DIET QUALITY AND BLOOD LEAD LEVELS IN URUGUAYAN FIRST GRADERS

Abstract

Objective:

Foods are a potential source of lead but also contain nutrients that counteract the intestinal absorption of lead. Translatable evidence is needed to better understand the relationship between whole diets and blood lead levels (BLLs) in children. In this cross-sectional study we investigated the association between dietary variety, adequacy, moderation, and overall diet quality with children’s BLLs.

Methods:

First graders (~ 7 years) from Montevideo, Uruguay were evaluated in years 2009–13 (Phase I, n=303) and 2015–19 (Phase II, n=443). Lead was measured in fasting blood via atomic absorption spectrometry and primary caregivers completed two non-consecutive 24-hour recalls. Indices of dietary variety (maximum score of 20), adequacy (maximum 40), moderation (maximum 30), and overall diet quality were calculated from food and nutrient intakes averaged over two recall days. The indices were adapted from the Diet Quality Index-International based on dietary recommendations for children. Missing covariate data were imputed. The associations between diet quality measures and BLLs were modeled separately in Phase I and II with multivariable linear models, testing for sex differences via interaction terms and stratified models.

Results:

The median[IQR] BLLs were 3.8[2.6, 4.9] and 1.3[0.7, 3.6] µg/dL in study Phases I and II, respectively. All diet quality scores were less than 60% of the maximum values and daily intakes of key food groups fell below recommended levels. There was little evidence of an association between measures of diet quality and children’s BLLs overall. On the other hand, higher diet variety, adequacy, and overall diet quality were associated with slightly higher BLLs among girls in Phase I only.

Conclusions:

In the context of low lead exposure and food consumption patterns that fall short of recommended levels, the relation of diet quality to children’s BLLs may depend on lead exposure levels.

Introduction

Although much progress has been made globally in reducing children’s blood lead levels (BLLs)¹, lead exposure affects millions of children across the world^2,3. There are no safe BLLs and levels <5 µg/dL are associated with a range of adverse neurobehavioral effects, including lower IQ and behavioral problems^4–6. Among different sources of lead exposure for children, food is of special interest. Children are vulnerable to dietary lead exposure because they could absorb more than 40% of the lead they ingest⁷ and this level could increase with iron deficiency^8,9. A recent scoping review highlighted the presence of lead in natural or minimally processed foods worldwide¹⁰. Vegetables, roots, tubers, and cereals were most affected, but other food groups (pulses, seeds and nuts, milk, eggs, fruits, meat and fish) also had lead in excess of the minimum limits (MLs) set for those foods by international standards¹⁰. In the U.S., consumer advocate groups have reported the presence of lead in commercially prepared as well as homemade baby foods^11–13. The Total Diet Study conducted by the U.S. Food and Drug Administration (FDA) has shown that some foods contain relatively high levels of lead (e.g., the mean concentration of lead in a 100 g serving of spinach was 0.6 µg)¹⁴.

The FDA also sets the maximum daily intake of lead from the diet, known as the Interim Reference Level (IRL). This level has been determined to be 2.2 µg/day for children^15,16 and would result in BLLs in excess of 3.5 µg/dL, the current population value set by the Centers for Disease Control and Prevention (CDC). Given the potential for dietary exposure to lead, reduced consumption of contaminated foods seems like an obvious approach, but any intervention must be based on a proper understanding of the relationship between whole diets and BLLs, and among people of varying ages and lead absorption levels, because the same foods may be excellent sources of critical nutrients^17,18.

It is well known that some nutrients prevent lead absorption, thereby influencing the level of lead circulating in the blood. For example, the CDC recommends that children consume a diet high in iron, calcium, and vitamin C as a secondary approach to prevent elevated BLLs. Nevertheless, the CDC has also acknowledged that the effectiveness of these recommendations has not been formally tested¹⁹. A deeper analysis indicates that the recommendations are based on the estimated consumption of nutrients rather than whole foods or diet patterns²⁰. But children eat foods rather than isolated nutrients; diets are mixtures of multiple nutrients that interact with each other. A recent study based on the National Health and Nutrition Examination Survey showed that higher consumption of cereals and milk was associated with lower BLLs among 12–36 month old US children²¹. In school children from Uruguay, higher consumption of dairy products was associated with lower BLLs²².

The role of food consumption in relation to lead exposure is complex, and focusing on individual foods or the intake of individual nutrients alone may be too simplistic, resulting in contradictory information and making dietary choices difficult²³. Instead, considering whole diets acknowledges this complexity and may better inform dietary recommendations to prevent or lower lead exposure in young children. One way to encompass the diet (as a whole) is to measure dietary patterns or quality. Our group has shown that data-driven diet patterns reflecting the consumption of nutrient-rich foods were associated with lower urinary arsenic but not lower BLLs^22,24. On the other hand, scored diet quality indices, such as such as the Healthy Eating Index²⁵, the Mediterranean Diet score ²⁶, or the Diet Quality Index-International (DQI-I)²⁷ reflect adherence to specific dietary recommendations designed to promote health. A study in Spanish children aged 4–5 years showed no association between scores on a relative Mediterranean Diet (rMED), the alternative Mediterranean Diet (aMED), or the components of either score with urinary lead concentrations in these children²⁸. While diet quality measures can be translated into recommendations for a healthy lifestyle and facilitate cross-setting comparisons, there are no studies linking diet quality to children’s BLLs. The objective of this cross-sectional study was to assess the association between dietary quality and BLLs. We used a modified version of the DQI-I to assess the diet quality of elementary school children from Montevideo, Uruguay. The DQI-I was developed for use across cultures, focusing on four characteristics of a healthy diet – its variety, adequacy, moderation, and overall balance. We hypothesized that overall DQI, as well as the individual components of the DQI, would be negatively associated with BLLs among children. Based on sex differences observed in the effect of BLLs²⁹ on developmental outcomes, we hypothesized that the aforementioned associations would differ by sex.

Methods

Study Design

This study leveraged data from the Salud Ambiental Montevideo (SAM) cohort in Montevideo, the capital of Uruguay. The cohort is being conducted among communities considered at risk for metal exposure based on previous epidemiological studies or knowledge of the pediatric population served by the Clinic for Environmental Contaminants at the Pereira Rossell Hospital. Participant evaluations occurred during two periods, in 2009 – 2013 (Phase I) and 2015 – 2019 (Phase II), and are described in detail elsewhere³⁰. The first phase encompassed private schools, while the second phase also included public schools because the study received the long-sought permission to work within the state educational system. A five-level socioeconomic status (SES) categorization is used to describe the SES of the underlying population served by schools in Uruguay: 1 represents high SES and 5 very low SES. Schools classified as SES 3–5 were targeted for this study. Children were invited to participate through the school, and in later years, television, radio, newspaper ads, and postering in public spaces were also employed. All first-grade children who regularly attended school were eligible, the sole exclusion criterion being a previous diagnosis of lead poisoning (defined as BLL >45 µg/dL). No child was excluded based on this criterion.

Interested caregivers attended an informational meeting where the rationale and procedures of the study were explained in detail. General information on childhood lead exposure like common sources, route of exposure, absorption, and health effects was provided at these meetings. Questions were answered and consent forms were provided. The voluntary nature of the study was stressed to the caregivers, as was the confidentiality of their information. Caregivers had the option of providing their consent at the end of the meeting or taking the forms home to discuss the study with other family members. The study received approval from institutional review boards or ethics commitees at the University at Buffalo, Catholic University of Uruguay, and University of the Republic of Uruguay.

Once caregivers agreed to participate in the study, they attended a meeting at the school (Phase I) or research center (Phase II) to fill out questionnaires about socio-demographic characteristics of the family, the child’s medical history, parenting ideas, parental stress, and the home environment. Following this, children participated in multiple study visits, one of which consisted of anthropometric measurements and the collection of a fasting blood sample.

For this study, samples from the two study phases were kept separate to account for differences in study recruitment as well as the secular trends in BLLs that occurred between 2009 and 2019³⁰.

Study sample

In Phases I and II, 367 and 515 children and their caregivers, respectively, were evaluated. For this analysis, only children who were fully enrolled (had caregiver consent to participate, had completed at least one study assessment, and were not a sibling of another study child), and had complete laboratory data and dietary recalls were included in the analysis. Younger children of each sibling group were excluded. Figure 1 outlines the sample selection. Furthermore, while the sample had missing values for covariates, multiple imputation by chained equations (MICE) was utilized to impute these missing values, as described in the statistical approach section.

Figure 1. Flow chart of sample selection for the study on diet quality and blood lead levels among Uruguayan first-graders.

¹Fully enrolled=the child was invited to participate, a caregiver provided consent, the child completed at least one evaluation, and the child was not a sibling of another study child;

²BLL=blood lead level.

Measures

Anthropometric measurements

Anthropometric measurements were taken by trained pediatric nurses or nutritionists. Child height was measured in triplicate to the nearest of 0.1 cm, using a portable stadiometer (Seca 214, Shorr Productions, Colombia, MD) and an average was taken. Children were weighed in triplicate to the nearest 0.1 kg using a digital scale (Seca 872, Shorr Productions, Colombia, MD) while wearing light clothing like school uniforms. The three measurements were averaged, and the final weight was calculated by subtracting standard weights of children’s clothing worn at the time of measurement. Based on the average weight (corrected for clothing) and average height measurements, the child’s BMI was calculated and expressed as kg/m².

Biological sample collection

After an overnight fast, a venous blood sample (3 mL) was collected from each child by a trained phlebotomy nurse, using a 25-gauge safety butterfly blood collection set (Vacutainer, Becton Dickinson, Franklin Lakes, NJ) in heparin-coated tubes (Vacutainer, Becton Dickinson, Franklin Lakes, NJ). Time of collection was noted, and the whole blood tubes were stored on ice in a cooler for the remainder of the clinic visit, after which they were transported to CEQUIMTOX (Specialized Center for Chemical Toxicology) at the Faculty of Chemistry at the University of Republic (Montevideo, Uruguay) and stored at −20°C until analysis. An additional 3 mL of venous blood was drawn into a serum tube with a clot activator and separator gel (Becton Dickinson, Franklin Lakes, NJ). A drop of this sample was used for hemoglobin measurement before it was allowed to clot.

Whole blood lead analysis

Depending on the volume of whole blood available, BLLs were measured at the CEQUIMTOX using Atomic Absorption Spectrometry (AAS, VARIAN SpectrAA-55B) with flame or graphite furnace ionization techniques. The graphite furnace utilizes a lower volume of specimen and was used if the volume of blood available for analysis was below 2 mL. If more whole blood was available, then the flame AAS (FAAS) method was used. However, BLLs for all Phase II samples were measured by graphite furnace AAS (GFAAS). The limit of detection (LOD) was as follows and reflects the attempt of the laboratory to keep up with the trends in children’s BLLs over time. For FAAS, it was 2.5 µg/dL in years 2009–2010 and 1.8 µg/dL in years 2011–13. For GFAAS, the LOD was 2.0 µg/dL in years 2009–10, 0.8 µg/dL in years 2011–13, 1.0 µg/dL in years 2015–16 and 0.4 µg/dL in years 2016–19. Values below the LOD were assigned a value equal to the LOD divided by the square root of 2. Analytical conditions were validated with standard quality assurance/quality control (QA/QC) procedures. The laboratory participates in accuracy certification through the Interlaboratory Program for Quality Control for Lead in Blood, Spain (PICC-PbS, 2001). For 95.1% of the 184 interlaboratory program blood samples, analytic results were CV<2%. CEQUIMTOX also participates in the U.S. CDC Lead and Multi-Element Proficiency Program.

Hemoglobin measurements

Hemoglobin was measured at the time of the blood draw from a drop of venous blood using a portable hemoglobinometer (HemoCue Inc, Lake Forest, CA). Quality control checks were performed daily using standard HemoCue controls (low, medium, high) provided by the manufacturer.

Assessment of Dietary Intake

Dietary intake was assessed through two 24-hour dietary recalls conducted by five trained nutritionists with the child and mother or another caregiver familiar with the child’s diet. The child contributed to the recall by recounting the food consumed at school or other occasions the caregiver was not present. The first recall took place at the school or research center on the day of the blood draw and the second took place over the phone without prior appointment (median [5%, 95%] time elapsed between the recalls was 19 [13, 47] days). The nutritionist made a maximum of three attempts to complete the telephone interview, and if unanswered or the primary caregiver was not available, the recall was not conducted.

A detailed list of all the foods and beverages the child consumed within the previous 24-hour period was collected. Information was obtained about the name of the meals, time, and place of consumption, amounts of foods consumed or food portions, food preparation methods, recipe ingredients and brand name of commercial products. Food models and household measurement cups were used during the in-person interview to help quantify the foods and beverages consumed. Intake of solid foods was recorded in grams and of drinks in milliliters. Questions were also asked about the intake of vitamin and mineral supplements. Neutral probing questions such as “Did your child eat/drink anything on the way home from school yesterday?” “Did he/she have anything before going to bed?” were asked to get accurate dietary information. All foods were assigned a unique code and entered, along with amounts consumed, into a database that contained the nutrient composition of 359 typical Uruguayan foods and preparations. These intakes were averaged over the two recall days.

Food groups

Serving sizes of each food item were determined in consultation with a registered dietitian and a public health nutritionist, and were guided by the UDSA Pyramid³¹, and MyPlate³². Because many individual foods did not represent normal distributions and were consumed by small numbers of children, twenty-five food groups were created based on similarities and frequency of consumption of individual foods that were averaged across 2 days of intake. These groups were: 1) fruit, 2) dark leafy vegetables, 3) red and orange vegetables, 4) beans and peas, 5) other vegetables, 6) potatoes, 7) breads, 8) grains, 9) pasta, 10) red meats, 11) white meats, 12) processed meats, 13) eggs, 14) milks, 15) cheeses, 16) yogurt, 17) soy products, 18) fats and oils, 19) milk-based desserts, 20) sweets, 21) pastries, 22) potato chips and fried potatoes, 23) sweetened beverages, 24) pizzas and dinner pies, 25) sauces and condiments. Each food item reported by children/caregivers was placed in one of the 25 groups.

Diet Quality Index

The quality of children’s diets was determined according to measures similar to the DQI-I²⁷, which was originally developed for U.S. and Chinese adults. The original DQI-I consists of four components: dietary variety, adequacy, moderation, and balance. Given that the DQI-I was being adapted for use in children, certain modifications were made, as described below. The DQI-I scoring system as modified for use in this study is shown in Online Table 1. For consistency with how the DQI-I was derived and to facilitate comparison with international samples, the assessment and score calculations for diet quality in this study are based on the recommendations for the U.S. population, but it is important to highlight that in 2020, the Uruguayan Ministry of Public Health published specific energy and nutrient intake recommnedations for the Uruguayan population³³. These are listed for children aged 6–9 years (except for calcium and iron which are given for 4–8 yeear olds) in Online Table 2.

The score for dietary variety was based on consuming foods from multiple groups and having protein from different sources. The five food groups considered were meat/fish/eggs, dairy/beans, fruits, vegetables, and grain. Higher number of food groups represented in the diet received higher number of points (e.g., having all 5 food groups scored 15 points). Furthermore, consuming protein from multiple sources such as red meat, white meat, dairy, beans, eggs, and fish received higher number of points (e.g., 3 sources or more was equivalent to 5 points). The maximum possible score on this component was 20 points.

Scores for dietary adequacy were based on the degree to which children met the daily recommendations in the 2020 – 2025 Dietary Guidelines for Americans³⁴, as described in more detail in Kim et al., 2003²⁷. The maximum possible score on this component was 40 points.

Due to lack of key input variables, the dietary balance component of the DQI-I could not be calculated. Specifically, the fatty acid ratio, PUFA:MUFA:SFA of the balance component²⁷ could not be calculated because a comprehensive database on values of polyunsaturated fatty acids (PUFA), monounsaturated fatty acids (MUFA) or saturated fatty acids (SFA) in the foods consumed in Uruguay is not available.

Finally, the dietary moderation component of DQI-I was calculated based on certain modifications to the original measure. Children’s intake of saturated fat, cholesterol, sodium, or empty-calorie foods was unavailable in this study due to the absence of comprehensive nutrient databases for foods consumed in Uruguay. To calculate a moderation component, available data on the consumption of sweets (candy, chocolate), pastries, sweetened/carbonated drinks, fries, chips, and savory snacks, as well as total fat intake were used instead (see Online Table 1 for details). The maximum possible score for this component was 30 points.

Statistical analysis

All statistical analyses were performed using Stata 18 (College Station, TX). The outcome (dependent) variable, BLLs, was treated as continuous and was untransformed for this analysis to aid interpretation. The predictor (independent) variables were the continuous dietary adequacy, variety, moderation index scores, and the overall diet quality index, which were calculated as described above.

Multiple imputation

The study covariates were missing between 1 and 37 observations (0.33 – 12.2%) in Phase I, and 2 – 79 observations (0.45 – 17.8%) in Phase II (see Online Table 3 for details). Multiple imputation via chained equations (MICE) in Stata was used to impute the missing information. Values for blood hemoglobin, drinking water source, BMI, maternal age, crowding, parental smoking status, maternal education and employment, and presence of children below 5 years of age were imputed using regressions (logit, ologit, etc.) appropriate to the distribution of each variable, the augment option to aid in the imputation of categorical variables, and 10% burn-in. On the other hand, child sex and age, child BLL and value <LOD, sex by BLL interaction term, year of enrollment, child’s school, and diet quality indices were fully observed. Fifty imputed data sets were created and combined using Rubin’s rules for use in all analyses described below.

Descriptive analyses

Continuous variables across the imputed datasets were summarized as mean ± SE or median[IQR], and frequencies were used for categorical variables. Scatterplots of BLLs against the diet quality scores were used to check the linearity assumption between predictors and the outcome.

To further understand the associations between diet quality and BLLs, we examined the Spearman correlation coefficients between individual food groups and BLLs. Analyses were performed for the absolute amounts consumed (measured in grams for solids and milliliters for liquids), and for the number of servings consumed (number of servings = amount consumed/ serving size).

Regression analysis

Separate regression models were fit for Phase I and Phase II enrollment. In Phase I, BLLs were modeled in ordinary least squares (OLS) regression as a function of variety, adequacy, moderation and overall DQI. The models included a cluster option to account for potential correlations within schools and differences among schools. The models for Phase II did not account for school due to a broader community enrollment. OLS regressions were also used in this analysis.

Based on the current literature, the following variables were considered as potential confounders: child sex (male/female), child age (continuous), child hemoglobin level (continuous), current smoking status of the parents (no/yes), maternal age (continuous), maternal education (continuous), crowding (a continuous variable calculated as the number of people living in the house divided by number of rooms), family possessions (count variable ranging 0–5), and source of drinking water (tank/tap unfiltered, tap filtered, bottled plus other source)^35–37. The final covariate selection was based on a DAG (Online Figure 1) and included child age and sex, child’s blood hemoglobin, maternal age and education, household crowding and possessions. Additionally, an indicator variable (no/yes) of whether the BLL value was below the analytical LOD. Effect modification by sex was explored by adding an interaction term to the above models. Stratified analyses were conducted where interactions were statistically significant at an alpha level of p <0.1.

Results

Participant characteristics

Participants in Phases I and II of the study were similar with respect to age (81.1 – 83.0 months), hemoglobin level, BMI, and maternal age (Table 1). With respect to nutritional status, on average, children had healthy hemoglobin values and indices of body fatness. Several characteristics differed between the phases, with the earlier study period having more boys, somewhat higher maternal education, and higher proportion of employed mothers. On the other hand, the later phase had more smoking caregivers and more households used unfiltered water for drinking. The median BLL was higher in Phase I than Phase II (3.8 µg/dL and 1.3 µg/dl, respectively), reflecting the secular trends in this population.

Table 1:

Characteristics of schoolchildren participating in the study on diet quality and blood lead levels 2009–2019, based on samples with imputed covariate data.

Characteristic	Mean ± SE, Median [IQR] or %
	Phase I (2009–13) N=303	Phase II (2015–19) N=443
Characteristics of the study child
Child’s age (months)	81.1 ± 0.4	83.0 ± 0.3
% Female	43.6%	49.9%
Child BLL (µg/dL)	3.8 [2.6, 4.9]	1.3 [0.7, 3.6]
Child Hemoglobin (g/dL)	13.2 ± 0.1	13.3 ± 0.04
Child BMI, kg/m2	16.9 ± 0.1	16.8 ± 0.1
Dietary variety score (max. is 20)  ≥15 points1	11.4 ± 0.2 31.7%	9.7 ± 0.2 20.8%
Dietary adequacy score (max. is 40)  ≥26 points1	22.3 ± 0.3 36.0%	21.0 ± 0.2 18.5%
Dietary moderation score (max. is 30)  ≥15 points1	11.5 ± 0.3 26.7%	14.1 ± 0.3 56.0%
Total diet quality score (max. is 90)  ≥51 points1	45.3 ± 0.5 26.7%	44.8 ± 0.5 30.2%
Characteristics of the parents/household
Maternal age, years	33.3 ± 0.4	33.7 ± 0.3
Maternal education, years	9.1 ± 0.1	8.0 ± 0.1
Mother employed %	69.1%	54.5%
Parent smokes %	53.8%	64.5%
% children <5 years in household	33.5%	38.3%
Household inhabitants/rooms	1.95 ± 0.04	2.3 ± 0.1
Source of drinking water
Tank/Unfiltered tap	31.0%	60.4%
Filtered Tap	20.7%	2.1%
Bottled/Other source	48.5%	37.5%

¹For comparison, the score represents the 75^th percentile of the distribution in Phase I.

Diet quality and dietary intakes

Diet quality scores are shown in Table 1 and indicate that dietary variety and adequacy scores were higher in Phase I than Phase II. Furthermore, a greater proportion of children in the earlier compared to later phase achieved higher scores (31.7% vs. 20.8% children had variety scores ≥15 points and 36.0% vs. 18.5% had adequacy scores ≥26 points). On the other hand, 56% of children in Phase II (vs. 26.7% in Phase I) had scores ≥15 points on dietary moderation. The overall dietary quality was very similar across the study years.

To further understand the quality of the diet in this population sample, dietary intakes within broad food groups (vegetables, fruit, grain, meat/eggs and dairy) were compared between study phases for boys and girls separately (Table 2). Overall, boys consumed more food than girls, and the participants in Phase I consumed greater mean amounts of food within the five groupings compared to those in Phase II. More information on the quantities and servings consumed within 25 more detailed food groupings are shown in Online Tables 4-7.

Table 2:

The consumption (given as quantities in grams or milliliters estimated from two non-consecutive 24-hour recalls) of vegetables, fruits, grains, meats/eggs and dairy among Uruguayan boys and girls in study years 2009–13 and 2015–19.

	Children enrolled in 2009–13 (Phase I of study) Given in g or mL					Children enrolled in 2015–19 (Phase II of study) Given in g or mL
Food group	N	Mean (SD)	Med	Min	Max	N	Mean (SD)	Med	Min	Max
Vegetables
Male	170	56.38 (76.31)	25	0	395	219	49.25 (66.53)	15	0	300
Female	133	51.9 (73.65)	25	0	412.5	224	41.19 (62.65)	9	0	400
Fruits
Male	170	128.74 (142.98)	100	0	950	219	104.32 (124.26)	75	0	1017.5
Female	133	137.19 (135.11)	100	0	682.5	224	106.4 (117.79)	75	0	672.5
Total grains
Male	170	128.83 (82.79)	111.25	0	360	219	111.88 (91.26)	90	0	640
Female	133	93.82 (70.25)	75.0	0	395	224	109.27 (81.97)	90	0	385
Meat/eggs
Male	170	116.14 (89.31)	102.5	0	450	219	84.73 (80.27)	65	0	380
Female	133	96.7 (78.75)	85	0	422.5	224	81.56 (78.84)	67	0	547.5
Dairy
Male	170	419.37 (196.44)	412.5	0	1065	219	324.06 (178.57)	332	0	800
Female	133	355.61 (167.11)	378	0	955	224	310.48 (176.74)	306.05	0	878

Dietary variety and adequacy were moderately inter-correlated (Spearman’s rho=0.55, p<0.01), and all diet quality components were correlated to the overall score (rho ranging 0.53–0.78, p<0.01; Online Table 8). The intercorrelation in the intake of food groups was weak (Online Table 9).

Compared to the daily intakes recommended by the U.S. Department of Agriculture, the mean number of servings from all food groups was generally low in this sample (Figure 2).

Figure 2. Mean number of food group servings consumed by Uruguayan first-graders evaluated in 2009–13 (study Phase I) and 2015–19 (Phase II) relative to the daily levels derived from the MyPlate recommendations for children¹.

¹The lower and upper level of recommended intake is the same for dairy.

Relationship between diet quality and children’s BLLs

There was little evidence of an association between any of the diet components or overall diet quality and BLLs in the study children (Table 3). These generally null relationships were consistent between study phases, despite differences in the ranges of BLLs observed between the phases. The single statistically significant association was with dietary variety and BLLs in Phase I of the study, where children with higher variety scores had slightly higher BLLs (0.03 [0.01, 0.06]). The overall findings are consistent with null correlations between the consumption of the five main food groups (vegetables, fruit, grains, meats/eggs/dairy) and BLLs (Online Table 7).

Table 3:

The association between the diet quality its components with blood lead concentrations of children aged ~7 years living in Montevideo, Uruguay (2009–2019).

	β [95% CI]1
	Phase I (2009–2013)2,3	Phase II (2015–2019)3
Dietary Variety
Unadjusted	0.01 [−0.03, 0.06]	-0.03 [−0.07, 0.01]
Covariate-adjusted	0.03 [0.01, 0.06]*	−0.01 [−0.04, 0.03]
Dietary Adequacy
Unadjusted	0.02 [−0.02, 0.06]	−0.004 [−0.04, 0.03]
Covariate-adjusted	0.03 [−0.02, 0.09]	0.005 [−0.03, 0.04]
Dietary Moderation
Unadjusted	−0.01 [−0.09, 0.07]	0.02 [−0.01, 0.05]
Covariate-adjusted	−0.01 [−0.07, 0.05]	0.01 [−0.01, 0.04]
Diet Quality Index - Modified
Unadjusted	0.005 [−0.03, 0.04]	−0.001 [−0.02, 0.02]
Covariate-adjusted	0.01 [−0.02, 0.04]	0.004 [−0.01, 0.02]

¹Models based on imputed data with samples of 303 and 443 in Phase I and Phase II.

²Standard errors around the estimate in Phase I account for clustering by school.

³Models adjusted for child’s sex, age, blood hemoglobin (g/dL), maternal age (continuous), maternal education (continuous), crowding in home (a continuous variable calculated as number of people living in the house divided by number of rooms), and an indicator of lead value below the limit of detection (yes/no).

^*p<0.05.

To explore potential differences by sex, interaction terms between the individual components of DQI (or the overall DQI) and sex were added to the appropriate covariate adjusted model. Based on these models, any interactions with p-values <0.1 were further explored in stratified models. Accordingly, interaction terms were statistically significant for sex and dietary adequacy (p=0.02), variety (p=0.03) and overall DQI (p=0.02) in Phase I the study (Table 4). In Phase II, the interaction term between sex and variety was also below the criterion p-value (p=0.093). In stratified models, girls with higher dietary variety, adequacy, and overall diet quality scores had slightly higher BLLs in Phase I. Nevertheless, as in models for the full sample, the magnitude of these associations was very modest. In boys, diet quality metrics were unassociated with BLLs in Phase I. Conversely, in phase II there was no evidence of differences by sex.

Table 4:

Select associations between diet quality and blood lead concentrations of children aged ~7 years living in Montevideo, Uruguay (2009–2019), stratified by sex (stratification performed when p-value of an interaction term between sex and diet quality was <0.1).

		β [95% CI]1
	Phase I (2009–2013)2,3		Phase II (2015–2019)3
	Dietary Variety
Girls	132	0.07 [0.05, 0.09]*	222	0.02 [−0.03, 0.08]
Boys	171	0.01 [−0.03, 0.04]	221	−0.04 [−0.09, 0.01]
	Dietary Adequacy
Girls	132	0.10 [−0.005, 0.21]#	-	-
Boys	171	−0.002 [−0.04, 0.04]	-	-
	Dietary Moderation
Girls	-	-4	-	-
Boys	-	-	-	-
	Modified DQI
Girls	132	0.04 [0.005, 0.08]*	-	-
Boys	171	-0.01 [−0.04, 0.03]	-	-

¹Models based on imputed data with samples of 303 (Phase I) and 443 (Phase II).

²Standard errors around the estimate account for clustering by school.

³Models adjusted for child’s sex, age, blood hemoglobin (g/dL), maternal age (continuous), maternal education (continuous), crowding in home (a continuous variable calculated as number of people living in the house divided by number of rooms), and an indicator of lead value below the limit of detection (yes/no).

⁴Stratified models not performed due to ns. interaction term between sex and diet quality.

^*p-value for models within strata <0.05;

^#p<0.1

Discussion

There is a growing focus on children’s diets as a source of lead exposure. The associations between the intake of individual nutrients and BLLs have been used to make widely accepted recommendations on dietary approaches to control lead exposure in children¹⁹. More recent findings on the content of lead (and other heavy metals/metalloids like arsenic) in baby foods have prompted discussions on the elimination of certain foods from children’s diets. The potential risk of such eliminations is the reduced intake of beneficial nutrients that foods deliver and that are essential for children’s growth. Therefore, there is a need for research that considers holistically how healthy diets relate to BLLs in children. Relating indices of diet quality to BLLs is an important approach due to its translational potential and comparability across settings but to date it remains underutilized. Using data from a cross-sectional study, we investigated how different aspects of diet quality (variety, adequacy, moderation) and an overall quality index are associated with BLLs in school-age children from Montevideo, Uruguay. We took advantage of the secular changes occurring in children’s BLLs in the years 2009–19, and split the children into two groups allowing us to examine how diet quality may be related to BLLs in periods of higher (2009–13; median BLL = 3.8 µg/dL) and lower (2015–19, median BLL = 1.3 µg/dL) lead exposure. We found little evidence of an association between diet quality and children’s BLLs in either study period. We did find small differences between the sexes in Phase I, with more evident associations in girls than boys. Interestingly, in prior analyses of data from Phase I, we found some evidence for an inverse association between the intake of dietary zinc and calcium, but not iron or vitamin C, and children’s BLLs²².

There are few studies to which we can compare our findings. Recently, Notario-Barandiaran and colleagues found that the relative and alternate Mediterranean Diet scores (rMED and aMED, respectively) were not associated with urinary lead concentrations in Spanish children aged 4–5 years²⁸. In the same study sample, scores on the general and healthful pro-vegetarian dietary patterns were also not related to urinary lead levels³⁸. In both studies, the adherence to the Mediterranean and pro-vegetarian diet patterns was fairly low. Similarly, in our study, we had low intake of recommended nutrients and foods. In a previous analysis of data from Montevideo on children enrolled between years 2009 and 2013, we found that higher scores on a nutrient dense dietary pattern derived through a data-driven (not index-based) approach and characterized by higher consumption of leafy green and orange vegetables, among other foods, were associated with higher BLLs in children²².

Other studies conducted among adults lend additional context to the interpretation of our findings. For example, Wang and colleagues found no link between a prudent dietary pattern and BLLs in older men³⁹. In pregnant British women, the data-driven confectionery diet pattern was associated with lower likelihood of having a BLLs ≥5 μg/dL⁴⁰. On the other hand, traditional, healthy, processed, or vegetarian patterns were not related to women’s BLLs.

In the analysis stratified by sex we found evidence of differential association of diet quality and BLLs between girls and boys. Specifically, higher variety, adequacy and overall diet quality scores were related to slightly higher BLLs in girls but not boys. This was only evident in the early phase of the study. The reason for these sex differences is unclear. While the girls in Phase I of the study had somewhat higher BLLs (mean difference of 0.53 μg/dL), boys had ~ 1 point higher scores across the different aspects of diet quality. Boys also had higher consumption in the five major food groups identified in this study. Possibly, the combination of higher BLLs and lower diet quality scores in girls contributed to the observed differences in associations. Prior studies in children^28,38 adjusted for sex but did not investigate sex differences in the association between dietary quality or patterns and lead exposure metrics, so there is no basis for a cross-study comparison. The absence of sex differences in Phase II of the study may be due to markedly lower mean and narrower distribution of BLLs, as well as lower diet variety and adequacy scores in years 2015–19 than 2009–13. It is also important to point out that diet variety and adequacy were moderately correlated (rho 0.55) and contributed to the total diet quality scores; thus, the differential associations between boys and girls observed for these metrics cannot be viewed as wholly independent.

To further understand the associations between diet quality and BLLs, we examined the Spearman correlations between the servings consumed within the five broad food groups and BLLs. In short, food servings were uncorrelated with BLLs across study years, except for grains (rho = 0.10, p<0.01). These findings suggest that specific food groups are not driving the associations between diet quality and BLLs. The grains group included items like breads, pasta, and rice in addition to cereals. By law, Uruguayan wheat flour is fortified with iron and folic acid; all derivative products (baked goods, pasta, etc.) made from local flour would therefore be fortified. Other items, like cereals¹⁰, rice and rice products⁴¹ can be sources of lead, although lead levels in foods specifically consumed in Uruguay are as yet uknown. Combining fortified, unfortified foods and grains that potentially have higher lead content into one group might mask the correlation between the grains group and BLLs in this analysis.

A discussion of the overall dietary quality in this sample is warranted as context for the study’s findings. Using a modified version of the DQI-I we observed that diet quality in Uruguayan first graders was generally low. In both 2009 – 2013 and 2015 – 2019, variety, adequacy, moderation and overall DQI scores were less than 60% of the maximum possible scores. Low scores for diet quality are not unique to Uruguayan children. A study of children aged 8 – 10 years participating in the Quebec Adipose and Lifestyle Investigation in Youth (QUALITY) study reported average DQI-I scores of 57.9 of possible 100⁴². Similarly, average HEI score for children aged 5 – 8 years in the NHANES 2015 – 2016 cycle was 55 of possible 100; only children 2 – 4 years and adults over 60 years had HEI scores over 60. Like the DQI, scores below 60% of the maximum on the HEI are considered poor³⁴. Despite the wealth of information available on healthy foods, nutrition, and the effect of diet on health and disease, there is still great need to convert this information into practices that improve health outcomes.

The findings of this study point to the complexity in the relationship between daily diets and lead exposure. On the one hand, diets may be sources of exposure, as amply demonstrated in prior research^10,43,44. As we do not currently have information on lead contamination in the Uruguayan food supply, we cannot describe the extent of contamination. However, it is important to point to technical norms or decrees that assign the permissible lead level in drinking water at 0.03 μg/L⁴⁵ and maximum limits of lead (2.0 mg/kg) in foods generally, with additional limits established specifically for fats and oils (0.1 mg/kg), sweetened chocolate (1.0 mg/kg), and citrus fruit juices (0.3 mg/kg)⁴⁶. More detailed guidelines for 50 different food types (exclusive of baby foods) were also established in 2011 for the MERCOSUR region, which includes Argentina, Brazil and Paraguay in addition to Uruguay.⁴⁷ On the other hand, nutrient dense diets provide iron, calcium and other nutrients that compete for the absorption of lead in the gut^8,48,49. The diet quality scores calculated in this study are related to nutrients believed to compete for intestinal absorption or recommended by the CDC as approaches to manage elevated BLLs in children. For example, in study years 2009–13, the adequacy, variety, moderation, and total DQI scores were correlated with dietary iron intake at rho 0.29–0.49 and vitamin C intake at rho 0.12–0.39. Variety, moderation and total DQI were correlated more modestly with zinc (rho 0.15–0.25), and moderation and total DQI with calcium at rho of 0.21.

The mostly null associations observed in this study between diet quality and BLLs are contrary to our hypotheses; they do not suggest that higher quality diets may protect children by lowering BLLs. The two periods under study represented different exposure scenarios, where the median BLL in Phase I was three times higher than in Phase II. While the two periods differed on exposure levels, there was consistency in the overall findings, with the possible exception of the link between diet variety and BLLs in Phase I. On the other hand, sex differences were observed in Phase I but absent in Phase II. The complex nature of diets—as a source of contaminants and beneficial nutrients—makes it challenging to reconcile our findings. Although low statistical power is a possible explanation for the overall null findings, we believe the sample size for this study is quite robust, with each period comprising ~300–440 participants. The overall exposure scenario may be key. On the one hand, when lead exposure is lower overall, diets may neither contribute to nor moderate BLLs. In turn, in settings or scenarios with higher lead exposure, diets may play a more important role; for example, lead contamination in foods together with other sources, could increase a child’s BLL. This supposition would need to be tested in population groups with different levels of lead exposure and/or varied exposure sources. Furthermore, our findings are based on a sample with diets that do not meet current dietary recommendations; possibly, when diets are sub-optimal, they cannot be expected to moderate children’s BLLs. Moreover, metrics that measure overall diet quality may not reflect the key aspects of dietary intakes that would be related to lower lead exposure. Given these complexities, our results serve to motivate additional research in population groups with other levels of diet quality and lead exposure metrics, particularly utilizing longitudinal designs.

Our findings must be interpreted within study strengths and limitations. With respect to strengths, this study utilized a version of the DQI-I that was modified to suit the study population to assess the quality of the children’s diets. By substituting saturated fat, cholesterol, sodium, and empty calorie foods which made up the original moderation component of the DQI-I with sweets, pastries, carbonated drinks, and fries/chips/savory snacks we have a setting-appropriate measure of adherence to dietary guidelines. These substitutions also allowed us to overcome the limitations of working with incomplete or outdated nutrient databases often inherent to lower-resource settings. The calculation of indices of dietary variety, adequacy and moderation allowed us to investigate different aspects of diet quality in relation to heavy metal exposure. Furthermore, the focus on diet quality rather than individual nutrients or even food groups accounted for the synergy between foods and provided a more holistic picture of children’s diets in association with BLLs. To increase the representativeness of dietary intakes, two 24-hour recalls were conducted over weeks apart and the average of those recalls was used in this analysis. Additionally, food aids and models were used to aid in recall and quantification of the amounts of foods and beverages that were consumed. BLLs were assessed using validated methods following a rigorous quality control process. Multiple variables were assessed as potential confounders and the multivariable model accounted for a comprehensive set of covariates. We conducted separate analyses for two study periods that reflected secular changes in lead exposure. Finally, multiple imputation of missing covariate data yielded robust sample sizes.

On the other hand, we acknowledge the following limitations. Importantly, the cross sectional nature of the study precludes any conclusions about causal relationships between diet quality and BLLs. Second, our modified scoring system for the moderation aspect of the DQI had not been tested previously and is not validated. However, diet quality indices that are currently validated for use in children require very specific information on nutrients⁵⁰ that are not readily available for diets consumed in Uruguay or other lower-resource settings. Scoring for the variety and adequacy components of the DQI-I is based on the number of servings from various food groups. We used the USDA Food Pyramid to define serving sizes, but other definitions might have yielded different results. Future studies should also gauge the dietary intakes of nutrients and foods against specific recommendations³³ established for the Uruguayan population. For dishes that were unique to Uruguay we identified similar American dishes and used the serving size definition for the American equivalent; again, reflecting the limitation of working with nutrient databases in under-resourced settings. With respect to 24-hour recalls, we acknowledge their limitations, including the possibility to miss the consumption of seasonal foods. To overcome some of the issues related to this tool, we used two non-consecutive days of recall spaced at least 2 weeks apart and averaged those values. Another caveat is that we excluded “pizzas and dinner-pies” (e.g., stews, sandwiches and pizza) when we narrowed the original 25 food groupings to the five food groups used to calculate variety scores. This decision might have contributed to some misclassification. On the other hand, accounting for these foods would have involved breaking the dishes down to their individual ingredients; without firm knowledge of quantities used, this process would also be error prone. For lead exposure, there was a change in laboratory methods as well as limits of detection over the 10 years of the sudy. However, adjusting the statistical analyses for a variable that indicated if a lead value was below the LOD and conducting the analyses separately for the two phases of the study accounted for some of the variability due to lab methods. Additionally, we were unable to control for other sources of lead exposure such as soil or dust, as these variables were not measured. On the other hand, we previously showed that water lead levels were not a major contributor to BLL in the study children ⁵¹.

In conclusion, in the context of food consumption patterns that fall short of the recommended levels and in a cohort of children with relatively low BLLs, there was limited evidence of an association between diet quality overall or dietary variety, adequacy and moderation and BLLs among school children. Nevertheless, these relationships may depend on the overall lead exposure scenario.