in this selected group. In addition, even in patients without monogenic FH, detection of significant atheroma is likely to lead to clinical reclassification and treatment beyond that indicated by risk factor based scoring (Framingham or QRISK) given the increased CVD risk identified in these individuals.31 ,32

This study is limited by its retrospective nature and also by the emerging management of patients with FH as well as being derived from a tertiary referral centre population. Data have been aggregated for ultrasonic angiology from previous semiquantitative scanning for stenoses and atheromatous plaques to which the addition of formal measurement of cIMT is a recent NHS innovation (2008). Similarly, CAC scanning has only been routinely available since 2008. This study is preliminary and retrospective and ideally a comprehensive prospective study should be done to validate this algorithm and its cost-effectiveness.

Many patients in this cohort presented before 2008 and thus recent imaging data have not been used as statin treatment reduces the progression of cIMT33 and thus would underestimate the utility of cIMT in identifying cases of FH. Median cIMT in patients with genetic FH in this study at 0.60±0.15 mm was considerably less than that reported in European series (0.70–0.80 mm)34 but similar to that in a decade younger Brazilian cohort (0.65 mm).35 This could reflect differences in protocols and standardisation of cIMT scanning36 or the need to develop UK-specific cIMT reference ranges. CAC scores are not affected by statin therapy37 and have been reported from cohorts of patients with FH in the USA and Europe. The median CACS of 280 Agatston units in this UK cohort was greater than the 51–87 reported in asymptomatic Dutch cohort studies.38 ,39 There is debate as to whether cIMT or CACS is the better imaging modality for atheroma detection. CACS are highly reproducible and consistently add to underlying risk factors to risk stratification unlike cIMT.40 However, as the detection of CACS requires calcifying plaques, significant atheroma involving cholesterol-rich plaques alone can be missed. A prospective cohort study showed that cIMT in younger patients where CACS was absent predicted future CVD events.41 In a FH population, 21% had plaques but no CACS and only 54% of all plaques showed calcification.38

NHS resources are limited. Screening programmes are underway to detect patients with CVD risk from which data on total cholesterol levels will be gathered.42 The addition of a family history of CVD at clinical review allows potential cases of FH to be identified.6 To identify those requiring specialist assessment and genetic testing, this study suggests that priority should be given to those with TX and/or a personal history of CHD before age 60, evidence of increased atheroma burden (>75th centile) on imaging by cIMT or CACS. Carotid IMT is the simpler, cheaper technique and is better for detection of atheroma in younger patients. It can also be performed in a primary care setting using portable equipment but is not widely available in the UK. However, the investment in the acute chest pain management strategy43 means that CACS, while more expensive, is more widely available and it is better standardised. If FH screening becomes a primary care-driven service then cIMT is likely to be more commonly used.

References

View Abstract

Footnotes

  • Contributors Patient data records: ASW, MAC. Clinical data acquisition: REJC, ASW. Biochemical data acquisition: MG. Carotid ultrasonography data: STP. Coronary calcium data: RP. Genetic data acquisition: CG. Data analysis: REJC, ASW. Manuscript writing: REJC, ASW. Guarantor: ASW.

  • Funding None.

  • Competing interests None.

  • Ethics approval Restrospective audit of anonymised clinical data from hospital databases.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.