Lead exposure and indices of height and weight in Uruguayan urban school children, considering co-exposure to cadmium and arsenic, sex, iron status and dairy intake

Abstract

Child growth depends on complex factors including diet, nutritional status, socioeconomic, and sanitary conditions, and exposure to environmental chemicals. Lead exposure is known to impair growth in young children but effects in school-age children are less clear. The effects of co-exposure to low-level lead and other toxic metals on child growth are not well understood. We examined cross-sectional associations of blood lead (BLL) with growth indices (Z scores of body mass index for age, BAZ, and height for age, HAZ) in Uruguayan urban school children (n=259; ~7y). Potential differences in these associations in children with lower vs. higher urinary inorganic arsenic metabolites (U-As), urinary cadmium (U-Cd), sex (42% girls), iron deficiency (ID, 39% children), or intake of dairy foods below recommended levels were examined. BLL was measured using AAS, U-As using HPLC-HGICP-MS, and U-Cd using ICP-MS. Dietary information was obtained by two 24-hour recalls completed by caregivers. Children’s linear growth was within age and sex-appropriate reference values. Overweight (BAZ > 1 ≤ 2 SD) was found in 20.1%, and obesity (BAZ > 2 SD) in 18.5%, of children. Ranges (5^th, 95^th percentile) of biomarker concentrations were: BLL, 0.8 −7.8 μg/dL; U-Cd, 0.01–0.2 μg/L, and U-As, 4.0–27.3 μg/L. BLL was inversely associated with HAZ (β [95% CI]: −0.10 [−0.17, −0.03]) in covariate-adjusted models. Although this association was slightly more pronounced in girls, children without ID, and children with lower U-As, there was little evidence of effect modification due to overlapping CIs in stratified models. BLLs were not associated with BAZ, except for a suggestion of a negative relationship in girls (−0.10 [−0.23, 0.02]) but not boys [0.001 [−0.11, 0.12]). Our findings indicate that exposure to low levels of lead was associated with lower HAZ in apparently normally growing urban school children. Larger future studies should help elucidate if these associations persist over time and across populations.

2. INTRODUCTION

Child growth and development depend on complex factors including diet and nutritional status, socio-economic, demographic, and sanitary conditions, and exposure to environmental chemicals such as toxic metals (Bellinger, 2012; Kordas et al., 2007). Lead and other toxic metals or metalloids, such as cadmium and arsenic, are known to impair child growth and neurodevelopment (Bellinger et al., 2017; Lin et al., 2011; Rahman et al., 2017; Wright and Baccarelli, 2007). Since these toxicants are ubiquitously distributed in the environment, particularly in emerging-market countries (Horton et al., 2013), their co-occurrence may impair child growth in synergistic ways (Kortenkamp et al., 2007). Although some studies have examined the associations between multiple toxic metals and children’s anthropometric indices (Choi et al., 2017; Gardner et al., 2013; Signes-Pastor et al., 2020), the impact of these exposures on child growth is still poorly understood. Moreover, as previously proposed (Kordas, 2017), the effects of toxic metals on child growth and development need to be examined simultaneously with dietary patterns and nutritional status, and socio-demographic characteristics.

Of all toxic metals, lead is the best known to impair child growth from prenatal and postnatal exposure even at low blood lead levels (≥ 1 μg/dL) (Bellinger et al., 2017). Associations between low-level lead exposure and growth have been extensively studied in infants and pre-school children (Bellinger et al., 2017; Choi et al., 2017; Gardner et al., 2013; Gleason et al., 2016). In contrast, less is known regarding these associations in school-age children (Burns et al., 2017; Deierlein et al., 2019; Kerr et al., 2019; Signes-Pastor et al., 2020). Some studies suggest that impaired growth in lead-exposed children may be sex-specific (Burns et al., 2017; Deierlein et al., 2019; Signes-Pastor et al., 2020), but findings are inconclusive. Therefore, it is important to examine relationships between exposure to lead, and other toxicants, separately in boys and girls.

Arsenic and cadmium are also known to impair the development of children. Several studies have shown that when examined alone, arsenic (Gardner et al., 2013; Gilbert-Diamond et al., 2016; Rahman et al., 2017) and cadmium (Kippler et al., 2012; Lin et al., 2011) are linked to poorer growth in young children. Early-life cadmium exposure has been related to diminished bone maturation in school-age children (Sughis et al., 2011) and to changes in bone-related biomarkers, with differences noted between boys and girls (Igra et al., 2019).

Few studies thus far have focused on multiple metal exposure and child growth (Choi et al., 2017; Gardner et al., 2013). A recent study tested the association of lead, mercury, fluoride, manganese and selenium on growth indices of 6–11 year old U.S. children (Signes-Pastor et al., 2020). In this study, blood lead was negatively and blood manganese was positively associated with BMI, waist circumference, standing height, and upper arm length, and these effects varied according to the concentrations of the other metals in the mixture, suggesting potential interactive effects.

Children’s exposure to lead and other toxic metals has been a health concern in Uruguay (Kordas et al., 2010; Kordas et al., 2016; Kordas et al., 2018b). Despite reductions in children’s blood lead levels during the last decade, associations have been found between blood lead and blood hemoglobin (Kordas et al., 2011), and neurodevelopment (Frndak et al., 2019; Kordas et al., 2015), and diet and nutritional status (Kordas et al., 2018a) in this vulnerable population. In previous studies of this population, children who had iron deficiency (ID) had higher blood lead compared to children without ID, and children consuming more dietary calcium and dairy foods had lower urinary lead compared to those consuming less (Kordas et al., 2018a). These relationships are consistent with the known effects of ID and dietary calcium on intestinal absorption of lead and could play a moderating role in the link between lead exposure and growth. The aim of the present study was to examine the relationship between exposure to low levels of lead and anthropometric indices in Uruguayan urban boys and girls attending first grade of elementary school. We hypothesized an inverse association between blood lead levels and children’s height for age and body mass index for age, with stronger associations in children with ID, lower dairy foods intake, as well as higher levels of arsenic or cadmium.

3. METHODS

3.1. Study setting, participant recruitment and study sample

The study, consisting of multiple cross-sectional samples recruited over a four year period, was conducted between July 2009 and August 2013. We targeted different neighborhoods of Montevideo, Uruguay, with documented or suspected metal exposures among children, as described previously (Kordas et al., 2016; Kordas et al., 2018b). Initially, 357 children aged ~7 years from 11 private elementary schools, and their parents were recruited for the study after providing informed consent. Approval was granted by ethics committees at the Pennsylvania State University, Catholic University of Uruguay, Chemistry Faculty at the University of the Republic of Uruguay (UDELAR), and the University at Buffalo.

3.2. Socio-Demographic Survey

Caregivers who agreed to participate and provided informed consent were invited to a meeting at the school to complete a questionnaire that covered topics including: socio-demographic characteristics of the family (parental education and occupation; potential occupational exposures; household possessions; spending on food and clothing; number of people, including children < 5 years of age, living in the house; size of the house [number of rooms], medical history of the child (history of hospitalizations; anemia; previous blood lead tests; iron supplementation), caregiver views on parenting (role of schools and teachers in child rearing; views of discipline and children’s behavior; etc.), and their home environment (frequency of cleaning activities; frequency of keeping windows open in summer and winter; time spent by children outdoors; etc.). Questionnaires were self-administered, but research staff were available to assist and provide clarification while caregivers filled out the survey.

3.3. Dietary Intake Measures

The diet of study children was measured using two 24-hour recalls conducted with a caregiver familiar with the child’s diet; the child contributed to the recall, particularly with regards to foods consumed at school. The first recall was conducted on the day of the blood draw and the second was completed over the phone, without prior appointment, at least 2-weeks later on either a weekday or a weekend. If the nutritionist could not reach the caregiver by phone on three separate occasions, the second recall was not conducted (n=13). To facilitate estimation of portion sizes, food models (Vázquez and Witriw, 1997) and household measurement cups were used during the first interviews with caregivers and children (Nutrición., 2002). Neutral probing questions were asked (ex. “Did your child eat/drink anything on the way home from school yesterday?”) to obtain more accurate recalls. Information collected included: the type of food, time and place of consumption, amount of food consumed, food preparation methods, recipe ingredients, brand names of commercial products, and vitamin or mineral supplements taken. The foods or beverages measured were assigned a unique code and entered, along with amounts consumed, into a database that contained the nutrient composition of 342 typical Uruguayan foods and preparations. Food intake was recorded in grams for foods and milliliters for drinks.

Intake of dairy foods (milk, yogurt [part-skim and skim], yogurt drink, smoothie, cheese [mozzarella, ricotta, fresh cheese], pudding and ice cream) was expressed in cup-equivalents per day and combined into a single variable. Conversion values from amounts to cup-equivalents for the dairy foods were obtained from the USDA Food Composition Database and the USDA MyPlate guidance.

3.4. Anthropometric Measures

Height was measured in triplicate to the nearest 0.1 cm using a portable stadiometer (Seca 214, Shorr Productions, Colombia, MD) by a trained pediatric nurse. Weight was measured in triplicate to the nearest 0.1 kg using a digital scale (Seca 872, Shorr Productions, Colombia, MD). Children wore light clothing like school uniforms but not shoes. The three measurements for height and weight were averaged. Standard weights of the clothing were subtracted from the child’s average weight. Body mass index (BMI, kg/m²) was calculated from weight and height. Z-scores for height-for-age (HAZ) and BMI-for-age (BAZ) were generated using the WHO AnthroPlus software. BAZ was used because it accounts for the child’s height. Because taller children weigh, on average, more than shorter children of the same sex and age, accounting for height is quite important, and gives an indication of slimness for a given body frame.

3.5. Biosampling

A 3 ml fasting blood sample was collected by a phlebotomy nurse using a 25-gauge blood collection set with a butterfly needle (Vacuntainer, Becton Dickinson, Franklin Lakes, NJ) into heparin coated tubes (Vacutainer, Becton Dickinson, Franklin Lakes, NJ) for blood lead determination. An additional 3 ml of blood was collected into a tube with a clot activator and separator gel for serum separation (Becton Dickinson, Franklin Lakes, NJ). The serum tube was left to stand for 45 minutes and then centrifuged for 10 minutes at 3000 RPM. Whole blood was transported on ice to the CEQUIMTOX, Toxicology Laboratory in the Faculty of Chemistry at the University of the Republic of Uruguay in Montevideo, where it was stored at −20°C until analysis. Serum samples were stored at −20°C at the Research Center, Catholic University of Uruguay, until they were shipped to Pennsylvania State University’s Department of Nutritional Sciences for further storage (at −20°C) and analysis.

First void urine samples were collected at home, on the morning of the blood draw, in polyethylene collection cups, and brought into school by children and their parents. The collection cups were previously washed with 10% HNO₃ and rinsed with deionized water, to reduce contamination with toxic metals. Parents were asked to capture urine mid-stream. The urine samples were transported on ice to the Catholic University of Uruguay on the day of collection, where they were processed and stored at −20°C. The specific gravity of urine samples was measured using a portable refractometer (PAL 10S, Atago Inc, USA) on the day of the collection. Urine samples were shipped on dry ice to the Karolinska Institute, Stockholm, and were stored at −20°C until trace element analysis.

3.6. Laboratory analyses

Blood lead (BLL) concentrations, indicative of recent exposure, were measured at the CEQUIMTOX using Atomic Absorption Spectrometry (AAS, VARIAN SpectrAA-55B), with either flame (FAAS) or graphite furnace (GFAAS) ionization techniques depending on sample volume. Up to 2012, the GFAAS was used in those blood samples for which the volume was below 2 mL (n= 186). From 2013 on, GFAAS was the only technique used to analyze all samples (n=69). The detection limits for the AAS techniques were 1.8 μg/dL and 0.8 μg/dL for FAAS and GFAAS respectively. Analytic conditions were validated with standard quality assurance and quality control (QA/QC) procedures (Parsons and Chisolm, 1997). CEQUIMITOX has a successful participation in Interlaboratory Programs for Quality Control for Lead in Blood as those from Spain (PICC-PbS, 2001) and the U.S. Centers for Disease Control and Prevention (CDC) Lead and Multi-Element Proficiency Program (US LAMP). For 95.1% of the 184 blood samples from Interlaboratory Programs for Quality Control, analytic results for lead in blood had CV < 2%.

Hemoglobin was measured in a drop of blood from the serum collection tube immediately after the draw, using a Hemocue 201+ portable hemoglobinometer (Hemocue, LakeForest, CA). The instrument was calibrated daily using the standard controls (low, medium, high) provided by the manufacturer. Ferritin, a marker of iron stores, was measured in serum using one of two methods, according to manufacturer instructions: 1) an immunoradiometric assay (Coat-A-Count Ferritin IRMA; SIEMENS Diagnostic Products, USA) and 2) an enzyme immunoassay (Spectro Ferritin, RAMCO Laboratories, Texas, USA). The ELISA assay was used when the laboratory was no longer able to handle radioactive materials. Intra- and inter-assay coefficients (CV) for the duplicate runs were 4.2% and 9.5%, respectively, for the IRMA method and 1.7% and 7.6% for the ELISA method; CV below 10% is typically considered acceptable. A correction factor was calculated to convert ELISA to IRMA assay values, as described previously (Kordas et al., 2018a). Serum C-Reactive Protein (CRP), a general marker of inflammation, was analyzed in duplicate using an ELISA technique (Erhardt et al., 2004). Serum control samples (Liquicheck, Bio-Rad) were used as standards. Intra-assay and inter-assay CVs were 4.9% and 8%, respectively. Based on WHO guidelines (WHO, 2001; WHO, 2020), anemia was defined as hemoglobin less than 11.5 g/L and ID as CRP-adjusted serum ferritin of less than 15 ng/mL.

The sum of arsenic metabolites in urine (U-As) was measured as the concentration of inorganic arsenic (I-As) and its methylated metabolites (MMA, DMA) in urine using high-performance liquid chromatography (HPLC) online with hydride generation (HG) and inductively coupled plasma mass spectrometry (ICP-MS), as described previously (Gardner et al., 2011; Li et al., 2008). The sum of arsenic species reflects the ongoing inorganic arsenic exposure from all sources, including drinking water and food. The limit of detection was 0.1 μg/L for inorganic arsenic (III) and MMA, 0.2 μg/L for DMA, and 0.3–0.5 μg/L for inorganic arsenic (V). Seven samples (2.1%) were below limit of detection for inorganic arsenic (III) and 26 (7.9%) were below the limit of detection for inorganic arsenic (V). Reported values were used in statistical analyses.

Cadmium in urine (U-Cd), a marker of long-term cadmium exposure (life-long), was measured using an Agilent 7700x ICP-MS (Agilent Technologies, Tokyo, Japan), equipped with octopole collision/reaction cell technology to minimize spectral interferences, as described in more detail elsewhere (Björklund et al., 2012; Kippler et al., 2007). The samples were analyzed in two batches, with limit of detection of 0.001 μg/L. No batch differences were observed. A single value was below the limit of detection and was used as given in the statistical analysis.

To compensate for the variation in dilution of the urine samples, U-As and U-Cd concentrations were adjusted to the average specific gravity of urine (1.024).

3.7. Statistical Methods

3.7.1. Exclusions and complete-case sample

A total of 259 participants were included in the complete-case analysis. Participants excluded from analysis (n=98) were those with missing data on variables representing the exposure, outcome or model covariates. Missingness of serum ferritin value (n=14) was used as an exclusion criterion in addition to the aforementioned covariates for models stratified by iron status.

3.7.2. Descriptive Statistics

Descriptive statistics were generated to characterize the complete-case (n=259) and excluded (n=98) samples. Median (5^th and 95^th percentiles) values were calculated for the following non-normally distributed continuous variables: U-Cd, U-As and SF corrected for C-reactive protein. Wilcoxon Rank Sum Tests evaluated differences between the complete-case sample and the children who were excluded. Mean ± SD were calculated for the following normally distributed variables: age, BLL, HAZ, height, weight, BAZ, body mass index, maternal education, hemoglobin, and average energy consumption. T-Tests were used to evaluate differences between the two groups. Percent/frequency of response by level of categorical variables was calculated for the following variables: sex, household crowding status, household possessions score below median (4 items), presence of stunting (HAZ < −2 SD), overweight (BAZ > 2 ≤ SD) and obesity (BAZ > 2 SD), anemia (blood hemoglobin <11.5 g/dL), and low iron stores (serum ferritin corrected for CRP <15 ng/mL). Chi-Square tests were used to evaluate differences between the two samples.

3.7.3. Linear Regressions

Separate ordinary least squares (OLS) regressions were generated to examine the relationship between BLL and each HAZ and BAZ. Unadjusted models were followed by covariate-adjusted models. Covariates were selected based on previous literature and included maternal education (years), blood hemoglobin, a measure of household crowding (number of people in the household divided by number of rooms), household possessions, and current parental smoking status (yes/no). Models were also adjusted for method of lead measurement (AAS/GFAAS) to account for the difference in limit of detection. Models containing BAZ as the dependent variable were additionally adjusted for energy consumption (kcal).

The models described above were repeated with stratification by median U-As (9.9 μg/L), median U-Cd (0.06 μg/L), sex, iron status (deficient vs. not deficient), and the level of intake of dairy foods (< vs. ≥2.5 cup-equivalents per day), based on current recommendations (CNSUP, 2004; USDA/DHHS, 2010). Beta-coefficients and 95% confidence intervals were reported for each OLS regression.

4. RESULTS

The 259 children in our study (58% boys) were on average 7 y old (5.5 to 8.7 y) and had adequate average height for age (HAZ) and BMI for age (BAZ). Overweight (BAZ > 1 ≤ 2 SD) was found in 20.1% and obesity (BAZ > 2 SD) in 18.5% of the children. Only one child was stunted (HAZ < −2 SD). There was virtually no anemia in the sample, but 39% of children had low iron stores indicative of ID (CRP-adjusted serum ferritin <15 ng/mL). Average daily energy intake from the diet was consistent with the energy requirements for this age group (2000 kcal/d). Median dairy intake was 1.9 cups per day; based on the two recalls, it appears that approximately 70% of the children do not meet the dietary guidelines for dairy foods (2.5 cups per day). Some children (22%) were living in crowded homes and 53% with currently smoking caregiver(s). Children excluded from the study were slightly older (by less than 2 months), but other characteristics were similar to those of children included in the complete-case analysis (Table 1).

Table 1.

Characteristics of 259 children and families participating in the study in comparison to those excluded due to missing data.

Characteristic	Included in study	Excluded from study		p-value1
	Mean ± SD, % Or Median (5%, 95%)	N	Mean ± SD, % Or Median (5%, 95%)
Age (months)	80.9 ± 6.4	92	82.4 ± 6.3	0.05
Sex (%)
Male	57.5%	46	48.9%	0.15
Female	42.5%	48	51.1%
Either parent currently smokes (%)	52.9%	26	55.3%	0.76
Mother’s education (years)	9.2 ± 2.7	75	8.7 ± 2.5	0.22
Household is crowded (%)	22.4%	7	14.9%	0.25
Household possessions score below median of 4 items (%)	55.2%	27	54.0%	0.88
Blood Pb concentration (μg/dL)	4.2 ± 2.1	56	3.8 ± 2.0	0.17
Urinary Cd concentration (μg/L)	0.06 (0.01, 0.17)	67	0.06 (0.02, 0.14)	0.91
Sum of As species in urine (μg/L)	9.9 (4.0, 27.3)	69	10.1 (4.3, 29.4)	0.65
Height (cm)	122.2 ± 6.0	67	123.0 ± 4.9	0.28
Height for age Z score (SD)	0.5 ± 1.1	65	0.5 ± 1.0	0.84
Weight (kg)	25.5 ± 5.9	68	25.8 ± 5.0	0.68
BMI for age Z score (SD)	0.7 ± 1.3	65	0.7 ± 1.4	0.84
Hemoglobin (g/dL)	13.2 ± 1.1	63	13.1 ± 1.1	0.48
<11.5 g/dL (%)	< 0.1%	3	< 0.1%	--
Serum ferritin (adjusted for C-Reactive Protein) (ng/mL)	18.0 (2.8, 45.6)	59	18.0 (3.4 – 44.0)	0.91
Low Iron Stores (%)	39.2%	24	40.1%	0.83
Dairy Intake (Servings, Cup Equivalents/d)	1.9 (0.5, 4.2)	62	2.1 (0.8, 5.1)	0.60
Energy Consumption (kcal/d)	2,208 ± 564	62	2,123 ± 544	0.28

¹Based on T-Tests (t), Wilcoxon Rank Sum Test (z), or Chi-Square Test (chi), as appropriate.

The mean ± SD of BLL was 4.2 ± 2.1 μg/dL. Approximately 25% of children presented blood lead levels at or higher than 5 μg/dL. Only 6 children had BLL at or over 10 μg/dL, and the highest blood lead concentration was 13.2 μg/dL. Urinary cadmium and arsenic concentrations were [median (5%−95%)]: 0.06 (0.01–0.17) μg/L for U-Cd and 9.9 (4.0–27.3) μg/L for U-As (Table 1). Biomarkers of metal exposure were modestly correlated: U-Cd with BLL (ρ = 0.16, p=0.01) U-Cd with U-As (ρ = 0.35, p<0.001), BLL with U-As (ρ = 0.07, p=0.27). There were no differences in BLL between boys and girls, between children stratified by median U-Cd and U-As, or by intake of dairy foods lower or above 2.5 cup-equivalents per day (Table 2). Children with ID had higher blood lead (4.9 ± 2.3 μg/dL) than those without ID (3.8 ± 1.8 μg/dL) (p<0.001) (Table 2). However, children with ID did not differ from those without ID on mean HAZ (0.5 ± 1.1 vs. 0.5 ± 1.1 SD) and BAZ (0.6 ± 1.3 vs. 0.7 ± 1.3 SD).

Table 2.

Blood lead concentrations (μg/dL) by sex, levels of other metals and nutritional factors (n=259).

Characteristic	N	Mean ± SD	P-value1
Sex
Boys	149	4.0 ± 1.9	0.11
Girls	110	4.5 ± 2.3
Urinary cadmium2
< 0.06 μg/L	129	4.0 ± 2.0	0.11
≥ 0.06 μg/L	130	4.4 ± 2.2
Urinary sum of arsenic species2
< 9.9 μg/L	131	4.1 ± 2.1	0.29
≥ 9.9 μg/L	128	4.4 ± 2.1
Iron deficiency3
No	149	3.8 ± 1.8	<0.001
Yes	96	4.9 ± 2.3
Dairy consumption4
< 2.5 cup-equivalents/d	188	4.3 ± 2.2	0.51
≥ 2.5 cup equivalents/d	71	4.1 ± 1.9

¹Based on t-test;

²Urinary metal concentration adjusted for specific gravity of urine;

³Defined as CRP-adjusted serum ferritin < 15 ng/mL;

⁴Based on dietary recommendations for children (CNSUP, 2004; USDA/DHHS, 2010).

Table 3 shows metal biomarker concentrations, the frequency of iron deficiency and of children meeting the dairy intake according to sociodemographic characteristics of the sample. Some group differences were noted. Mean BLLs were higher among children whose parents smoked at the time of the study and among households considered crowded. Older children had higher prevalence of ID. Median concentrations of U-Cd differed somewhat by age. Finally, more boys than girls met age-appropriate recommendations for dairy consumption.

Table 3.

Metal biomarkers, iron deficiency and dairy intake status according to sociodemographic characteristics of study sample (n=259).

	N	Child U-As (μg/L) Median [5%, 95%]1	Child U-Cd (μg/L) Median [5%, 95%]1	Child BLL (μg/dL) Mean ± SD2	Child has iron deficiency (%)3	Meets dairy intake recommendation (%)3
Child age (mo)
71 – 78	94	9.6 [3.4, 28.9]	0.05 [0.01, 0.17]	3.9 ± 2.1	25.6	22.3
79 – 83	82	10.9 [4.0, 27.3]	0.07 [0.01, 0.15]	4.3 ± 2.2	44.2	31.7
84 – 105	82	9.5 [5.4, 32.8]	0.06 [0.03, 0.14]**	4.5 ± 1.9	50.0***	28.9
Sex
Girls	110	9.6 [4.4, 24.7]	0.05 [0.01, 0.14]	4.0 ± 1.9	36.9	21.8
Boys	149	9.9 [4.2, 30.3]	0.09 [0.02, 0.17]	4.5 ± 2.3	40.8	31.5*
Parent currently smokes
No	122	10.2 [4.1, 26.1]	0.06 [0.01, 0.17]	3.9 ± 1.8	41.9	26.2
Yes	137	9.6 [3.9, 27.3]	0.06 [0.2, 0.16]	4.5 ± 2.3**	36.7	28.5
Household is crowded
No	201	9.7 [4.1, 27.7]	0.06 [0.01, 0.15]	4.1 ± 2.0	38.9	28.4
Yes	58	10.1 [3.9, 26.1]	0.06 [0.02, 0.17]	4.8 ± 2.2**	40.0	24.1
Luxury household possessions
Less than 4	143	9.6 [4.9, 24.0]	0.06 [0.01, 0.14]	4.3 ± 2.2	39.0	24.5
4 or more	116	10.5 [3.6, 30.3]	0.06 [0.02, 0.18]	4.1 ± 2.0	39.4	31.0
Maternal education (years)
4 – 8	152	9.9 [4.2, 32.8]	0.06 [0.02, 0.17]	4.3 ± 2.1	39.6	30.3
9 – 11	60	9.6 [4.4, 23.0]	0.06 [0.01, 0.19]	4.2 ± 2.1	37.0	21.7
12 – 17	47	9.6 [3.1, 25.3]	0.06 [0.01, 0.15]	4.0 ± 2.1	40.4	25.5

¹Differences in medians across categories tested with a Kruskal-Wallis rank test;

²Differences in mean ± SD across categories tested with a t-test or ANOVA;

³Differences in frequencies across categories tested with a Chi² test;

^***p-value <0.01;

^**p-value <0.05;

^*p-value <0.1.

Covariate-adjusted models of associations between BLL and Z scores of anthropometric indices in all study children, as well as stratified by As and Cd levels, sex, ID and dairy intake, are shown in Table 4. Considering all children, HAZ was negatively associated with BLL: each 1 μg/dL was related to a 0.10 SD lower height for age Z score. In stratified analyses, each 1 μg/dL of BLL was negatively associated with HAZ in girls (0.10 SD lower), in children with U-Cd above and below the median (0.12 and 0.10 SD, respectively), in children with U-As below median levels (0.14 SD), in children without ID (0.14 SD), and in those with dairy consumption lower than 2.5 cup-equivalents/day (0.10 SD lower). However, the 95% confidence intervals for the estimates across the two strata of all these variables overlapped considerably, suggesting little evidence of effect modification. BLL was not associated with BAZ in the full sample of children (Table 4); nevertheless, among girls, there was a suggestion of a negative association (−0.10 [−0.23, 0.02]) that did not reach statistical significance, possibly due to a relatively small sample size (n=110) or the large prevalence of overweight/obesity (~40%). There was no evidence of an association among boys (0.001 [−0.11, 0.12]).

Table 4.

Covariate-adjusted associations between blood lead concentrations, HAZ and BAZ among 6–8 y old children from Montevideo, Uruguay, stratified by sex, arsenic and cadmium exposure, iron deficiency, and diary intake.

Characteristic	N	HAZ1 ß [95% CI]	BAZ1 ß [95% CI]
All children	259	−0.10 [−0.16, −0.03]***	−0.05 [−0.14, 0.02]
Stratified analyses:
Sex
Girls	110	−0.10 [−0.19, −0.01]**	−0.10 [−0.23, 0.02]*
Boys	149	−0.06 [−0.16, 0.04]	0.001 [−0.11, 0.12]
Urinary cadmium (μg/L)
< 0.06 μg/L2	129	−0.12 [−0.23, −0.01]**	−0.07 [−0.20, 0.06]
≥ 0.06 μg/L	130	−0.10 [−0.19, −0.01]**	−0.04 [−0.15, 0.07]
Urinary sum of arsenic species
< 9.9 μg/L2	131	−0.14 [−0.25, −0.04]***	0.005 [−0.11, 0.12]
≥ 9.9 μg/L	128	−0.06 [−0.15, 0.02]	−0.09 [−0.20, 0.03]
Iron deficiency3
No	149	−0.14 [−0.25, −0.04]***	−0.01 [−0.13, 0.12]
Yes	96	−0.07 [−0.17, 0.02]	−0.07 [−0.20, 0.05]
Dairy consumption3
< 2.5 cup-equivalents/d	188	−0.10 [−0.18, −0.02]**	−0.04 [−0.14, 0.06]
≥ 2.5 cup-equivalents/d	71	−0.13 [−0.27, 0.01]*	−0.08 [−0.24, 0.08]

¹Models adjusted for: maternal education (years), hemoglobin (g/dL), home density, household possessions value, current parental smoking status (yes/no), method of lead measurement (AAS/GFAAS) and energy consumption (kcal) (only for BAZ);

²Median concentration;

³Defined as CRP-adjusted serum ferritin < 15 ng/mL;

⁴Based on current recommendations (CNSUP, 2004; USDA/DHHS, 2010);

^*p < 0.1;

^**p<0.05;

^***p<0.01.

5. DISCUSSION

Studies relating child growth and co-exposure to multiple toxic metals are scarce in school-age children. These studies are particularly relevant at the levels of exposure commonly present in market-emerging countries (Horton et al., 2013). In a cross-sectional study of school-age children in Uruguay, we found an inverse association between low-level exposure to lead and height for age in school children who otherwise appeared to be healthy and growing within normal anthropometric indices. Although we noted some differences in the association between BLL and height for age among boys and girls, between children with lower and higher urinary As concentrations, and children with and without ID, there was little evidence of effect modification by these factors. BLL was not associated with BMI for age, although the estimated relationship among girls suggest that further investigation of sex-differences is warranted.

Blood lead was inversely associated with height for age but there was no association with body-mass-index for age, consistent with results of studies in children of similar age in Greece ((Kafourou et al., 1997); mean blood lead 12.3 μg/dL); Pakistan ((Rahman et al., 2002); mean blood lead 16.1 μg/dL), and Korea ((Min et al., 2008); mean blood lead 2.4 μg/dL). In fact, lead exposure appears to affect linear growth since the most consistent associations with BLL found in studies of children at different ages (1 to 18 y) have been with reduced height (Ballew et al., 1999; Burns et al., 2017; Deierlein et al., 2019; Gleason et al., 2016; Ignasiak et al., 2006; Kafourou et al., 1997; Kerr et al., 2019; Min et al., 2008; Rahman et al., 2002; Selevan et al., 2003; Zhou et al., 2020), reduced head circumference (Ballew et al., 1999; Choi et al., 2017; Kafourou et al., 1997), and reduced limb length (Ignasiak et al., 2006; Kerr et al., 2019; Signes-Pastor et al., 2020). In contrast to results from our study, others did report a negative relationship between blood lead and indices of body weight in children (Choi et al., 2017; Deierlein et al., 2019; Ignasiak et al., 2006; Signes-Pastor et al., 2020; Zhou et al., 2020), further emphasizing the need to investigate these relationships in future studies, particularly focusing on sex differences.

Lead exposure impairs linear growth possibly by impacting skeletal development. Most of the body burden of lead is in bone, thus bone cells are immediate targets for lead toxicity. Lead inhibits osteoblast activity (Long et al., 1990) and perturbs the synthesis or secretion of bone matrix components (Hamilton and O’Flaherty, 1995) and bone cell responses to hormonal regulation (Hamilton and O’Flaherty, 1995). Lead also inhibits epiphyseal growth plate chondrocytes (Hicks et al., 1996). Moreover, lead may impair growth by inhibition of the hypothalamic-pituitary-growth axis resulting in lower serum insulin-like growth factor (Fleisch et al., 2013), and by inhibition of the hypothalamic-pituitary-gonadal axis resulting in delayed pubertal onset in girls (Selevan et al., 2003) and boys (Williams et al., 2010). Our findings on standing height are consistent with an effect of lead exposure on bone development. Given that most children in our study (74.3%) had BLL lower than 5 μg/dL, our results suggest that exposure to low levels of lead may hinder linear growth in otherwise normally growing children. Moreover, we found a somewhat stronger association of lead on height in girls than in boys (although confidence intervals were overlapping). A longitudinal study in the U.S. enrolled girls at 6–8 y and followed them up to 14 y (Deierlein et al., 2019). At 7 y of age, the difference in height was −2.0 cm for girls with blood lead ≥ 1 μg/dL compared to those with lower blood lead; this difference persisted, although attenuated, with advancing age. In contrast, in a study of girls and boys < 6 y of age, blood lead was negatively associated with height-for-age Z score only in boys (Zhou et al., 2020). The potential differential effects of lead exposure on growth between the sexes have not been extensively studied and need further replication and clarification because they could contribute to delays in puberty.

When stratified by iron status, children with ID had higher BLL than those without ID. This is consistent with results from previous studies in different settings (Bradman et al., 2001; Carvalho Rondo et al., 2011; Dislavo et al., 2009). Decreased iron reserves have been shown to upregulate the expression of DMT1 in mucosal cells (Martini et al., 2002), which in turn favor intestinal lead absorption (Bannon et al., 2002) thus contributing to increased BLL. It is reasonable to hypothesize that a better iron status protects children from toxic effects of lead.

However, in our study there was little evidence of an effect modification in the association between BLL and HAZ by iron status. This is an unexpected result, and suggests that children are at risk for the potential growth impairment effects of lead irrespective of iron status. Besides iron reserves, poor status or imbalance of other micronutrients essential for growth such as zinc might interact with lead exposure and impair linear growth of children. Additionally, several caveats are worth considering with respect to our findings. First, this was a cross-sectional study, which did not permit us to establish whether lead exposure or iron deficiency developed first in children. Second, both BLLs and serum ferritin represent short-term biomarkers; serum ferritin in particular is responsive to changes in iron supply in the diet. Without a history of earlier lead exposure or iron deficiency, it is unclear whether the effects of lead on growth occurred earlier, and the current BLL is simply correlated with exposures occurring in earlier life. In this vein, none of the study children had anemia, yet earlier studies in this population suggest that anemia, due to severe iron deficiency, is fairly common at younger ages (Queirolo et al., 2010). We did not have information on nutritional status of other micronutrients. For all these reasons, it is difficult to draw definite conclusions regarding the interplay of iron status, lead exposure, and child growth. Finally, our sample size was relatively small overall (n=259) and especially among children with low iron stores (n=96). Larger studies are needed to confirm or dispute these findings.

An inverse association between dairy intake and BLL has been found in preschool children highly exposed to lead (Li et al., 2019) and in adolescents with low lead exposure (Kim et al., 2017). This can be explained by the competition between calcium and lead for intestinal absorption based on animal and human adult studies (Kordas, 2017). A higher consumption of calcium and dairy foods was inversely associated with lead levels in a previous study in Uruguayan children (Kordas et al., 2018a). In the present study, there was no difference in BLL between children with dairy intakes below or above 2.5 daily servings and we found little effect modification in the lead-height for age relationship by the level of intake of dairy foods. The potential benefits of dairy food consumption in reducing lead toxicity/absorption should be addressed with additional research in other age and population groups.

There is conjecture (Kortenkamp et al., 2007) and now growing evidence that multiple metals can impact health outcomes like children’s development (Claus Henn et al., 2014), due to their synergistic interactions even at fairly low levels. However, the effects of co-exposure to lead and other toxic metals such as cadmium and arsenic on child growth are largely unknown, particularly in school-age children. Cadmium, like lead, accumulates in bone and cadmium exposure has been inversely associated with the plasma concentrations of the growth factor IGF-1 in 9 y old children from Bangladesh (Igra et al., 2019), as well as lower bone mineral density and proportion of minerals such as zinc, copper and iron in female rats exposed from weaning into adulthood (Brzózka et al., 2005). An in vitro study of human osteoblasts exposed to a range of doses of lead or cadmium showed that both metals, particularly cadmium, were cytotoxic to these bone cells via the production of reactive oxygen species and increased oxidative stress (Al-Ghafari et al., 2019). We have previously shown that low-level urinary arsenic in the same population of children was associated with higher biomarkers of oxidative stress (Kordas et al., 2018b) and other studies indicate that arsenic exposure is related to lower osteoblast differentiation and altered bone mineral density and microstructure (Wu et al., 2014). A mixture consisting of arsenic, cadmium and lead, administered during in the early postnatal period synergistically inhibited osteoblast viability and differentiation, and resulted in shortened tibias and growth plates in female rats (Abbas et al., 2013). All three metals affect bone or general molecular processes in the body through similar or common mechanisms that make synergistic effects of co-exposure on children’s growth plausible.

In our study, children with higher U-Cd and higher U-As had somewhat higher BLL concentrations but the exposure to these toxicants was generally low, so the group differences in BLLs were not statistically significant. The negative association between BLL and HAZ did not differ in children with U-Cd below or above median levels. The association was somewhat stronger in children with lower U-As but ultimately, with overlapping confidence intervals, there was limited evidence of effect modification by these toxicants in the relationship between BLL and height for age. A study of school-age children selected from the general U.S. population, noted potential interactions between manganese but not lead with other metals within the mixture (selenium, fluoride, manganese, lead, mercury) being investigated on standing height and upper arm length (Signes-Pastor et al., 2020). They did not include arsenic or cadmium in their study, preventing direct comparison with our findings. We also looked into studies in adult populations to inform the interpretation of our stratified analyses but that literature focuses mainly mineral density (BMD) and fractures. The effects of metal co-exposures on these endpoints appear to be mixed, and may depend on the level and duration of exposure. For example, in a study of adults from China with relatively high levels of environmental exposure, women with both high lead and high cadmium exposure had higher likelihood of low bone mineral density than women with high exposure to only one of the metals (Chen et al., 2014). On the other hand, among Spanish adults, low-level U-As and U-Cd were not associated with bone mineral density (Galvez-Fernandez et al., 2020). There is clearly an important need to further investigate the effects of toxicant co-exposures on child growth.

In previous studies, prenatal cadmium exposure was inversely associated with height for age Z score in 5 y old children In Bangladesh (Gardner et al., 2013), but there are few other studies to inform the relationship between Cd and body growth in children. U-Cd is a marker of lifetime exposure, and this timing suggests that synergistic effects on linear growth, together with chronic lead exposure, are possible. In all of our study children the levels of the urinary biomarkers of cadmium, as well as arsenic, were low compared to other populations (Gardner et al., 2013; Igra et al., 2019). U-As concentrations in our study were associated with recent dietary intakes, particularly of rice (Kordas et al., 2016), suggesting that this marker of recent exposure, if measured at a single time point, may not capture well a potential effects of arsenic on children’s body size. Furthermore, the half-life and the timing of exposure that U-As and BLL represent may be mismatched. On the other hand, U-As were inversely associated with height for age Z score in 5 y old children In Bangladesh (Gardner et al., 2013), albeit at much higher levels of exposure. Additional studies are needed to investigate any potential synergistic effects of As, Cd and BLL exposure on child growth.

Our study should be interpreted in light of some limitations. First, due to our study design (cross-sectional samples), the observed relationships between metal biomarkers and anthropometric measurements of the children cannot be interpreted as causal, but point to new research directions for understanding and addressing complex determinants of suboptimal child growth. We have no information on previous exposure history of the study children but have evidence that lead exposure has been gradually declining in Uruguay (Kordas et al., 2010). BLLs in children also generally track over time (Deierlein et al., 2019), suggesting that those with higher levels at the time of our study likely had higher levels at younger ages. The effects of lead on anthropometric endpoints appear consistent from 7 to 14 years of age, although diminish slightly over time for height (Deierlein et al., 2019). Second, while over one third of the children recruited into our study needed to be excluded due to missing information, we found that they did not differ from the complete-case sample on key characteristics, thus reducing the possibility of biased findings. Third, another statistical consideration that may affect the interpretation of our findings is that we did not account for multiple comparisons by adjusting the significance level. We conducted ~30 different regression models, increasing the possibility of committing type 1 error. Nevertheless, the fact that our findings on the association between BLL and height for age were consistent with previous literature, that we found no evidence of the hypothesized effect modification, and no associations between BLL and other endpoints increases confidence in our findings. Fourth, dietary characteristics were based on recalls over 2 days and may not represent typical intakes. Furthermore, the range of foods in our database was relatively small, especially compared to the USDA database. Nevertheless, other considerations increase our confidence in our data. For example, the mean energy intake of ~2200 kcal suggests that under-reporting of children’s diets by their caregivers was not a major issue; we were specifically interested in dairy intake and believe we were able to capture the majority of dairy foods consumed in early school years; dairy foods and especially milk and yogurt constitute important staples of children’s diets, suggesting that 24 hour recalls are more likely to represent typical intakes of these foods compared to other items. Fifth, most biomarkers represented short or very short-term exposure metrics, from days (U-As) to several weeks (BLL). Only U-Cd represented long-term exposure but the levels of this biomarker were comparatively low. We have previously discussed the importance of metal biomarker choice in interpreting associations with molecular endpoints/processes (Kordas et al., 2018b). These considerations are important for more distal health endpoints as well.

As a strength of the study, we compared children’s anthropometric parameters to sex- and age-appropriate international growth reference curves established for school-age children and adolescents (de Onis et al., 2007), calculating HAZ and BMI Z-scores. Based on these calculations we were able to conclude that on average, the children’s linear growth was as expected, although we observed a marked prevalence of overweight/obesity. Nevertheless, we found that even among apparently healthy growing children, lead exposure was associated with lower height for age. An additional strength of our study is that using stratified models, we were able to test for effect modification in the relationship between lead and anthropometric measures by sex, iron status, and dairy intake, variables that could potentially influence children’s growth and lead exposure. We also accounted for co-exposure to other metals, investigating potential effect modification by arsenic and cadmium exposure. Furthermore, all models were adjusted for sociodemographic factors and for total dietary energy intake, all known to affect both metal exposure and growth. Although the study used different matrices (urine, blood), the biomarkers selected for modeling represented gold-standard indicators of metal exposure.

In conclusion, exposure to low levels of lead was associated with lower height for age in apparently normally growing urban school-age children in Uruguay. The relationships between exposure to low levels of toxic metals and child growth are complex and should be examined while accounting for other pertinent factors.