INTRODUCTION
Low density lipoprotein cholesterol (LDL-C) is an important risk factor for cardiovascular diseases (CVD), such as myocardial infarction and stroke. Additionally, LDL-C levels have been used as treatment guidelines to determine whether an individual needs therapeutic lifestyle changes or interventions such as drug treatment [
1,
2]. Although it is important to accurately assess LDL-C through direct measurements in routine clinical practice, LDL-C levels have generally been estimated using the Friedewald formula [
3]. In the Korea National Health Screening Program (KNHSP), Friedewald-derived LDL-C estimates are calculated when triglyceride (TG) levels are lower than 400 mg/dL [
4]. In 2013, of 11,380,246 participants whose lipid profile was tested in the KNHSP, there were 236,436 persons (2%) with TG levels above 400 mg/dL [
5]. This means that the LDL-C levels of 98% of participants in the 2013 KNHSP were not directly measured, but indirectly measured using the Friedewald formula.
The Friedewald formula calculates LDL-C as LDL-C = total cholesterol – high density lipoprotein cholesterol (HDL-C) – (TG/5), using directly measured values of total cholesterol, HDL-C, and TG [
6]. The last term, (TG/5), is an estimate of very low density lipoprotein cholesterol (VLDL-C). The inaccuracy of the Friedewald formula when TG levels are above 400 mg/dL was first demonstrated by Friedewald et al. [
6]. A number of later studies have cautioned against relying on the Friedewald formula even when TG levels are under 400 mg/dL, suggesting that the Friedewald formula may still underestimate LDL-C and; thereby, CVD risk when TG levels are less than 400 mg/dL. Several studies have shown that CVD risk is especially underestimated when TG levels are high [
7,
8] and when the Friedewald-estimated LDL-C is low [
9,
10]. Recently, Martin et al. [
11] analyzed the concordance in guideline risk classification by the Friedewald estimates and by LDL-C direct estimates in 1,340,614 adults. They found that LDL-C levels were underestimated in 19,677 of 33,106 persons (59.4%) with TG levels of 200 to 399 mg/dL and a Friedewald estimate of less than 70 mg/dL of LDL-C.
Notwithstanding the cost-effectiveness of the Friedewald formula for LDL-C estimation, underestimating LDL-C levels may lead to misclassified CVD risk assessments and to major societal costs in the future from the poor prevention and management of CVD. Nonetheless, the Friedewald formula is commonly used to estimate LDL-C in routine practice worldwide. To this end, we validated the accuracy of the Friedewald formula using nationwide data from the Korea National Health and Nutrition Examination Survey (KNHANES). We also analyzed the agreement between the Friedewald estimates of LDL-C (LDL-C
F) and directly measured LDL-C levels (LDL-C
D), and the concordance between these measurements in determining the LDL-C risk classification used in clinical practice. In this study, we adopted the LDL-C risk classification outlined in the revised 2015 Korean Guidelines for the Management of Dyslipidemia [
12]. The revised guidelines reflect those presented in the National Cholesterol Education Program Adult Treatment Panel III (NCEP-ATP III) revised in 2002 [
13], where LDL-C values were newly classified into six groups. To our knowledge, no studies have investigated the degree of concordance in the revised LDL-C risk classification between LDL-C
F and LDL-C
D in the Korean population.
DISCUSSION
We performed a comparative analysis between the Friedewald estimates and directly-measured LDL-C values in 6,454 subjects with TG levels lower than 400 mg/dL. Additionally, the concordance in the LDL-C risk classification between LDL-CF estimates and LDL-CD values was evaluated. We found that higher TG levels and lower HDL-C levels were each associated with a greater underestimation of LDL-C when the Friedewald formula was used. We also observed two groups of subjects who were more likely to be misclassified into lower LDL risk categories: (1) high-risk subjects with either high TG or low HDL-C and (2) subjects with a low Friedewald estimate and a high TG level.
Several studies have shown that even when TG levels are less than 400 mg/dL, higher TG levels increase the degree to which the Friedewald formula underestimates LDL-C [
7,
8,
11,
17,
18]. In their study that measured LDL-C levels in 38,243 Korean subjects, Jun et al. [
7] found that the Friedewald formula significantly underestimated LDL-C levels in individuals with high TG levels. Likewise, our findings show that the Friedewald formula underestimated LDL-C, particularly in subjects with higher TG levels, and overestimated LDL-C in subjects with lower TG levels (
Table 2). Taking into account the underestimation and overestimation errors, Hwang et al. [
17] suggested that the Friedewald formula could be used without introducing estimation errors in subjects with TG levels of 36 to 298 mg/dL. Since underestimating LDL-C risk may lead to false-negative diagnoses in subjects with undiagnosed conditions, it might cause serious consequences in terms of health outcomes.
The underestimation of the Friedewald formula was more pronounced in high-risk subjects, such as those with hypertriglyceridemia or hypo-HDL-cholesterolemia. We found that the proportion of LDL-C underestimation increased as TG levels in subjects increased. Compared with subjects with TG levels of 200 to 299 mg/dL, subjects with TG levels of 300 to 399 mg/dL had a greater proportion of underestimation (31% vs. 45.6%) (
Table 4). Thus, among the 838 patients (623 + 215) with hypertriglyceridemia, 291 (193 + 98, 34.7%) with TG levels greater than 200 mg/dL had an underestimated LDL-C level. These findings support those of previous studies, such as the results reported by Hur et al. [
18], who suggested that the cutoff for directly measuring LDL-C levels should be lowered to > 200 mg/dL from the current cutoff TG level of > 400 mg/dL. Of the total of 1,465 subjects with hypo-HDL-cholesterolemia, corresponding to a HDL-C level of less than 40 mg/dL, 284 subjects (19.4%) had an underestimated LDL-C value (
Table 4). Although HDL-C levels have been shown to negatively correlate with TG levels [
19], a low HDL-C level itself is also a risk factor for CVD.
A major revision in the 2015 Korean Guidelines for the Management of Dyslipidemia was the addition of a new risk category—very-high-risk subjects—including those with preexisting CVD. The guidelines go further to advise a LDL-C goal of less than 70 mg/dL in subjects categorized into this group [
12]. In our study, 62 of 112 subjects (55.4%) with TG levels 200 to 399 mg/dL and a LDL-C
F level lower than 70 mg/dL were misclassified into lower risk categories. This suggests that when LDL-C is measured using the Friedewald formula in very-high-risk subjects, such as those with hypertriglyceridemia, more than half will have an underestimated LDL-C risk.
Recently, the direct measurement of LDL-C has become widespread, but in large-scale health screenings, the Friedewald formula is still widely used worldwide. The KNHSP also conducts screenings using the Friedewald formula in individuals with TG levels less than 400 mg/dL. More than 98% of participants were examined using the Friedewald formula during the 2013 KNHSP, in which only 2% of participants had TG levels greater than 400 mg/dL. If we assume that an average of 70% of the total of 22,022,447 participants (20,815,338 for the general check-up and 1,207,109 for the transitional age check-up) of the 2016 KNHSP received a primary examination, and 98% of them had TG levels lower than 400 mg/dL, the total number of participants whose LDL-C would then be directly measured at the primary examination can be deduced to be 15,107,399 [
4]. As the cost of directly measuring LDL-C is around 6,790 Korean won per person, this would lead to an additional net expense of around 100 billion won. Since the total budget of the 2014 KNHSP, which includes screenings for cancer, was 1 trillion won, it would be unrealistic for the LDL-C of all participants of the KNHSP to be measured directly.
Importantly, we found that in very-high-risk subjects in whom LDL-C should be accurately measured, the Friedewald formula often led to an underestimation of LDL-C. Either an alternative to the Friedewald formula adapted to the Korean population should be developed to produce better estimates of LDL-C levels in this subset of individuals, or LDL-C should be measured directly in high-risk groups. Recently, Martin et al. [
20] developed a novel alternative method to improve the accuracy of LDL-C estimations. In their validation study of this method in Korean subjects, Lee et al. [
21] reported that the Martin method was associated with a significantly higher concordance in LDL-C risk classification. Of note, subjects with TG levels of 200 to 399 mg/dL were found to exhibit significantly improved concordance in their risk categorization when the Martin method was used instead of the Friedewald formula (77.2% vs. 62.0%,
p < 0.001). However, further studies investigating the accuracy of the Martin method in estimating LDL-C are needed using large-scale data, such as those from national databases. In general, the target for non-HDL-C can be calculated by adding 30 to the graded LDL-C treatment goals for each risk category. This comes from the assumption that the difference between non-HDL-C and LDL-C
F is about 30. In this study, the difference between non-HDL-C and LDL-C
F varied according to the TG levels of the subjects. The difference was greater than 30 for those with higher TG levels and less than 30 for those with lower TG levels.
One of the limitations of this study is that the risk factors for CVD were not analyzed in our sample population. Taking the LDL-C treatment goal by risk category into account, assessing LDL-C risk classification mismatches by category will help validate the Friedewald formula more stringently. Another limitation is that we included the subjects who took medications (statins, fenofibrate, omega-3 fatty acids, steroid, or others), that may potentially contribute to changes in lipid profile and discordances.
In addition to TG and HDL-C, studies have suggested that other lipid-related factors such as total cholesterol and non-HDL-C levels, comorbidities, age, and gender influence the discordance between Friedewald-estimated and directly measured values [
7,
21]. Finally, the long-term implications of underestimating or overestimating LDL-C using the Friedewald formula were not evaluated. A cost-benefit analysis investigating the cost incurred from directly measuring LDL-C and the societal cost or burden arising from erroneous Friedewald estimations and the relative benefits of direct measurements should be conducted.
CVD, such as myocardial infarction and stroke, are a major cause of death in Korea. In 2013, the leading cause of death was cancer (28.3%), followed by cerebrovascular disease (9.6%) and heart disease (9.5%) [
22]. Since LDL-C is a major modifiable risk factor for CVD, it is vital for its levels to be accurately measured in clinical practice. Yet, in the KNHSP, the LDL-C levels of subjects with TG levels lower than 400 mg/dL are not directly measured, but are indirectly measured using the Friedewald formula. In this study, we found that subjects at high risk for CVD, in whom an accurate estimation of LDL-C is especially important, exhibited a high frequency of underestimated LDL-C risk. Around 34.7% of subjects with hypertriglyceridemia (TG levels above 200 mg/dL) had an underestimated LDL-C level; the proportion of subjects with an underestimated LDL-C level was as high as 55.4% when the Friedewald estimate was lower than 70 mg/dL. To summarize, we conclude that LDL-C should be directly measured in these high-risk subjects. Further studies are also required to develop a more accurate alternative to the Friedewald formula.