https://orcid.org/0000-0002-6082-5683
Figures
Abstract
Current child blood lead (Pb) screening guidelines assume that blood lead levels (BLLs) are relatively stable over time, and that only youngest children are vulnerable to the damaging effects of lower-range BLLs. This study aimed to test the stability of lower-range (≤ 10 μg/dL) child BLLs over time, and whether lower-range BLLs diminished with age among children aged 6 months to 16 years living in a lower-income neighborhood with a density of pre-1986 housing and legacy contamination. Age, sex, family income, age of residence, and/or residence proximity to point sources of Pb, were tested as potential additional factors. Capillary blood samples from 193 children were analyzed by inductively coupled plasma mass spectrometry (ICPMS). Multiple imputation was used to simulate missing data for 3 blood tests for each child. Integrated Growth Curve models with Test Wave as a random effect were used to test BLL variability over time. Among N = 193 children tested, at Time 1 testing, 8.7% had the BLLs ≥ 5 μg/dL (CDC “elevated” BLL reference value at the time of data collection) and 16.8% had BLLs ≥ 3.5 μg/dL (2021 CDC “elevated” BLL reference value). Modeling with time as a random effect showed that the variability of BLLs were attributable to changes within children. Moreover, time was not a significant predictor of child BLLs over 18 months. A sex by age interaction suggested that BLLs diminished with age only among males. Of the additional environmental factors tested, only proximity to a major source of industrial or vehicle exhaust pollution predicted child BLL variability, and was associated with a small, but significant BLL increase (0.22 μg/dL). These findings suggest that one or two BLL tests for only infants or toddlers are insufficient for identifying children with Pb poisoning.
Citation: Del Rio M, Rodriguez C, Alvarado Navarro E, Wekumbura C, Galkaduwa MB, Hettiarachchi GM, et al. (2023) Stability of blood lead levels in children with low-level lead absorption. PLoS ONE 18(6): e0287406. https://doi.org/10.1371/journal.pone.0287406
Editor: Shailja Sharma, All India Institute of Medical Sciences - Jodhpur, INDIA
Received: December 12, 2022; Accepted: June 5, 2023; Published: June 23, 2023
Copyright: © 2023 Del Rio et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: A minimal anonymized data set necessary to replicate study findings along with SPSS syntax for modeling longitudinal models can be found in the Supporting Information. For more additional information or questions contact Michelle Del Rio at midelrio@iu.edu.
Funding: Funding for this research came from the U.S. Department of Housing and Urban Development (HUD), Grant # TXLTS0010-18 (to CS & GH), the U.S. Environmental Protection Agency (EPA)(to CS). The study was also supported by funds from the University of Texas at El Paso (UTEP) Center for Environmental Resource Management, the UTEP Border Biomedical Research Center (BBRC)(to CS), the National Institutes of Health, National Institute for Child Health and Human Development (R21HD060120)(to CS), and the J. Edward and Helen M.C. Stern Professorship in Neuroscience (to CS), UTEP. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. URLs of each funder website: U.S. Department of Housing and Urban Development: https://www.hud.gov/program_offices/healthy_homes/lbp/lts U.S. Environmental Protection Agency: https://www.epa.gov/ UTEP’s Border Biomedical Research Center: https://www.utep.edu/science/bbrc/ UTEP’s Center for Environmental Resource Management: https://www.utep.edu/cerm/ National Institutes of Health, National Institute for Child Health and Human Development: https://www.nichd.nih.gov/.
Competing interests: The authors have declared that no competing interests exist.
Introduction
There is scientific and clinical consensus that no level of lead (Pb) exposure is “safe” for children [1]. Even lowest-level childhood Pb exposure is associated with irreversible adverse health effects including poorer academic performance [2–5], lowered IQ [6, 7], stunted growth [8, 9], and increased behavioral problems [10–13]. Despite knowledge of the permanent health consequences of childhood Pb exposure, more than half of a million U.S. children continue to be exposed to dangerous lower-level Pb [14], and children living in lower-income and predominantly Black and/or Hispanic neighborhoods are disproportionately affected [15–20]. Based on frequencies of individual and community risk factors, it has been recently estimated that approximately 50% of U.S. children below the age of 6 have detectable levels of Pb in their blood [21]. Of note, the actual child BLL data needed to inform this estimate do not exist.
Before a public health problem can be solved, its incidence and prevalence must be correctly characterized. With regard to child Pb exposure, implicit assumptions that underlie current testing policies may be undermining our best efforts. In most states, screening guidelines test children once or twice at ages 1 and 2 years; in many fewer states, testing recommendations extend to children up to the age of 6 years. Despite multiple recalls, point-of-care devices are still widely used for one-time screening. When a child’s BLL is determined to be below a locally designated level, Pb exposure risk is assumed to be low, and children may never be tested again. If children’s BLLs in fact vary significantly over time, and do not decrease with age, these screening guidelines cannot be expected to adequately detect exposed children; from a public health perspective, our understanding of the scope of the problem is necessarily incomplete.
In a recent paper [22] we comprehensively reviewed current knowledge regarding the complexity of Pb absorption, transport, and disposition in children’s bodies, how these would be expected to produce variable BLLs over time, and how these factors can continue to influence BLLs through adolescence. Moreover, these mechanisms differ broadly for inhaled as compared to ingested Pb. For example, when Pb is inhaled, transport from the lungs into the circulatory system depends on many shifting factors including Pb particulate size, length of exposure, frequency of exposure, age-based respiratory rate, where in the lungs Pb particulates are deposited, and importantly, individual differences in the development of the lungs and alveoli [23–25]. On the other hand, when Pb is ingested, markedly different chemical, biological, bio-physiochemical factors in the gut, and maturity of the GI tract, as well as behavioral factors related to dietary intake, dietary deficiencies, and hand-to-mouth and play habits, influence absorption and thus levels of Pb detected in blood samples [26–28]. Further complicating predictions, when Pb-contaminated household dust is inhaled as well as ingested–a plausible scenario–absorption mechanisms can be additive, antagonistic, and/or synergistic. Finally, in the bloodstream, it is well established that absorption into red blood cells is influenced by individual differences in genetic predisposition, specifically, common (ALAD and HPEPT2) genetic variants [18]; and by normal fluctuations in iron, calcium, and zinc levels [29]. Additional age dependent factors can facilitate or oppose the distribution of Pb into tissues, as well as re-release of Pb into the circulatory system.
In these ways, current biological evidence suggests the very high likelihood that children’s BLLs should be assumed to vary over time. Very few if any studies, however, have focused attention on this question. This study aimed to test the stability of BLLs over time in a sample of children with BLLs < 10 μg/dL. Possible risk factors contributing to BLL variability were also examined including age, sex, living in a family with income at or below the U.S. poverty line, living in a home built before 1986; and living near point sources of Pb contamination. Given the extant biological evidence discussed above, we hypothesized that BLLs would vary significantly within children; that Pb exposure and Pb exposure risk would not diminish with age; and that older children would not differ from younger children in their vulnerability to Pb exposure.
Materials and methods
Ethics and consents
The data here presented were collected for a large research program that aimed to improve child Pb surveillance and home mitigation in high-risk neighborhoods in El Paso, TX. All methods, procedures, forms, and materials for the research program were reviewed and approved annually by The Institutional Review Board of the University of Texas, El Paso (# 1309985). Written informed consent to participate was obtained from parents prior to the study; written assent from each child was obtained immediately prior to BLL testing. All study forms and materials were available in Spanish and English versions. Researchers in this study were bilingual and throughout the study interacted with participants in their preferred language.
Study design
This was an analytical, observational, prospective longitudinal study using a convenience sample of children ages 6 months to 16 years old. Data were collected in neighborhoods deemed at “high-risk” for child Pb exposure due to possible legacy contamination, lower-socioeconomic income level, and with a preponderance of homes built before 1986. (While the federal residential Pb-paint law was enacted in 1978, leaded pipes and Pb solder continued to be used until 1986.) The community was predominantly of Hispanic or Latinx descent (82.2%), with 55.5% of children living below the U.S. poverty threshold. Previous studies of children in these neighborhoods have shown relatively high proportions of children with elevated BLLs [18–20, 30]. As of 2020, 21 of 36 zip codes in this region continued to be areas targeted by the Texas Department of State Health Services for child Pb screening [31]. The research program was conducted between January 2018 and March 2020, ending prematurely due to COVID shutdowns.
Recruitment, data collection, and analytics
Families were identified through parent outreach in two elementary schools from one independent school district, during community events, community health fairs, by door-to-door invitation, and by “word of mouth.” First time and follow-up testing occurred during scheduled home visits. Families were compensated for their participation with a $20 gift card to a retail store upon completion of forms and child screenings. During the first home visit, parents completed a family demographics, home, and child characteristics form, including questions about the child’s age, family composition, race/ethnicity, income, education, history of Pb testing, cognitive and/or behavioral diagnoses, home Pb paint and soil hazards, and child behaviors that could increase risk of Pb exposure.
Anthropomorphic measures (height, weight, waist and hip circumference, blood pressure) were measured prior to every blood sample draw, and BMI was calculated for children ≥ 2 years of age. The blood sampling procedure followed a strict, multi-step protocol to ensure the collection of uncontaminated samples. Children’s hands were first washed thoroughly with antibacterial soap and water and dried with paper towels, then thoroughly wiped with an industry-standard cloth designed to remove metals from the surface of the skin (D-Wipe TM, Esca Tech, Inc., Milwaukee, Wisconsin). Each finger-stick whole blood sample of 50-microliters (μL) was collected into a sterile EDTA (Safe-T-Fill®) or Heparinized (Safe-T-Fill®) capillary blood collection 125-milliliter tube. Samples were stored at 4 C° up to one week, until transported overnight to the laboratory for analysis. Data and blood samples were collected by specially trained undergrad and graduate-level students in environmental and health sciences under the supervision of the program manager. The supervising program manager (MDR) had approximately 7 years of experience using this procedure with children (a pediatric phlebotomist is not required for finger stick blood collection according to CDC guidelines).
All blood Pb analyses were conducted by the Department of Agronomy at Kansas State University (Dr. Ganga Hettiarachchi, Ph.D., laboratory head) using inductively coupled plasma mass spectrometry (ICP-MS, Agilent 7500cx, Santa Clara, CA) [32]. Blood Pb results were recorded in micrograms per deciliter (μg/dL). Standards were prepared using stock solutions of Pb; Bi-209 was used as the internal standard element for Pb. Lyphochek Whole Blood Metals Control, Level 1 #527, Level 2 #528, and Level 3 # 529 (Bio-Rad Laboratories, Inc., Hercules, CA, U.S.A.) were used as the reference (QA/QC) samples to confirm percent recovery. Pb recoveries were in the range of 95–105%. The ICPMS limit of detection (LOD) for Pb was 0.04 μg/dL. BLL values below the ICP-MS detection limit of ≤ 0.04μg/dL (9/393, 2.3%) were manually imputed using values from an online random number generator application that produced random numbers with two decimal places between the values of 0 and 0.05.
The original goal of the study was to test each child a minimum of 4 times every 3 to 4 months. The frequency of testing however was dependent on parents’ schedules and testing was interrupted by COVID-19 pandemic community shutdowns in March 2020, resulting in BLL tests every 3 to 6 months, and up to 3 tests for some children (sample size details provided below).
Study sample
From an initial recruited sample of 110 families with 223 children, 9 children did not provide assent at Time 1 BLL testing; 1 child was too young to participate (4.5 months); 2 children exceeded the enrollment age (> 16 years old); 3 children were not available during the scheduled test appointment; and 2 families could not be contacted after their initial enrollment (“lost to follow-up”). The sample that completed Time 1 baseline BLLs included 107 families and 206 children. Since this study focused on lower-range BLLs, children with initial BLLs > 10 μg/dL were excluded from analyses (13/206, 6.31% of tested children). The final BLL data sample included N = 193 children.
In order to improve consistency in time between tests, from 320 total BLL samples, 12 (0.04%) BLLs measured less than two months after a previous sample, were excluded. The final sample included N = 193 children with BLLs at Time 1; N = 86 children with BLLs at Time 2; and N = 29 children with BLLs at Time 3. This sample size was not large enough to meet basic statistical power expectations for longitudinal analyses, and multiple imputation (SPSS Multiple Imputation Function, Version 26, IBM, Armonk, NY) was used to complete missing values.
Multiple imputation methods
Through analysis of missing patterns, the missing data were determined to be missing at random (MAR) and had non-monotone and monotone patterns of missingness. Fully conditional specification methods (FCS), set to generate random numbers using the Mersenne Twister method, and active generator initialization with a fixed value of 200000, were used [33]. For multiple imputation, at least 95% relative efficiency is preferred [34, 35]. For this study, 20 iterations provided a relative efficiency of 98% in all Individual Growth Curve (IGC) model analyses. Following previous recommendations [36, 37] continuous variables that were included as predictors in later models were centered before running multiple imputation. For comparison purposes, all longitudinal models were conducted with the original dataset of N = 29 children with three time points, and with the multiple imputation dataset that yield three time points for N = 193 children.
Data management and analysis
All data were entered and checked for accuracy and missing values in Microsoft Excel before importing to SPSS (macOS Version 26 IBM, Armonk, New York), for descriptive and comparative analyses. The analyses followed previous recommendations for generalized linear regression with random effects [38]; IGC modeling [39]; and guidelines in SPSS technical materials [40]. As expected for child BLL data, histograms and the skewness metric (2.18) showed that children’s BLLs at Time 1 (N = 193) followed a positive skew. IGC modeling accommodated the non-normal distribution of the data and previous recommendations for IGC [39] suggested that it was an appropriate and efficient means for testing whether child BLLs varied significantly over time.
In this approach, analyses begin with an estimation of the amount of variability attributable to individual variation over time (only the significance of the intercept is tested). The amount of possible intraindividual variability is then estimated by calculating an interclass correlation coefficient and this estimate is compared to a criterium to assess whether an IGC model including TIME as a random effect is warranted. To examine possible additional predictors of BLL variability, 3 time-invariant predictor variables were defined and tested including living below the 2020 poverty line; living in a home built before 1986; and/or living near a polluting industry, interstate highway, or international bridge. All three predictor variables were binary variables and coded -1 = no, 1 = yes. For all statistical tests, an alpha level of .05 was used.
Sex and age were included as fixed effects in all models tested. The main analyses tested Level 1 and Level 2 models. The Level 1 model tested the amount of change in BLL within children (intraindividual differences) from all time points. Results were used to calculate an Intraclass Correlation Coefficient (ICC) to estimate the amount of variability attributable to individual differences; an ICC less than 0.25 signaled large intraindividual differences, in which case IGC models would not be likely to perform better than a traditional ANOVA [39]. The Level 2 (growth curve) model assessed significant variations in individual trajectory changes in BLL by time. This model tested whether child BLLs changed predictably (either increased or decreased) as a function of time. Level 1 and Level 2 models were calculated using both the original (3 test waves, N = 29) and imputed (3 test waves, N = 193) data sets. (Pooled estimates of fixed effects were given in SPSS for N = 193 dataset.) Non-significant Level 2 test wave models were not submitted to tests of higher-order change with additional predictors. The results of these first tests guided subsequent models examining whether additional non-time variant factors might predict child BLLS (poverty, older home, living near pollution source).
Results
Demographic and clinical characteristics
Demographic and clinical characteristics of the sample are shown in Table 1. There were approximately equal proportions of males and females, and their BLL distributions were similar. The sample predominantly self-identified as Hispanic/Latino/Mexican American.
Mean family size was approximately 5 members and median reported annual household incomes were between $20,000 and $25,000, below the U.S. median household income for 2020.
Table 2 shows family home characteristics. The median built year of homes was 1969 and ranged between 1900 and 2016. A majority of children (70.5%, 136/193) of children lived in a home built before 1986 and 51.3% (99/193) lived near a polluting industrial site, heavily trafficked international bridge crossing, and/or major interstate highway. Of those living in older homes, 52.2% (71/136) also lived near a source of Pb contamination.
Descriptive analyses of child BLLs
As shown in Table 3, at all time points, the rates of children with BLLs exceeding the two most recent CDC BLL reference values (≤ 3.5 μg/dL, Oct. 2021 and ≤ 5 μg/dL, Jan 2012 –Sep 2021) were two- to five-fold higher than the 2.5% expected value [41].
For descriptive purposes, the change in BLLs between test points was calculated by dividing the absolute BLL difference of two time points by the initial compared BLL value, and then multiplying the result by 100. The percentages were then categorized into three categories including no change, less than 50% change, and equal to or greater than 50% change; proportions for each were calculated. Table 4 shows the results of these calculations. Among children who completed two BLL tests (N = 86), only 1.2% of children (1/86) had no change in BLLs between Time 1 and Time 2; BLL change in 26.7% (23/86) was less than 50%; among 72.1% (62/86), BLL change was 50% or greater. A Fisher’s Exact test determined that there were no differences in proportions of BLL change between males and females, X2 (2, N = 86) = 1.73, p = .710 (75.0% vs 69.0% with BLL change of 50% or greater, for males and females respectively).
Among children with three test points (N = 29) the descriptive change estimates were similar. From Time 1 to Time 2, no children had zero change; for 37.9% (11/29) BLL change was less than 50%; for 62.1% (18/29) BLL change was 50% or more. Between Time 2 and Time 3, 6.9% (2/29) had no change; for 34.5% (10/29) BLL change was less than 50%; for 58.6% (17/29) BLL change was 50% or more. Between Time 1 and Time 3, no children had zero change; for 27.6% (8/29) BLL changed less than 50% change; for 72.4% (21/29) BLL change was 50% or more. Again, Fisher’s Exact Tests determined that proportion differences in BLL change between males and females were non-significant between Time 1/Time 2 and Time 2/Time 3 (X2 (2, N = 29) = 2.207, p = .332), and between Time 1 and Time 3, X2 (1, N = 29) = .901, p = .343. Linear graphs (Figs 1 and 2) show BLLs overtime for children with two time points (N = 86), and children with three time points (N = 29). As illustrated, a majority of BLLs changed between time points, but there was no clear direction of constant increases or decreases over time.
Older children (ages ≥ 7 years) did not differ from younger children (ages < 7 years) in their vulnerability to Pb exposure. A Pearson Chi-Square test in the original dataset determined that there were no differences in the rates of BLLs exceeding the most recent CDC BLL reference value (≤ 3.5 μg/dL) between younger and older children (X2 (1, n = 308 data points from all time points) = 0.047, p = 0.828). Another Pearson Chi-Square test in the imputed dataset replicated the result (X2, (1, n = 11888 data points from all time points) = 0.91, p = 0.763).
Longitudinal models
Level 1 model analysis was first calculated for the original dataset of children with three time points (N = 29) (Table 5, row 1). The variability within children was statistically significant, F (1, 29) = 95.69, p < .001; the mean BLL was 1.57 μg/dL (t = 9.78, p < .001, 95% CI for the difference was 1.24 to 1.89). The calculated Intraclass Correlation Coefficient was extremely low (0.03) and estimated that only 3% of all total variation in BLLs was due to between-child (interindividual) differences (97% of all total variation in BLLs was attributable to within child, “intra-individual” differences). These results indicated that an Individual Growth Curve for this sample was unlikely to perform better than a traditional ANOVA [39].
The Level 1 model was re-run using the imputed dataset of N = 193 and results are shown in Table 5, row 2. The results were similar. The imputed dataset Level 1 model showed that the variability within children was statistically significant, p < .001. On the other hand, the calculated Intraclass Correlation Coefficient increased to 0.32, as compared to the Level 1 model of the original dataset (0.03, N = 29). Thus, approximately 32% of all total variation in BLLs was due to between-child (interindividual) differences and approximately 68% of all total variation in BLLs was attributable to within child (intraindividual) differences. Thus, the Intraclass Correlation Coefficient using the imputed dataset met the suggested 0.25 threshold criteria [39] and a Level 2 Individual Growth Curve model with time as a random effect was calculated; the SUBJECTS option was used to model each subject’s set of random parameters for the repeated effect of time, with an unstructured covariance matrix [33, 39, 42].
Table 5 summarizes estimates of fixed effects for the models predicting the influence of time, age, and sex, on child BLLs. The Level 2 models for N = 29 and N = 193 datasets were not significant predictors of child BLLs. The interaction of age and sex for N = 29 dataset was not significant (Est. = -0.027, SE = 0.041, t = -0.66, and p = .514); the model results for the imputed dataset showed a significant interaction of age and sex (Est. = 0.056, SE = 0.023, t = 2.59, p = .010). Since the age and sex interaction predicted child BLLs, these predictors were included in models with a third predictor.
Subsequent individual models tested three additional potential predictors of child BLLs (Tables 6–8) controlling for age and sex. These included living below the poverty line, living in a pre-1986 home, and living near a known source of Pb emissions. In the first 2 models, only the age and sex interaction was significant (and only in the imputed dataset). The final model showed a small but statistically significant effect on child BLLs of living near an environmental source of Pb pollution (N = 193, Est. = 0.22, SE = 0.093, t = 2.41, p = .026). The estimate suggested that living near a Pb point source increased child BLLs by an estimated 0.22 μg/dL (95% CI = 0.04 to 0.41).
Discussion
Child Pb exposure is regarded by many as the longest-standing public health epidemic in U.S. history, and children living in lower-income settings are at a disproportionate risk of exposure and its many long-term health effects. We lack valid estimates of the true incidence and prevalence of the problem, and this undermines our capacity to initiate and sustain efforts that effectively manage the problem. It has been recently estimated that based on the frequencies of established risk factors, at any given time, over 50% of U.S. children are positive for Pb by current standards. We noted that the rates of children in this sample with elevated BLLs were 2- to 5-fold greater than the expected 2.5%. These and other accumulating results should be a clarion call to action.
Current testing practices and their underlying assumptions may be one factor contributing to the ongoing problem. Acceptance of one or two negative BLL tests at some point during early childhood is insufficient to assume that a child is at low or no risk of future Pb exposure. This study tested the implicit logical assumption of these policies, that child BLLs are relatively stable over time, and that child BLLs across the school-age years are not likely to change over time. This study focused on children with lower-level Pb absorption (BLLs ≤ 10 μg/dL) living in high-risk neighborhoods.
Longitudinal data collection for this study was interrupted by the COVID pandemic and all analyses were conducted with all of the available collected data, and a complete imputed dataset with 3 data points for each of 193 children. Simple descriptive analyses and IGC models with time as a random effect and controlling for sex and age, showed that children’s BLLs varied significantly and randomly. Supporting the results of the longitudinal models, we noted that a majority of children had equal to or greater than a 50% change in BLL at 3 time points. Moreover, BLLs across this wide age range (6 months to 16 years) did not increase or decrease with time, suggesting that children did not become less vulnerable to Pb exposure as they aged.
IGC models with time as a repeated measure were also calculated to test whether 3 additional factors–living below the poverty level, living in a pre-1986 home, and living near a known polluting source–might predict child BLLs, controlling for sex and age. For the models using the imputed dataset, the interaction of sex and age was consistently significant; only living near a known polluting source predicted child BLLs and the significant increase was small.
The variability of BLLs observed in our study support the strong influence on variability that might be expected from multiple shifting environmental, physiological, and behavioral factors, that together in complex ways, influence Pb exposure, absorption, distribution, and accumulation in children [22]. A critical aspect of current policy is the implicit assumption that only children below the ages of either 3 years (in a majority of states) or 6 years (in many fewer states) are at a level of Pb exposure risk to warrant testing. While behaviors such as crawling, teething, mouthing contaminated non-food items, and playing in soil, place younger children at greater risk of Pb exposure and absorption, there is no reason to suggest that older children are at so little risk that testing should not be required. Indeed, the 2012 CDC Lead Prevention Task Force recognized the lack of evidence for restricting screening for children beyond the age of 6 years, and discussed the critical importance of identifying behaviors and factors that may predict BLLs in older children, and in children with chronic low-level Pb absorption [43].
Previous studies examining the stability of BLLs over time in children with chronic low-level Pb absorption have rarely been conducted, and those that have been done, have focused only on infants. For example, in a very large study (N = 267,687) of children ages 1 to 3 years, BLL changes among children with BLLs < 10 μg/dL were studied over a 36-month period. Conducted in the 1990s, change in the BLLs of children screened within two weeks of their first or second birthdays, was predicted by time, age, and season [44]. The findings were difficult to interpret due to the overlap of the study with the phasing out of Pb paint, leaded gasoline, and of the residential use of Pb solder and Pb pipes. Perhaps not surprisingly, among these youngest children, BLLs showed a steady decrease.
Another perhaps less confounded but also very short-term (4 month) study analyzed BLLs from a relatively small sample of 42 children during the first two years of life and measured Pb concentrations and isotope ratios in blood, urine, household dust, and duplicate diet samples (bottle-fed vs. breastfed) for children with BLLs ≤ 10 μg/dL [45]. The geometric mean of BLLs from birth and at 4 months remained unchanged (0.92 and 0.99 μg/dL, respectively). After 4 months, the BLLs remained near-constant or slightly increased, depending on whether children were exposed to Pb during home renovations for one brief or prolonged exposure, and whether bottle or breastfed.
Neither of these studies addressed change in BLLs for children across childhood, and neither examined dangerous lowest-range BLLs that are most common today. Many more studies are needed to characterize BLL change over time (for example, over a minimum of 5 years with BLL testing at least every 4 months), in order to identify sources that explain chronic low-level Pb, and to understand how Pb is absorbed and distributed at different developmental ages.
While many more studies are needed to better understand how BLLs represent child Pb exposure across the childhood years, overall, these findings provide initial support for the notion that our current child BLL testing policies are in need of major revision. One or two BLL tests for only youngest children are inadequate for detecting shifts in Pb exposure that, based on the data presented here, appear to be common among children living in high-risk neighborhoods throughout the school-age years. Previously our group presented a feasible model and plan for bi-annual universal testing, which shows how current testing practices could be amended to accomplish bi-annual testing in high-risk neighborhoods using (capillary) finger-stick samples collected following strict methods for ensuring clean sample collection and analyzed by ICPMS or GFAAS with a limit of detection at ≤ 0.04 μg/dL [46].
An alternative to re-instating bi-annual universal testing for all children may be to identify through known risk factor modeling, groups of children who are likely to be currently exposed to Pb. A substantial amount of work in this area is needed. Current prediction models of child Pb exposure risk often include a combination of sociodemographic and housing characteristics, distance from point sources of Pb, and family behaviors to determine exposure risk in early childhood [47, 48]. They do not however adequately capture relatively complete combinations of environmental and endogenous Pb sources, nor their variability, given age-specific behaviors that may influence children’s exposures. It is perhaps not surprising that these models have proven to have low predictive power. Some progress has been made in the use of data from the National Health Administration and Nutritional Examination Survey (NHANES) for the prediction of state and local child BLLs, but thus far, even these models have succeeded in predicting only (approximately) 30% of exposures [49].
Given the importance of predicting child lead exposure with greater accuracy, future studies need to continue characterizing factors that explain lower range BLLs in high-risk children. Studies should consider other key social, economic, and environmental factors that are not often considered in risk prediction models or for explaining the variability of children’s Pb exposure during childhood. These could include for example factors such as food insecurity, housing instability, and exposure to tobacco products [50].
Limitations
Data collection for this study was interrupted by COVID shutdowns. During the initial weeks of shutdown, we expected that data collection would resume within one or two months, if not weeks. We shifted our activities to virtual venues in the hope of promoting family engagement and maintained monthly or bi-monthly contact with all families via telephone. As the pandemic wore on, we developed educational videos on Pb prevention that were posted on our study website. By the time it was safe to return to the community, our study team had mostly graduated and moved on to other positions, and the study could not be resumed. As a result, multiple imputation had to be used to allow for regression analyses and, while not ideal, the close similarity of results obtained with the original as compared to imputed data sets gave some confidence that the results from the imputed dataset were plausible (Table 5). Nonetheless, we interpret these results with caution. A less impactful limitation concerned the pre-pandemic challenges of completing BLL tests every 3 months with families in which nearly all of the parents worked full time. While the statistical models used readily accommodated unequal timing between test waves, a dataset with more equal time intervals would allow for a greater range of hypotheses regarding child BLL change over time.
Summary
This study suggested that BLL variability is substantial over time, and did not diminish with age, in children ages 6 months to 16 years, and living in neighborhoods at high-risk of Pb exposure. The findings further suggested that in high-risk neighborhoods, child Pb sources posed shifting risks for children, and thus children require ongoing BLL monitoring throughout the school age years. Given the amount of variability observed, bi-annual routine BLL testing should be a minimum goal, to ensure detection of shifting lead exposure for children living in neighborhoods at high risk of Pb exposure, and/or for those living near known industrial sources of Pb emissions and/or interstate highways. A financially and practically feasible approach for re-instituting universal testing has been described in detail [46].
Acknowledgments
The authors would like to acknowledge Alexander Obeng; Diana Moreno; Crystal Costa; Carlos Chavarria; and Jaleen Avila for their important contributions to recruitment for this study.
References
- 1.
Centers for Disease Control and Prevention. Lead Poisoning Prevention. 2022 Sept 2 [cited 2022 Nov 23]. In: CDC Lead [Internet]. Atlanta: CDC 2022. Available from: https://www.cdc.gov/nceh/lead/prevention/default.htm
- 2. Bellinger DC, Stiles KM, Needleman HL. Low-level lead exposure, intelligence and academic achievement: A long-term follow-up study. Pediatrics. 1992 Dec 1;90(6):855–61. pmid:1437425.
- 3. Miranda ML, Kim D, Galeano MAO, Paul CJ, Hull AP, Morgan SP. The relationship between early childhood blood lead levels and performance on end-of-grade tests. Environ Health Perspect. 2007 Aug;115(8):1242–7. pmid:17687454
- 4. Zhang N, Baker HW, Tufts M, Raymond RE, Salihu H, Elliott MR. Early childhood lead exposure and academic achievement: evidence from Detroit public schools, 2008–2010. Am J Public Health. 2013 Mar;103(3):e72–77. pmid:23327265
- 5. Sorensen LC, Fox AM, Jung H, Martin EG. Lead exposure and academic achievement: evidence from childhood lead poisoning prevention efforts. J Popul Econ. 2019;32(1):179–218.
- 6. Lanphear BP, Hornung R, Khoury J, Yolton K, Baghurst P, Bellinger DC, et al. Low-Level Environmental Lead Exposure and Children’s Intellectual Function: An International Pooled Analysis. Environ Health Perspect. 2005 Jul 1;113(7):894–9. pmid:16002379
- 7. Mazumdar M, Bellinger DC, Gregas M, Abanilla K, Bacic J, Needleman HL. Low-level environmental lead exposure in childhood and adult intellectual function: a follow-up study. Environ Health. 2011 Mar 30;10(1):24. pmid:21450073
- 8. Gleason KM, Valeri L, Shankar AH, Hasan MOSI, Quamruzzaman Q, Rodrigues EG, et al. Stunting is associated with blood lead concentration among Bangladeshi children aged 2–3 years. Environ Health. 2016 Nov 4;15:103. pmid:27809911
- 9. Raihan MJ, Briskin E, Mahfuz M, Islam MM, Mondal D, Hossain MI, et al. Examining the relationship between blood lead level and stunting, wasting and underweight- A cross-sectional study of children under 2 years-of-age in a Bangladeshi slum. PLoS ONE. 2018 May 24;13(5):e0197856. pmid:29795643
- 10. Bellinger DC. Neurological and behavioral consequences of childhood lead exposure. PLoS Med. 2008; 5(5). Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2689677/ pmid:18507501
- 11. Braun JM, Hornung R, Chen A, Dietrich KN, Jacobs DE, Jones R, et al. Effect of Residential Lead-Hazard Interventions on Childhood Blood Lead Concentrations and Neurobehavioral Outcomes: A Randomized Clinical Trial. JAMA Pediatr. 2018 Oct 1;172(10):934–42. pmid:30178064
- 12. Bellinger D, Leviton A, Allred E, Rabinowitz M. Pre- and postnatal lead exposure and behavior problems in school-aged children. Environ Res. 1994 Jul;66(1):12–30. pmid:8013435
- 13. Goodlad JK, Marcus DK, Fulton JJ. Lead and Attention-Deficit/Hyperactivity Disorder (ADHD) symptoms: a meta-analysis. Clin Psychol Rev. 2013 Apr;33(3):417–25. pmid:23419800
- 14. Ruckart PZ, Jones RL, Courtney JG, LeBlanc TT, Jackson W, Karwowski MP, et al. Update of the Blood Lead Reference Value—United States, 2021. MMWR Morb Mortal Wkly Rep. 2021 Oct 29;70(43):1509–12. pmid:34710078
- 15. Yeter D, Banks EC, Aschner M. Disparity in risk factor severity for early childhood blood lead among predominantly African-American Black children: The 1999 to 2010 US NHANES. Int J Env Res Public Health. 2020;17(5). pmid:32121216
- 16. Egan KB, Cornwell CR, Courtney JG, Ettinger AS. Blood lead levels in U.S. children ages 1–11 Years, 1976–2016. Environ Health Perspect. 2021;129(3):037003. pmid:33730866
- 17. Benson SM, Talbott EO, Brink LL, Wu C, Sharma RK, Marsh GM. Environmental lead and childhood blood lead levels in US children: NHANES, 1999–2006. Arch Env Occup Health. 2017;72(2):70–8. pmid:26942652
- 18. Sobin C, Flores-Montoya MG, Gutierrez M, Parisi N, Schaub T. δ-Aminolevulinic acid dehydratase single nucleotide polymorphism 2 (ALAD2) and peptide transporter 2*2 haplotype (hPEPT2*2) differently influence neurobehavior in low-level lead exposed children. Neurotoxicol Teratol. 2015 Jan 1;47:137–45.
- 19. Sobin C, Gutierrez M, Alterio H. Polymorphisms of delta-aminolevulinic acid dehydratase (ALAD) and peptide transporter 2 (PEPT2) genes in children with low-level lead exposure. Neurotoxicology. 2009 Nov;30(6):881–7. pmid:19723536
- 20. Alvarez J, Del Rio M, Mayorga T, Dominguez S, Flores-Montoya MG, Sobin C. A comparison of child blood lead levels in urban and rural children ages 5–12 years living in the border region of El Paso, Texas. Arch Environ Contam Toxicol. 2018 Nov 1;75(4):503–11.
- 21. Hauptman M, Niles JK, Gudin J, Kaufman HW. Individual- and community-level factors associated with detectable and elevated blood lead levels in US children: results from a national clinical laboratory. JAMA Pediatr. 2021;175(12):1252–60. pmid:34570188
- 22. Del Rio M, Sobin C, Hettiarachchi G. Biological factors that impact variability of lead absorption and BLL estimation in children: Implications for child blood lead level testing practices. Journal of Environmental Health. 2022; Forthcoming.
- 23. Kastury F, Smith E, Lombi E, Donnelley MW, Cmielewski PL, Parsons DW, et al. Dynamics of lead bioavailability and speciation in indoor dust and X-ray spectroscopic investigation of the link between ingestion and inhalation pathways. Environ Sci Technol. 2019 Oct 1;53(19):11486–95. pmid:31460750
- 24. James AC, Stahlhofen W, Rudolf G, Köbrich R, Briant JK, Egan MJ, et al. Annexe D. deposition of inhaled particles. Ann ICRP. 1994 Sep 1;24(1–3):231–99.
- 25. Bailey MR, Ansoborlo E, Guilmette RA, Paquet F. Updating the ICRP human respiratory tract model. Radiat Prot Dosimetry. 2007;127(1–4):31–4. pmid:18192668
- 26. Deshommes E, Tardif R, Edwards M, Sebastien S, Prevost M. Experimental determination of the oral bioavailability and bioaccessibility of lead particles. Chem Cent J. 2012 Nov 22;6:138. pmid:23173867
- 27. Ruby MV, Davis A, Kempton JH, Drexler JW, Bergstrom PD. Lead bioavailability—dissolution kinetics under simulated gastric conditions. Environ Sci Technol. 1992 Jun;26(6):1242–8.
- 28. Liu W, Feng H, Zheng S, Xu S, Massey IY, Zhang C, et al. Pb toxicity on gut physiology and microbiota. Front Physiol. 2021 Mar 4;12:574913. pmid:33746764
- 29. Liu J, McCauley L, Compher C, Yan C, Shen X, Needleman H, et al. Regular breakfast and blood lead levels among preschool children. Environ Health. 2011 Apr 1;10:28. pmid:21457535
- 30. Sobin C, Parisi N, Schaub T, Gutierrez M, Ortega AX. δ-Aminolevulinic acid dehydratase single nucleotide polymorphism 2 and peptide transporter 2*2 haplotype may differentially mediate lead exposure in male children. Arch Environ Contam Toxicol. 2011 Oct;61(3):521–9.
- 31.
Texas Health and Human Services. Workbook: Blood Lead Testing and Elevated Blood Lead Levels in Texas Children [Internet]. 2020 [cited 2022 Mar 23]. Available from: https://tabexternal.dshs.texas.gov/t/BLS/views/BloodLeadTestingandElevatedBloodLeadLevelsinTexasChildren2013-2019/Story?:iid=1&:isGuestRedirectFromVizportal=y&:embed=y
- 32. Bazzi A, Nriagu JO, Linder AM. Determination of toxic and essential elements in children’s blood with inductively coupled plasma-mass spectrometry. J Environ Monit JEM. 2008 Oct;10(10):1226–32. pmid:19244647
- 33.
SPSSMixed.pdf [Internet]. [cited 2022 Jul 27]. Available from: https://csass.ucsc.edu/self-study%20courses/SPSSMixed.pdf
- 34. Seaman SR, Hughes RA. Relative efficiency of joint-model and full-conditional-specification multiple imputation when conditional models are compatible: The general location model. Stat Methods Med Res. 2018 Jun;27(6):1603–14. pmid:27597798
- 35. Rubin DB. Multiple imputation after 18+ years. J Am Stat Assoc. 1996;91(434):473–89.
- 36. Von Hippel PT. How to impute interactions, squares, and other transformed variables. Sociol Methodol. 2009;39(1):265–91.
- 37. Enders CK, Baraldi AN, Cham H. Estimating interaction effects with incomplete predictor variables. Psychol Methods. 2014;19(1):39–55. pmid:24707955
- 38. Golub MS, Sobin CA. Statistical modeling with litter as a random effect in mixed models to manage “intralitter likeness.” Neurotoxicol Teratol. 2020 Feb;77:106841.
- 39. Shek DTL, Ma CMS. Longitudinal data analyses using linear mixed models in SPSS: concepts, procedures and illustrations. ScientificWorldJournal. 2011 Jan 5;11:42–76. pmid:21218263
- 40.
IBM Docs [Internet]. 2021 [cited 2022 Jul 27]. Available from: https://prod.ibmdocs-production-dal-6099123ce774e592a519d7c33db8265e-0000.us-south.containers.appdomain.cloud/docs/en/spss-statistics/28.0.0?topic=SSLVMB_28.0.0/spss/advanced/idh_mixl.html
- 41.
Centers for Disease Control and Prevention. Blood Lead Levels in Children [Internet]. Childhood Lead Poisoning Prevention. 2022 [cited 2023 May 25]. Available from: https://www.cdc.gov/nceh/lead/prevention/blood-lead-levels.htm
- 42.
Heck RH, Thomas SL, Tabata LN. Multilevel and longitudinal modeling with IBM SPSS. 2nd ed. New York: Routledge; 2013. 462 p.
- 43.
Advisory Committee on Childhood Lead Poisoning Prevention. Low level lead exposure harms children: A renewed call for primary prevention [Internet]. 2012 p. 65. Available from: https://www.cdc.gov/nceh/lead/acclpp/final_document_030712.pdf
- 44. Haley VB, Talbot TO. Seasonality and trend in blood lead levels of New York State children. BMC Pediatr. 2004 Jun 4;4(1):8. pmid:15180903
- 45. Manton WI, Angle CR, Stanek KL, Reese YR, Kuehnemann TJ. Acquisition and retention of lead by young children. Environ Res. 2000 Jan;82(1):60–80. pmid:10677147
- 46. Sobin C, Guitierrez-Vega M, Flores-Montoya G, Del Rio M, Alvarez J, Obeng A, et al. Improving equitability and inclusion for testing and detection of lead poisoning in U.S. children. Milbank Q. 2022; Forthcoming.
- 47. Wheeler DC, Boyle J, Raman S, Nelson EJ. Modeling elevated blood lead level risk across the United States. Sci Total Env. 2021;769:145237. pmid:33493912
- 48. Sadler RC, LaChance J, Hanna-Attisha M. Social and built environmental correlates of predicted blood lead levels in the Flint water crisis. Am J Public Health. 2017;107(5):763–9. pmid:28323469
- 49. Schultz BD, Morara M, Buxton BE, Weintraub M. Predicting blood-lead levels among U.S. children at the census tract level. Environ Justice. 2017 Sep 14;10(5):129–36.
- 50. Del Rio M, MacDonald Gibson J. Factors influencing inequities in lead exposure in United States children: A systematic review. Int Public Health J. 2023;15(3). Forthcoming.