Chemosphere 68 (2007) 824–831

www.elsevier.com/locate/chemosphere

Calculation of serum ‘‘total lipid’’ concentrations for the adjustment

of persistent organohalogen toxicant measurements
in human samples
John T. Bernert *, Wayman E. Turner, Donald G. Patterson Jr., Larry L. Needham
Division of Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control and Prevention,
4770 Buford Highway, NE Atlanta, Georgia 30341, United States

Received 13 October 2006; received in revised form 12 February 2007; accepted 16 February 2007
Available online 3 April 2007

Abstract

Persistent organohalogen toxicants such as 2,3,7,8-tetrachlorodibenzo-p-dioxin or polychlorinated biphenyls measured in human

serum are often expressed on a lipid weight basis, most commonly by dividing the toxicants’ concentration by the weight of total lipids
in the sample. Therefore, the manner in which this lipid adjustment is calculated may influence the final reported result. Gravimetric total
lipid assays have been used, but they are time-consuming and sometimes may be ill-defined. Consequently, alternative methods using
enzymatic assays have been developed based on summing the individual lipid species measured. Recent reports, however, have suggested
that significantly different total lipid results may be obtained when using alternative formulae in a summation approach. In this report,
we summarize the results obtained from lipid measurements of nearly 900 samples made as part of a study of a group of older American
men (mean age 62 years), and we compare our total lipid estimates obtained by using both our standard and ‘‘short’’ formula (the latter
based on total cholesterol and triglycerides only) with results obtained using the recently proposed alternative formulae. Our findings
indicate that both our long and short formulae provide similar estimates of serum total lipid concentrations, and that differences
observed in lipid estimates when using the newer alternative summation methods may reflect differences in how the term ‘‘total lipid’’
is defined, especially with regard to the need to include the contribution of the weight of the cholesterol ester fatty acids in the calculation.
Published by Elsevier Ltd.

Keywords: Lipids; 2,3,7,8-TCDD; POPs; Enzymatic; Summation; Normalization

1. Introduction ‘‘Total lipids’’ is not a well-defined entity, but it is generally

understood to represent the weight of all of the hydropho-
The exposure of people to persistent organohalogen tox- bic lipid components in the serum as determined gravimet-
icants is often assessed by measuring the toxicants in blood rically following a well-defined and carefully conducted
serum. Because of their apolar and lipophilic character, solvent extraction and washing protocol. Gravimetric lipid
these toxicants are typically associated with the lipid matrix analyses thus represent the historically accepted standard
in tissues (Brown and Lawton, 1984; Patterson et al., 1988), for these measurements, but gravimetric assays are time-
and thus the results of the toxicant analysis may be consuming and ultimately defined only in terms of the
expressed per weight of lipid rather than on a whole weight methodology used to conduct them. For these reasons,
or a volume basis. In serum, different lipid or lipoprotein an alternative approach based on the analysis of individual
concentrations might be used for this adjustment, but com- lipid species followed by a summation of the results may be
monly the results are expressed per weight of ‘‘total lipids.’’ preferred.
We have previously defined and evaluated a method for
*
Corresponding author. Tel.: +1 770 488 7911; fax: +1 770 488 0181. total lipid estimates based on the enzymatic analysis of four
E-mail address: [email protected] (J.T. Bernert). individual lipid species and their summation (Akins et al.,

0045-6535/$ - see front matter Published by Elsevier Ltd.

doi:10.1016/j.chemosphere.2007.02.043
 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831 825

1989) and have shown that these results agreed well with who had a medical examination as part of a study con-
gravimetric assays based on a standard modified Folch ducted in 2002. Serum was processed when the blood sam-
extraction (Folch et al., 1957; Ullman and McCluer, 1977). ples were collected and then stored at 60 °C until the
Those analyses included direct measurements of free choles- samples were analyzed in 2005. The mean age (±Standard
terol and of phospholipids. However, we have also shown Deviation) of the participants in the study was 62 ± 7.2
that a reasonable total lipid estimate can be obtained when years.
only total cholesterol and triglycerides are available by
assuming a constant percent esterification of cholesterol in 2.2. Lipid analysis
serum and by estimating the phospholipid content from the
total cholesterol concentration (Phillips et al., 1989). Because All assays were made using an Hitachi Model 912 clinical
of the additional assumptions involved, the latter results analyzer. Total cholesterol was analyzed with the Roche
might be expected to be slightly less reliable than those Cholesterol/HP method (Roche Diagnostics, Indianapolis,
obtained by using the full expression. Cheek and Wease IN), an enzymatic colorimetric determination using choles-
(1969) reported that the prediction of gravimetric total lipid terol esterase and cholesterol oxidase. Free cholesterol used
values by using a summation approach was less accurate the Wako Free Cholesterol C method (Wako Chemicals
when a calculated rather than a measured phospholipid con- USA, Inc., Richmond, VA), an enzymatic colorimetric
centration was used. Furthermore, Phillips et al. (1989) assay which uses cholesterol oxidase and peroxidase but
found that the short formula was somewhat less successful omits cholesterol esterase. Triglycerides were analyzed with
than the long formula in correcting for postprandial changes the Roche Triglycerides/GPO method without blanking,
in total lipids. In that study the mean increase in total lipids and phospholipids were measured by using the Wako Phos-
following a test meal was estimated as 16% by the short for- pholipids B enzymatic colorimetric method. The Wako
mula , but it was 20% when using the long formula which phospholipids method uses phospholipase D and choline
incorporated small increases in the phospholipids. Thus, oxidase for the analysis, so it measures specifically the major
including measured phospholipid data in the calculation is choline-containing phospholipids including lecithin, lyso-
preferable when it is feasible to do so. Nevertheless, the lecithin, and sphingomyelin (Takayama et al., 1977). All
‘‘short formula’’ is more convenient, and in some cases only analyses were subjected to standard quality assurance pro-
total cholesterol and triglyceride data may be available. In cedures, and all reported results were from runs found to
exposure studies in which the population is not in the imme- be in control by standard statistical methods.
diate postprandial state we believe the short formula can be
useful and will provide group mean total lipid estimates that 2.3. Calculations
are very similar to those obtained by using the long formula.
Recently, alternative short formulae for calculating total Serum total lipids (TL) were estimated in each case by
lipids, which also are based on the use of enzymatic total summing the individual lipid values, including cholesterol
cholesterol and triglyceride measurements, have been sug- esters, free cholesterol (FC), triglycerides (TG), and phos-
gested (Covaci et al., 2006; Rylander et al., 2006). These pholipids (PL). Cholesterol in blood exists in both nones-
alternative approaches were reported to yield good total terifed form and in a form esterified to fatty acids. For
lipid estimates, but they were found to result in signifi- the cholesterol contribution to total lipids, we estimated
cantly different total lipid values compared with results the esterified cholesterol component as the difference
obtained by using our previously defined short formula between TC and free cholesterol. We then multiplied that
(Covaci et al., 2006). In this report, we summarize the value by a factor representing the ratio of the mean molec-
results we have obtained for total lipid estimates in a recent ular weight of esterified cholesterol divided by the molecu-
study involving nearly 900 men, and we compare our lar weight of unesterified cholesterol (i.e., 648.25/386.67
results obtained using both our long and short expressions = 1.677). The TL estimate is then obtained by summation
with those obtained using the recently proposed alternative (Akins et al., 1989) as: TL = 1.677*( TC  FC) + FC +
formulae. We conclude that our short formula provided a TG + PL.
valid and reliable estimate of serum total lipids in this When only TC and TG values are available, two addi-
study, and that differences observed when using the newer tional assumptions are made. First, we assume that serum
alternative equations in comparison with our short formula cholesterol averages 73% in the esterified form (with a CV
may reflect misunderstandings in the definitions of certain of about 7.7%). Second, we assume that the phospholipid
components in the total lipid expressions. concentrations may be linearly related to total cholesterol
by the expression: PL = (0.766 * TC) + 62.3 mg/dl, based
2. Experimental on the regression of PL on TC. The 81 people used to
develop this cholesterol–phospholipid regression expres-
2.1. Subjects sion as previously described by Phillips et al. (1989)
included the 50 individuals in an earlier study of Patterson
Results reported in this article were obtained during the et al. (1988), plus additional people with serum lipid data
analysis of blood samples from a group of 899 older men available from the same study who had not been included
826 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831

in the Patterson report because of other missing values Table 1

such as adipose measurements that were required for that Lipid concentrations of serum samples (N = 899)
study. Demographic data for the people in that study were Analyte Mean ± SD Percentiles
not available for one participant. Of the remaining 80 (mg/dl)
5th Percentile 95th Percentile
people, 31 (39%) were female and 3 (4%) were black. Their (mg/dl) (mg/dl)
ages ranged from 19 to 70 with a mean (±SD) of Total cholesterol 187 ± 34.9 134 247
43.8 ± 13.5 years. Free cholesterol 51.9 ± 9.5 37.7 68.7
Based on an assumed cholesterol ester percentage of Triglycerides 148 ± 78.0 62 294
Phospholipids 209 ± 31.4 161 265
73% and the regression analysis between total cholesterol Total lipidsa 636 ± 127.9 465 867
and phospholipid concentrations described above, a ‘‘short Total lipidsb 636 ± 122.5 467 861
form’’ of the TL expression could then be derived (Phillips Total lipidsc 555 ± 118.4 411 783
et al., 1989) as follows: Total lipidsd 526 ± 117.8 380 748
a
By CDC long formula – Akins et al. (1989).
TLS ¼ 1:677  ð0:73TCÞ þ ð0:27  TCÞ þ TG þ ð0:766  TCÞ þ 62:3 b
By CDC short formula (TC & TG only) – Phillips et al. (1989).
TLS ¼ ðð1:23 þ 0:27 þ 0:77Þ  TCÞ þ TG þ 62:3 c
By the formula of Covaci et al. (2006).
d
TLS ¼ ð2:27  TCÞ þ TG þ 62:3 mg=dl By the formula of Rylander et al. (2006).

If phospholipid data are available and only the free choles-

terol value is missing, then an alternative equation based the individual values obtained from TC and TG alone by
on TC + TG + PL can be derived in a similar manner: using the short formula vs. the total lipid estimate obtained
TLS2 ¼ 1:677  ð0:73TCÞ þ ð0:27  TCÞ þ TG þ PL with the full expression. The percentage difference between
the two estimates remained less than ±10% among all 899
TLS2 ¼ ð1:494  TCÞ þ TG þ PL mg=dl
participants. Regression of the short formula results on
those obtained by using the long formula resulted in an
For comparative estimates by the formula proposed by
adjusted r2 of 0.985.
Covaci et al. (2006), we used the equation:
When we applied the equations suggested by Covaci
TL ¼ ð1:12  TCÞ þ ð1:33  TGÞ þ 148 mg=dl et al. (2006) or Rylander et al. (2006) to these data, we
obtained lower total lipid estimates. For example, the mean
Estimates using the approach of Rylander et al. (2006) used
total lipid concentration for these data using our short for-
the expression:
mula was 636 ± 122 mg/dl (n = 899), which was identical to
TL ¼ 1:3  ðTC þ TGÞ þ 90 mg=dl the value estimated by using all of the individual lipid values
Thuresson et al. (2005) previously published an equation in the long formula. We have previously demonstrated a
for the measurement of TL that was very similar to the close correspondence between the long formula estimate
equation of Rylander et al. The estimate proposed by of TL and measurements by gravimetric analysis. However,
Thuresson was: application of the formula of Covaci et al. (2006) produced
a much lower mean TL estimate of 555 ± 118 mg/dl. An
TL ¼ 1:28  ðTC þ TGÞ þ 96 mg=dl even lower mean value of 526 ± 118 mg/dl was obtained
by using the formula of Rylander et al. (2006). Thus, these
Because of their similarity, results obtained using the alternative calculations produced values that averaged
Thuresson equation will be nearly the same as those ob- approximately 15% lower than did those we measured using
tained using the expression proposed by Rylander. either our short or long formula.
All calculations and statistical analyses were made using Similar results were obtained when these various expres-
SAS software, Version 9.0, SAS Institute, Research Trian- sions were applied to the mean lipid concentration values
gle Park, NC. Regression models were developed assuming from the reference pools previously assessed by Akins
error in both dimensions. et al. (1989). The results summarized in Table 2 indicate
that for each of the three reference pools, both our long
3. Results and short formula equations produced estimates similar
to the gravimetric value obtained for that pool. By con-
Mean results for each of the four serum lipid analytes trast, the results obtained by using the equations of Covaci
measured in the current study are given in Table 1 together et al. (2006) or Rylander et al. (2006) produced lower esti-
with the mean TL concentration as measured by both the mates that more closely approximated the simple summa-
long and the short formulae. The mean TL value estimated tion value (TC + TG + PL) for the three pools as
by using our short formula was identical to the result indicated in the last column of Table 2. These simple sum-
obtained with the long formula and also demonstrated a mation values underestimate total lipids because they do
similar variance relative to the full expression. The two esti- not incorporate a contribution from the fatty acids present
mates remained similar over the entire range of values in the cholesterol ester fraction. The same results were seen
observed. This agreement can be seen in Fig. 1, which plots (Table 3) when these equations were applied to three cur-
 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831 827

1600

Total Lipids by the Short Formula (mg/dl)

1400

1200

1000

800

600

400

200

0
0 200 400 600 800 1000 1200 1400 1600
Total Lipids by the Long Formula (mg/dl)

Fig. 1. Comparison of serum total lipid concentrations estimated by using the full and the ‘‘short’’ (cholesterol and triglycerides only) expressions. Total
lipids by the long- and the short-formula were estimated by using our equations TL and TLS, respectively, as described in the Section 2. The dotted line is
the regression line for these data; the solid line is the identity line.

Table 2
Analysis of reference pools
Pool Gravimetric CDC Covaci et al.c Rylander et al.d TC + TG + PLe
value (mg/dl) Long forma Short formb (mg/dl) (mg/dl) (mg/dl)
(mg/dl) (mg/dl)
Seronorm 991 ± 28 (n = 11) 960 880 707 693 821
Precilip 652 ± 18 (n = 10) 642 629 584 550 558
CDC 546 ± 15 (n = 23) 535 546 458 429 446
Notes: Pools are two commercial and one CDC reference pools; from Akins et al. (1989). Seronorm (TC = 278; FC = 73; TG = 186; PL = 357). Preclip
(TC = 167; FC = 42; TG = 187; PL = 204). CDC QC (TC = 175; FC = 44; TG = 86; PL = 185). Gravimetric values were determined by Akins et al.
(1989).
a
CDC Long form – Akins et al. (1989).
b
CDC Short form – Phillips et al. (1989).
c
Covaci et al. (2006).
d
Rylander et al. (2006).
e
TC + TG + PL is the simple sum of the three reference lipid values for each pool.

Table 3
Additional current reference pools and other samples
Samplea Total Triglycerides Phospholipids CDC Covaci et al.d Rylander TC + TG + PLf
cholesterol (mg/dl) (mg/dl) (mg/dl) et al.e (mg/dl) (mg/dl)
Total lipids long Total lipids short
(mg/dl) formb (mg/dl) formc (mg/dl)
Pool A 167 57 189 502 498 411 381 413
Pool B 284 115 313 858 822 619 609 712
Pool C 193 42 207 537 542 420 396 442
Group 1 213 140 243 701 686 573 549 596
Group 2 242 145 228 735 757 612 593 615
Group 3 191 80 196 561 576 468 442 467
a
Pool A – Wako control serum I, Normal. Code No: 410-00102; Lot 30H4; Pool B – Wako control serum II, Abnormal. Code No: 416-00202; Lot
31H4; Pool C – Precipath U Plus (Roche). 12149443 122; Lot 166448 (all lipid values as reported by the manufacturer; FC = 27%TC assumed for
Precipath). Group 1 = Canino, Italy population, n = 20, Dougherty et al. (1987), Group 2 = Nurmijarvi, Finland population, n = 21, Dougherty et al.
(1987), Group 3 = Beltsville, MD USA population, n = 21, Dougherty et al. (1987).
b
CDC long form – Akins et al. (1989).
c
CDC short form – Phillips et al. (1989).
d
Covaci et al. (2006).
e
Rylander et al. (2006).
f
TC + TG + PL is the simple sum of the three lipid values for each sample.
828 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831

rently available lipid reference materials and to three small et al. (2006) and of Rylander et al. (2006). However, we
populations taken from the study of Dougherty et al. also considered an alternative possibility that the term
(1987). In each case, the equations of Covaci and of Rylan- ‘‘Chol’’ in the equations of Covaci and of Rylander refered
der reflected only the simple sum of the three lipid species not to the concentration of total cholesterol that is com-
rather than the true total lipid value. monly understood by that term, but rather referred to a
Because each equation using only TC and TG involved previously adjusted concentration using assumed values
the same cholesterol and triglyceride values while estimat- for the customary percentage of esterified cholesterol and
ing the phospholipid concentration, the bias between the an assumed mean molecular weight of 571. That is, the
newer methods and our short formula has been proposed value might have been corrected before application in the
to result from this estimate of phospholipids (Covaci total lipid expression to account for the cholesterol ester
et al., 2006). However, we could not confirm that bias with contribution. This possibility might be inferred from the
our data. The estimated phospholipid concentration for the brief method descriptions in both Rylander et al. (2006)
samples in this study obtained from the equation relating and Covaci et al. (2006). However, if such an adjustment
PL and TC that was used in deriving our short formula was made, then what those authors refer to as the samples’
as described in the Experimental section produced good total cholesterol content would in fact be quite different
approximations of the actual measured values as indicated from what is commonly understood by that term in clinical
in Fig. 2. The mean phospholipid concentration estimated analyses. For example, if an adjusted value for TC is devel-
by that formula was 206 ± 27 mg/dl (n = 899), which was oped in advance by setting TC# = 571/386.67 * TC, or
in agreement with the mean measured concentration of 209 TC# = 1.4767 * TC, and total lipids for our data are then
± 31 mg/dl. When we used the expression for predicting estimated using this adjusted concentration of TC# in an
the phospholipid concentrations suggested by Covaci equation of TL 0 = (1.12 * TC#) + (1.33 * TG) + 148, then
et al. (2006) (PL = (0.12 * TC) + (0.33 * TG) + 148 mg/ a mean concentration of 656 ± 129 mg/dl would result.
dl), a mean concentration of 220 ± 27 mg/dl was obtained This result is similar to although slightly higher than the
that was approximately 5% higher than the mean measured estimate obtained using our equations. However, this
value. The expression proposed by Rylander et al. (2006) of approach would require a two-step calculation rather than
PL = 0.3 * (TC + TG) + 90 yielded similar agreement with the use of a single, integrated equation as described by
an estimated mean of 191 ± 28 mg/dl, or about 9% lower Covaci et al. (2006) and by Rylander et al. (2006), and there
than the measured phospholipid mean value. These differ- is no indication that such an adjusted cholesterol value was
ences are not sufficient to account for the consistent bias actually used by those authors.
reported (Covaci et al., 2006) between the two newer equa- These two possibilities are summarized in Fig. 3. Line 1
tions and the results obtained with our short formula. shows the TLS estimate obtained from the current data by
The likely explanation for the observed bias between using our short formula equation and, as previously noted,
these total lipid measurements is the exclusion of choles- shows good agreement with the long formula. Line 2 shows
terol ester fatty acids from the calculations of Covaci the estimate obtained by using the expression of Covaci
Calculated Phospholipid Concentration (mg/dl)

400

360

320

280

240

200

160

120

80

40

0
0 40 80 120 160 200 240 280 320 360 400
Measured Phospholipid Concentration (mg/dl)

Fig. 2. Relation between the measured concentration of serum phospholipids and the concentration estimated by regression analysis from the total
cholesterol concentration. The calculated phospholipid concentrations were estimated from the expression phospholipids = (0.766 * TC) + 62.3 mg/dl,
where TC is the total cholesterol concentration.
 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831 829

Estimated Total Lipids by Short Formulae (mg/dl)

1600

1400 4
3
1200

1000 1

800
2
600

400

200

0
0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Total Lipids by the Long Formula (mg/dl)

Fig. 3. Estimation of serum total lipids using different equations and assumptions for the calculation. Line 1 was obtained from the CDC ‘‘short’’ formula
(TLS in the Section 2); line 2 used the equation of Covaci et al. (7); line 3 is the same as line 2 but uses a previously ‘‘adjusted’’ total cholesterol
concentration (TC#) in the equation, calculated as TC# = 1.4767 * TC, to incorporate a term for the weight of the cholesterol esters; line 4 is the result
obtained using this adjusted cholesterol concentration (TC#) with the CDC short formula (TLS).

et al. (2006) that consistently underestimates the total lipid cholesterol in membranes and in serum lipoproteins is well
concentration and that appears to represent the simple sum established. By contrast, the relation between triglycerides
of (TC + TG + PL). However, if Covaci et al. (2006) had and phospholipids is relatively poor. This difference is
actually used an adjusted cholesterol value in their work not surprising since both cholesterol and phospholipids
rather than the customarily defined TC concentration, then are enriched in the LDL and HDL fractions of serum
a better estimate would be obtained from their expression as where they tend to associate as they do in membranes. Tri-
indicated by line 3. Finally, if that same adjusted cholesterol glycerides are most abundant in the VLDL fraction and in
value was then used inappropriately in the equation for the the chylomicra when the latter are present (Christie, 1982).
CDC short formula, line 4 would result with an erroneously The cholesterol–phospholipid ratio in serum and in specific
high total lipid estimate. Line 4 significantly overestimates lipoprotein fractions has been noted to vary over a limited
the total lipid concentration because a correction term for range in most cases (Cornwell et al., 1961). When we
cholesterol esters has been effectively entered twice. regressed phospholipids on total cholesterol for the 899
samples in this study, an r2 value of 0.697 was obtained,
4. Discussion which indicated that TC concentrations accounted for
approximately 70% of the variance observed in the PL val-
We report the results from the analysis of serum lipids in ues. By contrast, regression of PL on TG gave an r2 of only
899 samples analyzed as part of a recent study involving a 0.214. Thus, in our short formula for TL estimates, TG
group of older American men. We found excellent agree- represents only itself, whereas the necessary estimate of
ment between total lipid estimates for each sample deter- PL concentration is developed solely from the TC content
mined by a long formula calculation using each of the that it is more closely related to.
four lipid species in comparison with those obtained by Regardless of the method used to calculate total lipid
using a short formula based only on the total cholesterol concentrations from enzymatic lipid analyses, uncertainties
and triglyceride values. Since the long formula was previ- will always result from the unavoidable assumptions that
ously validated in both individual samples and in reference are made. These assumptions not only include an assumed
materials relative to gravimetric measurements, we con- constant percentage of esterified cholesterol in the blood
clude that both the long formula (Akins et al., 1989) and and a predictive relation between TC and PL when only
our short formula (Phillips et al., 1989) results obtained cholesterol and triglyceride concentrations are measured,
in this study are reasonable estimates of the total lipid con- but they also include the assumptions made in the enzy-
tent of these samples. matic assays themselves. For example, such assays typically
Our short formula includes an estimate for the phospho- require a hydrolysis of the lipids prior to measurements,
lipid concentration in a sample obtained from the regres- and this hydrolysis is assumed to be complete in all cases.
sion of PL on TC as defined in an earlier study (Phillips Furthermore, to convert a measured value for glycerol fol-
et al., 1989). This regression analysis did not include tri- lowing a hydrolysis of triglycerides, for example, one must
glycerides because of the physiological relation among assume a mean molecular weight for the triglycerides that
these lipid species. A relation between phospholipids and will depend on the actual fatty acid composition present
830 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831

in that sample. In addition, depending on how blanks are lipids when polychlorinated biphenyls were measured
handled, partial glycerides and free glycerol may or may (Phillips et al., 1989). These results indicate that our mea-
not be included in these measurements. surements of serum total lipids should be reliable in prop-
Similarly, enzymatic phospholipid assays typically mea- erly adjusting toxicant concentrations in samples. Thus, if
sure choline specifically and thus include only the main alternative total lipid estimates provide significantly differ-
serum phospholipids (phosphatidycholine, sphingomyelin, ent results, then similar comparisons should be made to
and lysophosphatidylcholine) in the analysis; minor phos- confirm their effectiveness.
pholipids such as phosphatidyl-ethanolamine are excluded. Several authors have noted concerns about the quality
Glycolipids are also excluded, but because of their rela- and interpretation of lipid adjustment measurements for
tively polar character, their exclusion is appropriate for this toxicant analyses. Ryan and Mills (1997) reported that
application. As with triglycerides, an average molecular the selection of different solvents in the extraction of blood
weight for phospholipids based on typical fatty acid com- lipids for gravimetric analysis could lead to substantial dif-
positions (and a common phospholipid species distribu- ferences in the results, potentially resulting in highly biased
tion) must be assumed in these calculations as well. Thus, toxicant measurements and body burden estimates. Ersbo-
the possibility exists that the assumptions that work well ell et al. (1990) noted that the variability in blood lipid
when applied to one population might be less reliable when measurements in assays of toxicants in WHO-coordinated
applied to a different population with, for example, a rad- interlaboratory studies has been unsatisfactory, requiring
ically different diet. The influence of such fatty acid compo- the elimination of lipid adjustments in those studies, with
sitional variation is probably limited, however, since, as results reported only on a whole weight basis. Ersboell
noted by Akins et al. (1989) in discussion of the cholesterol et al. (1990) further proposed that defined methods for
ester correction factor, even marked differences in fatty the lipid analyses should be established for use in future
acid composition of the cholesterol esters resulted in rela- studies. We agree that establishing a common basis for
tively limited deviations in the cholesterol factor and would measuring total lipids for toxicant adjustments would be
thus not be expected to be influential. useful, and we believe that common reference materials
In this study, comparison of total lipid results obtained should be developed and used for comparative purposes.
using the recently proposed alternative formulae of Rylan- Otherwise, determining exactly what is being compared
der et al. (2006) or of Covaci et al. (2006) with the results among studies by different groups can be difficult, and dif-
obtained with our approach, using either our data or lipid ferences reported in lipid-adjusted toxicant concentrations
data from other sources, suggest that the measurements of in such studies may at least partly represent only differences
Rylander and Covaci are not actually of the traditionally in the calculation methods used for the lipid adjustment
defined total lipids, but rather are estimates of the simple term.
summation of (TC + TG + PL). This calculation differs
from our and other previous measurements of serum total Acknowledgement
lipids in not accounting for the weight of the fatty acid
component of the cholesterol esters. If this simple summa- We thank Pam Olive, MT (ASCP) for performing all of
tion calculation was used, then the total lipid estimates the enzymatic serum lipid analyses on the participants in-
obtained by use of these alternative formulae would be cluded in this study.
expected to be consistently lower than our estimates, and,
consequently, their lipid-adjusted serum toxicant concen- References
trations would appear to be higher.
Lipid adjustments of nonpolar toxicant concentrations Akins, J., Waldrep, K., Bernert, J., 1989. The estimation of total serum
are ultimately a purely practical consideration meant to lipids by a completely enzymatic ‘summation’ method. Clin. Chim.
provide the best possible estimate of the toxicant’s body Acta 184, 219–226.
Brown, J., Lawton, R., 1984. Polychlorinated biphenyl (PCB) partitioning
burden. Many different approaches may possibly provide
between adipose tissue and serum. Environ. Health Persp. 61, 11–20.
better or worse adjustments, and the value of an approach Cheek, C.S., Wease, D.F., 1969. A summation technique for serum total
can only be established by a direct examination of the lipids. Comparison of methods. Clin. Chem. 15, 102–107.
results. Our summation estimate of total lipids has been Christie, W.W., 1982. Lipid Analysis, Second ed. Pergammon Press, New
shown to agree well with traditional gravimetric analyses York.
using a modified Folch extract (Akins et al., 1989). In addi- Cornwell, D., Kruger, F., Hamwi, G., Brown, J., 1961. Studies on the
characterization of human serum lipoproteins separated by ultracen-
tion, we have previously found that adjustment for serum trifugation in a density gradient. I. Serum lipoproteins in normal,
total lipids using our approach results in close agreement hyperthyroid and hypercholesterolemic subjects. Am. J. Clin. Nutr. 9,
between the lipid-adjusted serum TCDD concentrations 24–40.
and the lipid-adjusted concentrations of TCDD in adipose Covaci, A., Voorspoels, S., Thomsen, C., van Bavel, B., Neels, H., 2006.
Evaluation of total lipids using enzymatic methods for the normali-
samples from the same individuals (Patterson et al., 1988),
zation of persistent organic pollutant levels in serum. Sci. Total
with a slope statistically indistinguishable from 1.0. Simi- Environ. 366, 361–366.
larly, our lipid adjustment approach was found to effec- Dougherty, R., Galli, C., Ferro-Luzzi, A., Iacono, J., 1987. Lipid and
tively compensate for the postprandial rise in serum phospholipid fatty acid composition of plasma, red blood cells, and
 J.T. Bernert et al. / Chemosphere 68 (2007) 824–831 831

platelets and how they are affected by dietary lipids: a study of normal Ryan, J.J., Mills, P., 1997. Lipid extraction from blood and biological
subjects from Italy, Finland and the USA. Am. J. Clin. Nutr. 45, 443– samples and concentrations of dioxin-like compounds. Chemosphere
455. 34, 999–1009.
Ersboell, A., Nygren, M., Yrjanheikki, E., 1990. WHO-Coordinated Rylander, L., Nilsson-Ehle, P., Hagmar, L., 2006. A simplified precise
interlaboratory quality control studies on levels of PCBs, PCDDs, and method for adjusting serum levels of persistent organohalogen
PCDFs in human milk and blood. Organohalogen Compounds 4, 169– pollutants to total serum lipids. Chemosphere 62, 333–336.
174. Takayama, M., Itoh, S., Nagasaki, T., Tanimizu, I., 1977. A new
Folch, J., Lees, M., Sloane Stanley, G., 1957. A simple method for the enzymatic method for determination of serum choline-containing
isolation and purification of total lipids from animal tissues. J. phospholipids. Clin. Chim. Acta 79, 93–98.
Biol.Chem. 226, 497–509. Thuresson, K., Bergman, A., Jackobsson, K., 2005. Occupational expo-
Patterson, D., Needham, L., Pirkle, J., Roberts, D., Bagby, J., Garrett, sure to commercial decabromodiphenyl ether in workers manufactur-
W., Andrews, J., Falk, H., Bernert, J., Sampson, E., Houk, V., 1988. ing or handling flame-retardant rubber. Environ. Sci. Technol. 39,
Correlation between serum and adipose tissue levels of 2,3,7,8- 1980–1986.
tetrachlorodibenzo-p-dioxin in 50 persons from Missouri. Arch. Ullman, M., McCluer, R., 1977. Quantitative analysis of plasma neutral
Environ. Contam. Toxicol. 17, 139–143. glycosphingolipids by high performance liquid chromatography of
Phillips, D., Pirkle, J., Burse, V., Bernert, J., Henderson, L., Needham, L., their perbenzoyl derivatives. J. Lipid Res. 18, 371–378.
1989. Chlorinated hydrocarbon levels in human serum: effects of
fasting and feeding. Arch. Environ. Contam. Toxicol. 18, 495–500.