Clinical background
Familial hypercholesterolaemia (FH) is an autosomal dominant disorder with a prevalence of approximately 1 in 200 to 1 in 500 in the population. Untreated, heterozygous FH leads to premature coronary artery disease, often before the age of 55 in men and before 60 in women. The diagnosis is difficult for two reasons. First, cholesterol levels in FH overlap in part with those seen in polygenic hypercholesterolaemia, particularly in younger individuals. Second, genetic testing is not always available, and 10 to 40 per cent of patients with a clinical diagnosis of FH have no demonstrable causative mutation in the known FH genes [4].
The US (MEDPED) diagnostic criteria were developed precisely to address the first difficulty. The central idea of the tool is that the interpretation of a cholesterol value must be weighed against the probability of FH before the sample was taken, the so-called prior probability. A total cholesterol of 310 mg/dL means something entirely different in a first-degree relative of a confirmed FH patient than in a person from the general population. In the derivation cohort, only 4 per cent of people from the general population had FH at that level, compared with 95 per cent of first-degree relatives of known cases [1].
Applying the US (MEDPED) criteria
The tool uses three variables: total cholesterol, age and the degree of relatedness to a known FH case. The diagnosis is supported when the total cholesterol reaches or exceeds an age- and relatedness-specific threshold according to the following matrix:
| Age band | First-degree relative | Second-degree relative | Third-degree relative | General population |
|---|---|---|---|---|
| <20 years | 220 | 230 | 240 | 270 |
| 20–29 years | 240 | 250 | 260 | 290 |
| 30–39 years | 270 | 280 | 290 | 340 |
| ≥40 years | 290 | 300 | 310 | 360 |
All values in mg/dL. Reaching or exceeding the applicable threshold supports a diagnosis of FH.
The derivation cohort consisted of FH families in Utah, USA, in whom the diagnosis was validated by molecular genetic testing [1]. The authors applied Bayesian principles to prior probabilities to derive two sets of thresholds: one for relatives of confirmed cases and one for the general population. The thresholds are lower for closer relatives because the prior probability is higher, and they rise with age because cholesterol levels in the population increase with age independently of FH.
Interpretation in practice
The tool gives a binary outcome: the patient either reaches the applicable threshold or does not. There are no point steps or risk bands.
The patient reaches the threshold: A diagnosis of FH is supported. This warrants referral to a lipid clinic, initiation of lipid-lowering treatment according to current guidelines and the offer of genetic testing where available. If a mutation is confirmed, cascade screening of first-degree relatives should be offered.
The patient does not reach the threshold: A diagnosis of FH is not supported by this criterion. This does not exclude FH, particularly in young relatives in whom the cholesterol level has not yet risen to its maximum. The patient should be followed up with repeat lipid profiles, and if clinical signs emerge (tendon xanthomas, premature coronary artery disease in the family), more complete diagnostic criteria should be considered, for example the Dutch Lipid Clinic Network (DLCN) or Simon Broome.
Validation and performance
In an Asian cohort from Malaysia (n = 755, LDL cholesterol ≥4.0 mmol/L, recruited from specialists and health screening), US MEDPED was compared with DLCN as the reference [2]. US MEDPED identified only 105 of 415 DLCN-positive cases, corresponding to a sensitivity of 25.3 per cent. Specificity was 98.8 per cent and the positive predictive value 98.1 per cent. Among those classified as "not FH" by US MEDPED, 91.4 per cent of those whom DLCN classified as "definite FH" were false negatives, illustrating that the tool misses a substantial proportion of truly affected patients when used in a mixed specialist population. The Simon Broome criteria performed better in the same population, with a sensitivity of 51.1 per cent.
In an Australian cohort of 885 patients referred to a lipid clinic for genetic testing, four clinical criteria were compared against a demonstrated FH-causing mutation [3]. US MEDPED had an odds ratio of 10.5 for predicting a mutation, comparable to DLCN definite (OR 9.4) and Simon Broome definite (OR 11.7). The Youden index was 0.457 for MEDPED, second best after DLCN definite (0.487). The AUC was significantly higher for DLCN definite and MEDPED than for the other criteria. In this selected lipid clinic population, MEDPED therefore balanced sensitivity and specificity for mutation prediction better than several alternatives, but still less well than DLCN.
In summary, US MEDPED performs well as a specific screening tool with a high positive predictive value, but with low sensitivity in populations in which the prevalence of FH is not high enough to compensate for the high thresholds.
Limitations
The tool rests on total cholesterol alone and includes neither LDL cholesterol, clinical signs (tendon xanthomas, corneal arcus), a family history of premature coronary artery disease nor genetic testing. This makes it less powerful than DLCN or Simon Broome, which integrate several dimensions.
The lower thresholds for relatives presuppose that the index case is confirmed molecularly or clinically. Using the relative rows for a patient whose "index case" is merely a relative with a high cholesterol, without confirmed FH, overestimates the probability of FH.
The thresholds were derived in a population in Utah with a degree of founder effect and were validated molecularly in the 1990s. Transferability to other populations is not self-evident, and the Malaysian validation showed low sensitivity in an Asian population [2].
The tool is not recommended in current European guidelines. The 2013 EAS consensus statement recommends the DLCN criteria for diagnosis and cascade screening of first-degree relatives with LDL cholesterol measurement, not the MEDPED thresholds [4]. Canadian practice likewise recommends Simon Broome or DLCN over MEDPED [5].
Finally, the thresholds are given in mg/dL. In Swedish clinical practice cholesterol is reported in mmol/L, which requires conversion (1 mmol/L total cholesterol ≈ 38.7 mg/dL).
References
- Williams RR, Hunt SC, Schumacher MC, et al. Diagnosing heterozygous familial hypercholesterolemia using new practical criteria validated by molecular genetics. Am J Cardiol 1993. PMID: 8328379
- Abdul-Razak S, Rahmat R, Mohd Kasim A, et al. Diagnostic performance of various familial hypercholesterolaemia diagnostic criteria compared to Dutch lipid clinic criteria in an Asian population. BMC Cardiovasc Disord 2017. PMID: 29037163
- Chan DC, Pang J, Hooper AJ, et al. A Comparative Analysis of Phenotypic Predictors of Mutations in Familial Hypercholesterolemia. J Clin Endocrinol Metab 2018. PMID: 29408959
- Nordestgaard BG, Chapman MJ, Humphries SE, et al. Familial hypercholesterolaemia is underdiagnosed and undertreated in the general population: guidance for clinicians to prevent coronary heart disease: consensus statement of the European Atherosclerosis Society. Eur Heart J 2013. PMID: 23956253
- Turgeon RD, Barry AR, Pearson GJ. Familial hypercholesterolemia: Review of diagnosis, screening, and treatment. Can Fam Physician 2016. PMID: 26796832
