• atherosclerosis
• CHOLESTEROL
• DIAGNOSIS

There is robust evidence based on epidemiological, genome-wide association studies, Mendelian randomisation as well as randomised controlled trials (RCTs) implicating low-density lipoprotein cholesterol (LDL-C) as a causative factor in atherosclerotic cardiovascular disease (CVD).1 This is consistent with lipoprotein turnover modelling where, in terms of lipoproteins, low-density lipoprotein (LDL) particles are approximately seven times greater in number and remain in the circulation seven times longer than triglyceride (TG) containing particles in most individuals. So given its importance, what is the best way of assessing LDL-C? The short answer is to measure apolipoprotein B (ApoB), and the long answer has recently become more complicated given the increase in the number of LDL-C equations and improvements in direct LDL-C assays.

Evidence from a meta-analysis of epidemiological studies and RCTs on statins has shown that ApoB as a CVD risk predictor outperforms both LDL and non-high-density lipoprotein cholesterol (HDL-C).2 3 This discordance is magnified in those with hypertriglyceridaemia, where less cholesterol will be included in LDL and more in remnant particles. In a recent primary/secondary population study, LDL-C and non-HDL-C become non-significant CVD risk markers when ApoB is taken into account.4 This demonstrates that the CVD risk of LDL-C and non-HDL-C is more accurately predicted by particle number rather than the cholesterol content of the ApoB particles. This is reflected in recent guidelines where ApoB has been recommended for CVD risk assessment and a recognition that measurement can be adequately standardised to meet requirements for patient care.5 6

ApoB-100 is carried on a variety of atherogenic particles: LDL, lipoprotein(a), intermediate-density lipoproteins, very low-density lipoproteins and remnant particles. ApoB-48 is solely carried on chylomicrons; it is important to note that most of the commonly available ApoB assays are immunoassay (nephelometry/turbidimetry) and measure total ApoB (which includes both ApoB-100 and ApoB-48). That being said, even in postprandial samples in healthy individuals, the number of chylomicrons is usually less than 1%.7 However, a reference method using liquid chromatography-multiple reaction monitoring-mass spectrometry to measure ApoB-100 has been proposed, which may help eliminate this overlap.8 9 Issues to the introduction of ApoB-100 include the lack of widespread availability, assay cost in comparison with non-HDL-C or LDL-C (both of which can be calculated) and the fact that LDL-C is so ingrained in both guidelines and medical assessment.

Clearly, a significant adjustment/education would need to occur in order for this to be adapted, and a recent paper has sought to facilitate this process by generating a regression equation to convert ApoB results into an LDL-C .10 There are, however, some dissenting voices, with some studies suggesting that apoB adds little over non-HDL-C or LDL-C in populations with lower CVD risk.11 12

The original paper that has led to the widespread use of the Friedewald equation (FRE) is now 50 years old and continues to be the choice of providing LDL-C for many laboratories.13 The National Institutes of Health (Bethesda, USA)-based study had 448 subjects, 96 normal, 204 with type II hyperlipoproteinemia and 148 with type IV hyperlipoproteinemia. Fasting samples (12–14 hours) were used and the FRE was compared with ultracentrifugation (UC), the gold standard of LDL measurement. The correlation with UC was good for the former two groups but less so for TG >4.5 mmol/L, mainly in the latter group. One point to remember about the original FRE paper is that only correlation plots were used as opposed to difference plots or consideration on whether patients would be misclassified based on CVD risk assessment. In the accompanying best practice, the authors remind us the FRE also loses accuracy with TGs above 1.69 mmol/L and with LDL-C of <1.8 mmol/L.14 At this level, FRE underestimates LDL and may result in unnecessary treatment adjustment. Meta-analysis of RCTs suggests that mortality benefits from lipid-lowering therapy are limited below LDL-C of 2.6 mmol/L, but non-fatal CVD benefits continue to accrue at levels below this.15

In order to address some of the deficits associated with FRE, a number of alternative equations to calculate LDL-C have been developed, which are listed in supplemental table 1 in the accompanying best practice article.14 Two equations in particular have emerged to the forefront, the Martin-Hopkins equation (MHE) and the Sampson equation (SE).16 17 MHE has a clear advantage for low LDL-C; however it is complicated by requiring a 180 cell table which uses very patient-specific ratios of TG to non-HDL-C.18 Interestingly, the fasting status of the patients were unknown and those with TG above 4.5 mmol/L were excluded.16 On the other hand, the SE shows more accuracy for estimation of LDL-C in hypertriglyceridaemia but starts to lose accuracy above TGs of 6.5 mmol/L.19 What is underappreciated from the authors’ experience is that the FRE can generate negative LDL values. This is likely due to the negative bias with hypertriglyceridaemic samples and has also been seen in the MHE but rarely with the SE.17

Different international lipid guidelines use different lipid parameters to input into pooled cohort equations to calculate CVD risk to determine the need for lipid-lowering therapy for primary prevention.14 As discussed in the accompanying best practice article, while the US and European guidelines may use LDL-C, the UK guidance uses total cholesterol:HDL-C ratio for CVD risk assessment. The challenge with assessing an individual patient risk with such risk equations is that they are predominated by age, such that CVD risk maybe overestimated in elderly patients and underestimated in younger patients.20 To help address this gap, a combination of pooled cohort equation and RCT data has been proposed, and offering an individual patient a 10-year CVD risk could be balanced by consideration of whether they are eligible for a relevant RCT for lipid-lowering therapy.20

Accurate measurement using a single direct LDL-C assay may have been assumed to be the more precise method for assessing LDL-C compared with FRE, not just for CVD risk but also in assessment of those with potential familial hypercholesterolaemia. At least six such assays have been available since the 1990s; however, reagent costs and inaccurate results in patients with hypertriglcyeridaemia have meant that many labs have continued to use FRE.21 22 There have been some improvements in assay technology, and one such assay has recently been shown to be superior to both FRE and MH for CVD risk prediction in a Framingham outcome study.23

In conclusion, FRE is now 50 years old and is likely to be sufficient for those at low risk of CVD for primary prevention to provide input into the pooled cohort equation. For those with hypertriglyceridaemia, diabetes and metabolic syndrome or to guide therapy on lipid-lowering therapy for those at high risk of CVD, the FRE lacks precision and accuracy. The use of novel LDL-C formulae should assist with this, in addition to the use of additional biomarkers such as ApoB, which has been recommended by the European Society of Cardiology guidelines as an alternative to LDL-C for screening, diagnosing and management of dyslipidaemia for this cohort.24 Further studies on new equations or assays should go beyond correlation and focus on the number or degree of patient reclassification or clinical outcome studies to help translate the benefits of modifications in laboratory testing into clinical practice.