Performance of Calculated and Directly Measured Low-Density Lipoprotein Cholesterol in a Pediatric Population

Abstract

Objectives:

An assessment of methods for the accurate measurement of low-density lipoprotein cholesterol (LDL-C) at decreased concentrations has not yet been carried out. We evaluated the performance of the Friedewald equation, a direct enzymatic assay, and a novel equation for determining LDL-C levels in a pediatric population with elevated triglycerides and reduced LDL-C levels.

Methods:

LDL-C concentrations of 127 pediatric patients were determined by the Friedewald equation, a direct enzymatic assay, and a novel equation. The bias of each approach was assessed at selected LDL-C cutoffs and after stratifying samples by triglyceride content. The concordance of each approach, relative to the reference method, was determined at LDL-C cut-points of less than 70, 70 to 99, and 100 to 129 mg/dL.

Results:

The Friedewald equation substantially underestimated pediatric LDL-C concentrations below 100 mg/dL in the presence of elevated triglycerides. The Ortho Clinical Diagnostics (Raritan, NJ) direct LDL assay was positively biased at low LDL-C levels. The novel equation most effectively reduced the bias of the Friedewald equation at all LDL-C concentrations and increased the concordance of sample classification to the reference method.

Conclusions:

The novel equation should be used for accurate measurement of pediatric LDL-C when the concentration is below 100 mg/dL in the presence of elevated triglycerides (150-399 mg/dL).

Materials and Methods

Estimating low-density lipoprotein with the Friedewald equation (LDL-F) is the most widely used method for determining low-density lipoprotein cholesterol (LDL-C) concentrations and offers a convenient, cost-effective, and time-saving approach relative to the reference method (LDL-C with ultracentrifugation [LDL-UC]). Limitations to the formula’s use, however, include the failure to accurately quantify LDL-C in nonfasting individuals, when triglycerides (TGs) exceed 399 mg/dL (to convert to mmol/L, multiply by 0.01129), and in the presence of type III hyperlipidemia.¹ Recent studies also indicate that the Friedewald equation is constrained by a large negative bias when LDL-C levels fall below 70 mg/dL (to convert to mmol/L, multiply by 0.02586) and that this effect is amplified with increasing TG levels (>150 mg/dL).^2,3

For patients undergoing lipid-lowering therapy, a reasonable alternative to the Friedewald equation is to measure LDL-C by a direct homogeneous method (direct LDL). These assays are automated, do not require fasting, and provide timely results. Previous studies have highlighted the application of direct LDL-C assays to adult populations by demonstrating that they meet the performance standards set by the National Cholesterol Education Program (NCEP) and may serve to stratify patients into cardiovascular disease risk categories better than LDL-F.^4,5

For pediatric populations, accurate LDL-C measurement at low concentrations is needed because of increased screening for dyslipidemia. The 2011 National Heart, Lung, and Blood Institute guidelines suggests an initial nonfasting lipid panel be obtained for all patients aged 9 to 11 years, with a follow-up fasting lipid panel for those with abnormal findings.⁶ The rationale for lipid screening is earlier identification and treatment of children with familial hypercholesterolemia (FH) and those at risk for metabolic syndrome. For children diagnosed with FH, the therapeutic aims are to reduce LDL-C to 130 mg/dL or less or by 50% or more.⁷ If these goals are met, LDL-C concentrations may fall into the levels that cannot be accurately quantified by the Friedewald equation. Furthermore, the trends in pediatric dyslipidemia in the United States are a pattern of normal to mild LDL-C increases with moderate to severe TG elevations.⁶ Therefore, clinicians relying on Friedewald-calculated LDL-C for monitoring of treated high-risk patients or as a first-line screen for dyslipidemia chance misclassifying children into inappropriate NCEP risk categories.

Results

Study Population

A potential solution to monitor high-risk children is to use a direct homogeneous assay, but very little information exists regarding the performance of these methods at low LDL-C levels.⁸ In 2013, Martin et al⁹ proposed determining LDL with a novel equation (LDL-N) that uses an adjustable factor, based on non–HDL-C and TGs, to correct for the contribution of cholesterol from very low-density lipoprotein cholesterol (VLDL-C) in LDL-C estimation. The novel equation was reported to improve the concordance of estimated LDL-C to the reference method at concentrations of less than 70 mg/dL, particularly in the presence of high levels of triglycerides. Despite being derived and validated in a large cohort, less than 1% of participants for LDL-N development were younger than 18 years. Furthermore, the novel equation has not been independently validated in a pediatric population.

Analytical Methods

Another option to correct LDL-C estimation is to modify the Friedewald equation, using an alternative static factor to represent VLDL-C, by empirical derivation. Previous studies using a similar tactic reported improved accuracy in LDL-C estimation, but this type of approach has not been evaluated in children.^10-12

In this study, we report on the performance of two calculations (LDL-F, LDL-N) and a direct homogeneous method (direct LDL) for quantifying LDL-C in pediatric patients, having decreased LDL-C levels and elevated TGs, relative to the reference method. In addition, we assessed a modified Friedewald equation using a static factor of 7.5 (LDL-C_7.5) in the denominator of the VLDL-C portion of the equation.

Percent bias of calculated and directly measured low-density lipoprotein cholesterol in pediatric samples.

Data Analysis

Discussion

Bias of Calculated and Measured LDL-C at Low Concentrations in Pediatric Samples

The percent bias of the Friedewald equation, the novel equation, and direct LDL is presented in ❚Figure 1❚. A negative bias was observed for Friedewald-estimated LDL-C at all concentrations. At LDL-C levels of 100 mg/dL or less, the bias of the Friedewald equation was −21.6%. Inaccuracy was accentuated to −29.6% at an LDL-C cutoff of 70 mg/dL or less and decreased at LDL-C levels more than 100 mg/dL (−6.8%). Relative to the reference method, the direct LDL assay was positively biased (10.4% and 10.8%) when LDL-C was 100 mg/dL or less and 70 mg/dL or less, respectively. For concentrations of more than 100 mg/dL, the direct assay’s performance improved to a bias of 4.5%. The novel equation proposed by Martin et al⁹ proved to be more accurate than LDL-F and the direct LDL assay at all LDL-C concentrations. A bias of −0.2% was present when LDL-C concentrations were 100 mg/dL or less, and this was increased to −1.2% when only samples 70 mg/dL or less were included. Finally, for specimens having an LDL-C level of more than 100 mg/dL, the bias of LDL-N was 0.5%.

The bias of the Friedewald equation progressively increased at higher TG concentrations for all LDL-C levels. This effect was more pronounced at an LDL-C of 70 mg/dL or less and 100 mg/dL or less, where the negative bias increased from −19% to −36.3% and −14.6% to −28%, respectively, with a rise in TGs ❚Figure 2❚ A similar trend was observed for the direct LDL assay, although to a lesser degree than with the Friedewald equation. At LDL-C concentrations of 70 mg/dL or less and 100 mg/dL or less, elevated TG levels resulted in a bias increase of 6.8% and 5.3%, respectively, for the direct LDL method. For LDL-N, the bias did not worsen as TGs went from 150 to 199 mg/dL to 200 to 399 mg/dL at all LDL-C concentrations. For LDL-C less than 70 mg/dL, the bias slightly increased with TG content, going from −1.9% to 3%. However, for LDL-C less than 100 mg/dL, the novel equation’s bias decreased from −2.3% to 1.5%.

The Friedewald equation correctly classified all patients with an ultracentrifugation LDL-C value of less than 70 mg/dL ❚Table 2❚. However, the concordance of LDL-F was poor between 70 and 99 mg/dL, where 58% of samples were incorrectly estimated to be less than 70 mg/dL. In this same LDL-C range, the direct LDL assay and the novel equation misclassified 25% and 22%, respectively. For LDL-C levels between 100 and 129 mg/dL, LDL-F incorrectly classified 13 (72%) of 18 patients, while the direct LDL assay and LDL-N reclassified eight (44%) of 18 and 10 (56%) of 18 patients, respectively.

The LDL-C_7.5 equation was the most accurate method for LDL-C estimation at levels of 70 mg/dL or less (−0.3% bias), but a slight positive bias of 1% was observed at an LDL-C of 100 mg/dL or less ❚Table 3❚ and ❚Table 4❚. In contrast to the Friedewald equation and the direct LDL assay, the bias of LDL-C_7.5 did not increase in the presence of higher TG levels. In fact, bias decreased at all LDL-C cutoffs in samples containing 200 to 399 mg/dL TGs relative to 150- to 199-mg/dL concentrations. An evaluation of the concordance of LDL-C_7.5 to LDL-UC risk group classification revealed that LDL-C_7.5 performed better than LDL-N at all cut-points. In comparison to the direct LDL assay, LDL-C_7.5 was a more accurate method for classifying patients at lower LDL-C levels but not at a cut-point of 100 to 129 mg/dL.

This is the first study that we know of to evaluate the performance of a direct LDL assay and the novel equation in a pediatric population with low LDL-C and elevated TGs. This work addresses the need for accurate LDL-C measurement at concentrations near 70 mg/dL, which is warranted by recent updates to the therapeutic goals for children with familial hypercholesterolemia or at risk for metabolic syndrome. The most important finding of our study is that the novel equation substantially reduces the bias of the Friedewald equation at low LDL-C concentrations in pediatric patients. Furthermore, LDL-N was not subject to large increases in bias in samples with elevated TGs, as was noted with the other methods.

As expected, the Friedewald equation produced a large negative bias at low LDL-C concentrations, which increased with higher TG levels. On average, Friedewald-estimated LDL-C was 10 mg/dL lower than the reference method at an LDL-C of 70 mg/dL or less and TG concentrations of 150 to 199 mg/dL; this difference increased to −20 mg/dL at TG concentrations of 200 to 399 mg/dL. An important consequence of this underestimation is illustrated by the high misclassification rate present at LDL-C cut-points of 70 to 99 mg/dL and 100 to 129 mg/dL. Previous reports point out that the negative bias of the Friedewald equation stems from its reliance on a static factor (ie, TG/5) to represent cholesterol from VLDL-C and therefore does not account for interindividual variation in this ratio.^2,13,14 In fact, this shortcoming was acknowledged in the original Friedewald study, and alternative LDL-C calculations have been suggested but not widely accepted.^1,10-12

We next evaluated a direct LDL method to determine if this provided a more accurate measure of LDL-C in the lower ranges. These assays have the advantage of being available from various manufacturers and not requiring fasting, which is particularly attractive for pediatric LDL-C measurement. Our results indicate that the direct LDL assay provided a better estimation of LDL-C at reduced levels than the Friedewald equation but still produced considerable bias. Interestingly, the bias was also noted to slightly increase in samples having relatively higher TG content. This was surprising granted the reported application of these assays for hypertriglyceridemic patients, and we suspect this effect is due to turbidimetric interference. Although other studies have characterized the performance of direct LDL assays in children, they have focused on hyperlipidemic populations or comparisons of fasting vs nonfasting specimens.^15-17 These pediatric studies, as well as others in adults, report mixed results on the utility of direct LDL assays relative to the Friedewald equation but have failed to specifically consider their application at LDL-C levels below 70 mg/dL or with elevated TGs.^18-20 Overall, our findings demonstrate that the Vitros direct LDL assay is a more accurate option than the Friedewald equation at decreased LDL-C levels, but this approach is limited by a positive bias and may therefore misclassify patients into higher risk categories. Together with the extra cost associated with the direct LDL assay, we conclude that this approach is not appropriate for our pediatric population.

Influence of TG Concentration on the Bias of LDL-F, Direct-LDL, and LDL-N

The novel equation proved to be more accurate than both the Friedewald equation and the direct LDL assay. This was most evident at an LDL-C of 100 mg/dL or less, where a bias of −0.2% was observed, but the improved accuracy was apparent at all LDL-C concentrations assessed. As expected, this translated into enhanced concordance at cut-points of less than 70 mg/dL and 70 to 99 mg/dL. In addition, the novel equation was not subject to a large increase in bias with elevated TGs, as were the Friedewald equation and the direct LDL assay. Although a relative increase in bias did occur at an LDL-C of 70 mg/dL or less with higher TG concentrations, this equated to a net increase in bias of 1.1%. Furthermore, for samples having an LDL-C of 100 mg/dL or less (n = 103), our results show a decrease in bias with concomitant TG elevation. The improved LDL-C estimation with the novel equation is the result of its ability to account for a variable proportion of cholesterol from VLDL-C. An adjustable factor, ranging from 3.1 to 11.9, is substituted into the equation based on non–HDL-C and TG levels. The adjustable factor therefore functions as a sliding scale, allowing for modification of the TG/VLDL-C ratio in cases where the Friedewald equation is under- or overestimating this value.

Concordance of LDL-F, Direct LDL, and LDL-N at Clinically Relevant LDL-C Cutoffs

Although proposed only 3 years ago, the novel equation has already received a fair amount of attention in validation studies aiming to improve LDL-C estimation. In the first report published, more than 23,000 patients (91% adults) were included in a comparison of the novel equation and the Friedewald equation to the reference method.²¹ Similar to our findings, this study reported an increased concordance by the novel equation at cut-points of less than 70 mg/dL and 70 to 99 mg/dL. In addition, the authors noted that there was a clear improvement in estimated LDL-C values using the novel equation relative to the Friedewald equation. Following this study, two groups in Korea evaluated the novel equation in adult populations and also reported increased accuracy relative to the Friedewald equation.^22,23 Despite these optimistic initial reports, a perspective that should be considered is the difficulty that a laboratory may encounter in implementing the novel equation. For example, adjusting and validating a laboratory information system (LIS)–based calculation to include the 180-cell strata reported by Martin et al⁹ is a daunting task. Smaller laboratories may not have the resources, knowledge, or personnel to implement the novel equation and would likely be better served by another option for low LDL-C measurement.

Percent bias of calculated and directly measured low-density lipoprotein cholesterol at increasing triglyceride concentrations.

The modified LDL-C equation (LDL-C_7.5), which uses a static TG/7.5 ratio, yielded the lowest bias of all methods at an LDL-C of 70 mg/dL or less. The TG/7.5 ratio is based on the median adjustable factor determined by rearranging the novel equation to solve for x using the reference method LDL-C values of 70 mg/dL or less (ie, x = TG/(non–HDL-C – LDL-UC). In contrast, the median adjustable factor from the novel equation was 7.6 in samples with an LDL-UC of 70 mg/dL or less. This calculation was tested in our study for two reasons. First, incorporating the median factor determined by the reference method should maximally reduce bias since it is derived from the gold-standard measurement procedure. Second, the use of an equation with a static ratio, rather than an adjustable factor, provides the benefit of ease of implementation for LIS calculation. As opposed to the novel equation, the LDL-C_7.5 equation requires only a single modification to the Friedewald equation, making it a less demanding alternative. Furthermore, laboratories may empirically determine the ideal ratio for their own population by performing a simple method comparison and calculating the median adjustable factor. For laboratories serving both adult and pediatric populations, a solution would be to reflex to the modified equation based on age. Despite the limitations associated with using an equation that relies on a static factor, the LDL-C_7.5 equation provides a much improved estimate of LDL-C at concentrations of 70 mg/dL or less relative to the Friedewald equation and offers the additional benefit of simple integration into LISs.

For accurate LDL-C measurement at levels below 100 mg/dL, in samples with TGs between 150 and 399 mg/dL, laboratories serving pediatric populations should consider using the novel equation. Institutions that cannot integrate the novel equation into their LIS have the option of using the modified equation as a simple alternative to reduce bias. However, because the modified equation does not account for interindividual variability in VLDL-C content, the novel equation should be used whenever possible. In our hospital, where we have a dedicated LIS support team, we have begun the process of implementing the novel equation. Our plan is to provide LDL-N values along with LDL-F and a disclaimer cautioning clinicians about the underestimation of LDL-C by the Friedewald equation.

LDL-C, low-density lipoprotein cholesterol.

Conclusion

Bias and Concordance of Calculated LDL-C Using a Static TG/VLDL-C Ratio of 7.5

References

Author notes