Current Initiatives for Detecting Patients with Familial Hypercholesterolemia in Primary Care

Abstract
Databases of primary care practices and pathology providers in Queensland and South Australia were screened using variable cut-off levels for total cholesterol (TC) and low-density lipoprotein cholesterol (LDL-C) to detect patients with possible familial hypercholesterolemia (FH). Over a 12- month period, 1,728,805 lipid tests were performed in the Queensland population of 4,332,739. Of these, 52.4% included LDL-C measurement. The proportion of the population having lipid testing increased from 12% at age <25 years to 93% at age >74 years. Females aged <35 years were tested almost twice as often as males and with similar frequency in those aged >44 years. For all ages, the proportion of patients with LDL-C >251 mg/dL was 0.13%. This proportion declined with increasing age. Dutch Lipid Clinic Network (DLCN) scores were assigned to LDL-C levels of 691 patients in primary care practices, and 57.8% had TC levels >290 mg/dl (possible FH). Among 1,057 patients retrieved from pathology provider and primary practice databases, 82.6% of those with TC levels >290 mg/dl had type II hyperlipidemia [HLP], 8.5% had types III and IV HLP and 1.8% had types I and V HLP. Database interrogation conveniently detected large numbers of patients with potential FH at minimal cost. In contrast, cascade family screening (CFS) of index patients with FH Morocco detected 12 cases of FH Morocco in near relatives. Screening was time-consuming, costly and relatively inefficient, with almost 50% of relatives either refusing to be tested or being non-contactable. These data suggest that online prospective flagging of pathology results – with DLCN scoring and clinical confirmation of the diagnosis - may provide a convenient means for identifying patients with possible FH in primary care. Tracing of relatives of FH cases through published family histories also could facilitate screening.

Introduction
Only an estimated 1% of FH patients are being diagnosed worldwide, although this percentage varies considerably in different countries.^1-5 This gap needs to be addressed by developing appropriate management strategies.⁴ With a prevalence of 1:500, there are about 666,000 and 44,000 individuals with FH, living in about 222,000 and 15,000 U.S. and Australian families, respectively. Recent studies have suggested an even higher prevalence, and these numbers could be considerably greater.⁶ FH patients, therefore, need to be mostly managed by primary care physicians.^4,5 With a focus on primary care, we performed two studies to compare different strategies for detecting patients with possible FH, with the aim of reducing the future burden of coronary heart disease (CHD).

The Severe Hypercholesterolemia In Primary Care (SHIP) study aimed to provide data on the prevalence of severe hypercholesterolemia (SHC) and its lipoprotein subtypes in primary care. It investigated databases of primary care physicians and pathology providers to retrieve patients with sufficiently elevated levels of TC and LDL-C. Such patients have a variety of lipoprotein disorders. Those with predominant elevation of LDL-C levels are at high risk of CVD, while those with predominant hypertriglyceridemia (HTG) are at risk of acute pancreatitis. Others have variable CVD risk, depending on the lipoproteins present to excess. It is important to distinguish these different lipoprotein subtypes, because management strategies differ.

Patients with type II HLP also may be screened for FH using either Make Early Diagnosis to Prevent Early Deaths in Medical Pedigrees with FH (MEDPED-FH), Simon Broome or DLCN diagnostic criteria (Table 1).⁴ DNA analysis also can be performed to confirm the diagnosis.⁴

The Barossa Family Heart Study (BFHS) is based in rural South Australia, where German Lutherans first settled in 1842. They brought not only skills in farming and winemaking, but also the LDL receptor (LDLR) mutation FH "Morocco" (LDLR_00037, c.682G>T [exon4, pGlu228X] according to FH mutation nomenclature), resulting in the synthesis of a severely truncated (by 75%) LDLR protein. This mutation occurs relatively frequently in Germany.⁷ Other Lutherans migrated in the mid-1800s to Texas, Nebraska and South Africa, and also may have imported the FH Morocco gene to these sites.^8,9,10

Many Lutheran families in Australia and the U.S. published family histories, often through Lutheran Church publishing houses, which allow identification of people with premature deaths who may have inherited FH.

The classic method for FH case detection is by cascade family screening (CFS).⁴ This method involves screening of TC and/or LDL-C levels of near relatives of patients with FH to detect the disorder at as early a stage in life as possible.

Methods

SHIP Study
Selection of SHC patients involved computer interrogation of patient records of general practitioner (GP) surgeries and lipid profiles of local pathology providers. De-identified data were obtained for ethical reasons. Lipid data from the Queensland population were adjusted to exclude repeat tests, with the latest test being included for analysis.

Primary care patient records were interrogated using either the PEN clinical auditing tool or PARIS, a software interrogation tool provided by Merck Sharp & Dohme (Australia).¹¹ Pathology providers used their private software to identify two groups of patients with hypercholesterolemia (HC): moderate HC (TC>274 mg/dL plus LDL-C >174 mg/dL) or severe HC (LDL-C >=251 mg/dL). The latter represented patients with possible FH.¹²

Patients with severe HC defined as TC>290 mg/dl were retrieved from two metropolitan and four rural primary practices using PEN or PARIS auditing tools. This TC cut-off is used in the UK to detect patients with possible FH, those for whom family screening is recommended.⁵ Lipoprotein subtypes were determined by a modification of the Fredrickson classification, and LDL-C levels given point scores according to DLCN criteria (Table 1).^4,13 The likelihood of a mutation causing FH is related to the DLCN point score, such that positive LDLR mutation rates are ~5% in those with scores <3, ~12% in those with scores 3-5, ~25% in those with scores 6-8, and ~90% in those with scores >8.¹⁴

Barossa Family Heart Study
The FH Morocco mutation was initially detected in two Barossa residents, both sixth-generation descendants of Georg Eckert, who migrated from Silesia (in modern-day western Poland, near the German border) to South Australia in 1849.⁸ CFS was performed on near relatives of these index patients. Questionnaires were sent requesting details of family history of premature CHD or death, personal history of premature CHD, family physician details, highest recorded cholesterol level, current cholesterol-lowering treatment, and number of first-degree relatives. DNA consent forms were sent for completion and return by reply-paid post.

Lipids were measured by standard automated techniques in fasting blood, and standard DNA techniques and primers were used to detect FH Morocco using an automated analyzer and DNA isolated from peripheral blood.¹⁴

The Eckert family history was examined for ages of death in males and females before the statin era, making a diagnosis of presumed FH Morocco in those dying prematurely (age <55 years in men and <65 years in women), excluding those recorded with deaths under age 30 and those with a recorded non-CVD cause such as pneumonia, infection or accident. Other deaths were presumed to be to the result of non-FH Morocco-related causes.

Griffith University and Bellberry Human Research Ethics Committees provided ethical approval.

Results

SHIP Study
Queensland population: lipid testing and prevalence of SHC
During a 12-month period, the ageand gender-specific proportions of the Queensland population (total 4.3 million) referred for lipid profiles increased progressively by about 10% per increasing decade of age, from 11.4% (ages <25 years) to 91.8% (ages >74 years), with an average of 38.3% for all ages (Table 2).

The proportion of patients with moderate HC (TC>274 mg/dL plus LDL-C >174 mg/dL) at age <25 years was 1.8% of lipid tests that included LDL-C, and increased with age to peak (4.6%) at ages 45-54 years. The proportion for all ages was 3.6% (N=32,640).

The proportion of patients at all ages with severe HC (LDL-C >251 mg/dL) was 0.13% of lipid tests that included LDL-C, 35% below that expected for FH (0.2%) (Table 2). The proportion of these patients with probable FH was lower than expected for FH in those ages 45-64 years (0.14-0.15%) but was as expected for FH (about 0.2%) in those ages <45 years. The observed proportions were considerably lower than expected for FH in those ages >65 years (0.03-0.12%), particularly in males.

Primary care: prevalence of severe hypercholesterolemia
Including two metropolitan and four rural primary care practices, the SHIP Study identified 726 patients with TC >290 mg/dL using PARIS interrogation software; 49.5% were men and mean age was 61.3 years; 644 (89%) were on lipid therapy with mean (mg/dL) lipids: TC 193, triglycerides (TG) 146, HDL-C 54 and LDL-C 112; 82 (19%) were not on therapy with mean lipids: TC 309, TG 195, HDL-C 62 and LDL-C 193.

One rural practice interrogated its patient database (N= 17,612) for total cholesterol (TC) levels. For all ages, TC was recorded in 30%, of whom the proportions of patients with TC <154 mg/dL, 154–208 mg/dL, 212–290 mg/dL and >290 mg/ dL were 38%, 53%, 7%, and 2% (N= 138), respectively. The latter were patients with potential FH.

Lipoprotein subtypes in patients with severe hypercholesterolemia
The Figure shows the proportions of patients in South Australian and Queensland primary care and pathology practices who had TC>290 mg/dL, according to type of hyperlipidemia (HLP) and using arbitrary levels of TG based on the original Fredrickson classification.¹³ During a period of one year, the proportions of 754 patients in primary care databases with types IIA, IIB, III plus IV, IV and I plus V HLP were 53%, 18%, 2%, 5% and 1%, respectively, with some variation between practices and whether patients were being treated with lipid-modifying drugs. Similar proportions were observed for the five-year pathology provider database and for patients with SHC overall (Figure 1). Dutch Lipid Clinic Network (DLCN) scores DLCN scores for LDL-C levels of >330, 250-329, 190-249 and 155-189 mg/dL are 8, 5, 3 and 1 points, respectively (Table 1). Diagnoses of definite FH, probable FH and possible FH require >8, 6-8 and 3-5 points, respectively.⁴ The proportions of 691 patients with TC>290 mg/dL and DLCN scores of 8, 5, 3 and 1 points were 8.2%, 15.7%, 33.9% and 27.8%, respectively, with missing data recorded in 5.1%.

Barossa Family Heart Study
Family history analysis
According to the Eckerts of Tauer,⁸
premature deaths to 1979, including those resulting from FH Morocco, occurred in 27 direct descendants of Georg Eckert, with a mean age of 45.2 +/- 7.8 (SD) years in 14 men and 57.3 +/- 9.1 years in 13 women. This gender difference was statistically significant (p<0.01). There also were six female spouses with premature deaths (mean age 49.2 +/- 9.6 years); these also may have had FH Morocco, because they had the same ethnic (German/Lutheran) and community backgrounds.

Presumed non-FH-affected direct relatives (those without premature death) had mean ages of death of 69.76 +/- 8.7 years in men (N=17) and 72.9 +/- 5.4 years in women (N=8). There were 26 presumed non-FH-affected spouses with mean ages of death 71.12 +/- 9 years in men (N=17) and 76.18 +/- 7.4 years in women (N=11).

Cascade family screening and DNA analysis
Of the 151 people included in CFS, 107 (71%) were near-relatives of the two index cases with FH Morocco and 44 were unrelated (Table 3). DNA testing was performed in 39 people, of whom 23 (59%) were related to the index cases. The majority of the latter (70%) had FH Morocco, and the remaining relatives were negative for FH Morocco. All individuals who were not related to the index cases (N=16) were negative for FH Morocco. One relative of index case No. 2 refused to consent for DNA analysis; 11 relatives of index case No. 1, 27 relatives of index case No. 2 and one unrelated individual did not reply to telephone calls, emails or letters and were classified as not responding. The proportion of non-responders was 13% of those screened. DNA testing has yet to be performed in 53 people (of whom 33 are near relatives of the index cases).

Cases of FH Morocco had a severe FH phenotype with mean untreated total cholesterol 355.1 +/- 34.7 (SD) mg/ dL (N=7). Two males had premature CHD, one receiving two angioplasties and stents between the ages of 45 and 55, in addition to coronary artery bypass graft CABG surgery at age 60. The other had a myocardial infarction at age 55.

Discussion
These studies provide further information on the prevalence of severe hypercholesterolemia and possible FH in primary care. In the SHIP study, the high proportion of lipid testing in older patients may reflect current guidelines for lipid-lowering drug therapy, which relate more to co-morbidity and absolute CVD risk than to lipid levels. Lipid testing of younger patients may have been for FH screening since the proportion of patients with LDL-C >251 mg/dL was ~0.2%, approximating the expected prevalence of FH in most populations.⁴ The proportion of lipid profiles with moderate HC (TC>274 mg/dL and LDL-C >174 mg/ dL) was considerably higher than expected for FH (1-8-4.6% vs. 0.2%), suggesting these cut-off levels may not be high enough for FH diagnosis.

The proportion of patients with HC and severely elevated LDL-C levels consistent with probable FH (>290 mg/dL) was similar to that expected for FH in the age groups <45 years (Table 2). Reasons for the greater proportion of women than men in the age groups >45 years, especially in those >74 years of age, are speculative and may be related to hormonal changes, differences in lipid-lowering therapy, differences in survival, compliance or factors such as secondary causes of HLP.

Controlling LDL-C levels is important in patients with type II HLP; nevertheless, significant numbers of uncontrolled or untreated patients were detected, suggesting significant gaps in management that require addressing in future studies.

The TC cut-point >290 mg/dL for identifying potential patients with FH also detected 17.4% of cases with lipoprotein subtypes other than type II, which are associated with lower CVD risk. Most of these had mixed HLP (8.5%), and a small percentage had chylomicronemia (1.8%) (Figure 1). Management of these patients requires consideration of specific secondary causes (especially glucose intolerance and alcohol intake) in addition to specific therapy.¹⁵

In the Barossa Family Heart Study, examination of published family histories was helpful in estimating longevity before the statin era, although exact causes of death usually were not recorded. Most early German immigrants had a normal life expectancy, and those with presumed FH had about a 20-years-shorter lifespan. These results suggest that systematic examination of published family histories may detect premature deaths related to presumed FH, and facilitate CFS in their descendants.

CFS required a large time commitment with a low return because of the numbers of relatives who could not be contacted for a variety of reasons (including change of address, change of name and relatives unknown to index cases). Some hypercholesterolemic relatives of FH Morocco index cases had no detectable LDLR mutation. These data do not support CFS as the primary means for FH case detection, but this may reflect the small sample numbers involved and the limited research resources available, preventing detection of other possible mutations.⁴

Similar results were obtained in the recent MEDPED-FH program in Australia, in which many relatives of FH patients could not be contacted for a variety of reasons, and cascade family screening was difficult to perform effectively.¹⁶

Conclusions
Cascade family screening was found to detect relatively small numbers of potential FH cases in comparison with screening of primary practice and pathology provider databases. Dutch Lipid Clinic Network scoring of LDL-C levels showed that 58% of 691 patients with severe HC had possible FH. Published family histories of index FH cases could facilitate FH detection. The results of the SHIP and BFHS studies suggest that online prospective flagging of patients with severe HC (TC>290 mg/ dL) by pathology providers may provide an efficient means for identifying patients with possible FH in primary care. Flagging also may alert primary care physicians to perform online Dutch Lipid Clinic Network scoring for confirmation of FH diagnosis, in addition to the need for cascade family screening.

The cost-effectiveness of FH detection in these studies needs to be determined, with the aim of convincing governments to fund national programs for FH detection and treatment, which are required in many countries, including Australia and the U.S.

Acknowledgments
The authors wish to acknowledge the contributions of Mrs. Sheila Storrs for performing family screening and coordinating research, and Dr. Pukar Thapa and Mr. Dino Peterson for data analysis and management.

Disclosure statement: Dr. Hamilton-Craig is a member of the Lipid Advisory Boards of Merck Sharp & Dohme, AstraZeneca, Abbott and Amgen (Australia). Dr. Kostner is a member of the Lipid Advisory Boards of Merck Sharp & Dohme, Abbott and Amgen (Australia). Dr. van Bockxmeer has no disclosures to report. Dr. Michaelides has no disclosures to report.