American Journal of Epidemiology
Copyright © 2000 by The Johns Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 152, No. 9
Printed in U.S.A.
Patterns and Determinants of Blood Lead During Pregnancy
Irva Hertz-Picciotto,1 Margaret Schramm,2 Margaret Watt-Morse,2 Kim Chantala,3 John Anderson,4 and John
Osterloh5
The pattern of blood lead during pregnancy was investigated in a cohort of 195 women who, between October
1992 and February 1995, entered prenatal care at Magee-Womens Hospital in Pittsburgh, Pennsylvania, by
week 13 of pregnancy. Blood was drawn as many as five times, once in each of the first two trimesters and a
maximum of three times in the third trimester. Blood lead determinations were made by atomic absorption
spectrophotometry. Potential sources or modifiers of lead exposure were collected by interviews, including
sociodemographic, pregnancy history, occupational, and lifestyle data. Results confirmed a previously reported
U-shaped curve in blood lead concentration during pregnancy as well as findings that blood lead levels increase
with age, smoking, lower educational level, and African-American race and decrease with history of
breastfeeding and higher intake of calcium. Additionally, interactions were found between time since last
menstrual period and both maternal age and calcium. Specifically, older mothers showed steeper increases in
blood lead concentrations during the latter half of pregnancy than did younger mothers, and intake of calcium
had a protective effect only in the latter half of pregnancy, an effect that became stronger as pregnancy
progressed. These findings provide further evidence that lead is mobilized from bone during the latter half of
pregnancy and that calcium intake may prevent bone demineralization. Am J Epidemiol 2000;152:829–37.
bone and bones; calcium; lead; lead poisoning; nutrition; pregnancy
During pregnancy, substantial quantities of calcium are
required for fetal bone growth. If maternal dietary sources
are insufficient, bone demineralization may occur, with consequent release of lead. Patterns of blood lead during pregnancy may provide clues about this process. Rothenberg et
al. (7), who enrolled about 200 Mexico City women at public prenatal clinics during their first trimester, found a
decline in blood lead between weeks 12 and 20, followed by
a rise that continued throughout the remainder of pregnancy.
The initial decline is believed to be due to pregnancyinduced plasma volume expansion (fluid mass increases
more than cell mass) (8), but the subsequent rise could be
due to either increased absorption or mobilization through
osteoclastic resorption of lead stored in bone, or both. In the
Rothenberg et al. study, blood lead concentrations ranged
from 1.0 to 35.5 µg/dl, with a geometric mean of 7.0 µg/dl.
In the present study, we sought to replicate the findings
from Mexico City (7) in a population of pregnant women in
the United States, where lower blood lead levels prevail (9,
10). By following the patterns of blood lead during pregnancy in a population with few sources of ongoing exposure
and by analyzing factors potentially related to bone stores of
lead and their mobilization, we addressed potential endogenous sources of blood lead in pregnant woman.
Although lead has a half-life of about 45 days in the
bloodstream, most of the body burden is stored in bone,
where the half-life is estimated to be decades (1). Recent
evidence suggests that lead stored in bone is not fully
sequestered but rather may become bioavailable during periods of increased bone resorption, for example, in the postmenopausal period or during pregnancy. Cross-sectional
studies show higher lead levels in post- than premenopausal
women after adjustment for age (2–4). This post- versus premenopausal difference was greater among nulliparous than
parous women, suggesting that prior pregnancies had
depleted the lead stores in this latter group, resulting in less
lead to be mobilized after menopause (2). Symanski and
Hertz-Picciotto (4) showed that the postmenopausal effect
was strongest in the first 4 years after menopause, when
bone loss is most rapid (5, 6).
Received for publication July 23, 1999, and accepted for publication January 12, 2000.
Abbreviations: BMI, body mass index; CDC, Centers for Disease
Control and Prevention; LMP, last menstrual period.
1
Department of Epidemiology, School of Public Health, University
of North Carolina, Chapel Hill, NC.
2
Department of Obstetrics and Gynecology, Magee-Womens
Hospital, University of Pittsburgh, Pittsburgh, PA.
3
Carolina Population Center, University of North Carolina,
Chapel Hill, NC.
4
Department of Nutrition, School of Public Health, University of
North Carolina, Chapel Hill, NC.
5
Department of Laboratory Medicine, University of California,
San Francisco, San Francisco, CA.
Reprint requests to Dr. Irva Hertz-Picciotto, Department of
Epidemiology, CB#7400, School of Public Health, McGavranGreenberg Hall, University of North Carolina, Chapel Hill, NC 27599
(e-mail: [email protected]).
MATERIALS AND METHODS
Study population
This study enrolled a cohort of healthy pregnant women
who initiated prenatal care during the first trimester at MageeWomens Hospital or one of two auxiliary clinics in Pittsburgh,
829
830
Hertz-Picciotto et al.
Pennsylvania, between October 1992 and February 1995.
Women were excluded if they were less than 18 years of age,
did not speak English, did not plan to carry the pregnancy to
term, were not African American or White, were past 13
weeks since their last menstrual period (LMP), or had preexisting medical conditions such as chronic hypertension, diabetes, or psychoses. A total of 753 women were enrolled initially. Of these women, 183 were excluded for eligibility
reasons, and 72 randomly selected White women were
excluded from further data collection to maintain a 50:50
racial balance, leaving a cohort of 498 (figure 1). Blood specimens were collected once in the first and second trimesters,
as many as three times in the third trimester, and at delivery.
We focused on women who completed the entire protocol,
delivered at Magee-Womens Hospital, and did not miscarry.
Within this group, we selected a subset of women (n 195)
for blood lead determinations by stratifying on calcium status
and oversampling those with a low (<1,000 mg/day) or high
(>2,000 mg/day) dietary calcium intake. Since women with a
low calcium intake would be most likely to mobilize minerals
from bone, this sampling strategy was designed to maximize
our statistical power and precision in evaluating changes in
blood lead during pregnancy. The possible drawback to this
strategy is that women whose calcium intake has been low
over a long time period might have already mobilized their
bone stores of lead, which could lead to bias and/or loss of
precision in measuring changes induced by pregnancy.
Lead determinations
Blood lead concentration was determined once in each of
the first and second trimesters and as many as three times in
the third trimester, as well as from maternal and umbilical
cord blood at delivery. Blood was drawn by venous phlebotomy into trace metal Vacutainers (Becton-Dickinson and
Company, Franklin Lakes, New Jersey) containing EDTA,
stored temporarily at 4˚C at Magee-Womens Hospital, and
sent in monthly batches via overnight delivery to the San
Francisco General Hospital Metals Laboratory, University
of California, San Francisco. This laboratory has 20 years of
experience in blood lead analysis; participates in the Centers
for Disease Control and Prevention (CDC)/American
Association of Clinical Chemists blood lead proficiency
testing program, with certification by the US Occupational
Safety and Health Administration; and analyzes hospitalbased, program-based (Childhood Lead Prevention
Programs), and research samples.
Blood lead analyses were performed by graphite furnace
atomic absorption spectrophotometry with deuterium background correction (11). Duplicate aliquots were prepared for
specimens, calibrators, and controls. Initially, each duplicate
preparation was analyzed twice and the quadruplicate
results averaged, but differences between the means of the
quadruplicate determinations and the means from the first
duplicate determinations were negligible and nonsystematic; thenceforth, we proceeded with duplicate determinations only. All samples for the same woman were analyzed
in the same run. Transfer tubes and reagent plasticware were
precleaned with low-lead acid and water washes. Samples
were prepared in a class 1,000 clean room that contained a
high-efficiency particulate air filter. Quality control blood
samples included spiked banked blood and CDC proficiency
test samples. Quality control values were assayed by using
prior repetitive analysis or CDC mean assignments. Both
FIGURE 1. Construction of the cohort of healthy pregnant women from those who initiated prenatal care at Magee-Womens Hospital in
Pittsburgh, Pennsylvania, 1992–1995.
Am J Epidemiol Vol. 152, No. 9, 2000
Blood Lead in Pregnancy
types of quality control blood samples were analyzed for
every two to six subject samples. Interassay precision was 6
percent (coefficient of variation) at 2.0 µg/dl.
Interview data
Each woman was interviewed twice by a trained interviewer: once between 8 and 18 weeks and the second time
between 31 and 41 weeks of pregnancy. The two structured
questionnaires took on average 35 and 40 minutes, respectively, to complete. Most women were interviewed in person; however, since some had already spent a long day at the
clinic, some were interviewed by telephone or the in-person
interview was stopped and later completed by telephone
(n 49 for the first and n 9 for the second interviews).
Interviewers collected information on potential sources or
modifiers of lead exposure and on factors that might confound or modify associations with maternal blood pressure,
infant weight, and length of gestation. The potential sources
or modifiers of lead exposure included age, prepregnancy
weight, reproductive and breastfeeding histories, residential
history, occupational history, use of ceramic cookware or
Grecian Formula hair care products (Combe Incorporated,
White Plains, New York), calcium intake (both dietary and
supplemental), education and income, hobbies involving
lead exposure, occupations and hobbies of persons with
whom the woman lived (past and present), smoking, and
alcohol consumption.
Questionnaires were reviewed for completeness prior to
data entry. Data collection forms (interviews, visit forms,
medical record abstractions, etc.) were entered in duplicate
and were verified. Range and consistency checks were then
performed. After thorough data cleaning, an analysis file
was created with all relevant variables.
The date of the LMP was based on self-report in the interview. If ultrasound results and the self-report disagreed by
more than 14 days, the ultrasound date was used.
Prepregnancy body mass index (BMI; kg/m2) was calculated by using self-reported height and prepregnancy
weight. For six women with missing prepregnancy weight,
we imputed values by using weight gain and prepregnancy
weight in strata based on age and race, the strongest predictors of prepregnancy weight in these data. For each woman
who was missing prepregnancy weight, we first calculated
the mean daily weight gain as of the first prenatal visit for
nonmissing women in her stratum; this value was then multiplied by days since LMP at her first prenatal visit and was
subtracted from the weight recorded at that visit.
Education was coded ordinally in three levels. “Low”
included those women aged 20 years or more who had less
than a high school education. “Medium” included those who
earned a high school diploma at any age as well as those
aged less than 20 years without one. “High” was reserved
for those women with some college education.
The occupational history requested start and stop dates
(or ages at the beginning and end of the job), job title, work
activities, and industry. This information was used to code
each position by using Standard Industry Classification
codes and codes published by the Labor Force Statistics
Am J Epidemiol Vol. 152, No. 9, 2000
831
Branch of the US Department of Health and Human
Services (12). A list of Standard Industry Classification
codes with any potential for lead exposure, obtained from
the National Institute for Occupational Safety and Health
(13), was reviewed to classify jobs and calculate the number
of months of potential job-related lead exposure for each
woman. Besides occupational history, the questionnaire
included specific queries about major occupations (battery
manufacturing, welding, etc.) and hobbies (jewelry making,
etc.) known to involve lead exposure; the woman was asked
whether she or anyone with whom she lived had ever
worked in these occupations or engaged in these hobbies.
We created a total number of such jobs and/or hobbies 1) for
her and 2) for others with whom she lived.
A food frequency table, which included portion sizes, was
used at both interviews to obtain women’s current sources of
calcium: these items (table 1) contain a substantial portion
of the calcium in the adult US diet (14). Dietary intake of
calcium was calculated by using DIETSYS version 3.0 software developed by Block et al. (15). Supplemental calcium
was determined from the weekly intake of prenatal vitamins,
calcium supplements, and calcium-containing antacids. The
calcium content of specific brands reported was estimated
by using manufacturers’ product information from the
1992–1995 editions of the Physicians’ Desk Reference (16).
(A complete listing by year, brand name, and product is
available from the authors or at their Web site:
http://www.cpc.unc.edu/projects/lead/). Thus, daily calcium
intake from the diet, antacids, calcium supplements, and
prenatal vitamins was calculated. Indicators of lifetime calcium intake were constructed from self-reported milk consumption during childhood, adolescence, previous pregnancies, and lactational periods and from lifetime consumption
of antacids and calcium supplements.
TABLE 1. Food items* used to assess dietary calcium intake
in a cohort of pregnant women in Pittsburgh, Pennsylvania,
1992–1995
Whole milk, 2% milk, 1% or skim milk
Flavored yogurt
Hard cheeses, medium cheeses, soft cheeses
Ice cream
Eggs
Doughnuts, cookies, cake
White bread, rolls and crackers
Whole wheat, rye, and other dark breads
Cornbread, grits, tortillas
Pizza
Mixed dishes with cheese
Small fish (with bones): sardines, salmon
Okra
Greens (mustard, turnip, collard, beet)
Broccoli
* Listed items from a food frequency table represent the major
contributors to calcium in the US diet.
832
Hertz-Picciotto et al.
Statistical analyses
Initially, we conducted simple bivariate linear regressions
for each visit to screen our large pool of potential predictors
of blood lead. Those factors showing some association with
the outcome (broadly defined as a p value of <0.25 for more
than two visits) and for which data were complete were then
selected for possible inclusion in multivariate models.
Factors that met this criterion included mother’s age,
prepregnancy BMI, smoking, caffeine intake, total calcium
intake, race, education, alcohol consumption, and number of
servings of canned foods per month.
Blood lead values were plotted against week of pregnancy to enable visual inspection of the longitudinal pattern.
The resulting graph suggested a nonlinear relation; in
regression models, we therefore examined splines as well as
a simple quadratic term for time since LMP.
We next fit longitudinal models predicting all blood lead
measurements taken before labor, which included as many
as five determinations for each woman. A longitudinal, multivariate fixed effect model was fit by using all factors that
met our inclusion criteria (described above), along with time
since LMP. Longitudinal analysis accounting for intraindividual correlation was implemented with SAS software (17)
by using both the PROC MIXED and PROC GENMOD
procedures, the latter of which uses generalized estimating
equations. As both procedures produced similar results, further model building was conducted with generalized estimating equations. A marginal (or “population-average”)
regression model was selected because we were not interested in estimating subject-specific parameters.
Several different covariance structures were examined.
The two with Akaike’s Information Criterion values indicating the best fit were the autoregressive first order and the
exchangeable (sometimes referred to as compound symmetry) covariances. Since prenatal visits were not spaced
equally, the autoregressive structure (in which the covariances decline rapidly for nonadjacent visits) seemed less
plausible; therefore, exchangeable covariances were used.
Variances were assumed to be homoscedastic. As residuals
for some visits did not show Normality, two transformations
of blood lead were tried: square root and logarithmic. The
square root transformation yielded Normal errors, but since
the predictive importance of the risk factors was virtually
the same regardless of transformation, this paper reports
results based on the log transformation. This choice was
based on interpretability; it allows presentation of the percentage change in blood lead level for a given increase in the
predictor variable. For comparison, we also present the
model with untransformed blood lead values.
Nonlinear relations were explored by using splines and
other transformations of the predictor variables. Interactions
were also examined. Because of the changing kinetics of
lead during pregnancy, we hypothesized that the predictors
of lead in the early half of pregnancy would differ from the
predictors of blood lead concentrations later in pregnancy.
Thus, for factors thought to affect the rate of maternal bone
resorption in response to fetal skeletal growth, we constructed interactions with time since LMP. A comparison
across the five visits (pregnancy weeks 6–13, 14–27, 28–32,
33–36, and ≥37) of coefficients in cross-sectional bivariate
regressions, as well as inspection of splines, suggested interactions of time since LMP with mother’s age at enrollment
and with total calcium intake.
Models were run with and without two blood lead values
outside the main distribution—8.45 and 5.70 µg/dl—which,
based on published norms (9), were likely to be true measurements. This paper presents results for models that
included these outliers, but their exclusion had minimal
influence on parameter estimates.
RESULTS
Those women who did not complete the study, that is, who
miscarried, dropped out, or delivered elsewhere (n 129),
were slightly more likely to be African American than White
(55 vs. 49 percent) and were younger and hence less likely to
have completed high school than those who did complete the
study. Interview data for a subset of women who dropped out
or delivered elsewhere indicated a smaller percentage who
drank alcohol as compared with completers but no differences with regard to smoking, previous pregnancy history,
breastfeeding, income, employment, time since LMP at entry
into the study, and body mass. Among those who completed
the study, the total cohort differed little from the subset for
whom lead determinations were conducted (table 2). The
main difference was that the subgroup with lead determinations had a higher calcium intake because we oversampled
those at both extremes of this distribution.
As shown in figure 2, 878 measurements from 195 pregnant women indicated low blood lead concentrations relative to levels at which human health effects have been
demonstrated (for instance, above 10 µg/dl, lead poses a risk
to cognitive development in children (18)). Two measurements of more than 5 µg/dl are not shown in this figure. A
U-shaped curve characterizes blood lead level as a function
of weeks of pregnancy; this curve was also present when the
outlying observations were included.
The best predictive models are presented in table 3. Each
coefficient for model 1 and model 2 represents the change in
the blood lead level and the change in the log blood lead
level, respectively, in µg/dl for a one-unit increase in the x
variable, unless otherwise indicated. The last column shows
the percentage change in blood lead for a specified increase
in the x variable based on model 2. As shown on the graph
(figure 2), time since LMP at blood draw was associated
nonlinearly with blood lead level. Splines were initially used
for time since LMP, but a quadratic term centered at 20
weeks was found to give an equally good fit with fewer
parameters. Mother’s age at enrollment in the study also
modified the effect of time since LMP; therefore, a product
term was introduced to reflect this relation: the older a
woman was, the steeper the increase in blood lead as pregnancy progressed through the last trimester (figure 3); concomitantly, the later in pregnancy, the greater the difference
associated with older maternal age. For example, from week
20 to week 40 of pregnancy, blood lead levels in women
with a low calcium intake increased 25 percent at age 18
Am J Epidemiol Vol. 152, No. 9, 2000
Blood Lead in Pregnancy
TABLE 2. Sociodemographic and lifestyle characteristics (%)
of pregnant women who completed the full study protocol
and of the subset for whom blood lead values were
determined, Pittsburgh, Pennsylvania, 1992–1995
Completers
(n = 369)
Lead
group*
(n = 195)
Age (years)
<20
>30
19
14
20
13
Education: <12th grade
17
15
Race: African American
48
50
Pregnancies
First
Previous spontaneous abortion
24
24
25
22
Ever breastfed
16
15
Smoking
Ever
Current
55
31
58
30
Drank alcohol before pregnancy
75
82
Household income ($)
<10,000
10,000–25,000
>25,000
49
36
15
50
34
16
32
27
Body mass index (kg/m )
<20
20–<24
24–<28
≥28
12
36
21
31
11
40
19
30
Calcium intake (mg/day)
≤600
>600–1,000
>1,000–2,000
>2,000
11
17
45
26
17
22
19
42
Entered study by gestational week 9
50
47
Characteristic
Currently employed
2
* Women whose dietary calcium intakes were high or low were
oversampled.
years, 37 percent at age 23 years, 65 percent at age 33 years,
and 99 percent (i.e., a doubling) at age 43 years.
Low educational level was associated with not only
higher blood lead levels but also a higher prevalence of
smoking (data not shown); when either of these variables
was removed, the standard error for the coefficient of the
remaining one was smaller. We left both factors in the model
since lead is established to be present in cigarette smoke,
and education is likely to be a surrogate for other factors
besides smoking that are associated with lead. AfricanAmerican women had higher blood lead levels than White
women did. Despite higher BMI among African-American
women, their higher blood level was independent of any
association between BMI and blood lead. In fact, blood lead
increased with BMI only for women whose BMI was less
than 24 kg/m2, which falls into the category of normal body
weight for height. No association with blood lead was found
Am J Epidemiol Vol. 152, No. 9, 2000
833
when prepregnancy BMI was in the overweight or obese
range.
Total number of months of breastfeeding infants from
previous pregnancies was inversely associated with blood
lead. The magnitude of the coefficient was unchanged when
we removed from analysis two women with over 40 months
of breastfeeding. Although bivariate analyses suggested a
similar association of parity with blood lead, the high correlation of parity and months of breastfeeding made it difficult
to distinguish independent contributions from these two
variables.
Higher calcium intake was associated with lower blood
lead levels in the period from week 20 to the end of pregnancy. Stepwise, progressive differences were seen in four
groups: from ≤600 mg/day, to >600–1,000 mg/day, to
>1,000–2,000 mg/day, and to >2,000 mg/day (data not
shown; ordinal coding was adopted after monotonic and
similar increases were established across these four categories). Thus, even well above the Recommended Dietary
Allowance for calcium, some benefit was observed during
the latter half of pregnancy. The contrast in blood lead levels between the highest and lowest calcium intakes is shown
in figure 3 for African-American women aged 18 and 38
years, with mean, middle, or modal levels of other variables.
As shown in this figure, calcium intake, maternal age, and
time since LMP all influenced blood lead levels even at
these low concentrations. Somewhat smaller effects are
shown in table 4, which presents predicted blood lead concentrations according to smoking status, education, breastfeeding, BMI, and race. Thirty months of previous breastfeeding had a greater impact than smoking, race, or a
20-unit change in BMI.
Higher caffeine consumption was also predictive of
increased blood lead levels (model not shown) and was
associated with smoking (positively) and with race (African
Americans consumed far less caffeine). As a result, inclusion of caffeine decreased the smoking coefficient and
increased the race coefficient.
DISCUSSION
Among studies examining lead over the course of pregnancy, many have not reported lead measurements for the
same women or for all three trimesters (19–22). Knight et al.
(23) measured blood lead throughout pregnancy, but not all
women contributed data to all three trimesters. Rothenberg
et al. (7), who did include measurements in all three
trimesters for the same women in the cohort from Mexico
City, demonstrated a clear U-shaped curve, with a nadir at
20 weeks. Swedish data for a similar-sized cohort showed
lower blood levels but included measurements at only two
time points before delivery: week 10 and week 32 (20, 21).
According to our curves and those of Rothenberg et al.,
these two data points would be expected to show similar
levels and, in fact, they do.
The women in the Mexico City study (7) had substantial
concurrent external exposures from leaded gasoline and use
of glazed ceramic cookware fired at low temperatures. Since
the pregnant women we studied had no ongoing substantial
834
Hertz-Picciotto et al.
FIGURE 2. Scatterplot of 892 blood lead (Pb-blood) measurements taken at different times since last menstrual period from 195 pregnant
women, Magee-Womens Hospital, Pittsburgh, Pennsylvania, 1992–1995. For each woman, blood lead levels were determined from three to five
different blood drawings (after exclusion of values with missing reports on calcium intake, the final model included 3 women with two, 9 women
with three, 56 women with four, and 127 women with five lead measurements). Two values of blood lead concentrations >5 µg/dl are not shown
but were included in the models.
exposure to environmental sources of lead, increased
absorption of lead from the environment would likely have
played a smaller role. Thus, the fact that a previously
observed nonlinear relation over the course of pregnancy
was confirmed for US women in the 1990s appears to support a role for mobilization of lead stored in the mineral
matrix of the skeleton. Since lead has long been known to
cross the placenta and to accumulate in numerous fetal
organs, including the brain, in proportion to the growth of
those organs (24), endogenous maternal lead could pose a
risk to early fetal development.
Gulson et al. (25) estimated the contribution of skeletal
lead to blood lead, based on isotopic characterization of
blood lead samples, in 13 recent immigrants to Australia
from eastern Europe. Several isotopes of lead are present in
the environment, and their distribution in the body reflects
the sources of that person’s body burden. As eastern
European lead sources have a higher ratio of 206Pb to 204Pb
(25), the relative change in the isotopic ratio from a prepregnancy blood sample to samples taken during pregnancy may
provide information about the contribution of skeletal lead
(representing long-term exposure) to blood lead. On the
basis of a population whose blood lead levels were similar
to those in our population, Gulson et al. estimated that about
31 percent (range, 13–65 percent) of the changes in blood
lead levels during pregnancy were due to skeletal stores of
lead.
Other evidence from our analysis may support the interpretation that the late pregnancy rise in blood lead results, at
least partially, from transfer of lead from bone into the circulating blood. For example, the greater effect of time since
LMP on blood lead among older mothers would logically
follow, since those born earlier were exposed to leaded
gasoline for more years of their lives. While older women
might preferentially absorb more lead than younger women
do, we are not aware of toxicokinetics data supporting this
type of effect in women aged 35–44 years. Estimates suggest that if absorption of dietary lead doubled, blood lead
would increase about 10 percent (0.2 µg over a 2 µg level)
(25); the increase we observed between weeks 20 and 40 of
pregnancy was of greater magnitude in women more than 35
years of age (figure 3).
Another of our findings, the protection associated with
increased calcium intake during the latter half of pregnancy,
is consistent with the hypothesized bone kinetics, although it
could also reflect decreased absorption of lead. In lactating
women, calcium excretion was reduced and bone resorption
was increased, while intestinal absorption did not increase
(26). The protective effect we observed from breastfeeding of
previous infants may have resulted from partial unloading of
bone stores of maternal lead during prior pregnancies and lactation, thus depleting the reservoir from which to draw lead
during the current pregnancy. However, a recent study of lactation showed neither a change in blood lead during the first
6 months postpartum nor any association between bone
changes and blood lead changes (27). In contrast, immigrants
to Australia appear to show greater mobilization of skeletal
lead during lactation than during late pregnancy (28).
Am J Epidemiol Vol. 152, No. 9, 2000
Blood Lead in Pregnancy
TABLE 3.
835
Final regression models for prediction of lead and log(blood lead) in pregnancy*,†
Variable
Intercept§
Model 1: Prediction of
blood lead:
β (SE‡)
Model 2: Prediction of
log(blood lead):
β (SE)
% Change
in blood
lead
1.67 (0.10)
0.41 (0.06)
0.37 (0.12)
0.19 (0.06)
(Gestational age¶)
0.023 (0.004)
0.013 (0.003)
(Figure 3)
Interaction: (woman’s age) ×
(gestational age)2
0.014 (0.005)
0.007 (0.003)
(Figure 3)
Smoker (vs. nonsmoker)
0.18 (0.09)
0.11 (0.05)
12.0
–0.12 (0.07)
–0.066 (0.041)
–6.4
0.16 (0.09)
0.11 (0.05)
11.5
Woman’s age (per 10 years, at 20
weeks’ gestation)
2
Education (3 levels# adjusted for
age)
Black race (vs. White)
21.0
Body mass index (increase per
kg/m2)
≤24
>24
0.055 (0.027)
–0.002 (0.011)
0.035 (0.022)
0.000 (0.011)
4.0
0.0
Total months breastfed (for each
6 months)
–0.082 (0.038)
–0.045 (0.027)
–4.0
0.003 (0.034)
–0.006 (0.022)
–1.0
–0.028 (0.013)
–0.019 (0.008)
(Figure 3)
Calcium** (per level, at 20 weeks)
Interaction: (calcium after week
20) × (gestational age)
* Based on a population-average longitudinal model; fit by using the PROC GENMOD procedure in SAS
software (SAS Institute, Inc., Cary, North Carolina).
† Each beta coefficient represents the change in the blood lead level (model 1) or the log blood lead level
(model 2) (µg/dl) for a one-unit increase in the x variable, unless otherwise indicated.
‡ SE, standard error.
§ Blood lead level (model 1) or log of the blood lead level (model 2) (µg/dl) at 20 weeks’ gestation for a White
nonsmoker aged 23 years with a body mass index of 24, who is at the lowest educational level, has never
breastfed, and currently consumes ≤600 mg/day of calcium.
¶ Centered at 20 weeks of gestation and coded as the number of 28-day months from week 20 for both the
main effect and its interactions with woman’s age and calcium intake after week 20. This coding, for a quadratic
centered variable, describes the U-shaped relation with gestational age; the positive coefficient for the interaction
with maternal age indicates an increasingly steep U as women age (interaction depicted in figure 3), representing
a stronger association of gestational age with blood lead in older compared with younger women.
# Coded as follows: 0, did not complete high school and aged ≥20 years; 1, either completed high school or
aged <20 years and did not complete high school; 2, some college education.
** Coded ordinally at four levels: ≤0.6, >0.6–1, >1–2, >2 g/day (interaction depicted in figure 3). This
interaction indicates that, for example, at 36 weeks of gestation, blood lead level is 23% lower for a calcium intake
of >2 g/day compared with ≤0.6 g/day.
We attempted to explore whether some part of the increase
in blood lead after week 20 of pregnancy could be an artifact
associated with an increase in hematocrit, but very few
women had hematocrit measurements toward the end of
pregnancy. Using the data available, which included an early
third trimester measurement (weeks 28–32) for about half of
the 195 women, we found that adjustment for hematocrit did
not alter the substantive findings. Interestingly, a study of
primates demonstrated a reduction in first trimester bone
mobilization (29). If applicable to humans, this result would
imply that not all of the blood lead decrease in the early half
of pregnancy is attributable to the increase in plasma volume.
This study, like all epidemiologic investigations, has
some limitations. Many of our variables were based on selfreports. Thus, the lack of association between blood lead
and BMI above the value of 24 may be a result of poor qualAm J Epidemiol Vol. 152, No. 9, 2000
ity data for prepregnancy weight if women who weighed
less provided more accurate reports than heavier women
did. Calcium intake was also measured with error, since
food frequency information is only approximately correct.
Nevertheless, the broad categories we created for calcium
intake likely provided a reasonable rank-ordering of subjects (particularly given the wide range of intake), thus
allowing estimation of an impact on blood lead. Another
limitation stems from the fact that few women reported
drinking alcohol during pregnancy, using Grecian Formula
hair care products, having had occupations or hobbies that
could have involved lead exposure, or living with persons
who had such occupations or hobbies; consequently, our
power to detect associations with these factors was very low.
Our results confirm established predictors of blood lead
in nonoccupationally exposed populations, including
836
Hertz-Picciotto et al.
FIGURE 3. Fitted curves for blood lead (Pb-blood) levels of pregnant women aged 18 and 38 years who initiated prenatal care at MageeWomens Hospital in Pittsburgh, Pennsylvania, in 1992–1995 and whose calcium intake levels were high (>2,000 mg/day) or low <600 mg/day).
Values of other variables were held constant, as follows: education, middle level (completed high school or aged <20 years and did not complete high school); body mass index (kg/m2, 24); nonsmoker; race, African American; did not breastfeed previously.
smoking, lower socioeconomic status, African-American
race, and older age (4, 9, 10, 30–32). Our ability to
observe these associations even at very low exposures
supports the quality of our blood lead determinations and
confirms that high-quality blood lead determinations can
distinguish real trends from noise even at today’s low
exposures. It also indicates that even as exogenous sources
of lead exposure have declined, known risk factors continue to have measurable effects. Another point of concern
is that while breastfeeding infants from previous births
was protective for the mother and hence for the fetus of
the current pregnancy, the implication for her previous
children, who absorbed some of her lead burden, is one of
potential harm.
In conclusion, we confirmed a U-shaped pattern of blood
lead concentration across pregnancy. Our research adds further evidence that lead stored in the skeleton may be accessible during pregnancy, particularly and increasingly during
the third trimester. The associations we observed with smoking, socioeconomic status, and age suggest that much of the
so-called background levels of lead may be preventable.
Finally, our study indicates that calcium intake may provide
some protection not only near the Recommended Dietary
Allowance but well above it.
TABLE 4. Predicted blood lead value* at 36 weeks’ gestation, by education, smoking status, months of previous breastfeeding,
race, and body mass index, for a cohort of pregnant women in Pittsburgh, Pennsylvania, 1992–1995†,‡
Months
breastfed
White race
African-American race
Educational
level
Smoking
status
Low
High
Low
High
Nonsmoker
Nonsmoker
Smoker
Smoker
0
0
0
0
1.8
1.6
2.0
1.7
(1.5,
(1.4,
(1.7,
(1.5,
2.1)
1.8)
2.3)
2.0)
2.1
1.8
2.3
2.0
(1.7,
(1.6,
(2.0,
(1.8,
2.4)
2.0)
2.6)
2.3)
2.1
1.8
2.3
2.0
(1.7,
(1.5,
(1.9,
(1.7,
2.5)
2.1)
2.8)
2.4)
2.0
1.7
2.2
1.9
(1.7,
(1.5,
(1.9,
(1.6,
2.4)
2.0)
2.6)
2.3)
2.3
2.0
2.6
2.2
(2.0,
(1.8,
(2.2,
(2.0,
2.7)
2.2)
2.9)
2.5)
2.3
2.0
2.6
2.2
(1.9,
(1.7,
(2.1,
(1.9,
2.8)
2.3)
3.0)
2.6)
Low
High
Low
High
Nonsmoker
Nonsmoker
Smoker
Smoker
30
30
30
30
1.4
1.2
1.6
1.4
(1.1,
(1.0,
(1.2,
(1.1,
1.9)
1.6)
2.1)
1.8)
1.6
1.4
1.8
1.6
(1.2,
(1.1,
(1.4,
(1.2,
2.2)
1.9)
2.4)
2.1)
1.6
1.4
1.8
1.6
(1.2,
(1.1,
(1.3,
(1.2,
2.3)
1.9)
2.5)
2.2)
1.6
1.4
1.8
1.6
(1.2,
(1.1,
(1.3,
(1.2,
2.1)
1.8)
2.3)
2.0)
1.8
1.6
2.0
1.8
(1.4,
(1.3,
(1.5,
(1.4,
2.4)
2.0)
2.7)
2.3)
1.8
1.6
2.0
1.8
(1.3,
(1.2,
(1.5,
(1.3,
2.5)
2.1)
2.8)
2.4)
BMI§ = 20
BMI = 30
BMI = 40
BMI = 20
BMI = 30
BMI = 40
* Values are expressed as blood lead level in µg/dl (95% confidence interval).
† Age is held at 28 years and calcium intake at 1,000–2,000 mg/day.
‡ Predicted values are based on the model for log blood lead shown in table 3.
§ BMI, body mass index (kg/m2).
Am J Epidemiol Vol. 152, No. 9, 2000
Blood Lead in Pregnancy
ACKNOWLEDGMENTS
This work was supported by grant 1 R01-ES05738 from
the National Institute of Environmental Health Sciences.
16.
17.
18.
REFERENCES
1. Rabinowitz M. Toxicokinetics of bone lead. Environ Health
Perspect 1991;91:33–7.
2. Silbergeld E, Schwartz J, Mahaffey K. Lead and osteoporosis:
mobilization of lead from bone in postmenopausal women.
Environ Res 1988;47:79–94.
3. Grandjean P, Nielsen G, Jorgensen P, et al. Reference intervals
for trace elements in blood: significance of risk factors. Scand
J Clin Lab Invest 1992;52:321–37.
4. Symanski E, Hertz-Picciotto I. Blood lead levels in relation to
menopause, smoking, and pregnancy history. Am J Epidemiol
1995;141:1047–58.
5. Elders PJ, Netelenbos JC, Lips P, et al. Accelerated vertebral
bone loss in relation to the menopause: a cross-sectional study
on lumbar bone density in 286 women of 46 to 55 years of age.
Bone Miner 1988;5:11–19.
6. Nilas L, Christiansen C. Rates of bone loss in pregnant
women: evidence of accelerated trabecular bone loss after the
menopause. Eur J Clin Invest 1988;18:529–34.
7. Rothenberg S, Karchemer S, Schnaas L, et al. Changes in serial blood lead levels during pregnancy. Environ Health
Perspect 1994;102:876–80.
8. Taylor D, Lind T. Red cell mass during and after normal pregnancy. Br J Obstet Gynaecol 1979;86:364–70.
9. Pirkle J, Brody D, Gunter E, et al. The decline in blood lead
levels in the United States. The National Health and Nutrition
Examination Surveys (NHANES). JAMA 1994;272:284–91.
10. Pirkle J, Kaufmann R, Brody D, et al. Exposure of the US population to lead, 1991–1994. Environ Health Perspect 1998;
106:745–50.
11. Miller D, Paschal D, Gunter E, et al. Determination of lead in
blood using electrothermal atomisation atomic absorption
spectrometry with a L’vov platform and matrix modifier.
Analyst 1987;112:1701–4.
12. 1990 census of population and housing, alphabetical index of
industries and occupations. Washington, DC: US Department of
Commerce, Economics and Statistics Administration, Bureau of
the Census, January 1992. (Publication no. 1990 CPH-R-3).
13. Unpublished provisional data as of 7/1/90. NIOSH, National
Occupational Exposure Survey (1981–83). Cincinnati, OH:
US Department of Health and Human Services, Public Health
Service, Centers for Disease Control, National Institute for
Occupational Safety and Health, Division of Surveillance,
Hazard Evaluations and Field Studies, Surveillance Branch,
Hazard Section. (Telephone number: 513-841-4491).
14. Cummings SR, Block G, McHenry K, et al. Evaluation of two
food frequency methods of measuring dietary calcium intake.
Am J Epidemiol 1987;126:796–802.
15. Block G, Hartman AM, Dresser CM, et al. A data-based
Am J Epidemiol Vol. 152, No. 9, 2000
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
837
approach to diet questionnaire design and testing. Am J
Epidemiol 1986;124:453–69.
Physicians’ desk reference for nonprescription drugs. Oradell,
NJ: Medical Economics Company.
SAS Institute, Inc. SAS/STAT software: changes and enhancements through release 6.12. Cary, NC: SAS Institute Inc, 1997:
1167.
Preventing lead poisoning in young children. A statement by
the Centers for Disease Control—October 1991. Washington,
DC: US Department of Health and Human Services, Public
Health Service, Centers for Disease Control, 1991.
Bonithon-Kopp C, Huel G, Grasmick C, et al. Effects of pregnancy on the inter-individual variations in blood levels of lead,
cadmium and mercury. Biol Res Pregnancy Perinatol 1986;7:
37–42.
Lagerkvist B, Ekesrydh S, Englyst V, et al. Increased blood
lead and decreased calcium levels during pregnancy: a
prospective study of Swedish women living near a smelter. Am
J Public Health 1996;86:1247–52.
Lagerkvist B, Soderberg HA, Nordberg G, et al. Biological
monitoring of arsenic, lead and cadmium in occupationally and
environmentally exposed pregnant women. Scand J Work
Environ Health 1993;19(supp 11):50–3.
McMichael AJ, Vimpani GV, Robertson EF, et al. The Port
Pirie cohort study: maternal blood lead and pregnancy outcome. J Epidemiol Community Health 1986;40:18–25.
Knight E, Spurlock B, Edwards C, et al. Biochemical profile of
African American women during three trimesters of pregnancy
and at delivery. J Nutr 1994;124:943S–53S.
Barltrop D. Transfer of lead to the human foetus. In: Barltrop
D, Burland WL, eds. Mineral metabolism in paediatrics. A
Glaxo symposium. Oxford, England: Blackwell Scientific
Publications, 1969:135–51.
Gulson B, Jameson C, Mahaffey K, et al. Pregnancy increases
mobilization of lead from maternal skeleton. J Lab Clin Med
1997;130:51–62.
Specker B, Vieira N, O’Brien K, et al. Calcium kinetics in lactating women with low and high calcium intakes. Am J Clin
Nutr 1994;59:593–9.
Osterloh J, Kelly T. Study of the effect of lactational bone loss
on blood lead concentrations in humans. Environ Health
Perspect 1999;107:187–94.
Gulson B, Mahaffey K, Jameson C, et al. Mobilization of lead
from the skeleton during the postnatal period is larger than during pregnancy. J Lab Clin Med 1998;131:324–9.
Franklin CA, Inskip MJ, Baccanale CL, et al. Use of sequentially administered stable lead isotopes to investigate changes
in blood lead during pregnancy in a nonhuman primate
(Macaca fascicularis). Fundam Appl Toxicol 1997;39:109–19.
Mahaffey K, Annest J, Roberts J, et al. National estimates of
lead blood levels: United States, 1976–1980: associated with
selected demographic and socioeconomic factors. N Engl J
Med 1982;307:573–9.
Blood lead levels for persons ages 6 months–74 years: United
States, 1976–80. Vital Health Stat 11 1984;11:1–59.
Weyermann N, Brenner H. Factors affecting bone demineralization and blood lead levels of postmenopausal women—a
population-based study from Germany. Environ Res 1998;
76A:19–25.