Spatial clustering of toxic trace elements in adolescents around the Torreón, Mexico lead–zinc smelter

Abstract

High blood lead (BPb) levels in children and elevated soil and dust arsenic, cadmium, and lead were previously found in Torreón, northern Mexico, host to the world’s fourth largest lead–zinc metal smelter. The objectives of this study were to determine spatial distributions of adolescents with higher BPb and creatinine-corrected urine total arsenic, cadmium, molybdenum, thallium, and uranium around the smelter. Cross-sectional study of 512 male and female subjects 12–15 years of age was conducted. We measured BPb by graphite furnace atomic absorption spectrometry and urine trace elements by inductively coupled plasma-mass spectrometry, with dynamic reaction cell mode for arsenic. We constructed multiple regression models including sociodemographic variables and adjusted for subject residence spatial correlation with spatial lag or error terms. We applied local indicators of spatial association statistics to model residuals to identify hot spots of significant spatial clusters of subjects with higher trace elements. We found spatial clusters of subjects with elevated BPb (range 3.6–14.7 µg/dl) and urine cadmium (0.18–1.14 µg/g creatinine) adjacent to and downwind of the smelter and elevated urine thallium (0.28–0.93 µg/g creatinine) and uranium (0.07–0.13 µg/g creatinine) near ore transport routes, former waste, and industrial discharge sites. The conclusion derived from this study was that spatial clustering of adolescents with high BPb and urine cadmium adjacent to and downwind of the smelter and residual waste pile, areas identified over a decade ago with high lead and cadmium in soil and dust, suggests that past and/or present plant operations continue to present health risks to children in those neighborhoods.

INTRODUCTION

Children and adolescents in Torreón, an industrial city in the State of Coahuila in northern Mexico, have shown elevated blood lead (BPb) levels for at least three decades.^1–3 Documented sources of exposure to toxic metals and metalloids in Torreón include high levels of lead, cadmium, and arsenic in roadside dust in residential areas downwind and adjacent to the Met-Mex Peñoles plant, the largest lead–zinc–silver smelter complex in Latin America and the fourth largest in the world,⁴ and moderately high arsenic concentrations in drinking water throughout the city (8–75 µg/l).⁵ Although BPb levels in children have decreased substantially from 1999 (mean 17.0 µg/dl) to 2010 (mean 5.5 µg/dl)⁶ following a program of emission control, environmental remediation, and child lead surveillance at the turn of the century, lead exposure is still a major public health problem in Torreón.

As in other smelter towns,^7,8 the location of the smelter and local area characteristics such as the direction of prevailing winds might be important determinants of the geospatial distribution of elevated BPb among Torreón residents. The goal of this paper was to determine the spatial distribution of concentrations of toxic trace elements based on subject residence in a sample of adolescents living in Torreón, and to determine whether higher concentrations of any of the trace elements clustered around known or suspected industrial sources.

METHODS

Study Sample

Between August 2009 and June 2010, we performed the “Cuida tu Corazon” (“Take care of your heart”; C2C) study to evaluate the association of BPb and urine antimony (Sb), arsenic (As), cadmium (Cd), molybdenum (Mo), thallium (Tl), tungsten (W), and uranium (U) with cardiovascular and renal function measures (to be reported separately) among 512 males and females 12–15 years of age living in Torreón. Residential neighborhoods border the Peñoles smelter complex in the south of the city (Figure 1);³ The dominant wind pattern is from the north and the east.⁹ In 1999–2000, the Health Department of the State of Coahuila conducted a census of all children and adolescents residing within 1.5 km of the smelter complex. The original census continued annually from 2002. We selected children and adolescents from the census who had at least one BPb determination performed before 2004 and who would be 12–15 years of age at the time of the study (N = 6254). We divided the study population into five strata based on the initial BPb determination for each census subject (<10.0, 10.0–14.9, 15.0–19.9, 20.0–44.9, and ≥45.0 µg/dl), randomly selecting subjects within each stratum to achieve a target sample of 512 participants. The study was approved by the Institutional Review Boards of Juarez University of Durango State, the Johns Hopkins Bloomberg School of Public Health, and the New York State Department of Health’s Wadsworth Center (Albany, New York, USA). All subjects and their parents or legal guardians provided signed informed consent.

Figure 1.

Aerial view of the Peñoles Met-Mex metal smelter and surrounding communities, circa 1999. The view is to the south. The smelter waste pile occupies the center foreground, the smelter and offices are in the center background, behind the waste pile. Photo from: Programa para mejorar la calidad del aire en la región de la Comarca Lagunera 2010–2015. SEMARNAT (Secretaria de Medio Ambiente y Recursos Naturales), no date, no copyright. Accessed on 3 December 2012 from http://www.semamat.gob.mx/temas/gestionambiental/calidaddelaire/Documents/Calidad%20del%20aire/Proaires/ProAires_Vigentes/9_ProAire%20Comarca%20Lagunera%202010-2015.pdf. Color contrast values of photo have been adjusted to enhance detail.

Data Collection

Study questionnaires were administered by trained study interviewers in home interviews and clinic visits. We used clinic visits to collect additional data, including detailed physical measurements and blood and urine specimens for trace element analysis.

Blood specimens were collected in EDTA purple top tubes (Becton Dickinson, Franklin Lakes, NJ, USA) and refrigerated immediately. Spot urine specimens were collected in plastic containers that had been washed with 10% HNO₃ overnight and rinsed with 18.2 MΩ cm deionized water. Refrigerated blood and urine specimens were transported daily to the Laboratory of Toxicology of Juárez University, Durango State, where specimens were aliquoted and frozen at −80 °C pending analysis.

Laboratory Analyses

BPb concentrations were measured in duplicate at the Laboratory of Toxicology of Juárez University using a graphite furnace atomic absorption spectrometer equipped with longitudinal Zeeman-effect background correction (Analyst 800, PerkinElmer, Norwalk, CT, USA), according to the method described by Miller et al.¹⁰ Specimens with a coefficient of variation >5% (N = 11) were reanalyzed in duplicate until coefficient of variation <5%. The limit of detection (LOD) was 0.7 µg/dl and two subjects were < LOD. National Institute of Standards and Technology Standard Reference Material 955b, Lead in Bovine Blood (NIST; Gaithersburg, MD, USA) was used to ensure the accuracy and traceability of BPb measurements to international standards. The laboratory successfully participates in the BPb Inter Laboratory Program of Quality Control from the Facultad de Medicina, Universidad de Zaragoza, Zaragoza, Spain and in the Wisconsin State Laboratory of Hygiene’s Proficiency Testing Program for BPb.

Urine concentrations of Sb, total (unspeciated) As, Cd corrected for molybdenum oxide interference. Mo, Tl, W, and U were measured in the Trace Elements section of the Laboratory of Inorganic and Nuclear Chemistry at the New York State Department of Health’s Wadsworth Center. The analyses were performed using an ELAN DRC II inductively coupled plasma-mass spectrometer (PerkinElmer Life and Analytical Sciences, Shelton, CT, USA) equipped with Dynamic Reaction Cell technology.¹¹ Multielement calibration standards were prepared from a NIST-traceable stock solution (High Purity Standards, Charleston, SC, USA) and a six-point, calibration curve used for each element. Method accuracy was assessed by analysis of NIST Standard Reference Material 2670a, Toxic Elements in Urine (freeze-dried), and Standard Reference Material 2668, Toxic Elements in Frozen Human Urine. The laboratory successfully participates in several external quality assessment schemes for trace elements in urine, including those operated by the Centre de Toxicologie du Québec, Canada, The University of Surrey, UK and the University of Erlangen, Germany. The LODs for Sb, As, Cd, Mo, Tl, W, and U were 0.2, 1.1, 0.02, 1.0, 0.02, 0.06, and 0.001 µg/l, respectively. The corresponding percentages of participants with urine element levels <LOD were 76.0, 0.0, 0.8, 0.0, 0.4, 22.9, and 0.2%, respectively. Because of the high proportion of participants with urine antimony and urine tungsten concentrations <LOD, we do not present further statistical analyses of these elements.

Urine creatinine concentrations were measured by an enzymatic assay (Siemens Dimension Vista 1500; Siemens Medical Solutions USA, Malvern, PA, USA). Urine trace element concentrations were normalized to urine creatinine concentrations to account for variations in urine dilution in spot urine specimens and results were expressed as µg/g creatinine.

Water samples from 12 separate wellheads of the municipal water system serving the subject area were collected by the municipal water system authorities of Torreón County in 2010.¹² During the study data collection period, the Torreón city water system had no centralized water treatment system for distributing potable water. Thus, water from each well was distributed to specific neighborhoods and each wellhead water arsenic concentration served to represent arsenic in water delivered to those neighborhoods. Measurements of total arsenic in water were made by the Torreón municipal water system¹² using a standard method based on hydride generation atomic absorption spectrometry, following the procedure recommended by Perkin Elmer.¹³

We identified study participant residences on a geo-referenced Torreón street map as well as the metal smelter, the main pile of smelter waste, other known unofficial plant dumping areas (Garcia-Vargas, personal communication), a covered canal unused since 1957 that formerly transported liquid industrial waste from the metal smelter and urban waste, and rail and truck routes transporting ore to the plant, to identify possible geographic determinants of element concentrations in sample adolescents (Figure 2). We used a wind rose for 2011,⁹ the natural year closest to the measurement of elements in our sample, to determine prevailing winds. Predominant and strongest winds are from the arc north–northwest through the southeast. There was almost a complete absence of winds from the south and southwest.

Figure 2.

Geographically referenced map of Torreón, Coahuila. Blue dots represent the residence sites of study participants. The red cross represents the center of Peñoles Met-Mex smelter complex. North is up as shown in the wind-rose inset. Other possible sources for toxicological element exposure are as follows: () confinement of residual materials from Zn-Cd plant (Jarosite); () former open dump of smelter debris, now closed and covered; () ore truck transport route; () a former irrigation channel converted to conduct urban sewage and smelter industrial waste, now abandoned; () railway transport route for ores. The wind rose shows that winds are predominantly from the arc north–northwest to the east–southeast.

Statistical Analysis

Trace element concentrations were markedly right skewed and were natural log-transformed for statistical analysis. We calculated descriptive statistics for the elements and selected control variables stratified by sex. Sex differences in continuous variables were determined by t-test and in categorical variables by the Fisher’s exact test.

We constructed ordinary least squares (OLS) multiple regression models for log-transformed BPb and for log-transformed creatinine-corrected urine trace element concentrations using available demographic variables as predictors (apart from residence location). These variables included mean-centered age, sex, family income, subject former and current smoking, private or public school attendance, living at same address for at least 10 years, use of municipal water, commercial bottled water or municipal water with a filter for domestic water consumption, consumption of milk or cola beverage at least once per week, seafood consumption during the previous week, and municipal water arsenic concentrations. We included milk and cola intake because milk produced in the Torreón region was reported to contain arsenic¹⁴ and many soft drinks are bottled in Torreón, presumably using local well water, except for Coca-Cola, the largest selling brand in the city, which is bottled elsewhere. As some types of seafood are reported to contribute organoarsenic species, such as arsenbetaine, to the diet,¹⁵ we used a dichotomous variable to code seafood eating within the week before blood and urine sampling for trace element analysis. Although municipal water arsenic concentrations were used primarily to adjust spatial models of urine arsenic concentrations for known regional differences in water arsenic, we also used water arsenic as a proxy for other potential water-borne metal sources. We calculated Spearman correlations between water arsenic concentrations by well and measured trace element means in subjects grouped by well areas. We used Stata 11 and 12 (StataCorp, College Station, TX, USA) for all descriptive analyses and preliminary multiple regression modeling.

Spatial Analyses

We used GeoDa 1.4¹⁶ to construct spatial models. We calculated a distance-based spatial weight matrix with a cutoff value of 0.9 km using decimal degree geographic coordinates of subject homes provided by the Mexican Institute for Geographical and Statistical Information.¹⁷ We diagnosed OLS BPb and urine trace element models for the presence of two types of spatial dependence, spatial error, and spatial lag, using robust Lagrange Multiplier tests. Both types of dependence indicate a violation of the OLS assumption of residual independence; spatial lag dependence additionally indicates a violation of the assumption of independent observations of the dependent variable.

If model diagnostics indicated statistically significant spatial error or lag dependence, we reanalyzed the models adjusted for spatial error or spatial lag dependence to reduce global spatial clustering to statistical insignificance and eliminate OLS assumption violations, providing coefficients and standard errors of model variables uninfluenced by the above OLS violations.

We calculated model residuals, adjusted for spatial error or spatial lag where necessary, as estimates of element concentrations were not explained by model variables and not subjected to spatial clustering of model explanatory variables. These residuals were used to calculate Moran’s I,¹⁸ a statistic representing global spatial associations based on inverse spatial weighting derived from distances between subject residences.

We used local indicators of spatial association (LISA) statistics¹⁹ to measure the inverse distance-weighted association of trace element model residuals with the trace element residuals of neighboring subjects to identify local spatial clustering. As detailed in Anselin,¹⁹ these statistics classified study participants into four groups for each element distribution as follows: hot spots (subjects with high element levels relative to the sample living in close proximity to other subjects with high element levels); cold spots (subjects with low element levels living in close proximity to other subjects with low element levels); “cluster outliers” (high element level subjects living in close proximity to other subjects with low element levels); and other subjects, not otherwise identified in the LISA analysis. Hot spots had model residuals indicating model underestimation of element concentrations, and cold spots had model residuals indicating model overestimation of element concentrations. To simplify the interpretation, we mapped only significant hot- and cold spots (frequency-based probabilities ≤0.05, determined by 10,000 random permutations of the data set), with all other subjects grouped together. We also used LISA analysis on unadjusted trace element concentrations for comparison to the model residual analysis. We excluded no data points among the elements examined with spatial analysis.

We plotted contour maps of untransformed BPb and creatinine-adjusted urine trace element concentrations using the natural neighbor technique²⁰ to show spatial correspondence of trace element levels with LISA evidence of local spatial clustering. A digitized map of Torreón was overlaid on the combined LISA and contour map to locate hot- and cold spots. Data and map manipulation were performed in Surfer 8.09 (Golden Software, Golden, CO, USA).

To insure that individual subject addresses could not be recovered from the maps, all plotted subject residence locations were degraded to 10 m accuracy in addition to the best-case 7.9 m accuracy specified for single-frequency satellite-based GPS determination.²¹ In addition, symbols used to indicate subject residence on our maps were 40 m wide. All spatial statistical models, however, used the full accuracy of the GPS-determined coordinates in calculating spatial clustering.

RESULTS

Quality control data for urine trace element levels are shown in Supplementary Table 1. Distribution of urine trace element concentrations among study participants uncorrected by urine creatinine are shown in Supplementary Table 2. Spearman correlations among all trace elements are shown in Supplementary Table 3. Geometric mean (95% CI of mean) of BPb was 4.0 (3.8 and 4.2) µg/dl. Geometric means (95% CI of mean) of creatinine-corrected urine trace elements (arsenic, cadmium, molybdenum, thalium, and uranium, respectively) were 36.5 (35.1, 38.0), 0.23 (0.22, 0.24), 63.7 (61.4, 66.2), 0.27 (0.26, 0.28), and 0.04 (0.039, 0.045) µg/g. Urine trace element levels were significantly higher for males compared with females except for cadmium and molybdenum (Table 1). Other control variables showed no sex differences.

Table 1.

Descriptive statistics of sample by sex.

Veriablesa	Females (N = 250)			Males (N = 262)			P-valued
	Meanb	Count	SDc	Meanb	Count	SDc
Urine arsenic (As)	33.6		1.6	39.5		1.6	0.0001
Urine cadmium (Cd)	0.22		2.0	0.23		2.0	0.58
Urine molybdenum (Mo)	61.7		1.5	65.7		1.6	0.11
Blood lead (Pb)	3.5		1.7	4.6		1.7	<0.00005
Urine thallium (Tl)	0.25		1.7	0.29		1.7	0.0004
Urine uranium (U)	0.04		2.3	0.04		2.0	0.01
Age (in years)	13.9		1.2	14.0		1.1	0.26
School grade	6.9		1.3	6.9		1.3	0.94
Income (>$3000/month)		78			66		0.09
Smoke (within last month)		92			102		0.12
Occupational risk		133			132		0.54
Eat fish (≥1/week)		24			39		0.08
Drink milk (≥1/week)		134			130		0.38
Drink cola (≥1/week)		121			128		0.93

P-values for categorical variables with count summaries are calculated by comparison with the omitted dummy variable as follows: income: <$3000/month; smoke: never smoked; parental occupational risk for any trace element exposure: no occupational risk; eat fish, drink milk/cola: <1/week.

^aUrine trace element units µg element/g of creatinine (creatinine-corrected); blood Pb units µg of lead/dl of blood.

^bElement means are geometric means; other means are arithmetic means.

^cElement SD are geometric SD; other SD are arithmetic SD.

^dt-test P-values for variables with means; Fisher’s exact P-values for variables with counts.

Spearman correlations between well water arsenic and trace elements in subjects grouped by well areas were positive and significant (P ≤ 0.05) for creatinine-corrected urine As, Mo, Tl, and U (Table 2). Figure 3 shows the spatial distribution of subjects classified by well water As. Summaries of trace element models are found in Table 3. Only variables significant at P ≤ 0.05 are shown along with the type of spatial dependency adjustment used, if any. Full models are shown in Supplementary Table 4a–f. Model R² varied from 0.107 (Mo) to 0.239 (As). In multiple regression models, male sex was significantly associated with higher concentrations of trace elements except Cd and Mo, and increasing age was significantly associated with lower concentrations of trace elements except Cd.

Table 2.

Spearman correlations of total water As measured at 12 wellheads serving the sample area and subject trace element concentration means grouped by well service area.

Subject trace elementa	ρ	P-values
Urine As	0.860	<0.001
Urine Cd	0.435	0.158
Urine Mo	0.677	0.016
Blood Pb	0.544	0.068
Urine Tl	0.947	<0.001
Urine U	0.632	0.028

^aAll urine elements are creatinine corrected.

Figure 3.

Post map of subject residence location coded by water As concentration measured at wellheads servicing the subject area. Open circles are well locations, with same color code as the residence locations. Several wells had identical (nearest µg/I) As concentrations. Two wells with the lowest and two wells with the highest water As are combined on the map. See Figure 2 for more details.

Table 3.

Summary of final demographic models showing only variables significant at P≤0.05.

Elementsa	Final model	Significant variablesb	Model Rb
Urine arsenic	Spatial lag	+Sex (male)#	0.239
In(As)		−Agew
		+Occupational risk*
		−Private school*
		−Use bottled water#
		+Fishw
		+Milk#
Urine cadmium	Spatial lag	+Occupational risk*	0.127
In(Cd)		−Use bottled water**
		−Drink cola*
Urine molybdenum In(Mo)	OLS	−Agew	0.107
		−Family income (>$3000 Mexican Pesos/month)*
		+Arsenic in tap water#
Blood lead	Spatial error	+Sex (male)w	0.224
In(Pb)		−Age#
		−Income > $3000*
		−Private school*
		+Arsenic in Tap water#
		−Drink cola#
Urine thallium	Spatial error	+Sex (male)w	0.215
In(Tl)		−Agew
		−Use bottled water**
		+Arsenic in tap waterw
Urine uranium	Spatial lag	+Sex(male)**	0.160
ln(U)		−Agew
		−Not current smoker*
		−Private school#
		−Use bottled water#

Abbreviations: OLS, ordinary least squares; +, − sign of coefficient.

^*P<0.05;

^**P<0.01;

^#P<0.005;

^wP<0.0001.

^aAll urine elements are creatinine corrected.

^bStatistically significant non-informative categorical responses (“don’t know” “won’t say” “missing”) are not shown. P-values of categorical variables are from comparison with omitted categories: Sex (female), parental occupational risk for all trace elements (no occupational or hobby risk), water use (tap water from municipal lines), school (public school), family income (≤$3000 Mexican Pesos/month), lived at same address <10 years; smoker (never smoked), fish (<1 portion/week), milk and cola (<1 glass/week). All demographic models used the same set of variables, except for corrections of spatial dependencies where needed. See Supplementary Table 4a–f for complete models including all variables, coefficients and 95% confidence intervals.

Spatial Results

Figure 4a–f shows hot- and cold spot grouping (LISA maps) of all metals based solely on measured concentration. There are strong tendencies for hot spots to cluster in either the northwest or the southeast sectors with cold spots occupying the opposite pole. These metal concentration gradients are responsible for the significant Moran’s I statistic indicating spatial dependence of element concentrations. Clustering seen using measured element concentration can be owing to many factors, including point or area sources, prevailing winds, anisotropic spatial sample distribution, and sample characteristic distribution. Thus, for elements showing significant age, sex, and/or sociodemographic associations, uneven distribution of such sample characteristics can influence the apparent spatial clustering of higher and lower measured element concentrations.

Figure 4.

(a–f) Post maps of significant (frequency-based probability <0.05) hot () and cold spot () subject locations determined from measured (unadjusted by models) trace element concentrations. Black symbols were subjects that did not significantly cluster into hot or cold spots. The unfilled irregular polygon in the center-upper left sector represents the area occupied by the Peñoles smelter complex (southern part of the polygon) and the major waste pile from plant operations (northern part of the polygon). Other industrial plants occupy the arm of the polygon extending west–northwest of the Peñoles property.

Figure 5a–f contains combined contour post maps showing measured element concentration contours with LISA hot and cold spots based on model residuals. Global spatial correlations have been statistically removed from the model residual maps. The hot and cold spots remaining were not influenced by possible uneven distribution of other determined subject characteristics but could still be influenced by variables not explicitly in the model, including point and area element sources as well as prevailing wind patterns. Table 4 shows measured subject trace element concentrations of subjects identified with model residual hot and cold spots.

Figure 5.

(a–f) Contour-post maps of (a) arsenic, (b) cadmium, (c) molybdenum, (d) lead, (e) thallium, and (f) uranium in study subjects. Contour maps of measured trace element concentrations were formed by the natural neighbor technique. Some subject locations on the borders of the contour space fall outside the contour area, but elemental concentrations contribute to the contours within the contour area. All trace elements except blood lead (BPb) were creatinine-corrected urine concentrations. Post overlay of significant (frequency-based P<0.05) element hot () and cold spot () locations were determined from residuals of demographic models, corrected for global spatial dependence where needed (see Methods section). See Figure 4 legend for more details.

Table 4.

Measured element concentrations in subjects associated with LlSA hot and cold spot spatial clusters determined from residuals of demographic models.

Trace elements	N	Measured geometric mean (µg element/g creatinine)a	Geometric SD	Lowb	Highb
Urine arsenic
Hot	24	48.0	1.3	25.8	75.9
spots
Cold	8	21.5	1.4	12.1	31.2
spots
Urine cadmium
Hot	21	0.38	1.55	0.18	1.14
spots
Cold	21	0.12	1.49	0.06	0.25
spots
Urine molybdenum
Hot	0	-	-	-	-
spots
Cold	6	44.5	1.4	26.0	57.4
spots
Blood leada
Hot	35	6.9	1.4	3.6	14.7
spots
Cold	32	2.7	1.4	1.4	4.4
spots
Urine thallium
Hot	13	0.42	1.35	0.28	0.93
spots
Cold	6	0.18	1.28	0.13	0.25
spots
Urine uranium
Hot	6	0.08	1.30	0.07	0.13
spots
Cold	10	0.03	1.28	0.02	0.04
spots

Abbreviation: LlSA, local indicators of spatial association.

^aBlood lead units are µg/dl.

^bThe range of hot and cold spots for some elements overlaps because the clusters were based on model residuals and the descriptive statistics were based on measured element concentrations of those same subjects.

The highest levels of water As from municipal wells were found in the southeast (Figure 3), where the major grouping of measured arsenic hot spots was also found (Figure 4a). In adjusted models, no hot spots for As in subjects were found in the southeast, as water As was explicitly accounted for in the urine As model, along with lagged spatial dependence (Figure 5a).

Hot- and cold spot geographic locations for urine Cd (Figure 5b) and BPb (Figure 5d) substantially overlapped. Those hot spots were principally located downwind and immediately adjacent to the metal refinery waste pile in the north of the Peñoles property and the smelter operations area south of the property.

Tl and U hot spots occupied similar but non-overlapping locations in the southeast of the study area. Higher Tl levels (Figure 5e) were clustered downwind of the former sewage and industrial waste canal and ore transport rail lines. Higher U levels (Figure 5f) were clustered around the former plant open waste dump, now covered, and downwind of the rail and truck ore transport routes. Mo (Figure 5c) showed no model LISA hot spots.

DISCUSSION

In this cross-sectional study of BPb and urine trace elements in male and female adolescents residing in Torreón, Mexico, we found significant spatial clustering of higher BPb and urine Cd levels adjacent to and downwind of the Peñoles smelter complex. We also found significant spatial clustering of higher urine Tl and U adjacent to and downwind of the truck and rail route staging areas for delivery of raw materials to the smelter complex. The only spatial component considered in previous studies of trace element contamination in Torreón was crude distance to the smelter complex, indicating that the smelter was a possible point source for some elements.^4,22 Our study adds detailed location and distance spatial components and the evaluation of multiple trace elements to the study of toxic exposures in this geographic area.

Since the turn of this century, Peñoles has embarked on a program of limiting stack emissions from their refinery and enlarging critical bagging enclosures. They have also performed environmental remediation by street cleaning operations around the plant since May 1999, buying properties of residents from 1999 to 2001 in the areas where the highest BPb were found, relocating these residents to other areas, and planting the evacuated areas with trees. They continue to maintain a childhood lead surveillance program in cooperation with state health authorities. Nevertheless, over 86% of adolescents in BPb hot spots in the present study exceeded the CDC reference level of 5 µg/dl.²³

Before the Peñoles remediation program, two groups made independent measurements of soil Pb, Cd, and As in residential areas as a function of distance to the smelter.^4,22 Soil concentrations had combined ranges of 1,640–17,320 µg/g for Pb, 80–1,497 µg/g for Cd, and 50–570 µg/g for As. Soil element levels were higher in sites closest to the smelter. The CDC²⁴ performed a survey sample-design study of BPb in children aged 1–6 years and dust lead levels around the smelter complex in March 2001, after the start of cleanup efforts. In addition to finding significantly higher BPb in children closest to the plant, they also found lead dust in the children’s indoor play areas up to 216 µg/g and in their outdoor play areas up to 464 µg/g. Soil lead in the outdoor play areas within 1.5 km of the plant ranged from 25 to 2179 µg/g and soil lead beyond that distance from 24 to 589 µg/g.

Sparse rain (yearly average 23 cm) and frequent high winds (average maximum daily wind over 18 km/h) combine to produce dusty conditions in Torreón with dust storms typically preceding desert thunderstorms.²⁵ Total suspended particulates measured from the air quality monitoring network within Torreón throughout the year frequently exceed the Official Mexican Norm limiting the 24-hour average to <210 µg/m³.²⁶ Twenty-four hour air Pb concentration measured during April to December, 2007 at 6-day intervals was highest at the station about 1 km from the western border (downwind) of the Peñoles property, ranging between 0.2 and 1.1 µg/m³.²⁶ Twenty-hour hour air Cd and As measured at the same location varied between 0.01 and 0.14 µg/m³ and between 0.01 and 0.05 µg/m³, respectively. The US EPA set the National Ambient Air Quality Standard for Lead to 0.15 µg/m³ (rolling 3-month average) in 2008.²⁷

After adjusting our model of As for sociodemographic variables and the spatial pattern of As concentration in water, we found only a few subjects clustered for higher urine As. The small cluster location did not suggest wind-transported As from the smelter operation, even though previous studies^4,22 and the known association of As with some precious metal ores indicated the possibility of some plant contribution to As burden. Most of the total urine As in our study appeared to come from As in the water supply and diet. Nine water district wells providing water to the study area exceeded the EPA limit of 10 µg/I for As in drinking water and six exceeded the 25 µg/I Mexican norm. As we did not speciate urine As, some fraction of urinary As in our subjects may have come from dietary organoarsenic compounds. Despite including model adjustments for fish and milk consumption, such adjustments will only partially adjust total As for organoarsenic species. Measurement of speciated urine As would have provided better assessment of health risk from this metalloid.

Other limitations need to be considered in the interpretation of our data. All urine element determinations were made with spot urine specimens, which require correction for urine dilution and may be affected by diurnal variation in excretion. This was a cross-sectional study; longitudinal associations with seasonal wind and rainfall patterns were not possible.

In addition, the sample was chosen with the goal of providing roughly equal numbers of male and female children in selected earlier BPb ranges who previously participated in the long-term surveillance program that started in 1999. The children in the surveillance program and this study were self-selected by their parents and did not necessarily represent the population of children living in Torreón.

Our study also had several strengths, including use of standardized field data collection techniques, rigorous quality control of trace element measurements, and adequate power of the study to detect statistically significant results.

The published literature documented elevated levels of Pb, Cd, and As in soil, dust, and air close to the Peñoles plant before and during the cleanup efforts of the plant operators. Our data showed clusters of children with elevated BPb adjacent to and downwind of the Peñoles smelter complex over a decade after the start of the company’s efforts to reduce exposure to the surrounding residential population. Despite the remedial actions of the company and heightened population awareness of the problem, clusters of subjects with BPb in these areas >5 µg/dl likely indicate ongoing exposure. Furthermore, while Pb exposure in this population has received substantial attention, our data also indicate that the smelter complex directly or indirectly contributes to Cd, Tl, and U exposures.

Our findings are based on spatial statistical models of subject trace element concentrations and typical wind patterns in the area. The most current soil and dust samples are from 1999 to 2002, and these are not precisely located in reference to the plant. To determine the role of past and present plant operations in measured BPb and urine element hot spots, renewed soil sampling for trace element analysis in and around hot spot areas should be undertaken. Air particulate sampling for Pb and Cd on the plant perimeter would help determine whether current controls on stack emissions and refining residues are sufficient to prevent continued accumulation of elevated soil and dust trace element levels in affected neighborhoods.