American Journal of Epidemiology
Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health 2012.
Vol. 175, No. 7
DOI: 10.1093/aje/kwr375
Advance Access publication:
February 2, 2012
Original Contribution
Relation of Blood Cadmium, Lead, and Mercury Levels to Biomarkers of Lipid
Peroxidation in Premenopausal Women
Anna Z. Pollack, Enrique F. Schisterman*, Lynn R. Goldman, Sunni L. Mumford, Neil J. Perkins,
Michael S. Bloom, Carole B. Rudra, Richard W. Browne, and Jean Wactawski-Wende
* Correspondence to Dr. Enrique F. Schisterman, Epidemiology Branch, Division of Epidemiology, Statistics, and Prevention
Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, 6100 Executive Blvd., Room 7B03,
Rockville, MD 20852 (e-mail: [email protected]).
Initially submitted June 14, 2011; accepted for publication September 23, 2011.
Exposures to cadmium, lead, and mercury are associated with adverse health effects, including cardiovascular
disease, which may be promoted by lipid peroxidation. The authors examined cadmium, lead, and mercury in relation
to plasma levels of F2-8a isoprostanes (isoprostane), 9-hydroperoxy-10,12-octadecadienoic acid (9-HODE),
13-hydroxy-9,11-octadecadienoic acid (13-HODE), and thiobarbituric acid reactive substances (TBARS) in 252 women
from western New York State (2005–2007). Healthy premenopausal women were followed for 2 menstrual cycles,
with biomarkers of lipid peroxidation being assessed 8 times per cycle. Metals were measured at baseline in whole
blood. Linear mixed models were used to estimate the association between cadmium, lead, and mercury and lipid
peroxidation biomarkers. Median cadmium, lead, and mercury levels were 0.30 lg/L, 0.86 lg/dL, and 1.10 lg/L,
respectively. Blood cadmium, lead, and mercury were not associated with increases in isoprostane, TBARS, 9-HODE,
or 13-HODE levels. Isoprostane levels decreased 6.80% (95% confidence interval: 10.40, 3.20) per 1% increase
in mercury. However, after adjustment for a simulated strong confounding factor, such as precisely measured fish
consumption, the observed association was attenuated, suggesting that this unexpected association could be
attributable to unmeasured confounding. In this population of healthy premenopausal women with low exposure
levels, cadmium, lead, and mercury were not associated with elevated lipid peroxidation biomarkers.
cadmium; hydroxyl-octadecadienoic acid; isoprostane; lead; mercury; oxidative stress; thiobarbituric acid reactive
substances; women
Abbreviations: CI, confidence interval; EDTA, ethylenediaminetetraacetic acid; 9-HODE, 9-hydroxy-10,12-octadecadieneoic acid;
13-HODE, 13-hydroxy-9,11-octadecadieneoic acid; LOD, limit of detection; SD, standard deviation; TBARS, thiobarbituric acid
reactive substances.
Cadmium, lead, and mercury are nonessential metals and
are frequently detected in the US population (1). Even at low
levels of exposure, metals have been associated with adverse
health effects, pointing to their relevance as a public health
concern (2–4), particularly among women of reproductive
age (5, 6). Cadmium is inhaled primarily through smoking
cigarettes and ingested via consumption of shellfish, offal,
and dark, leafy vegetables (7). Lead exposure stems from
inhalation of lead in air, ingestion of lead paint-contaminated
dust, drinking water, and release of lead stores in bone (8).
Mercury exposure occurs predominantly as methylmercury
from fish consumption (9). Given widespread exposure, research is needed regarding the effects of metals on precursors of chronic disease, particularly cardiovascular disease,
at environmentally relevant levels of exposure.
At exposure levels commonly found in the general US population, metals are associated with hypertension (10, 11) and
cardiovascular disease (12). Oxidative stress may lie within
the biologic pathway linking metals and cardiovascular disease (13). Experimental evidence shows that metals contribute
to oxidative processes (14, 15), and oxidative stress levels are
important contributors to the development of chronic disease
645
Am J Epidemiol. 2012;175(7):645–652
646 Pollack et al.
(16, 17). Epidemiologic evidence is limited, and studying
oxidative stress biomarkers poses a particular challenge in
premenopausal women because of fluctuating hormone levels,
which may influence levels of oxidative stress (18, 19).
F2-8a isoprostanes (hereafter called isoprostane),
9-hydroxy-10,12-octadecadienoic acid (9-HODE), 13hydroxy-9,11-octadecadienoic acid (13-HODE), and thiobarbituric acid reactive substances (TBARS) are biomarkers
of different components of the lipid peroxidation process and
have been linked with chronic disease development. Isoprostane is derived from arachidonic acid and is stable (20, 21),
whereas 9-HODE and 13-HODE are linoleic acid peroxidation metabolites (22). The TBARS test represents an index
of systemic lipid peroxidation, albeit with methodological
concerns (23).
Given the paucity of research among women with low levels
of metals and the broad public health implications of lipid
peroxidation-related effects, our aim was to evaluate the effects
of metal exposure, specifically exposure to cadmium, lead,
and mercury, in relation to 4 biomarkers of lipid peroxidation
in a cohort of healthy premenopausal women.
MATERIALS AND METHODS
Study cohort
The BioCycle Study enrolled healthy, premenopausal
women (ages 18–44 years) to evaluate the relation between
biomarkers of oxidative stress and hormone levels over the
course of a regular menstrual cycle. Health status was ascertained by self-report. Inclusion criteria comprised a selfreported menstrual cycle length of 21–35 days for the past
6 months, not actively trying to conceive, and no history of
polycystic ovary syndrome; exclusion criteria included vitamin
supplement use (24). Women were followed prospectively for
1 (n ¼ 9) or 2 (n ¼ 250) menstrual cycles. The 9 women
followed for 1 cycle were recruited for a pilot substudy. A total
of 252 women had blood measures of metals and lipid peroxidation and were thus included in this analysis.
Recruitment and data collection occurred from 2005 to 2007
at the University at Buffalo (State University of New York)
in western New York State, under an Intramural Research
Program contract from the Eunice Kennedy Shriver National
Institute of Child Health and Human Development. A sample
size of 250 was needed to achieve 99% power to detect
a change in slope from 0.11 under the null hypothesis to
0.00 under the alternative hypothesis when the standard
deviation of the exposure was 1.00, the standard deviation of
the outcome was 0.36, and the 2-sided alpha level was 0.05.
All participants provided written informed consent prior
to participation. Under a reliance agreement, the National
Institutes of Health depends on the designated institutional
review board of the University at Buffalo for review, approval,
and continuing oversight of its human subject research for
the BioCycle Study.
Clinic visits
Women made clinic visits to the University at Buffalo
Women’s Health Research Center to provide blood samples
8 times per menstrual cycle, corresponding to early menstruation, the mid- and late follicular phases, 2 days around
the expected time of ovulation, and the early, mid-, and late
luteal phases. Visits took place in the morning in order to
obtain fasting blood samples and reduce diurnal variation.
Clearblue Easy fertility monitors (Inverness Medical, Waltham,
Massachusetts) aided in scheduling visits to appropriate
menstrual cycle phases (25).
Analysis of metal exposure
Whole blood was collected at the screening visit in
ethylenediaminetetraacetic acid (EDTA) purple-topped tubes
(prescreened for trace metals) provided by the Centers for
Disease Control and Prevention. Samples were analyzed for
levels of cadmium, lead, and mercury in whole blood by inductively coupled plasma mass spectrometry at the Division
of Laboratory Sciences of the National Center for Environmental Health, Centers for Disease Control and Prevention.
The limits of detection (LODs) for cadmium, lead, and mercury
were 0.20 lg/dL (25% < LOD), 0.25 lg/dL (0% < LOD),
and 0.30 lg/dL (12% < LOD). Values below the LOD were
reported by the laboratory. To minimize potential bias, values
below the LOD were not substituted (26). The interassay precisions (relative standard deviations) for cadmium, lead, and
mercury were 4.3%, 2.6%, and 3.2% at levels of 2.04 lg/L,
2.89 lg/dL, and 5.77 lg/L, respectively.
Biomarker analysis
Blood serum and plasma collection and handling protocols
were designed to minimize preanalytical variability (27).
Specimens were collected at each of the clinic visits (up to
8 visits) during each menstrual cycle and were delivered
to the processing laboratory, centrifuged at 1,500 3 g for
10 minutes, portioned into cryotubes, and frozen at 80°C
within 90 minutes of phlebotomy. Isoprostane, TBARS,
9-HODE, and 13-HODE levels were measured in EDTA
(15% K3EDTA) anticoagulated plasma. Isoprostane samples
were analyzed by gas chromatography-mass spectrometry
at the Molecular Epidemiology and Biomarker Research
Laboratory of the University of Minnesota (21, 28). An internal standard, [2H4]-isoprostane (>98% pure; Cayman
Chemical Company, Ann Arbor, Michigan), was used. Total
plasma 9-HODE, 13-HODE, and TBARS were measured at
the Oxidative Stress Research Laboratory of the University
at Buffalo. 9-HODE and 13-HODE were measured by high
performance liquid chromatography with diode array detection at 234 nm following mild alkaline hydrolysis of lipid
esters to yield total free fatty acids, including 9-HODE and
13-HODE (29, 30). HODE samples were delivered to the
laboratory and analyzed daily. The TBARS test measures
malondialdehyde, a 3-carbon aldehyde produced from hydrolysis of some lipid hydroperoxides (23). TBARS samples
were measured at an excitation of 535 nm and emission of
552 nm with an RF-5000U spectrofluorometer (Shimadzu
Scientific Instruments Inc., Columbia, Maryland), and TBARS
levels are expressed in nmol/mL (malondialdehyde) equivalents (29, 31). Collection and handling protocols were designed to minimize variability in exogenous factors (27, 32).
Am J Epidemiol. 2012;175(7):645–652
Blood Metal Levels and Lipid Peroxidation Biomarkers
The half-lives of biomarkers of lipid peroxidation are several
hours in duration (17). The interassay coefficients of variation
were 9.4%, 8.3%, 9.0%, and 9.2%, for isoprostane, TBARS,
9-HODE, and 13-HODE, respectively.
Collection of covariate data
At screening, women provided information on their health
and reproductive history and lifestyle. Trained staff recorded
height (meters) and weight (kilograms) to determine body
mass index (weight (kg)/height (m)2). Physical activity was
measured with the International Physical Activity Questionnaire (33). Intakes of whole foods (fish, shellfish, vegetables,
and grains) and nutrients (dietary iron) were assessed using the
general food frequency questionnaire developed by the
Nutrition Assessment Shared Resource of the Fred Hutchinson
Cancer Research Center (Seattle, Washington) at baseline
(6-month recall). Women recorded supplemental vitamin use
in daily diaries and according to the study inclusion guidelines
were not meant to be taking vitamin supplements. Occupation
was reported to determine possible occupational exposures
that might be associated with oxidative stress or metals.
Statistical analysis
Demographic characteristics were compared according
to tertiles of metals and isoprostane measured at the first
menstrual cycle visit during the first cycle under study, and
associations were assessed using t tests, chi-squared tests,
or Fisher’s exact tests where appropriate. Distributions of
oxidative stress biomarkers were checked for normality and
log-transformed.
Linear mixed models were used to evaluate the association
between exposure to metals and levels of oxidative stress,
while accounting for nonindependence between individual
measurements across the cycle and multiple cycles per woman.
Each log-transformed metal was analyzed in a separate model;
results are interpreted as the percent change in oxidative stress
level per 1% increase in metal (exposure) level. Random intercepts accounted for variation in baseline oxidative stress
levels between women. Covariate selection was determined
by a review of the literature and included age (years; continuous), body mass index (continuous), smoking status (current/
not current), and race (white, black, Asian, other). Models
for mercury were additionally adjusted for fish consumption
(servings/month; continuous). Measures of income, education,
physical activity, parity, dietary iron, shellfish, vegetables,
dietary selenium, dietary calcium, and total energy intake
were considered as potential confounders but did not appreciably alter the effect estimates. Because reproductive
hormones are potential mediators in the association between
metals and lipid peroxidation, they were not considered as
confounders. P values are 2-sided. Statistical analyses were
conducted in SAS 9.2 (SAS Institute, Inc., Cary, North
Carolina) and R (R Foundation for Statistical Computing,
Vienna, Austria).
We conducted a sensitivity analysis to evaluate the impact
of adjusting for fish consumption. We compared results of
this analysis with our final models, which were adjusted for
measured fish consumption from the baseline food frequency
Am J Epidemiol. 2012;175(7):645–652
647
questionnaire. Because fish consumption is known to be
measured with considerable error, the observed correlation
coefficients for the correlation between fish consumption
and mercury (q ¼ 0.31 and q ¼ 0.14) were attenuated.
Therefore, we generated a deattenuated variable to represent
actual fish consumption that was correlated more strongly
than measured fish consumption with mercury (q ¼ 0.55)
and isoprostane (q ¼ 0.50).
RESULTS
Median blood cadmium, lead, and mercury levels were
0.30 lg/L (interquartile range, 0.19–0.43), 0.86 lg/dL (interquartile range, 0.67–1.20), and 1.10 lg/L (interquartile range,
0.58–2.00), respectively. The demographic characteristics
of the study population varied somewhat by metal level
(Table 1). Older women had significantly higher levels of
cadmium and lead. Body mass index did not differ significantly by metal level. Parity and education did not differ
by metal level. White women had lower cadmium levels
(mean ¼ 0.32 lg/L; standard deviation (SD), 0.30) than blacks
(mean ¼ 0.36 lg/L; SD, 0.23) and Asians (mean ¼ 0.46 lg/L;
SD, 0.28). Asian women had higher lead (mean ¼ 1.50 lg/dL;
SD, 0.78) and mercury (mean ¼ 2.37 lg/L; SD, 2.12)
levels than whites (lead mean ¼ 0.95 lg/dL (SD, 0.65);
mercury mean ¼ 1.33 lg/L (SD, 1.14)) or blacks (lead
mean ¼ 0.92 lg/dL (SD, 0.30); mercury mean ¼ 1.36 lg/L
(SD, 0.99)). Current smokers had higher cadmium levels,
but smoking status was not associated with lead and mercury
levels.
Table 2 shows unadjusted and adjusted percent change in
lipid peroxidation level per 1% increase in metal exposure
derived using linear mixed models. We observed that blood
metal levels tended to be inversely associated with isoprostane levels, though generally without statistical significance.
In unadjusted models, each 1% increase in metal level was
associated with statistically significant decreases in isoprostane level. In adjusted models, cadmium and lead were no
longer statistically significantly associated with decreases in
isoprostane levels, while each 1% increase in mercury level
was associated with a statistically significant 6.80% decrease
(95% confidence interval (CI): 10.40, 3.20) in isoprostane
level. TBARS levels were consistently but not statistically
significantly increased in relation to metal exposure, although,
in unadjusted models, each 1% increase in mercury was associated with a statistically significant 3.49% (95% CI: 0.34, 6.64)
increase in TBARS. 9-HODE and 13-HODE were not
statistically significantly associated with metals.
The addition of a simulated confounder at various levels of
modest correlation between mercury and isoprostane (approximately 0.8 to þ0.8) shows that such an association could
be observed (see Web Figure 1 (http://aje.oxfordjournals.
org/)). However, the peak of Web Figure 1 demonstrates that
after adjustment for a highly correlated confounding factor
(approximately q ¼ 0.4–0.7 with mercury and q ¼ 0.4
to 0.7 with isoprostane), the association between mercury
and isoprostane was approximately null or weakly positive.
Specifically, additional adjustment for a simulated confounder
that was correlated with mercury (q ¼ 0.55) and isoprostane
Lead Level, mg/dL
Cadmium Level, mg/L
0.03–0.29 (n 5 122)
Characteristic
Mean (SD)
%
0.30–3.10 (n 5 127)
Mean (SD)
P Value
%
c
0.31–0.88 (n 5 123)
Mean (SD)
%
Mercury Level, mg/L
0.89–6.20 (n 5 126)
Mean (SD)
P Value
%
0.00–1.09 (n 5 120)
Mean (SD)
%
1.10–9.90 (n 5 129)
Mean (SD)
P Value
%
Age, years
25.8 (7.7)
28.7 (8.4)
0.006
25.7 (7.5)
28.8 (8.6)
0.003
27.1 (8.4)
27.4 (7.9)
0.77
Body mass indexd
24.0 (3.8)
24.2 (3.9)
0.70
24.5 (3.8)
23.7 (3.8)
0.12
24.5 (4.0)
23.7 (3.6)
0.09
Energy intake,
kcal/day
1,653 (400)
Selenium, lge
1,550 (391)
85.6 (37.6)
e
Vitamin E, mIU
Isoprostane, pg/mLf
0.04
94.3 (45.1)
0.11
1,630 (383)
1,571 (411)
89.1 (41.3)
0.24
90.8 (42.2)
0.71
1,640 (428)
1,563 (366)
91.7 (41.4)
0.13
88.2 (42.0)
0.52
9.4 (6.0)
10.6 (9.9)
0.27
9.9 (9.1)
10.1 (7.3)
0.84
8.7 (5.1)
11.3 (10.3)
0.01
50.5 (17.7)
48.1 (17.3)
0.28
52.6 (19.4)
46.1 (14.8)
0.004
52.8 (18.9)
46.0 (15.3)
0.003
13-HODE, lmol/Lf
0.26 (0.30)
0.26 (0.30)
0.99
0.27 (0.27)
0.24 (0.32)
0.47
0.27 (0.27)
0.25 (0.33)
0.52
9-HODE, lmol/Lf
0.22 (0.21)
0.20 (0.20)
0.42
0.22 (0.22)
0.19 (0.20)
0.37
0.21 (0.21)
0.20 (0.21)
0.58
0.92 (0.24)
0.86 (0.22)
0.04
0.89 (0.23)
0.89 (0.24)
0.94
0.89 (0.24)
0.89 (0.22)
0.89
TBARS, nmol/mL
f
Race
White
70
48
69
50
63
56
Black
17
23
19
20
20
20
Asian
7
21
5
24
10
19
Other
5
7
6
6
7
5
99
94
98
95
0.50
97
95
0.50
0.23
59
50
0.31
Nonsmoker/former
smoker
Physical activity
level
0.002
0.04
0.002
0.26
0.04
Am J Epidemiol. 2012;175(7):645–652
Low
11
9
58
52
Medium
28
43
36
36
33
38
High
61
48
6
13
7
12
Nulliparous
76
71
0.35
73
75
0.71
73
74
0.81
More than high
school
education
89
87
0.51
86
90
0.40
87
89
0.54
Abbreviations: 9-HODE, 9-hydroxy-10,12-octadecadieneoic acid; 13-HODE, 13-hydroxy-9,11-octadecadieneoic acid; SD, standard deviation; TBARS, thiobarbituric acid reactive substances.
Metal exposures were dichotomized at the median value.
b
Percentages may not sum to 100 because of rounding.
c
Difference between mean values (analysis of variance) for continuous variables; chi-squared test or Fisher’s exact test for categorical variables. All P values are 2-sided.
d
Weight (kg)/height (m)2.
e
Selenium and vitamin E levels were determined by baseline food frequency questionnaire.
f
Lipid peroxidation levels as measured at the first clinic visit.
a
648 Pollack et al.
Table 1. Characteristics of the Study Population According to Category of Metal Exposure,a BioCycle Study, Buffalo, New York, 2005–2007b
Blood Metal Levels and Lipid Peroxidation Biomarkers
649
Table 2. Change in Oxidative Stress Level (%) per 1% Increase in Blood Metal Level in a Linear Mixed Model
(n ¼ 252), BioCycle Study, Buffalo, New York, 2005–2007
Biomarker
and Model
Cadmium,
mg/L
95% CI
Lead,
mg/dL
95% CI
Mercurya,
mg/L
95% CI
7.48
13.49, 1.48
11.89
20.37, 3.42
9.22
13.16, 5.28
2.48
8.65, 3.68
4.67
13.35, 4.01
6.80
10.40, 3.20
Isoprostane
Unadjusted
b
Adjusted
9-HODE
Unadjusted
4.56
3.49, 12.61
2.40
8.94, 13.75
1.25
6.65, 4.14
Adjusted
3.90
5.03, 12.83
0.35
12.13, 12.82
0.92
6.56, 4.72
Unadjusted
7.78
1.26, 16.82
1.75
11.01, 14.51
3.54
9.62, 2.55
Adjusted
6.08
3.96, 16.12
4.36
18.36, 9.64
3.54
9.86, 2.78
Unadjusted
2.32
2.40, 7.03
6.08
0.57, 12.74
3.49
0.34, 6.64
Adjusted
1.32
3.95, 6.59
5.24
2.18, 12.67
3.08
0.20, 6.37
13-HODE
TBARS
Abbreviations: CI, confidence interval; 9-HODE, 9-hydroxy-10,12-octadecadieneoic acid; 13-HODE, 13-hydroxy-9,
11-octadecadieneoic acid; TBARS, thiobarbituric acid reactive substances.
a
For mercury, adjustment included adjustment for fish consumption (servings/month; continuous) as measured by
baseline food frequency questionnaire.
b
Results were adjusted for age, race (black, white, Asian, or other), body mass index (weight (kg)/height (m)2), and
smoking (yes vs. no/former).
(q ¼ 0.50), at levels of correlation similar to measured fish
consumption but more strongly correlated to account for
deattenuation, further diminished the association between
mercury and isoprostane to 3.87% (95% CI: 7.18, 0.55).
Since the simulated confounder was based on strong correlations, we calculated deattenuated correlation coefficients
for the correlations between mercury, isoprostane, and the
measured confounding factors to determine whether such
strong correlations were plausible, given our observed data
(34). We found that for reasonable amounts of error variance
in the measurement of dietary fish consumption (fish-mercury:
r2e ¼ 3.24; fish-isoprostane: r2e ¼ 1.56), in comparison with
the observed variance of fish consumption (2,028.41 servings/
year2; mercury, 1.77 lg/L2; isoprostane, 0.14 pg/mL2), the
observed correlations between fish consumption and mercury
(q ¼ 0.30) and fish consumption and isoprostane (q ¼ 0.14)
approximated the simulated levels (deattenuated q ¼ 0.50
and deattenuated q ¼ 0.50, respectively). Further, 43% of
participants reported no fish consumption.
DISCUSSION
Our findings indicate that low blood levels of mercury,
lead, and cadmium were not associated with increasing
levels of lipid peroxidation biomarkers (isoprostane, TBARS,
9-HODE, and 13-HODE) in healthy premenopausal women.
However, we did observe a modest inverse association between
mercury measured in blood and isoprostane level.
Our results suggest that at low levels of metals, lipid peroxidation biomarkers among healthy premenopausal women
are not increased by cadmium, mercury, and lead, a finding that is consistent with the available literature. A recent
Am J Epidemiol. 2012;175(7):645–652
population-based study in US adults over age 40 years found
that inflammatory biomarkers were not increased with lead
exposure, particularly among women (35). Biomarkers of
lipid peroxidation were not elevated in relation to lead and
cadmium in males with blood lead levels less than 40 lg/dL
(36). Further, an occupational study found a strong increase
in lipid peroxidation with blood lead levels above 35 lg/dL
but only a weak association with lower blood lead levels (37).
Most epidemiologic evidence that metals are positively associated with oxidative stress has come from occupational
studies with higher exposure (38, 39), suggesting that our
findings could be attributable to a threshold or nonlinear
effect. The maximum measured blood lead and mercury
levels in the BioCycle Study were 6.2 lg/dL and 9.9 lg/L,
lower than mean levels reported for most occupational cohorts,
though comparable to levels found among adults in the general US population (9, 40). Geometric mean blood cadmium,
lead, and mercury levels were 0.29 lg/L (7), 1.78 lg/dL (8),
and 1.02 lg/L (41), respectively, among reproductive-aged
women in the National Health and Nutrition Examination
Survey, as compared with 0.29 lg/L for cadmium, 0.91 lg/dL
for lead, and 1.04 lg/L for mercury in the BioCycle Study.
The association between isoprostane and mercury was
unexpected but did not persist among persons with lead
levels above the median. While it is possible that this finding
represents a true association, as nonlinear associations or
hormesis could explain such a finding at low levels of mercury, there are several alternative explanations. This association could be attributable to residual confounding resulting
from our use of a somewhat crude instrument (i.e., a food
frequency questionnaire) to assess fish consumption. Fish
intake is a recognized source of mercury exposure, and n-3
fatty acids found in fish have been associated with decreased
650 Pollack et al.
isoprostane levels (42, 43). Most mercury exposure among
non-occupationally exposed persons is likely to be methylmercury from fish (44), but we did not separately examine
different forms of mercury (inorganic, methyl), which may
exert different effects on oxidative stress biomarkers (45).
Because of measurement error, residual confounding from fish
consumption is likely; ideally, one would measure n-3 fatty
acid and methylmercury concentrations to preclude this bias.
Dietary intake data are subject to measurement error, which
may attenuate the observed correlations between fish consumption, mercury, and isoprostane (46). Therefore, fish
consumption (47) may be more strongly correlated with mercury and isoprostane than we observed, because of attenuation
from measurement error. The observed correlation between
fish consumption (measured at baseline by food frequency
questionnaire) and mercury was 0.30 (q ¼ 0.30), and the correlation between fish consumption and isoprostane was 0.14
(q ¼ 0.14). The sensitivity analysis that accounted for
possible measurement error by simulating a confounding
factor at deattenuated levels of correlation demonstrated the
plausibility of a strong confounding factor, measured with
error, to account for the unexpected association between
mercury and isoprostane.
The association between mercury and isoprostane represents
the total effect, or any direct effect of mercury on isoprostane plus any indirect effects that are mediated by estradiol.
In previous work, we observed an inverse association between estradiol and isoprostane (18) and a positive, though
not statistically significant, association between mercury and
estradiol (48), such that the indirect effect of mercury on isoprostane mediated by estradiol (the product of the estradiolisoprostane and mercury-estradiol associations) would be
inverse. Under assumptions of no interactions, linearity of
effect, and no unmeasured confounding, the sum of the direct
and indirect effects equals the total effects (49). It is plausible
that the indirect effects are stronger than the direct effect
between mercury and isoprostane, resulting in the total
effect’s being inversely associated as well.
Our study had several strengths, including the timing of
biomarker specimen collection, multiple measurements across
the menstrual cycle, and state-of-the-art measurement techniques for both metals and measures of oxidative stress. It is
unlikely that our findings are attributable to diurnal variation
in oxidative stress levels, which are generally considered
minimal (50). Our samples were processed rapidly and were
shipped in batches to minimize batch variability that may
occur when analyzing repeated samples among individuals.
Finally, the BioCycle Study recruited healthy women who
were selected to minimize known confounding factors, such
as adherence to a specific diet, which may affect oxidative
stress and metal levels.
However, our analysis had some limitations. Metal exposure was assessed only at the screening visit, and in some
cases women did not complete the study protocol in consecutive menstrual cycles (n ¼ 23). Thus, metals measured at
screening may not have reflected exposure during the second
cycle under study. We conducted subanalyses restricted to the
first menstrual cycle and to women who completed the study
protocol in consecutive cycles, and our findings were consistent (data not shown). BioCycle Study participants may have
a narrower range of lipid peroxidation and metal levels than
the general population, as the women were selected to be
healthy. This could have hindered detection of effects, particularly if associations occur beyond a threshold level. The
BioCycle Study had a low prevalence of current smoking
(n ¼ 16), and smoking contributes to both metal exposure
and oxidative stress (51). Despite the lack of a statistically
significant difference in isoprostane levels by smoking status,
participants who reported never smoking had slightly lower
mean isoprostane levels than those who reported smoking
daily to weekly (50.33 pg/mL vs. 54.93 pg/mL). We observed
expected trends in lipid peroxidation biomarkers. For example, obese women in BioCycle had higher mean isoprostane
levels than normal-weight women, as expected (70.4 pg/mL
(SD, 34.0) and 57.1 pg/mL (SD, 20.1), respectively) (52).
Moreover, lipid peroxidation biomarkers in plasma may not
represent oxidative damage in other tissues, thus limiting our
ability to observe oxidative damage in other parts of the body.
This is plausible, since mercury and cadmium are stored in
the kidney and may exert localized effects.
We demonstrated that biomarkers of lipid peroxidation
were not elevated in relation to increasing blood levels of
cadmium, lead, and mercury at the low exposure levels experienced by the general population. Our findings suggest
that isoprostane may decrease in relation to increasing levels
of mercury, although this finding may be explained by measurement error or by indirect effects’ being stronger than the
direct effect between mercury and isoprostane. To our knowledge, this was the first study to assess repeated measures of
multiple biomarkers of lipid peroxidation in relation to low
levels of metal exposure. Our data suggest that there is no
association between low, environmentally relevant levels of
cadmium, lead, and mercury and blood levels of TBARS,
9-HODE, and 13-HODE. However, in future studies, researchers may consider investigating these associations in
other populations, as this was a highly selected, healthy population of premenopausal women, and little is known about
potential effects of metals on oxidative stress biomarkers in
older women or children.
ACKNOWLEDGMENTS
Author affiliations: Epidemiology Branch, Division of
Epidemiology, Statistics, and Prevention Research, Eunice
Kennedy Shriver National Institute of Child Health and
Human Development, Bethesda, Maryland (Anna Z. Pollack,
Enrique F. Schisterman, Sunni L. Mumford, Neil J. Perkins);
Department of Epidemiology, Bloomberg School of Public
Health, Johns Hopkins University, Baltimore, Maryland
(Anna Z. Pollack, Lynn R. Goldman); Department of Environmental Health Sciences, School of Public Health, University
at Albany, State University of New York, Rensselaer, New
York (Michael S. Bloom); and Department of Social and
Preventive Medicine, University at Buffalo, State University of New York, Buffalo, New York (Carole B. Rudra,
Richard W. Browne, Jean Wactawski-Wende).
This research was supported by the Intramural Research
Program of the Eunice Kennedy Shriver National Institute
Am J Epidemiol. 2012;175(7):645–652
Blood Metal Levels and Lipid Peroxidation Biomarkers
of Child Health and Human Development, National Institutes
of Health, and by the Long-Range Research Initiative of the
American Chemistry Council.
The authors acknowledge Dr. Leila Jackson for leading
the ancillary study of metals.
Conflict of interest: none declared.
REFERENCES
1. Centers for Disease Control and Prevention. Fourth National
Report on Human Exposure to Environmental Chemicals.
Atlanta, GA: Centers for Disease Control and Prevention; 2009.
2. Bazan NG, Colangelo V, Lukiw WJ. Prostaglandins and other
lipid mediators in Alzheimer’s disease. Prostaglandins Other
Lipid Mediat. 2002;68-69:197–210.
3. Ikizler TA, Morrow JD, Roberts LJ, et al. Plasma F2-isoprostane
levels are elevated in chronic hemodialysis patients. Clin
Nephrol. 2002;58(3):190–197.
4. Oberg BP, McMenamin E, Lucas FL, et al. Increased prevalence
of oxidant stress and inflammation in patients with moderate
to severe chronic kidney disease. Kidney Int. 2004;65(3):
1009–1016.
5. Goldman LR, Shannon MW. Technical report: mercury in the
environment: implications for pediatricians. Pediatrics. 2001;
108(1):197–205.
6. Vahter M, Berglund M, Akesson A, et al. Metals and women’s
health. Environ Res. 2002;88(3):145–155.
7. Mijal RS, Holzman CB. Blood cadmium levels in women of
childbearing age vary by race/ethnicity. Environ Res. 2010;
110(5):505–512.
8. Lee MG, Chun OK, Song WO. Determinants of the blood lead
level of US women of reproductive age. J Am Coll Nutr. 2005;
24(1):1–9.
9. Mahaffey KR, Clickner RP, Bodurow CC. Blood organic
mercury and dietary mercury intake: National Health and
Nutrition Examination Survey, 1999 and 2000. Environ Health
Perspect. 2004;112(5):562–570.
10. Staessen J, Bruaux P, Claeys-Thoreau F, et al. The relationship
between blood pressure and environmental exposure to lead
and cadmium in Belgium. Environ Health Perspect. 1988;
78(6):127–129.
11. Yazbeck C, Thiebaugeorges O, Moreau T, et al. Maternal
blood lead levels and the risk of pregnancy-induced hypertension:
the EDEN Cohort Study. Environ Health Perspect. 2009;
117(10):1526–1530.
12. Navas-Acien A, Guallar E, Silbergeld EK, et al. Lead exposure
and cardiovascular disease—a systematic review. Environ
Health Perspect. 2007;115(3):472–482.
13. Steinberg D, Parthasarathy S, Carew TE, et al. Beyond
cholesterol. Modifications of low-density lipoprotein that
increase its atherogenicity. N Engl J Med. 1989;320(14):
915–924.
14. Stohs SJ, Bagchi D. Oxidative mechanisms in the toxicity of
metal ions. Free Radic Biol Med. 1995;18(2):321–336.
15. Ercal N, Gurer-Orhan H, Aykin-Burns N. Toxic metals and
oxidative stress part I: mechanisms involved in metal-induced
oxidative damage. Curr Top Med Chem. 2001;1(6):529–539.
16. Galanis A, Karapetsas A, Sandaltzopoulos R. Metal-induced
carcinogenesis, oxidative stress and hypoxia signalling. Mutat
Res. 2009;674(1-2):31–35.
17. Basu S. F2-isoprostanes in human health and diseases: from
molecular mechanisms to clinical implications. Antioxid
Redox Signal. 2008;10(8):1405–1434.
Am J Epidemiol. 2012;175(7):645–652
651
18. Schisterman EF, Gaskins AJ, Mumford SL, et al. Influence
of endogenous reproductive hormones on F2-isoprostane
levels in premenopausal women: the BioCycle Study. Am
J Epidemiol. 2010;172(4):430–439.
19. Karowicz-Bilinska A, Plodzidym M, Krol J, et al. Changes
of markers of oxidative stress during menstrual cycle. Redox
Rep. 2008;13(5):237–240.
20. Kadiiska MB, Gladen BC, Baird DD, et al. Biomarkers
of Oxidative Stress Study III. Effects of the nonsteroidal
anti-inflammatory agents indomethacin and meclofenamic
acid on measurements of oxidative products of lipids in
CCl4 poisoning. Free Radic Biol Med. 2005;38(6):711–718.
21. Milne GL, Yin H, Brooks JD, et al. Quantification of
F2-isoprostanes in biological fluids and tissues as a measure
of oxidant stress. Methods Enzymol. 2007;433:113–126.
22. Jira W, Spiteller G, Carson W, et al. Strong increase in hydroxy
fatty acids derived from linoleic acid in human low density
lipoproteins of atherosclerotic patients. Chem Phys Lipids.
1998;91(1):1–11.
23. Janero DR. Malondialdehyde and thiobarbituric acidreactivity as diagnostic indices of lipid peroxidation and
peroxidative tissue injury. Free Radic Biol Med. 1990;9(6):
515–540.
24. Wactawski-Wende J, Schisterman EF, Hovey KM, et al.
BioCycle Study: design of the longitudinal study of the
oxidative stress and hormone variation during the menstrual
cycle. Paediatr Perinat Epidemiol. 2009;23(2):171–184.
25. Howards PP, Schisterman EF, Wactawski-Wende J, et al.
Timing clinic visits to phases of the menstrual cycle by using
a fertility monitor: the BioCycle Study. Am J Epidemiol. 2009;
169(1):105–112.
26. Schisterman EF, Vexler A, Whitcomb BW, et al. The limitations
due to exposure detection limits for regression models. Am
J Epidemiol. 2006;163(4):374–383.
27. Browne RW, Bloom MS, Schisterman EF, et al. Analytical
and biological variation of biomarkers of oxidative stress
during the menstrual cycle. Biomarkers. 2008;13(2):
160–183.
28. Milne GL, Sanchez SC, Musiek ES, et al. Quantification of
F2-isoprostanes as a biomarker of oxidative stress. Nat Protoc.
2007;2(1):221–226.
29. Browne RW, Armstrong D. HPLC analysis of lipid-derived
polyunsaturated fatty acid peroxidation products in oxidatively modified human plasma. Clin Chem. 2000;46(6):
829–836.
30. Browne RW, Armstrong D. Simultaneous determination of
polyunsaturated fatty acids and corresponding monohydroperoxy and monohydroxy peroxidation products by HPLC.
Methods Mol Biol. 2002;186:13–20.
31. Armstrong D, Browne R. The analysis of free radicals, lipid
peroxides, antioxidant enzymes and compounds related to
oxidative stress as applied to the clinical chemistry laboratory.
Adv Exp Med Biol. 1994;366:43–58.
32. Browne RW, Bloom MS, Schisterman EF, et al. Analytical and
biological variation of F2-isoprostanes during the menstrual
cycle. Clin Chem. 2009;55(6):1245–1247.
33. Craig CL, Marshall AL, Sjöström M, et al. International
Physical Activity Questionnaire: 12-country reliability and
validity. Med Sci Sports Exerc. 2003;35(8):1381–1395.
34. Spearman C. The proof and measurement of association
between two things. By C. Spearman, 1904. Am J Psychol.
1987;100(3-4):441–471.
35. Songdej N, Winters PC, McCabe MJ Jr, et al. A population-based
assessment of blood lead levels in relation to inflammation.
Environ Res. 2010;110(3):272–277.
652 Pollack et al.
36. Kasperczyk A, Kasperczyk S, Horak S, et al. Assessment of
semen function and lipid peroxidation among lead exposed
men. Toxicol Appl Pharmacol. 2008;228(3):378–384.
37. Jiun YS, Hsien LT. Lipid peroxidation in workers exposed to
lead. Arch Environ Health. 1994;49(4):256–259.
38. Kasperczyk S, Kasperczyk A, Ostalowska A, et al. Activity of
glutathione peroxidase, glutathione reductase, and lipid peroxidation in erythrocytes in workers exposed to lead. Biol
Trace Elem Res. 2004;102(1–3):61–72.
39. Devi SS, Biswas AR, Biswas RA, et al. Heavy metal status and
oxidative stress in diesel engine tuning workers of central Indian
population. J Occup Environ Med. 2007;49(11):1228–1234.
40. Muntner P, Menke A, DeSalvo KB, et al. Continued decline
in blood lead levels among adults in the United States: the
National Health and Nutrition Examination Surveys. Arch
Intern Med. 2005;165(18):2155–2161.
41. Schober SE, Sinks TH, Jones RL, et al. Blood mercury levels
in US children and women of childbearing age, 1999–2000.
JAMA. 2003;289(13):1667–1674.
42. Nälsén C, Vessby B, Berglund L, et al. Dietary (n-3) fatty acids
reduce plasma F2-isoprostanes but not prostaglandin F2a in
healthy humans. J Nutr. 2006;136(5):1222–1228.
43. Mahaffey KR, Clickner RP, Jeffries RA. Adult women’s blood
mercury concentrations vary regionally in the United States:
association with patterns of fish consumption (NHANES
1999–2004). Environ Health Perspect. 2009;117(1):47–53.
44. Mozaffarian D, Rimm EB. Fish intake, contaminants, and
human health: evaluating the risks and the benefits. JAMA.
2006;296(15):1885–1899.
45. Berntssen MH, Aatland A, Handy RD. Chronic dietary
mercury exposure causes oxidative stress, brain lesions, and
altered behaviour in Atlantic salmon (Salmo salar) parr. Aquat
Toxicol. 2003;65(1):55–72.
46. Archer KJ, Dumur CI, Taylor GS, et al. A disattenuated
correlation estimate when variables are measured with error:
illustration estimating cross-platform correlations. Stat Med.
2008;27(7):1026–1039.
47. Trichopoulou A, Costacou T, Bamia C, et al. Adherence to
a Mediterranean diet and survival in a Greek population.
N Engl J Med. 2003;348(26):2599–2608.
48. Pollack AZ, Schisterman EF, Goldman LR, et al. Cadmium,
lead, and mercury in relation to reproductive hormones and
anovulation in premenopausal women. Environ Health
Perspect. 2011;119(8):1156–1161.
49. Kaufman S, Kaufman JS, MacLehose RF, et al. Improved
estimation of controlled direct effects in the presence of
unmeasured confounding of intermediate variables. Stat Med.
2005;24(11):1683–1702.
50. Basu S, Eriksson M. Lipid peroxidation induced by an early
inflammatory response in endotoxaemia. Acta Anaesthesiol
Scand. 2000;44(1):17–23.
51. Morrow JD, Frei B, Longmire AW, et al. Increase in circulating
products of lipid peroxidation (F2-isoprostanes) in smokers.
Smoking as a cause of oxidative damage. N Engl J Med. 1995;
332(18):1198–1203.
52. Furukawa S, Fujita T, Shimabukuro M, et al. Increased
oxidative stress in obesity and its impact on metabolic
syndrome. J Clin Invest. 2004;114(12):1752–1761.
Am J Epidemiol. 2012;175(7):645–652