Degree of Concordance Between Calculated
and Measured Low-Density Lipoprotein
Sikandar Hayat Khan1, Farah Sobia2, Rahat Shahid2, Syed Ohen Raza Shah Bukhari4, Syed Mohsin Manzoor4, Roomana Anwar4 and Muhammad Tariq4
1 Department of Pathology, Pakistan
2Department of Surgery, Pakistan
3Department of Medicine, Pakistan
4Medical Officer, Pakistan
Submission: July 10, 2020; Published: July 15, 2020
*Corresponding author: Sikandar Hayat Khan, Department of Pathology, PNS HAFEEZ Hospital, Pakistan
How to cite this article: Sikandar H K, Farah S, Rahat S, Syed O R S B, et al. Degree of Concordance Between Calculated and Measured Low-
Density Lipoprotein Cholesterol Methods. Curre Res Diabetes & Obes J 2020; 13(4): 555866.
Background: Low-density lipoprotein cholesterol (LDLc) remains a key lipid parameter and focus of clinical prevention targets for cardiovascular diseases. Although measured LDLc values are now available, many clinicians still rely on calculated method. Accordingly, several methods have evolved to overcome the shortcomings of previous models and improve risk prediction.
Objective: This study aims to compare the outcomes of various calculated and direct methods and to correlate biochemical risks with calculated and measured LDLc.
Place & duration of study: Jan-17 to April-8 at Naval hospital.
Subjects and methods: The final analysis included 230 subjects with available mLDLc measurements. A paired t-test was used to compare measured and calculated methods developed by Friedewald et al , Anandaraja et al , Martin et al , Hattori et al , Chen et al , Vujovic et al , de Cordova et al , Ahmadi et al , Puavilai et al . Subsequently, calculated parameters exhibiting non-significant associations were further evaluated using a Bland–Altman Plots.
Results: All calculated LDLc parameters except model of Ahmadi et al. exhibited moderate positive linear correlation with mLDLc. For the non-significant correlations of measured LDLc with Anandaraja-LDLc, Hatori-LDLc, and Cordova-LDLc, the level of agreement was demonstrated using Bland–Altman plots. Ahmadi et al method exhibited the strongest correlations with atherogenic small dense-LDLc, insulin resistance, and nephropathy.
Conclusion: Most calculated LDLc methods exhibited moderate correlations with measured methods; however, only results of Anandaraja & Hatori et al. [2,4] and de Cordova et al,  closely resembled measured methods.
In recent years, direct homogenous methods for measuring low-density lipoprotein cholesterol (LDLc) in clinical settings have been introduced. During the last decade, these techniques have improved in terms of both accuracy and precision , leading to their applications in large-scale clinical trials and reports of their
clinical advantages . At their core, these homogenous and direct LDLc methods aim to generate more reliable and clinically suitable data. However, the use of these methods is hindered by high costs, a reported lack of concordance between methods, and reproducibility issues [12,13].
It remains uncertain whether indirect calculation-based
measures of LDLc would provide similarly reliable and acceptable
clinical targets. However, the use of these measures into laboratory
systems would likely be both cost- and labor-effective and highly
precise. Following the introduction of an LDLc calculation-based
methodology by Friedewald & colleagues  (Friedewald-LDLc),
multiple authors attempted to address the discrepancies identified
in this approach. The original Friedewald equation incorporated
the total cholesterol (TC), high-density lipoprotein cholesterol
(HDLc) and triglyceride (TG) levels and divided the final variable
by 2.2 to calculate the LDLc value (mmol/L) . However, a review
of the literature suggests that this equation exhibits positive bias
at low TG values and would thus inflate the LDLc value even if the
levels of both TG and TC were within normal ranges .
The National Cholesterol Education Program (NCEP) has
defined strict targets for clinical laboratory measures, namely
a total error of <12% and precision and accuracy levels of
>96% . Unfortunately, the data suggest that the currently
available homogenous direct LDLc methods do not yet meet
these performance characteristics [13,14]. Accordingly,
biotechnological attempts to improve these homogenous methods
have been accompanied by attempts to address the defects and
shortcomings of calculated measures. For example, Martin et
al.  demonstrated the superiority of an adjustable TG: very
LDLc (VLDLc) ratio over the equation developed by Friedewald
& colleagues  Furthermore, de Cordova et al. recommended a
simpler LDLc formula, namely LDLc = 3/4 (TC – HDL) , while
Chen et al.  calculated LDLc (mg/dl) as Non-HDL-C x 90% - TG x
10% in an attempt to address the positive TG bias in patients with
normal TC levels . Similarly, Anadaraja & Vujovic et al. [2,6]
calculating LDLc (mg/dL) as 0.9*TC – 0.9*TG/5 – 28 or cLDLc = TC
– HDL-C – (TG/6.58), respectively [2,6]. Ahmadi et al.  perceived
that the Friedewald formula would overestimate the LDLc value in
patients with lower serum TG levels and developed the formula of
LDLc (mmol/L) = TC/1.19 +Triglycerides/0.81 – HDLc/1.1 – 0.98
. Still other researchers have similarly attempted to address the
limitations of existing LDLc calculation methods [4,9].
The above review suggests considerable variance among the
results of LDLc calculation-based methods used around the world.
These discrepancies have sowed confusion, as LDLc is defined as
the lipoprotein fraction isolated by ultracentrifugation with a
density of 1.019–1.063 g/ml . Moreover, evidence suggests
that LDLc can be further fractionated into more (e.g., small
dense LDLc [sdLDLc]) and less atherogenic fractions (e.g., large
buoyant LDLc [lbLDLc]). This characteristic has evoked questions
regarding which of the available equations could best identify with
atherogenic lipidemia and conventional biochemical risks . In
this study, we aimed to compare and correlate the LDLc fractions
calculated using various methods with the measured homogenous
LDLc results while considering the aforementioned evidence and
assuming the homogenous methods depict the actual CVD risk.
This study also aimed to correlate the homogenous and calculated
values with traditional biochemical risk factors to determine the
optimal method for risk prediction.
This cross-sectional study was conducted in the Departments
of Pathology and Medicine of PNS HAFEEZ Hospital from January
2017 to April 2018. Formal approval for this study was sought from
the ethical review board of this hospital. Patients who reported
for the evaluation of plasma glucose or lipid profiles while in a
medical fasting state in various Out Patient Departments (OPDs)
were invited to participate. All potential subjects were briefed
about the study, the use of their data, and the probable side
effects of phlebotomy, and those who agreed to participate were
asked to provide a signed consent form and their clinical details
and history as per the study questionnaire. Subsequently, the
participants underwent a baseline clinical examination, including
the collection of blood pressure and anthropometric data. At this
stage, the primary exclusion criteria were a chronic disease (e.g.,
diabetes, hypertension, or ischemic heart disease), pregnancy,
inappropriate medical fasting, or the use of medication for chronic
or acute medical disorders. The initial dataset included 232
subjects; however, two samples could not be analyzed during the
study due to technical mistakes.
From each subject, a 10-ml sample of blood was collected
and divided into tubes containing sodium fluoride, EDTA, or
no additives. The total cholesterol, triglycerides, and glucose
levels were analyzed using the CHOD-PAP, GPO-PAP, and GODPAP
methods, respectively. The LDLc and HDLc levels were
measured via cholesterol esterase methodology using an ADVIA
1800 clinical chemistry system. Glycated hemoglobin (HbA1c)
was measured using ion-exchange resin chromatography by
3RD generation ELISA on Elisys Uno instrument (HUMAN). Spot
measures of the urine-albumin-creatinine ratio (UACR) were
performed in 174 subjects via immuno-turbidimetry using the
ADVIA 1800 system. Serum insulin levels were analyzed via a
chemiluminescence method using the Immulite® 1000 device.
The Homeostasis Model Assessment for insulin resistance
(HOMA-IR) equation was calculated according to Mathew et al.
 The sdLDLc level was calculated as described by Srisawasdi
et al. . The various equations used to calculate LDLc are listed
in Table 1. The hospital lab daily performs internal quality control
AS PER Westgard’s rules for accepting or rejecting batches along
with participates regularly in National External Quality Assurance
Program of Pakistan (NEQAPP). Every attempt is made to keep
the analytical imprecision and accuracy targets within the NCEP
defined parameters. Errors, if any are addressed till rectification
as per Westgard’s methodology.
Data were initially entered into Microsoft Excel spreadsheets
(version 2007). All abovementioned calculated LDLc values and
conversions from units of mg/dl were performed using Excel and
subsequently transferred to SPSS (version 15) for the statistical
analysis. Descriptive statistics were calculated for age and sex
using SPSS analyze function. A Pearson’s correlation analysis was
used to calculate the correlations of measured LDLc values with
all calculated LDLc values and with biochemical parameters such
as the HOMA-IR (insulin resistance), HOMA%B (insulin sensitivity
surrogate), UACR, fasting plasma glucose, HbA1c, and sdLDLc.
Differences between the measured and various calculated LDLc
levels were calculated using paired sample t-test. Subsequently,
non-significant associations were further evaluated using Bland–
Altman plots with 95% confidence intervals (CIs). Finally, a
Pearson’s correlation was used to determine the correlation
coefficients for the relationships of all LDLc measures with the
aforementioned biochemical risk indicators. A p-value of <0.05
was considered as statistically significant.
Our dataset included 122 female and 108 male subjects with
a mean (+ SD) age of 46.50 (±11.91) years. The mean body mass
index and waist-to-hip ratio were 27.17 (±5.22) and 0.93 (±0.082),
respectively. Table 2 presents the results of a Pearson correlation
analysis between various calculated and measured LDLc values.
Notably, a paired sample t-test yielded non-significant differences
of the measured LDLc with the calculations obtained using the
Anandaraja-LDLc, Hatori-LDLc, and De Cordova-LDLc equations
(Table 3). Accordingly, we developed Bland–Altman plots for
those LDLc parameters which matched measured LDLc as shown
in Figures 1-3.
Table 4 lists the correlations of the calculated and measured
LDLc values with biochemical risk factors such as nephropathy,
insulin resistance and sensitivity, and various diabetic parameters.
A Pearson’s correlation analysis of risk yielded the highest
positive linear correlation coefficient (r) of 0.778 (p <0.001) for
the association between mLDLc and risk, followed by the formulas
suggested by Ahmadi et al,  0.662 (p <0.001) and de Cordova et
al,  (r = 0.470, p <0.001). The other calculated values exhibited
modest correlations with mLDLc, whereas the value calculated
using the Friedewald equation yielded the weakest correlation of
0.198 (p = 0.003). Our results further identified the LDLc value
calculated according to Ahmadi et al,  as the most strongly
correlated with the metabolic parameters of nephropathy (r =
0.139, p = 0.067) and insulin resistance (r = 0.259, p <0.001).
None of the other calculated values or the measured LDLc value
correlated significantly metabolic parameters.
Our study identified moderate correlations between directly
measured LDLc and most available equations. Interestingly, the
calculated LDLc according to Ahmedi et al,  correlated very
weakly with the measured LDLc; however, these calculated values
exhibited the strongest correlations with atherogenic lipid levels
(e.g., sdLDLc), insulin resistance, and nephropathy, whereas no
other calculated value correlated strongly with biochemical risk
factors. Furthermore, the Anandaraja-LDLc, Hatori-LDLc, and
Cordova-LDLc equations were not associated with measured
The “LDLc concept”, which was used to define the current
primary and second treatment targets, represents a progression
from various discoveries in the rapidly evolving field of “lipidology.” However, the existing literature suggests that although LDLc has
some utility, it is not a good biomarker of CVD. Ide et al. observed
that in patients with CVD, the LDL concentrations remained
stable even after the initiation of a polyunsaturated diet. However,
these authors observed an increase in large LDL particles and
concomitant decrease in sdLDL particles consequent to this dietary
intervention . The potential lack of association between an
elevated LDLc level and the atherogenicity is consistent with the
possibility that a normal LDLc level might mask an elevated level
of sdLDLc, a CVD risk factor . Therefore, we might attribute
the weak correlations between LDL measures and biochemical
risk biomarkers to the stronger association of the CVD risk
with the level of sdLDL particles, rather than the overall LDLc
It remains uncertain why only the LDLc value calculated
according to Ahmadi et al,  exhibited strong correlations with
atherogenic dyslipidemia, insulin resistance, and a marker of
surrogate nephropathy, even in the presence of normolipidemia.
This finding might be attributable to the well-known Asian
“Atherogenic lipid triad .” Given its strong correlation
with sdLDLc, this equation appears to be more suitable for
the individuals suspected to have atherogenic dyslipidemia,
particularly those in the Asian population. Consistent with our
findings Goel et al,  demonstrated that Indian patients with
coronary artery disease had higher levels of sdLDLc despite
normal levels of LDLc, suggesting that the former is a more
effective parameter when assessing CVD. We further observed
that the LDLc value calculated according to Anandaraja et al,
 exhibited a moderate and non-significant correlation with
the measured value. However, the exclusion of HDLc from this
equation is likely responsible for the lack of correlations between
the calculated value and biochemical CVD risks .
Our observation of the variability and minimal degree of
agreement between most calculated LDLc values with mLDLc
and the inability of these values to indicate insulin resistance,
atherogenic dyslipidemia, and nephropathy suggest the need for
a different diagnostic approach towards lipid disorders. Kinetic
studies have demonstrated that the regulation of lipids and
lipoproteins in the body is a dynamic process wherein cholesterol,
fatty acids, and apolipoproteins are exchanged and thus alter
the shapes, sizes, densities, and characteristics of lipoproteins
[19,25,26]. Given that these changes occur rapidly within the
plasma, our broader biotechnological understanding of the
interactions of lipids and lipoproteins leads us to suggest that the
analyses of LDLc and HDLc sub-fractions may provide a better
indicator of the CVD risk. Consistent with our suggestion, Collins
et al. identified multiple differences in proteins within LDLc using
liquid chromatography-tandem mass spectrometry . For
example, the vertical auto profile (VAP) clearly demonstrates a
shift from the conventional lipid profile to an evaluation of smaller
sub-fractions within the lipid profile, thus enabling a better
evaluation of the CVD risk that accounts for the residual risk and
atherogenic dyslipidemias . However, in our Pakistani cohort,
we found that the values calculated according to Ahmadi et al.
exhibited excellent correlations with multiple CVD risks.
Our study findings have significant clinical implications. First,
although this study aimed to compare the various published
formulas and thus identify a method superior to that of Friedewald
et al,  we found that the values yielded by most equations
did not correlate with the measured LDLc values. Moreover, we
highlighted poor correlations of biochemical risk factors with
the measured and all calculated LDLc values except the value
determined according to Ahmadi et al,  suggesting that this
equation may actually be useful in a Pakistani population. More
significantly, our study highlighted the importance of clarifying
the smaller LDLc fractions and improving the related lab tests
used for diagnosis and monitoring.
Still, we must highlight certain limitations of our study. First,
the study was conducted in an urbanized Pakistani population,
which has been reported to exhibit a higher degree of atherogenic
dyslipidemia [23, 24]. Accordingly, our findings must be validated
in other regions. Second, our study design was cross-sectional
and conducted using data collected in clinical settings. Therefore,
epidemiological studies are needed to confirm our findings.
Most of the LDLc values calculated existing equations
exhibited moderate correlations with the measured LDLc values.
By contrast, only the equations developed by Anandaraja, Hatori
& de Cordova et al, [2,4,7] yielded values that correlated nonsignificantly
with the measured values; in other words, these
equations yielded near-actual values. However, the formula
developed by Ahmadi et al. correlated most strongly with
atherogenic dyslipidemia, nephropathy, and insulin resistance in
our Pakistani cohort.
Ethical approval: The study “Degree of concordance between
calculated and measured low-density lipoprotein cholesterol
methods” was formally approved by hospital’s ethical review
committee. We confirm that our research work conforms to
“World Medical Association’s Declaration of Helsinki – Ethical
Principles for Medical Research Involving Human Subjects”
Signing of inform consent by participants: All subjects were
required to sign written informed consent before participation
into study. All included subjects were explained about the research
requirements and use of data along with confidentiality issues.
Availability of SPSS data & outputs: Subject data can be
provided on formal request.
Author’s contributions: SHK: (Corresponding Author)
Idea conception, Diagnostic lab work, Manuscript writing. FS:
Analysis of data, manuscript writing. RS: Patient selection, History and examination, data analysis and manuscript writing.
SORSB: Statistical methods application, manuscript writing. SMM:
Lab work, data output (SPSS) analysis, contribution to writing
manuscript. RA: data output (SPSS) analysis, contribution to
writing manuscript, MT: Overall study coordination, medical
writing. All study authors approved the final manuscript version.