Clinical Chemistry 60:12
1519–1523 (2014)
Validation of a Proposed Novel Equation for Estimating
LDL Cholesterol
Jeffrey W. Meeusen,1 Alan J. Lueke,1 Allan S. Jaffe,1,2* and Amy K. Saenger1
BACKGROUND: Aggressive LDL cholesterol (LDL-C)-
lowering strategies are recommended for primary and
secondary prevention of cardiovascular events. A newly
derived equation for LDL-C estimation was recently pub-
lished that addressed limitations in the commonly used
Friedewald LDL-C calculation method. The novel
method was reported to classify patients with superior con-
cordance to measured LDL-C compared to the Friedewald
method, particularly in patients with LDL-C ⬍70 mg/dL.
METHODS: We evaluated the performance of the novel
method within an independent cohort of 23 055 pa-
tients with LDL-C measured by the gold standard
␤-quantification reference method.
RESULTS: Overall Friedewald underestimated and the
novel method overestimated measured LDL-C. Both
estimations significantly deviated from the reference
method when LDL-C was ⬍70 mg/dL. Overall, the
Friedewald and novel calculations correctly classified
77% and 78% of patients, respectively. The largest dis-
crepancy in classification was observed in individuals
with measured LDL-C ⬍70 mg/dL. For this group the
novel calculation would reclassify 8.7% of patients as
⬎70 mg/dL compared to the Friedewald equation.
CONCLUSIONS: We compared both novel and Friede-
wald estimated LDL-C against the LDL-C reference
method; in contrast, the prior study relied on valida-
tion of a subset of samples by ␤-quantification to allow
the use of the vertical autoprofile method for LDL-C
measurement. We conclude that the novel method has
some benefits but it is unclear whether improvements
over the Friedewald calculation are substantive enough
to justify making the change in routine clinical practice
and to improve patient outcomes.
© 2014 American Association for Clinical Chemistry
1
Division of Clinical Core Laboratory Services, Department of Laboratory Medi-
cine and Pathology, Mayo Clinic, Rochester, MN; 2 Division of Cardiovascular
Diseases, Department of Medicine, Mayo Clinic, Rochester MN.
* Address correspondence to this author at: Mayo Clinic, 200 First St. SW,
Rochester, MN 55905. Fax 507-538-7060; e-mail jaffe.allan@mayo.edu.
Received May 27, 2014; accepted September 8, 2014.
Previously published online at DOI: 10.1373/clinchem.2014.227710
Lipids, Lipoproteins, and Cardiovascular Risk Factors
LDL cholesterol (LDL-C)3 is an important modifiable
risk factor for atherosclerotic cardiovascular disease
(1, 2 ). However, it is primarily a calculated measure via
the Friedewald equation and not measured directly.
The Friedewald equation uses total cholesterol (TC),
HDL cholesterol (HDL-C), and triglycerides (TG) to cal-
culate LDL-C [LDL-C ⫽ TC ⫺ HDL-C ⫺ (TG/5)]. Data
advocating other LDL-C quantitative methods have sug-
gested that this calculation [termed in this manuscript as
Friedewald calculated LDL-C (LDL-CF)] misclassifies
cardiovascular risk in many patients, particularly those
with hypertriglyceridemia (3– 6 ). Misclassification has
been attributed to the TG/5 ratio used in the Friede-
wald equation, which represents an estimation of
VLDL-C based on the average ratio of TG:cholesterol
present in VLDL lipoproteins. Inaccuracies in LDL-CF
at TG ⬎400 mg/dL were acknowledged by Friedewald
and colleagues in their original validation (7 ) and later
proven by DeLong et al. (8 ).
A recent study reported by Martin et al. has advo-
cated use of a newly derived equation to estimate
LDL-C that is intended to correct for this limitation in
the Friedewald calculation and improve LDL-C esti-
mation even when TG values are ⬍400 mg/dL (9 ). The
novel LDL-C (referred to here as LDL-CN) is calculated
using an adjustable factor empirically determined on
the basis of an individual patient’s TG and non-
HDL-C; this equation was derived using a large cohort
of patients with LDL-C measured by the vertical auto-
profile (VAP) method. Use of LDL-CN improved con-
cordance over the Friedewald equation compared to
measured LDL-C, particularly in patients with low
LDL-C values [LDL-C ⬍70 mg/dL (⬍1.8 mmol/L)].
However, changing longstanding approaches in a con-
sistent manner across the global healthcare system
might not be an easy task. Thus, advocacy for adoption
of new techniques requires independent verification in
multiple data sets, particularly when a variety of labo-
ratory methods are used, and the magnitude of the
© 2014 American Association for Clinical Chemistry
3
Nonstandard abbreviations: LDL-C, LDL cholesterol; TC, total cholesterol; TG,
triglycerides; LDL-CF, Friedewald calculated LDL-C; VLDL-C, very LDL-C; LDL-CN,
novel calculated LDL-C; VAP, vertical autoprofile; LDL-C␤Q, LDL-C measured by
␤-quantification; IQR, interquartile range; ACC/AHA, American College of Car-
diology/American Heart Association.
1519
benefit must be compelling. Accordingly, we evaluated
the performance of the LDL-CN estimation method Table 1. Demographic characteristics of the study
compared to the Friedewald equation method in an cohort.
independent patient population in which LDL-C was
Age, median (IQR), years 53 (40–64)
measured by the gold standard ␤-quantification refer-
a
ence method (LDL-C␤Q). Total, no. (%) 23 055
⬍11 years 962 (4)
Materials and Methods 11–18 years 1144 (5)
⬎18 years 20 625 (90)
All data were accessed in compliance with the Mayo Sex, no. (%)b
Clinic Institutional Review Board. Retrospective anal- Male 11 638 (51)
ysis of lipoprotein metabolism profiles between 2003 Female 11 003 (48)
and 2013 identified 23 055 individuals with LDL-C␤Q Cholesterol, mg/dL,c median (IQR)
and TG ⬍400 mg/dL. All samples were collected after a
Total 185 (155–220)
minimum of 8 h fasting. Serum TC and TG were mea-
HDL-C 44 (37–54)
sured on a Roche Cobas c501 (Roche Diagnostics).
Non-HDL 138 (109–171)
HDL-C, LDL-C␤Q, and VLDL-C concentrations were
determined as described previously (10 ). Briefly, 1 mL LDL-C␤Q 110 (87–139)
of serum was ultracentrifuged for 15 h at 75 000g. The VLDL-C 23 (15–35)
upper layer (containing chylomicrons and VLDL) and Triglycerides, mg/dL, median (IQR) 131 (89–196)
lower layer (containing HDL and LDL) were separated TG:VLDL-C 5.7 (4.5–7.6)
and transferred to unique aliquot tubes. Each layer was 5th–95th percentile, range 3.4–12.7
analyzed for cholesterol and TG. HDL-C was deter- 1st–99th percentile, range 2.7–22.5
mined following selective precipitation of LDL from Estimated LDL-C, mg/dL
the lower layer using a mixture of magnesium chloride Friedewald LDL, median (IQR) 108 (83–138)
and dextran sulfate. LDL-C␤Q is calculated as [(lower
Difference (LDL-CF ⫺ LDL-C␤Q) ⫺3 (⫺8 to 3)
layer cholesterol) ⫺ (HDL-C)] and VLDL-C is defined
Novel LDL, median (IQR) 112 (88–141)
as [(upper layer cholesterol) ⫺ (chylomicron choles-
Difference (LDL-CN ⫺ LDL-C␤Q) 0 (⫺5 to 6)
terol)]. The analytical performance of these lipid mea-
surements in our laboratory is directly certified by the a
Data on patient age was unavailable for 324 patients.
b
CDC Lipid Standardization Program. Interassay preci- c
Data on patient sex was unavailable for 414 patients.
To convert cholesterol (mg/dL to mmol/L) multiply values by 0.0259 and for
sion was determined in serum by repeat measurement triglycerides multiply by 0.0113.
daily for 20 days. The CV for TC was 1.8% at 93 mg/dL,
for HDL-C was 2.4% at 43 mg/dL, and for TG was 2.7%
at 85 mg/dL. LDL-C␤Q precision was determined by Results
repeat analysis (n ⫽ 37) of a pooled serum sample with
mean LDL-C of 109 mg/dL (SD, 4.2 mg/dL; %CV, Our cohort was predominantly middle-aged individu-
3.9%). Non-HDL-C was calculated as (TC ⫺ HDL-C). als [median 53 years; interquartile range (IQR) 40 – 64]
with 9% of patients ⬍18 years of age (Table 1). Overall,
The LDL-CF was: LDL-CF ⫽ (TC ⫺ HDL-C) ⫺ (TG/
TG was 20% higher (median, 143 mg/dL; IQR, 94 –234
5), and the novel method was calculated as: LDL-CN ⫽
mg/dL) and HDL-C was 20% lower (median, 43 mg/
(TC ⫺ HDL-C) ⫺ (TG / X), where X is an adjustable
dL; IQR, 35–53 mg/dL) in our population than in the
factor based on the 180-cell method described by Mar-
previously published derivation cohort (9 ). The
tin et al. (9 ). Concordance in classification by the Na- LDL-CN values were based on empirical TG:VLDL-C
tional Cholesterol Education Program Adult Treat- ratios ranging from 3.1 to 11.9 assigned according to
ment Panel III guideline cutoffs (11 ) was defined as the TG and non-HDL-C strata in a derivation cohort (9 ).
number of estimated LDL-C values appropriately clas- Analyzing our cohort in a similar manner, we identi-
sified relative to the number of measured LDL-C values fied a range of TG:VLDL-C ratios from 3.0 to 12.2 with
classified in each group using the reference method as a median TG:VLDL-C of 5.8 (see Table 1 in the Data
the gold standard. (To convert TC, HDL-C, LDL-C, Supplement that accompanies the online version of
and VLDL-C to mmol/L, multiply by 0.0259; for TG this report at http://www.clinchem.org/content/vol60/
multiply by 0.0113.) Statistical analyses were per- issue12). The median TC, non-HDL-C, and LDL-C␤Q
formed using JMP software (SAS Inc.). Statistical sig- concentrations were within 10% of the previously pub-
nificance was defined as a P value ⬍0.0001. lished derivation cohort (Table 1).

1520 Clinical Chemistry 60:12 (2014)

Comparison of Friedewald and a Novel LDL Estimation

The median difference for LDL-CF was ⫺4 mg/dL (IQR,

Fig. 1. Comparison of measured and estimated LDL-C
methods.
Median LDL-C is shown for each method according to
guideline categories as determined by LDL-C␤Q. I bars are
IQR.
Overall, LDL-CF underestimated LDL-C␤Q,
whereas LDL-CN tended to overestimate LDL-C␤Q.
The median LDL-C␤Q was 110 mg/dL (IQR 87–139),
the median LDL-CF was significantly lower at 108
mg/dL (IQR 83–138; P ⬍ 0.0001), and the median
LDL-CN was significantly higher at 112 mg/dL (IQR
88 –141; P ⬍ 0.0001; Fig. 1). The median difference
between LDL-CF and LDL-C␤Q for all samples was ⫺3
mg/dL (IQR, ⫺8 to 3) and the median difference for
LDL-CN was 0 mg/dL (IQR CI, ⫺5 to 6).
Both estimation methods significantly deviated from
the reference method when LDL-C␤Q was ⬍70 mg/dL.
⫺9 to 0) in this category compared to ⫺1 mg/dL (IQR,
⫺5 to 4) for LDL-CN. Consequently, the LDL-CN
method calculated fewer negative LDL-C concentra-
tions compared to the LDL-CF method (1 vs 9 patients,
respectively).
Concordance between estimated LDL-C and LDL-
C␤Q according to guideline-recommended cutoffs (11 ) of
70, 100, 130, 160, and 190 mg/dL was similar between the
2 estimations. LDL-CF correctly classified 17726 (77%)
patients, compared to LDL-CN, which correctly classified
17914 (78%) patients (Table 2). Most misclassified pa-
tients were within 1 NCEP category of LDL-C␤Q; only 88
LDL-CF and 115 LDL-CN patients were misclassified by 2
or more NCEP categories.
In considering the implications to patient care, it is
important to note that the current standard of practice
is not the gold standard LDL-C␤Q, but rather LDL-CF.
To account for this we performed a series of analyses in
which the estimated value was considered the reference
(Table 3). LDL-CF was marginally more concordant
compared to LDL-CN when LDL-C was ⬎100 mg/dL.
Concordance was significantly higher for LDL-CN
compared to LDL-CF when LDL-C was ⬍100 mg/dL
(P ⬍ 0.001). The greatest difference in concordance
between the 2 estimations was observed in patients
with LDL-C␤Q ⬍70 mg/dL. In this category, LDL-CF
correctly classified 2306 patients and LDL-CN correctly
classified 2089 patients. However, LDL-CF classified a
total of 3239 patients with values ⬍70 mg/dL whereas
LDL-CN classified only 2474 patients in this category,
providing relative concordances of 71.2% and 84.4% for
LDL-CF and LDL-CN, respectively, if the initial classifica-
tion were defined by the estimated value (Table 3).
Our final analysis assessed the changes that clini-
cians would expect to see in patient LDL-C values if the
novel method were to be implemented in routine prac-

Table 2. Performance of Friedewald and novel estimates of LDL-C compared to the reference method LDL-C␤Q.

LDL-C␤Q LDL-CF LDL-CN

a
LDL-C, mg/dL No. True positives False positives True positives False positives

ⱖ190 1138 988 167 1013 215

160–189 1941 1373 530 1441 691
130–159 4244 3063 997 3176 1312
100–129 6945 5110 1262 5246 1453
70–99 6292 4886 1440 4949 1085
⬍70 2495 2306 933 2089 385
Grand Total 23 055 17 726 5329 17 914 5141
a
To convert cholesterol (mg/dL to mmol/L) multiply values by 0.0259 and for triglycerides multiply by 0.0113.

Clinical Chemistry 60:12 (2014) 1521

 Table 3. Concordance in guideline classification by Friedewald vs novel estimates of LDL-C.
LDL-CF
LDL-C, mg/dL Concordant/totala
ⱖ190 988/1155
160–189 1373/1903
130–159 3063/4060
100–129 5110/6372
70–99 4886/6326
⬍70 2306/3239
Overall 17 726/23 055
a
Concordance was determined according to measured LDL-C␤Q.
tice rather than the Friedewald. Overall, 19 541 patients
(84.8%) would have been grouped in the same NCEP
category, 330 patients (1.4%) would have been catego-
rized in a lower NCEP category, and 3184 patients
(13.8%) would have been categorized in a higher
NCEP category (see online Supplemental Table 2). The
largest changes would be seen in patients with LDL-CF
⬍100 mg/dL, for which ⬎20% of patients were
grouped in a higher NCEP category by LDL-CN com-
pared to LDL-CF.
Discussion
In the present study, we compared both the Friedewald
and a novel estimated LDL-C against LDL-C␤Q. In
contrast to the prior study, which relied on validation
of a subset of samples by ␤-quantification to allow the
use of the vertical autoprofile method for LDL-C mea-
surement, we had ␤-quantification values on the entire
cohort. In our patient cohort, the novel method signif-
icantly overestimated LDL-C. Conversely, the Friede-
wald method tended to underestimate LDL-C.
Several differences between the previously pub-
lished cohort and the current study deserve acknowl-
edgement. The median TG:VLDL-C ratio derived from
our cohort is closer to the fixed factor of 6 proposed by
Delong and colleagues (8 ) than the 5.2 ratio reported
in the derivation cohort (9 ). The higher median TG
and lower median HDL-C suggest our cohort may have
been enriched with dyslipidemic individuals compared
to the published novel derivation cohort (9 ).
In our cohort, the overall concordance between
estimated and measured LDL-C according to NCEP
guideline classification was similar for both methods,
providing a 75%– 80% certainty of correct classifica-
tion regardless of the method used (Table 3). However,
as in the previous report, our study found significant
differences in the equation performance at LDL-C ⬍70
1522 Clinical Chemistry 60:12 (2014)
LDL-CN
% (95% CI) Concordant/total % (95% CI)
85.5 (83.4–87.5) 1013/1228 82.5 (80.3–84.5)
72.1 (70.1–74.1) 1441/2132 67.6 (65.5–69.6)
75.4 (73.5–77.3) 3176/4488 70.8 (68.7–72.7)
80.2 (78.3–84.7) 5246/6699 78.3 (76.4–80.0)
77.2 (74.8–78.5) 4949/6034 82.0 (80.7–84.1)
71.2 (69.1–73.1) 2089/2474 84.4 (82.5–85.7)
76.9 (75.2–79.4) 17 914/23 055 77.7 (76.0–79.6)
mg/dL. In this category, LDL-CN had a higher rate of
concordance with LDL-C␤Q and misclassified fewer
patients compared to the Friedewald calculation. Al-
though LDL-CF gave a higher absolute number of pa-
tients whose results were concordant with those of
LDL-C␤Q in the LDL-C ⬍70 mg/dL category, the over-
all percentage of concordant values was lower, and the
LDL-CF gave an increased number of false positives.
Consequently, certainty of correct classification as ⬍70
mg/dL is only 71% for the Friedewald method com-
pared to 84% for the novel method.
The novel method incrementally but significantly
improves LDL-C estimation when compared to the
Friedewald equation. In clinical decision-making, it is
important to note that LDL-C is a single component
among many factors that determine a patient’s risk of
cardiovascular disease. The issue of how to use LDL-C
in clinical practice has been controversial. It is not our
goal to extend the controversy in regard to the 2013
American College of Cardiology/American Heart As-
sociation (ACC/AHA) guidelines (12 ), but simply to
evaluate this novel method.
Recent ACC/AHA guidelines deemphasize titrat-
ing to specific LDL-C goals; however, specific LDL-C
values are recommended as criteria for considering
statins in 2 situations. First, patients with an LDL-C
⬎190 mg/dL are considered a statin benefit group. Es-
timating LDL-C by the novel method had negligible
effect on patients in this category compared to the Frie-
dewald method. Second, patients with a risk estimate
⬎7.5% and an LDL-C between 70 and 189 mg/dL are
considered a statin benefit group. According to our co-
hort, adoption of the novel method would place 14% of
patients in a higher LDL-C NECP classification (see
online Supplemental Table 2). Most of the recategori-
zation would occur for patients with an LDL-CF of ⬍70
mg/dL, leading to 25% fewer patients grouped in this
category.
Comparison of Friedewald and a Novel LDL Estimation

The benefit of using the novel calculation is mea-
surable yet incremental. Changing a calculation is a
simple matter in the modern laboratory. However, the
Friedewald method is well established and is likely the
most universally adopted and standardized calculation
in laboratory medicine. It remains to be discerned if the
benefits of the novel method justify its widespread
adoption.
Author Contributions: All authors confirmed they have contributed to
the intellectual content of this paper and have met the following 3 re-
quirements: (a) significant contributions to the conception and design,
acquisition of data, or analysis and interpretation of data; (b) drafting
or revising the article for intellectual content; and (c) final approval of
the published article.
Authors’ Disclosures or Potential Conflicts of Interest: Upon man-
uscript submission, all authors completed the author disclosure form.
Disclosures and/or potential conflicts of interest:
Employment or Leadership: A.K. Saenger, AACC and Roche
Diagnostics.
Consultant or Advisory Role: A.S. Jaffe, Beckman-Coulter, Alere,
Abbott, Roche, Radiometer, Critical Diagnostics, Trinity, ET Health-
care, Amgen, and theheart.org.
Stock Ownership: None declared.
Honoraria: None declared.
Research Funding: None declared.
Expert Testimony: None declared.
Patents: None declared.
Role of Sponsor: No sponsor was declared.
Acknowledgments: We thank Jean Hornseth, Kristen Roberts, and
Becky Leindecker for their excellent technical assistance.

1. Gaziano JM, Gaziano TA. What’s new with mea-
suring cholesterol? JAMA 2013;310:2043– 4.
2. Steinberg D. Thematic review series: the patho-
genesis of atherosclerosis: an interpretive history
of the cholesterol controversy, part III: mechanis-
tically defining the role of hyperlipidemia. J Lipid
Res 2005;46:2037–51.
3. Bachorik PS, Ross JW. National Cholesterol Edu-
cation Program recommendations for measure-
ment of low-density lipoprotein cholesterol: ex-
ecutive summary. The National Cholesterol
Education Program Working Group on Lipopro-
tein Measurement. Clin Chem 1995;41:1414 –20.
4. Nauck M, Warnick GR, Rifai N. Methods for mea-
surement of LDL-cholesterol: a critical assessment
of direct measurement by homogeneous assays
versus calculation. Clin Chem 2002;48:236 –54.
5. Martin SS, Blaha MJ, Elshazly MB, Brinton EA,
Toth PP, McEvoy JW, et al. Friedewald-estimated
versus directly measured low-density lipoprotein
cholesterol and treatment implications. J Am Coll
References
Cardiol 2013;62:732–9.
6. Oliveira MJ, van Deventer HE, Bachmann LM,
Warnick GR, Nakajima K, Nakamura M, et al.
Evaluation of four different equations for calcu-
lating LDL-C with eight different direct HDL-C
assays. Clin Chim Acta 2013;423:135– 40.
7. Friedewald WT, Levy RI, Fredrickson DS. Estima-
tion of the concentration of low-density lipopro-
tein cholesterol in plasma, without use of the
preparative ultracentrifuge. Clin Chem 1972;18:
499 –502.
8. DeLong DM, DeLong ER, Wood PD, Lippel K,
Rifkind BM. A comparison of methods for the
estimation of plasma low- and very low-density
lipoprotein cholesterol. The Lipid Research Clinics
Prevalence Study. JAMA 1986;256:2372–7.
9. Martin SS, Blaha MJ, Elshazly MB, Toth PP, Kwite-
rovich PO, Blumenthal RS, Jones SR. Comparison of
a novel method vs the Friedewald equation for
estimating low-density lipoprotein cholesterol
levels from the standard lipid profile. JAMA 2013;
310:2061– 8.
10. Ellefson RD, Jimenez BJ, Smith RC. Pre-beta (or
alpha 2) lipoprotein of high density in human
blood. Mayo Clin Proc 1971;46:328 –32.
11. National Cholesterol Education Program (NCEP)
Expert Panel on Detection, Evaluation, and Treat-
ment of High Blood Cholesterol in Adults (Adult
Treatment Panel III). Third Report of the National
Cholesterol Education Program (NCEP) Expert
Panel on Detection, Evaluation, and Treatment of
High Blood Cholesterol in Adults (Adult Treat-
ment Panel III) final report. Circulation 2002;106:
3143– 421.
12. Stone NJ, Robinson J, Lichtenstein AH, Bairey
Merz CN, Lloyd-Jones DM, Blum CB, et al. 2013
ACC/AHA guideline on the treatment of blood
cholesterol to reduce atherosclerotic cardiovascu-
lar risk in adults: a report of the American College
of Cardiology/American Heart Association Task
Force on Practice Guidelines. J Am Coll Cardiol
2014;63:2889 –934.

Clinical Chemistry 60:12 (2014) 1523