Open

Preface

This Practical Guide is part of Forever Healthy's  "Maximizing Health"  initiative that seeks to holistically review the world's leading medical knowledge on various health-related topics and turn it into actionable information.

Motivation

Numerous aspects of lipid metabolism have been implicated in the genesis of cardiovascular disease (CVD). Given the prevalence and mortality of CVD, it is essential to identify optimal biomarkers of lipid metabolism that may provide better risk prediction. Several lipid biomarkers have been proposed for monitoring, and several methods have been developed for measuring them. This Practical Guide aims to identify those markers, evaluate their relevance to CVD risk prediction, and determine the best lipid monitoring protocol available at the moment. 

Key Questions 

This Practical Guide seeks to answer the following questions:

• Which lipid parameters are useful for monitoring CVD risk?

• Which test methods are available?

• Which test method is best for each parameter?

• What is the reference range for each parameter?

• What is the total error (imprecision + inaccuracy)?

• Which factors affect each parameter?

• At what age should testing commence?

• How often should testing be repeated?

• What is the optimal lipid monitoring protocol based on the currently available evidence?

Impatient readers may choose to skip directly to the Presentation of results section for tips on practical application. 

Methods

A list of potential lipid parameters was compiled from medical and laboratory resources. A literature search was then conducted on PubMed for each parameter using the search terms shown in Table 1 and includes results available as of August 14, 2020. Titles and abstracts of the resulting studies were screened and relevant articles downloaded in full text. The references of the full-text articles were manually searched in order to identify additional studies that may have been missed by the search terms.

Table 1: Literature Search 

Summary of Recommendations made in Clinical Practice Guidelines

Several clinical practice guidelines (CPGs) on lipid monitoring and management have been published in recent years and the main recommendations are summarized in the table below:

Table 2. Summary of Guideline Recommendations  

Examination of the recommendations made in the current CPGs reveals that many "strong" recommendations are made on the basis of a weak level of evidence. Additionally, guidelines that were developed from the same evidence base provide differing recommendations. Approximately 50% of the recommendations made in all major clinical guidelines are based on the lowest level of evidence (primarily expert opinion) (Allan et al., 2015). All of the evidence underlying recommendations as to which patients should receive lipid profiling and which parameters to screen is low (based on expert opinion, observational or retrospective studies) (Mach et al., 2019).

A review and appraisal (according to AGREE II criteria) of 19 lipid-management-related guidelines determined that the CPGs are generally of good quality (Zhou et al., 2019). The majority of CPGs chose LDL-C as the primary target for lipid management and some recommend non-HDL-C and apoB as secondary targets. In general, the newer guidelines prefer a nonfasting sample for screening. The appraisal identified inconsistencies between the guidelines related to the timing/frequency of blood tests and the target biomarker (Zhou et al., 2019).  

Conflicts of interest are also a significant problem with many guideline expert panel members having close ties to pharmaceutical companies that produce cholesterol-lowering medications (chriskresser.com; Zhou et al., 2019).

Additionally, none of the current guidelines contain any mention of which test method is optimal or the level of error to be expected for the recommended biomarkers.

Review of Lipid Biomarkers 

Our literature search identified 21 lipid biomarkers ranging from those included in a standard lipid panel to those that are just entering clinical practice. For each biomarker, the main evidence for and against its use in CVD risk prediction is summarized. The available information on test types, standardization & error, reference values, factors affecting measurement, and frequency of testing for each parameter is also analyzed.

The review begins with markers that are typically reported on a standard lipid panel as well as markers that can be derived from it:

• Total cholesterol

• LDL-C

• HDL-C

• Triglycerides

• VLDL-C

• Non-HDL-C

Additionally, we review the newer, less commonly measured markers collectively known as "advanced lipid tests": 

• LDL-P

• LDL subfractions

• HDL-P

• HDL subfractions

• HDL functional assays

• Lp(a)

• ApoB

• ApoA-I

• ApoB/ApoA-I

• ApoC-III

• Remnant lipoprotein particles

• Ox-LDL

• Sterols

• Fatty acids

• Lipidomics 
  
  

Total cholesterol

Total cholesterol (TC) has been a major component of cardiovascular risk screening models for several years (Peters et al., 2016). In addition to its association with CVD risk, abnormal levels of cholesterol have been observed in Smith-Lemli-Opitz syndrome, type II diabetes, neurological diseases, and cancer (Li et al., 2019). 

Many high-quality studies have linked elevated TC levels and increased intraindividual variation of TC to CVD risk. A systematic review (n=884,416) found that a 39 mg/dL (1 mmol/L) increase in TC was associated with a 20% increase in the relative risk of CVD in women and 24% in men (Peters et al., 2016). The top quartile of intraindividual variation in TC has been associated with an over 60% increase in all-cause mortality, a 40% increase in myocardial infarctions, and a 50% increase in strokes (Simpson, 2019).

In the Framingham study, subjects with elevated serum total cholesterol (>275 mg/dl) had an increased risk of adverse outcomes, whether healthy or with CAD. Compared with persons with cholesterol levels <200 mg/dl (<5.17 mmol/liter), the risk ratios for patients with elevated cholesterol levels were 3.8 for reinfarction, 2.6 for CAD mortality, and 1.9 for overall mortality (Kannel, 1995). A large meta-analysis (n= 256,259) examining the correlation of TC with cardiovascular disease events found a hazard ratio (per 1 standard deviation increase) of 1.24 (Perera et al., 2015).

A 1% drop in TC is associated with a 2-3% CVD risk reduction (Holme, 1992), at least in those under 50 years of age. Reducing high TC has been shown to reduce morbidity and mortality in young and middle-aged men (Naito et al., 1992). 

However, TC is evidently not the whole story as the prevalence of cholesterol levels ≥240 mg/dl (≥6.21 mmol/liter) in persons who had sustained myocardial infarction was only 35–52% in men and 66% in women and 20% of myocardial infarctions occurred in people who had cholesterol levels <200 mg/dl (<5.17 mmol/liter) (Kannel, 1995). Several studies have also observed that individuals with low TC were just as atherosclerotic as those with high TC (Ravnskov et al., 2018).

Some studies have shown that TC performs worse than cholesterol subfractions (LDL-C, HDL-C, etc...) in terms of clinical validity, responsiveness to treatment changes, and detectability of long-term changes (Glasziou et al., 2014). 

Up to certain levels, cholesterol is beneficial for the human body. It has crucial roles in the building of cell membranes and as a precursor of steroid hormones (Li et al., 2019), and high TC has been inversely correlated with all-cause mortality in people aged >65 (Liang et al., 2017). Over the age of 50, falling cholesterol levels are associated with decreased longevity. Each 1 mg/dL drop in TC per year, was associated with an 11-14% increase in coronary and total mortality (Anderson et al., 1987). 

Test types

There are three main categories of analytical methods for cholesterol level determination: 1) classical chemical methods, 2) fluorometric and colorimetric enzymatic assays, and 3) analytical instrument approaches. While classical methods are simple and inexpensive to perform, multi-step procedures are required. Enzymatic assays are costly. Chromatographic and mass spectrometric methods are the most accurate and sensitive but are costly and require extensive sample pretreatment (Li et al., 2019). 

The National Institute of Standards and Technology has approved two methods, isotope-dilution mass spectrometry, and a modified Abell-Kendall protocol, as reference measurement procedures for blood cholesterol quantification (Li et al., 2019). Although the modified Abell-Kendall reaction is considered a standard reference-method, its use is not widely accepted for routine tests since highly corrosive agents are used (Li et al., 2019). For routine analysis, many labs use commercial quantification assay kits, handheld point of care testing devices, and automated analyzers based on enzymatic assays. 

Standardization & error

Among the lipid and lipoprotein analytes, TC was the first to undergo standardization (a process that makes the results comparable across measurement procedures ) and is now the most straightforward and well-standardized of the analytes (Warnick et al., 2008). 

The expected total error for TC is ± 8.9-10% ( Warnick et al., 2008 ) where the total error is the sum of errors in accuracy (closeness to the true value) and precision (closeness of repeat measurements). 

The National Cholesterol Education Program ( NCEP) performance goal for total error is ≤ 8.9%. That is, 95% of individual TC measurements should fall within 8.9% of the reference method procedure (RMP) value over repeated measurements (Warnick et al., 2008). However, a study on error levels of 5 methods for TC screening found that none of the methods met the NCEP performance recommendations. The errors led to a risk level misclassification in 5-18% of patients (Miller et al., 1993). 

Enzymatic assays are subject to additional challenges as they may not be strictly selective for cholesterol (react with other sterols). Some chemicals present in a sample (ascorbic acid or bilirubin) may consume hydrogen peroxide, creating bias during indirect cholesterol quantification ( Li et al., 2019 ).

Gas and liquid chromatography are widely accepted to be more reliable, sensitive, and accurate than the other methods (Li et al., 2019). 

On average, the intraindividual variation of TC is 5-6.4% (Warnick et al., 2008; da Silva et al., 2018). TC variation is estimated to be about 2.8% within-day, 7.9% within-month, and 3.9-10.9% over the long term (Naito et al., 1992). At the extreme, day-to-day cholesterol values in the same individual have been shown to vary by 15%, and an 8% difference has been identified within the same day (Naito et al., 1992). 

Reference values

Factors affecting measurement

Season

• Blood lipid levels exhibit mild seasonal variation (3-5%) with a peak in total cholesterol levels in the winter and a trough in the summer. One study found that the amplitude of seasonal variation of total cholesterol concentration was 3.9 mg/dL (0.10 mmol/L) in men and 5.4 mg/dL (0.14 mmol/L) in women (Uptodate.com; Naito et al., 1992).

Position

• Positional changes can affect cholesterol levels. Levels can decrease by as much as 15% in the recumbent (horizontal) position. As a result, hospitalized patients can be expected to have lower levels than outpatients (Mosby's Manual of Diagnostic and Laboratory Tests, 2017).

Sex & Age

• Adult males 20-45 years generally have higher levels than age-matched females. Women <20 years of age and postmenopausal women generally have higher levels than age-matched males. In both genders, the levels increase 1.5 mg/dL/year with levels in males plateauing at age 50 and in females increasing exponentially (Naito et al., 1992).

Diet

• May influence TC by about 10% (Naito et al., 1992). Maintain a stable weight and usual diet for 2 weeks before lipid assessment (Naito et al., 1992). 

Alcohol

• Excessive alcohol will slightly increase TC (Naito et al., 1992). 

Exercise

• Habitual vigorous activity causes a decrease of 0-15% but the effects may not be apparent for 3-6 weeks (Naito et al., 1992). 

Illnesses

• Certain illnesses can affect cholesterol levels. For example, patients with acute myocardial infarction (AMI) may have as much as a 50% reduction in cholesterol level for as long as 6 to 8 weeks ( Mosby's Manual of Diagnostic and Laboratory Tests, 2017 ).

Stress

• Mental stress can raise cholesterol by 10-50% in the course of half an hour (Ravnskov et al., 2018). 

Drugs 

• adrenocorticotropic hormone, anabolic steroids, beta-adrenergic blocking agents, corticosteroids, cyclosporine, epinephrine, oral contraceptives, phenytoin (Dilantin), sulfonamides, thiazide diuretics, and vitamin D may increase TC ( Mosby's Manual of Diagnostic and Laboratory Tests, 2017 )

• allopurinol, androgens, bile salt-binding agents, captopril, chlorpropamide, clofibrate, colchicine, colestipol, erythromycin, isoniazid, liothyronine (Cytomel), monoamine oxidase inhibitors, niacin, nitrates, and statins may decrease cholesterol ( Mosby's Manual of Diagnostic and Laboratory Tests, 2017 )

Test conditions

• Fasting is not required (Oxford Handbook of Clinical and Laboratory Investigation, 2017)

• The test can be performed on serum or using a whole blood spot (Oxford Handbook of Clinical and Laboratory Investigation, 2017)

Frequency 

Assuming 5% analytical imprecision and a mean biological variation of 6.5%, it was determined that duplicate analyses of 4 specimens per patient would be optimal to determine the true TC value (Naito et al., 1992). For practicality, the lipid standardization panel recommends two measurements taken at least one week apart but within two months (Naito et al., 1992).

Current guidelines by organizations such as the American Heart Association recommend measurements once every 4-6 years because recent evidence has suggested that, due to the weak signal–noise (SN) ratio of cholesterol levels, frequent screening and/or monitoring – as part of long-term management – more often captures measurement error rather than true changes (Perera et al., 2015). However, a detailed analysis showed that annual monitoring was both cost-saving and effective (Perera et al., 2015).

Total Cholesterol Summary Table 

LDL-C

It is widely accepted that low-density lipoprotein (LDL) has a causal role in CVD. Patient-level meta-analyses (n=170,000) by the Cholesterol Treatment Trialists’ Collaboration found that for every 40 mg/dL of LDL-C reduction with statins, there was a 10% relative decrease in all-cause mortality, 20% in coronary heart disease mortality, and a 22% proportional reduction in the risk of major cardiovascular events over a median 5 year period of treatment. Further analyses also showed reductions in the need for myocardial revascularization and ischemic stroke (17%) during each year the treatment was continued (Rached & Santos, 2020;  da Silva et al., 2018). 

Intraindividual LDL-C variability has also been linked to increased CVD risk. Each standard deviation of LDL-C variability is associated with around a 10-20% increase in CVD events. An analysis of 9 coronary intravascular US trials showed that coronary atheroma progression was associated with higher variability of LDL-C. In patients with Type II DM, being in the top quartile of LDL-C variability was associated with an approximately 10% greater carotid intima-media thickness (cIMT) (Simpson, 2019). 

Despite its wide acceptance as an ideal marker of CVD risk, several studies bring into question the causal role of LDL-C in CVD. It has been argued that if LDL-C is atherogenic, people with high LDL-C should have more atherosclerosis than those with low LDL-C. At least four studies have shown a lack of an association between LDL-C and the degree of atherosclerosis. In a study of 304 women, no association was found between LDL-C and coronary calcification (Ravnskov et al., 2018) and numerous Japanese studies have found that high LDL-C is not a risk factor for CVD mortality in women of any age (Ravnskov et al., 2018). 

I n a large American study, that included almost 140,000 patients with acute myocardial infarction, LDL-C at the time of admission was lower than average (Ravnskov et al., 2018). Additionally, despite having LDL-C levels at the goal defined by the guidelines, patients often present with elevated levels of other atherogenic lipoproteins (ApoB 25%, non-HDL-C 31.1%, and ox-LDL 49.2%) (Paredes et al., 2019) and suffer CVD events that thus correlate more strongly with other markers. 

A large meta-analysis comparing 10 lipid biomarkers found a hazard ratio of only 1.16 with elevated LDL-C for CVD events, amongst the lowest of the markers analyzed (Perera et al., 2015). Furthermore, the most important outcome – an increase in overall life expectancy – has never been mentioned in any cholesterol-lowering trial. As calculated from available information, statin treatment does not prolong lifespan by more than an average of a few days (Ravnskov et al., 2018). 

Test types

Ultracentrifugation (reference method)

• The combination of ultracentrifugation and heparin-Mn2+ precipitation, with the generic term ‘β quantification’ (BQ) has been proposed as the LDL-C reference method by the NCEP (Ramasamy, 2018;  Nakamura et al., 2014). The  BQ method includes the cholesterol in intermediate-density lipoproteins (IDL) and the cholesterol in lipoprotein(a) (Lp(a)) in addition to the cholesterol in the classical LDL fraction. Moreover, very-low-density lipoprotein (VLDL) remnants can be found in both the VLDL range and the IDL range, which is measured as LDL by BQ (Contois et al., 2011). Measuring LDL-C by ultracentrifugation is time-intensive and expensive.

Calculated LDL-C

• The Friedewald formula is used as the primary method to calculate LDL-C from TC, triglycerides (TG), and HDL-C provided the patient is sampled after an overnight fast, and the TG values are < 400 mg/dL (Ferraro et al., 2019). This formula calculates LDL-C in mg/dL by subtracting the sum of HDL-C and VLDL-C (TG/5) from TC.

Direct LDL-C

• By homogeneous methods, LDL-C is separated from other cholesterol fractions with the characteristics of surfactants, and LDL-C is directly measured using an automatic analyzer. Several different homogeneous LDL-C assays are used for routine clinical analysis (Yano et al., 2019). The potential advantages of direct measurement of LDL-C include the ability to measure LDL-C at high levels of TG, the ability to measure LDL-C in nonfasting samples, and the reduction of imprecision by a single measurement instead of a value that is calculated from 3 different results.

Standardization & error

The Clinical Laboratory Improvement Amendments of 1988 (CLIA) evaluation criteria recommended LDL values should be within ± 30% of the reference method  (Warnick et al., 2008). The NCEP criteria for LDL-C expect inaccuracy ≤ ±4% of the RMP, imprecision of ≤ ±4%, and a total error ≤ ±12% ( Warnick et al., 2008). 

The Friedewald calculated value of LDL-C has well-established limitations: 1) methodological errors may accumulate since the formula necessitates three separate analyses of TC, TGs, and HDL-C; and 2) a constant cholesterol/TG ratio in VLDL is assumed. With high TG values (>4.5 mmol/L or >400 mg/dL) the formula cannot be used (escardio.org). It has been suggested that Lp(a) cholesterol should also be subtracted for a more accurate determination of LDL-C contribution to cardiovascular risk (Ramasamy, 2018). Additionally, t he equation has not been validated for use in patients using statins or other lipid-lowering drugs (Preiss & Neely, 2015).  Calculated LDL-C exhibits an underestimation bias of about 5% compared with direct assays (Schaefer et al., 2016). 

Several attempts to modify the formula to achieve a more accurate measurement have been attempted. A study (n= >1.3 million) that replaced the fixed ratio of 5 for VLDL-C estimation with one of 180 adaptable ratios based on a patient's individual non-HDL-C and TG values (Martin's formula) found that i n patients with LDL-C <70 mg/dL and triglycerides of 150-199 mg/dL, LDL-C accuracy was 92% compared to 61% with the Friedewald equation. In those with TG levels of 200-399 mg/dL, the accuracy was 83% compared to only 40% with Friedewald calculation (Ferraro et al., 2019). The overall accuracy of the equation compared to direct ultracentrifugation (the RMP) was 92% in contrast to 85% accuracy for Friedewald estimation (Ferraro et al., 2019). 

However, there are inaccuracies using Martin's formula as well. It has been reported that Martin’s equation was not accurate for estimating LDL-C unless TG was between 300 and 400 mg/dl and another formula was recommended in its place (Cordova & Cordova formula) as an alternative when direct LDL-C measurement is not available (Cordova et al., 2020). 

The risk-based cut points for LDL cholesterol defined by NCEP are determined from epidemiological studies, many of which used calculated LDL-C rather than the reference method (Contois et al., 2011).

Direct LDL-C assays use proprietary chemical-based methods and are not necessarily reliable. They are not standardized and, in some cases, can be even less accurate than the Friedewald equation (Ferraro et al., 2019). The reaction specificities of homogeneous methods to LDL and VLDL subfractions have been shown to vary, with reduced reactivity to small dense LDL and nonspecific reactivity to VLDL.

Direct LDL-C assays have been reported to give falsely low LDL-C in patients with various causes of cholestasis (Ramasamy, 2018). One study reported that  7 of 8 direct LDL-C methods failed to show improved CVD risk score classification over the corresponding calculated LDL-C methods ( van Deventer et al., 2011 ).  Another study comparing direct LDL-C assessment found that the total error ranged from 13.5% in healthy individuals to -26.6% and + 31.9% in those with known cardiovascular disease or dyslipidemias (Ferraro et al., 2019). In contrast, a  comparison of four different homogeneous direct methods for LDL-C with the β quantitation reference method found that all had acceptable assay precision within the NCEP performance guidelines. However, all four methods exhibited a progressively poor performance as the TG concentrations increased, though the performance was superior to the Friedewald calculation (Ramasamy, 2018). 

Normal physiologic variation of LDL-C is reportedly between 7.8-8.2 % (da Silva et al., 2018).  Direct LDL-C assays show less intraindividual variability than the Friedewald equation (Simpson, 2019).

Reference values