Using the delayed spatial alternation task to assess environmentally associated changes in working memory in very young children

Abstract

Background:

Working memory (WM) is critical for problem solving and reasoning. Beginning in infancy, children show WM capacity increasing with age but there are few validated tests of WM in very young children. Because rapid brain development may increase susceptibility to adverse impacts of prenatal neurotoxicant exposure, such as lead, tests of WM in very young children would help to delineate onset of developmental problems and windows of susceptibility.

Purpose:

Our objective was to assess the feasibility of administering a Delayed Spatial Alternation Task (DSAT) to measure WM among 18- and 24-month old children enrolled in an ongoing longitudinal birth cohort study and compare DSAT performance with age and general cognitive development. We further explored whether prenatal lead exposure impacted DSAT performance.

Methods:

We assessed 457 mother-child pairs participating in the Programming Research in Obesity, GRowth, Environment and Social Stressors (PROGRESS) Study in Mexico City. The DSAT and Bayley Scales of Infant Development (BSID-III) were administered at 18- and 24-months. Lead was measured in maternal blood collected during pregnancy (MBPb) and in a subsample of children at 24-months (CBPb). We regressed DSAT measures on MBPb and CBPb, child sex, and maternal age, education, socioeconomic status, and household smoking. We compared DSAT performance to BSID-III performance with adjusted residuals.

Results:

24-month children perform better on the DSAT than 18-month children; 24-month subjects reached a higher level on the DSAT (3.3 (0.86) vs. 2.4 (0.97), p < 0.01), and had a higher number of correct responses (20.3 vs. 17.2, p < 0.01). In all DSAT parameters, females performed better than males. Maternal education predicted better DSAT performance; household smoking predicted worse DSAT performance. A higher number of correct responses was associated with higher BSID-III Cognitive scales at 18 months (r = 0.20, p < 0.01) and 24 months (r = 0.27, p < 0.01). MBPb and CPBb did not significantly predict DSAT performance.

Conclusion:

Improved performance on the DSAT with increasing age, the positive correlation with the BSID-III cognitive and language scales and the correlation with common sociodemographic predictors of neurodevelopment demonstrate the validity of the DSAT as a test of infant development.

1. Background

Executive function (EF) refers to a collection of basic cognitive processes including attentional control, cognitive inhibition, inhibitory control, cognitive flexibility and working memory (WM) that guide purposeful, goal-directed behavior (Cartwright, 2012; Dawson and Guare, 2010; Anderson, 2002; Anderson et al., 2001). As one of the core executive functions, WM is defined as the ability to memorize new information, hold it in short-term memory, concentrate, and manipulate the information to produce results (BAL, 1999; Smith and Jonides, 1997). WM contributes to overall measures of intelligence, though correlations between these constructs are inconsistent (Alloway and Alloway, 2010a; Colom et al., 2008; Tsatsanis, 2007; Wiguna et al., 2012). As with other EFs, WM is a dynamic system that emerges in infancy and expands with age. The early development of WM is evidenced by administration of the classical “A not B” search task, which is commonly used to study the effects of time delay on search behaviors in humans and nonhuman animals (Diamond, 1985, 1990; Espy et al., 1999). This task has been used to provide insight into the developmental timing of working memory in infants and young children (Diamond, 1995; Bell, 2012; Diamond, 2013; Cuevas et al., 2012; Reynolds and Romano, 2016). Despite widespread use in the field of psychology, A-not-B type tasks are underutilized in the field of children’s environmental health, limiting our understanding of environmental exposures and early executive functions such as working memory.

While it is recognized that early biologic and social factors interact to affect the development of WM (Diamond, 2009; Escalona, 1982), few studies focus on associations between environmental toxicant exposure and WM in very young children. Understanding the relationship between exposure and developmental outcomes such as working memory requires approaches that can assess both exposure and health at each relevant life stage (i.e., infancy, early childhood, preadolescence, etc.). Understanding whether effects of a developmental neurotoxicant (i.e., lead) are immediate or delayed will help delineate the underlying mechanisms. Delayed effects may suggest epigenetic mechanisms, while immediate effects may indicate subtle developmental trajectory changes. Environmental epidemiology researchers would benefit from tools that could assess EF and WM in early childhood in order to plot the longer term trajectory in longitudinal studies.

To illustrate this point, lead (Pb) is a well-known neurotoxicant (Fowler et al., 2007). Evidence suggests that developing nervous system of young children is more vulnerable to the adverse effects of lead exposure than adults (Needleman et al., 1979; Bellinger et al., 1987; Dietrich et al., 1987; Hu et al., 2006; Boucher et al., 2014). Prenatal and early childhood lead exposure has been repeatedly associated with cognitive deficits in children at school age, although the trajectory of that deficit has not been well studied (Lidsky and Schneider, 2003). Most studies focus on IQ, a global construct of cognition (Roy et al., 2009) and the association between lead exposure and EF including WM is understudied. Because previous studies associating lead exposure with EF and WM focused on older children (> 4 years of age) and measured EF and/or WM at a single timepoint, the nature of the association in children younger than 4 years of age, the developmental trajectory of the association, and the developmental windows of susceptibility are largely unknown (Canfield et al., 2004). Tests purporting to assess WM in very young children need to be assessed for their expected correlations with validated tests of infant general development and well-established predictors such as age, sex and parental education. The delayed spatial alternation tasks (DSAT), a classical A not B task, has been widely used in exposure studies in rodents and non-human primates to demonstrate the impact of lead exposure on working memory (Rice and Karpinski, 1988; Alber and Strupp, 1996; Rice, 1990; Levin and Bowman, 1988). While the DSAT has been used in psychological literature to understand the developmental trajectory of WM in typically developing children, the DSAT has not been used to assess lead-associated changes in WM in very young children.

The objective of this study was to investigate the feasibility and characterization of a DSAT task to assess WM in children at 18- and 24-months of age and to determine whether the DSAT reflects expected associations with age (Gathercole et al., 2004; Roman et al., 2014), sex, and parental education. We hypothesize that WM improves with age. To further characterize the task (Alloway and Alloway, 2010b), we compared DSAT performance to the Bayley Scales of Infant Development-Third Edition (BSID-III) cognitive, motor, and language scales, a comprehensive assessment of early childhood development (Bayley-III, 2006). Finally, we assessed the relationship between maternal blood lead concentrations and performance on the DSAT at 18- and 24- months of age.

2. Methods

2.1. Participants

Between 2007 and 2011, healthy pregnant women at 12–24 weeks’ gestation were recruited into The Programming Research in Obesity, GRowth, Environment and Social Stressors (PROGRESS) (Rosa et al., 2016) study through Mexico’s social security system, Instituto Mexicano del Seguro Social (IMSS). Participants were included if they were ≥18 years old, had access to a telephone, and planned to reside within Mexico City for the subsequent 3 years (Burris et al., 2013). Exclusion criteria included heart or kidney disease, steroids or antiepilepsy drug use, daily alcohol consumption, or pregnancy > 20 weeks’ gestation. In total, 1054 women agreed to participate in the parent study, 948 women who were followed until delivery and birthed a live infant; 760 returned for at least one subsequent study visit and remain actively followed. This study consists of 457 mother-child dyads with complete data for the DSAT at 18 and/or 24 months, a 2nd trimester maternal blood lead sample, and no missing data on other a priori covariates.

2.2. Measures of child development

Children, accompanied by their mothers, participated in neurodevelopmental follow-up visits at the study facility at 18 and 24 months. Visit assessments included the DSAT (Espy et al., 1999; Goldman et al., 1971) for WM and the BSID-III to measure cognitive, motor and language development. Assessments were conducted by experienced, trained child psychologists. For quality control/quality assurance, study psychologists were trained by the same senior psychologist and were blind to the child’s lead level. The test was always performed in the same research facility under the same conditions for all participants.

2.2.1. Delayed spatial alternation task

The DSAT version administered in this study is a modification of the classic A not B paradigm (Espy et al., 1999). It is a two-choice task in which the child has to alternately retrieve an object hidden underneath one of two identical opaque cups on a large testing board. The administrator places the object under one of the two cups in view of the child (“visible hiding”) then asks the child to point to the cup with the hidden object. Each opportunity the subject is given to identify the hidden object is considered a trial. If the child correctly selects the cup with the hidden object, the child receives a reward (e.g., a snack cereal), and the object is the placed into the alternate cup for the subsequent trial. After 4 consecutive successful trials, a 5-second time delay between hiding the object and subject retrieval is introduced. After 4 consecutive successful trials, 10- and 15-second delays are introduced. At any time, if the child fails 4 consecutive trials, the test reverts to the previous level until the child can complete 4 consecutive trials. The test is stopped if a child fails at 8 consecutive trials. The test continues for a maximum of 30 trials. If a subject passed the 4th condition (15 sond delay, “visible hiding”), the administrator will repeat the 4 levels (0, 5, 10 and 15 sond delays) in the “invisible hiding” condition. In this condition, the administrator hides the object in one of the two opaque cups outside of the subject’s view. No subject in our study progressed through the 0-delay trails in the “invisible hiding” condition (considered level 5). A detailed description of the DSAT is given in Appendix 1.

Parameters to assess performance on the DSAT included in these analyses are described in Table 1. They are as follows: Total number of correct trials (“corrects”), longest run of consecutive correct trials (“runs”), number of correct trials completed prior to the first error (“trials to error”), number of trials until first reversal in levels (“trials to reversal”), highest level achieved (“highest level achieved”) and greatest number of consecutive failed trials/errors (“consecutive errors”) (Table 1).

Table 1.

Description and summaries of Delayed Spatial Alternation Task (DSAT) variables among 18- and 24-month PROGRESS subjects.

DSAT Component	DSAT Parameter Description	18 m, N = 380 Median (Range)	24 m, N = 359 Median (Range)
Correct trials	Total number of correct responses (min, max = 1, 30)	18 (2–24)	21 (1–27)
Consecutive trials	Number of consecutive correct trials (min, max = 0, 30)	5 (1–12)	7 (1–18)
Highest level achieved	Highest level achieved (max = 5)	2 (1–5)	3 (1–5)
Trials to error	Number of trials to first error	3 (1–13)	6 (1–15)
Trials to reversala	Number of trials until subject reverses to previous level	21 (5–30)	23 (5–30)

^aIt is possible for subjects to proceed through DSAT without making a reversal. At 18 months 152 subjects did not revert to a previous level; at 24 months, 76 subjects did not revert to a previous level.

2.2.2. Bayley scales of infant development

The BSID-III scales were used to assess infant development at 18- and 24- months of age. The BSID-III are comprised of a series of standardized measurements used to assess cognitive, motor, and language development of infants and toddlers (Bayley-III, 2006). Tests were conducted by trained child psychologists.

2.3. Blood lead measurements

Second trimester maternal venous blood and child venous blood at 24 months of age were collected by sterile venipuncture during scheduled study visits. Trace metal Vacutainer tubes (Becton-Dickinson and Company, Franklin Lakes, NJ) containing EDTA were used to collect the samples. Samples were stored at 4 °C and then transported to the Trace Metals Laboratory at the Harvard T.H Chan School of Public Health and stored at −20 °C until analysis. Lead levels were measured via inductively-coupled plasma mass spectrometry (ICP-MS) (Elan 6100; PerkinElmer, Norwalk, CT). Five replicate measurements of each sample were taken and averaged. The recovery of the analysis quality control standards and spike samples was 90 %–110 %; limit of detection for the procedure was 0.02 μg/dL (Tamayo et al., 2017).

2.4. Covariate/additional data collection

Standardized questionnaires were administered to mothers at each study visit to collect sociodemographic information including maternal age at pregnancy, years of education, and self-reported household smoking exposure. Household smoking was determined based on the mother’s report that at least one household member smoked. 99.6 % (455/457) of participants reported not smoking during pregnancy. Maternal socioeconomic status (SES) was derived using the index created by the Asociación Mexicana de Agencias de Investigación de Mercados y Opinión Pública (AMAI Rule 13 × 6) (Rodosthenous et al., 2017; Stroustrup et al., 2016). A subset (72 %, 331/457) of participants completed the Home Observation for Measurement of the Environment (HOME) at 24 months postpartum (Caldwell and Bradley, 1984).

At 18 months, 380 subjects had complete data (mean age = 18.2 months +/−0.25) and at 24 months, 359 subjects had complete data (mean age 24.4 months +/−0.43). Both timepoints (18 and 24 month) DSAT assessments were available for 282 subjects.

2.5. Statistical approach

We performed descriptive analyses to determine the distribution of the exposure and outcome variables. Maternal blood lead levels were positively skewed and were thus modeled as tertiles, comparing the second and third tertile to the bottom tertile (reference), and as natural-log-transformed (Table 2). DSAT variables were considered as counts. A summary of DSAT descriptions and their medians and ranges is shown in Table 2.

Table 2.

Participant characteristics of PROGRESS cohort (N = 760) stratified by those who participated in the DSAT at 18 or 24 months and those who did not participate.

Characteristic	Participants (N = 457)a Mean (SD) or %	Non-Participants (N = 303) Mean (SD) or %	P-value*
Maternal age, years	45.7 (5.45)	27.47 (5.49)	0.59
Child Gender (% female)	50 %	44 %	0.09
SES Category
Low SES	53 %	52 %	0.97
Medium SES	37 %	38 %
Higher SES	11 %	10 %
Education, years	11.85 (2.76)	11.79 (2.84)	0.77
Household Smoking (% Yes)	30 %	33 %	0.29
HOME Scorea	31.9 (5.47)	31.36 (5.32)	0.29
Blood lead (μg/dL)
2nd trimester maternal bloodb	2.75 (1.93, 4.4)	2.93 (1.95, 4.26)	0.27
24-month child bloodc	2.21 (1.64, 3.3)	2.22 (1.58, 3.13)	0.96
BSID-III at 18 months
Cognitive	95.79 (10.26)	97.35 (10.05)	0.11
Motor	94.57 (9.88)	97.25 (9.13)	< 0.01
Language	88.67 (11.3)	89.6 (11.71)	0.39
BSID-III at 24 months
Cognitive	92.43 (8.76)	92.21 (8.19)	0.78
Motor	93.44 (9.3)	95.06 (9.73)	0.06
Language	90.05 (9.31)	88.86 (8.71)	0.16

^aMissing data for some covariates: HOME (n = 126), Child blood lead (n = 301).

^bMedian (p25, p75), significance testing done with median test.

^cp < 0.05.

We used the Wilcoxon signed-rank test to compare DSAT performance by age of administrations (18 and 24 months). We used Mann-Whitney U test to examine associations between DSAT performance and child sex (male/female) and Kruskal-Wallis test to examine associations between DSAT parameters and maternal education (high school, college, > college) and maternal SES (low, medium, higher).

To determine the association between the DSAT measures and the continuous BSID-III cognition, motor, and language scales, we performed multiple linear regressions for each DSAT measure and BSID-III scale (in separate regression models), adjusted for child sex, maternal age, education, SES, and household smoking. We then calculated standardized residuals from each model and created a scatterplot of these residuals with a Loess smoother to examine associations between DSAT and BSID-III scales. We used Cox-Snell residuals to model the number of trials until first error and number of trials until first reversal. We also ran paired Pearson correlation coefficients to assess the association of the residuals.

To examine associations between MBPb and DSAT, we used multiple linear regression for DSAT parameters including correct, longest run, and consecutive error and ordinal logistic regression for highest level achieved. Cox Proportional Hazards was used to model the number of trials until first error, and the number of trials until first reversal. All models were adjusted for covariates selected from bivariate associations and included child sex, maternal age, education, SES and household smoking.

In sensitivity analyses we included the HOME assessment to adjust for the quality of the home environment. We also conducted sensitivity analyses adjusting for child’s blood lead (CBPb) at 24 months. 24- month blood lead levels were available in a small subset of the study population (n = 156). To account for the non-normal distribution and the count-like nature of the DSAT parameters, we re-ran models using Poisson and negative binomial regression with robust standard errors. As modeling with alternative distributions did not impact the size or direction of associations, we present analyses as linear regressions. Statistical analyses were conducted using STATA 14 (StataCorp, www.stata.com) and R 3.4.2 (R foundation for Statistical Computing, cran.r-project.org).

3. Results

3.1. Study population

A summary of DSAT descriptions and their medians and ranges is shown in Table 1. Demographic characteristics of the cohort are shown in Table 2. Overall, most characteristics were similar between participating subjects and non-participating subjects. At 18- and 24-months, BSID-III motor scores were statistically higher in non-participants compared to participants (95.05 (9.723) vs 93.44 (9.3), p = 0.06; Table 2). Mother-child pairs who completed both the 18-and 24-month assessments (n = 282) had slightly higher HOME scores compared to mother-child pairs who only completed one assessment (HOME score for single assessment vs HOME score for both assessments, p = 0.07). Lead (Pb) was detected in 100 % (n = 457) of maternal blood samples collected at the 2nd trimester of pregnancy (Table 1). Median (25th percentile, 75th percentile) of MBPb was 2.75 (1.93, 4.4) ug/dL. Median (25th percentile, 75th percentile) of CBPb (collected at 24 months of age, n = 156) was 2.21 (1.64, 3.30) ug/dL (Table 1). Maternal and child blood lead were weakly correlated (Spearman’s r = 0.09, P < 0.05).

3.2. Demographic predictors of DSAT performance

The majority of subjects in both groups completed the maximum number of 30 trials of the DSAT assessment (92 % at 18 months and 98 % at 24 months). As demonstrated in Fig. 1, performance on the DSAT improved between 18 and 24 months in all DSAT parameters (Fig. 1). success at proceeding to the next level decreased as the time delay (0 s, 5 s, 10 s, 15 s) increased. At 18 and 24 months, all subjects successfully passed the ‘no time delay’ trial. At 18 months, only 11 % of subjects completed level 4 (4 successful trials at 15 s delay, “visible hiding” condition) while 42 % of 24 months subjects made it to through level 4. At 24 months, subjects performed better on nearly all measures of the DSAT including higher number of corrects (Wilcoxson signed rank z = 10.80, p < 0.01), more consecutive trials (z = 9.26, p < 0.01), more trials to first error (z = 10.18, p < 0.01), fewer trials to reversal (z = 2.05, p = 0.04), and higher level achieved (z = 9.65, p < 0.01). With the exception of the number of consecutive correct trials, subjects who performed the DSAT at both 18 and 24 months (n = 292) did not perform better than those who only completed the DSAT at 24 months (n = 88).

Fig. 1.

DSAT performance by age. Comparison of DSAT performance between subjects who completed the assessment at 18 and 24 months.

Solid bars represent median, error bars represent 25^th and 75^th percentile. This analysis focused only on participants who completed the DSAT at both 18 and 24-month study visits, n = 282.

* Wilcoxon signed-rank test; p < 0.01.

Overall, females outperformed males on all DSAT parameters (Fig. 2). Compared to males at 18 months, females had more correct trials overall (Mann Whitney U; z = −2.4, p = 0.02), more trials before making the first error (z = −1.82, p = 0.07), and more trials before a reversal (z = −2.68, p < 0.01), and overall achieved a higher level (z = −1.75, p = 0.08). At 24 months, females more consecutive correct trials (z = −1.90, p = 0.05) and completed more trials before an error (z = −2.39, p = 0.02). At 18 months, higher maternal SES was associated with a higher number of consecutive trials (Kruskal-Wallis χ² = 5.4, p = 0.06) and more trials until the first error (χ² = 7.2, p = 0.03). At 24 months, higher SES was not significantly associated with any DSAT parameters. Further, years of maternal education was not associated with DSAT performance at 18 or 24 months. Presence of household smoking was associated with a lower number of consecutive trials (p = 0.03) at 18 months.

Fig. 2.

DSAT performance by sex. Comparison of DSAT performance between female and male subjects at 18 months (A) and 24 months (B).

Solid bars represent median, error bars represent 25^th and 75^th percentile. Darker lines represent males and lighter grey lines represent females, 18 months subjects (A) and 24 months subjects (B).

Mann Whitney U; *p < 0.1, **p < 0.05.

3.3. Comparison of DSAT and BSID-III

At both 18 and 24 months, higher BSID-III cognitive scores were associated with better performance on all DSAT parameters. Fig. 3 includes the plot of associations between the residuals of the DSAT number of corrects and BSID-III cognitive scale at 24 months, demonstrating that higher BSID-II scores are associated with better performance on the DSAT. Pearson correlation coefficients showed that residuals from BSID-III Cognitive and Language scales were positively correlated with most DSAT parameters at both 18 and 24 months of age. The strongest correlations were observed between the 24-month BSID-III cognitive scale and DSAT corrects (r = 0.27, p < 0.05), and between the 24-month BSID-III language scale and DSAT corrects at 24 months of age (r = 0.25, p < 0.05) (Supplemental Table 4).

Fig. 3.

Loess scatterplot showing the association between residuals of DSAT Number of Corrects and BSID-III Cognitive Score at 24 months (n = 376; Pearson’s r = 0.27, p < 0.01).

3.4. Maternal blood Pb and DSAT outcomes

In unadjusted analyses, DSAT performance at 18 months did not differ by tertile of MBPb (Fig. 4). In multivariable linear regression analyses adjusting for covariates (maternal age, child sex, maternal education, maternal SES, and presence of household smoking), MBPb was not significantly associated with a child’s DSAT performance at 18 or 24 months of age, modeled as either tertiles or as natural-log-transformed. However, we observed a trend indicating higher MBPb and poorer DSAT performance at 18 months of age (Table 3). At 24 months of age, there was no consistent pattern of association between MBPb and DSAT performance.

Fig. 4.

Comparison of DSAT performance by tertile of second trimester maternal blood lead in children at 18 months of age^a (n = 380).

^aSolid bars represent median, error bars represent 25^th and 75^th percentiles.

Table 3.

Multivariable regression model results showing the association of second trimester maternal blood lead with child’s DSAT performance at 18 months of age^a.

Maternal Blood Pb, μg/dl	≤2.16	2.17–3.79	≥3.80	p-trend	ln-transformed
N	136	115	129		380
Correct trials, β (95 % CI)	0.00 (ref)	−0.57 (−1.67, 0.52)	−0.19 (−1.24, 0.87)	0.9	−0.14 (−0.84, 0.55)
Consecutive trials, β (95 % CI)	0.00 (ref)	−0.29 (−0.83, 0.24)	−0.38 (−0.9, 0.14)	0.19	−0.18 (−0.52, 0.16)
Highest level achieved, OR (95 % CI)b	1.00 (ref)	0.72 (0.45, 1.14)	0.87 (0.56, 1.37)	0.79	0.96 (0.71, 1.28)
Trials to error, HR (95 % CI)c	1.00 (ref)	1.14 (0.84, 1.55)	1 (0.73, 1.36)	0.84	0.98 (0.8, 1.2)
Trials to reversal, HR (95 % CI)c	1.00 (ref)	1.05 (0.71, 1.56)	0.83 (0.56, 1.23)	0.3	0.89 (0.69, 1.16)

^aAll models are adjusted for maternal age(years), sex of child (female vs male), maternal education (years), maternal SES (low, med, higher), and household smoking (no, yes).

^bOrdinal logistic regression results.

^cModeled as Cox Proportional Hazards.

3.5. Sensitivity analyses

Among subjects with the completed HOME assessment at 24 months, adjustment for HOME score at 24 months did not change the association between MBPb and DSAT. In a subset of subjects (n = 149), we examined the association between 24-month blood lead (CBPb) and DSAT. There was no statistically significant association between CBPb (modeled as tertiles and as natural-log-transformed) and DSAT parameters (Supplemental Table 1). Regression models using Poisson and negative binomial regression with robust standard errors gave similar results as linear regression models (Table 4).

Table 4.

Multivariable regression model results showing the association of second trimester maternal blood lead with child’s DSAT performance at 24 months of age^a.

Maternal Blood Pb, μg/dl	≤2.16 μg/dl	2.17–3.79 μg/dl	≥3.80 μg/dl	p-trend	ln-transformed
N	123	118	118		359
Correct trials, β (95 % CI)	0.00 (ref)	0.27 (−0.42, 0.96)	0.11 (−0.58, 0.79)	0.9	0.01 (−0.44, 0.46)
Consecutive trials, β (95 % CI)	0.00 (ref)	−0.25 (−0.93, 0.44)	0.38 (−0.3, 1.07)	0.16	0.25 (−0.2, 0.69)
Highest level achieved, OR (95 % CI)b	1.00 (ref)	1.45 (0.89, 2.34)	1.22 (0.76, 1.97)	0.61	1.2 (0.87, 1.65)
Trials to error, HR (95 % CI)c	1.00 (ref)	1.18 (0.9, 1.56)	0.89 (0.68, 1.17)	0.22	0.91 (0.76, 1.1)
Trials to reversal, HR (95 % CI)c	1.00 (ref)	0.83 (0.48, 1.43)	0.68 (0.38, 1.2)	0.19	0.75 (0.51, 1.1)

^aAll models are adjusted for maternal age(years), sex of child (female vs male), maternal education (years), maternal SES (low, med, higher), and household smoking (no, yes).

^bOrdinal logistic regression results.

^cModeled as Cox Proportional Hazards.

4. Discussion

Understanding the early development of WM would improve our understanding of how environmental factors influence brain development, giving researchers the ability to plot developmental trajectories at an early age and investigate the role of exposure timing. Such work would overcome barriers that limit our understanding of how WM might be influenced by prenatal exposure to environmental chemicals such as lead. In this paper, we examined the feasibility of assessing WM in 18-and 24-month old children using a DSAT assessment. Our study demonstrated the feasibility of administering the DSAT among young children; nearly all 18-month and all 24-month children in this study were able to complete the task without evidence of fatigue or frustration and the cohort provided a range of scores as expected. Additionally, performance on the DSAT improved with increasing age (Espy et al., 1999); 24-month old subjects performed better than 18-month subjects, suggesting that the test can detect expected improvements with age in very young children. Higher BSID-III cognitive and language scales at 24 months were associated with better DSAT performance. This provides a measure of assurance that the DSAT is capturing information on brain development that correlates with broad-band tests of general development. Better performance among females is also consistent with other studies of neurodevelopment including the Bayley Scales. In aggregate, these results suggest that the DSAT is a useful research tool to assess WM in very young children.

While our results were suggestive of an inverse association between prenatal lead exposure and DSAT performance at 18-months, results were not statistically significant and were not observed at the 24-month timepoint. It is possible that prenatal lead exposure may have only subtle impacts on WM at this stage of development and with our relatively small sample size, we may not have adequate power to detect associations with low-level lead exposure. Further, it is also plausible that the lack of association is because we only administered one task of working memory and did not measure the integrity of other supportive functions (i.e., attention, inhibitory control, motor coordination). Studies demonstrate that lead-associated poor performance on WM tasks later in life may arise from cumulative effects, or from developmental changes that are not expressed till later in life. Finally, the relative imprecision of any behavioral task administered in early childhood could tend to drive results toward a null finding due to random misclassification bias.

The DSAT is a test of spatial WM function in which the subject is rewarded for correctly locating a hidden object at alternating locations after an increasing time delay. A hallmark of WM processing is reduced accuracy as the time between responses increases (Horst and Laubach, 2012). In our study, success at completing the task decreased as the time delay (0 s, 5 s, 10 s, 15 s) increased. At 18 and 24 months, all subjects successfully passed the ‘no time delay’ trial. As the time delay increased, subjects made increasingly more errors and more subjects failed out of the task. At 18 months, only 11 % of subjects completed level 4 (4 successful trials at 15 s delay, “visible hiding” condition) while 42 % of 24 months subjects made it to through level 4. Literature demonstrates that infants beyond 5–6 months of age demonstrate basic yet immature aspects of WM, with significant improvement in these basic functions occurring from 5 to 6 months of age throughout childhood (Reynolds and Romano, 2016). Additional literature with preschool age and older children show the advancement of WM and other executive functions after four years of age (McGuigan and Núñez, 2006). Few studies have explored the development of WM and executive skills in children between one and three years of age (McGuigan and Núñez, 2006). These findings support previous research demonstrating that age predicts performance on WM tests (Espy et al., 1999; Carlson, 2005; Diamond and Doar, 1989).

In our study, DSAT performance was positively associated with cognition assessed using the BSID-III. The modest correlation is expected as the BSID-III taps into multiple domains of cognition and does not focus on WM, but nonetheless WM should correlate with general cognitive function. The association between DSAT and BSID-III remained generally consistent at 18 and 24 months. Our findings are consistent with other studies suggesting higher IQ scores (like the BSID-III, a test of general cognition) are associated with better performance on tests designed to measure EF, including WM. A study of 255 children aged 10–15 observed a positive association between higher IQ scores and virtual radial arm maze (VRAM) performance (Braun et al., 2012a). A study of 385 children age 4–12 reported that higher IQ scores were associated with better CANTAB scores (Luciana and Nelson, 2002). Spatial memory span, pattern recognition and the Tower of London (perfect solutions) were significantly, positively associated with non-verbal ability. Similarly, pattern recognition, ID/ED set-shifting and Tower of London were associated with verbal ability. A study of 720 children age 5–13 found that higher IQ scores were associated with more correct lever holds on the temporal response differentiation (TRD) task (Chelonis et al., 2004). There is a large body of evidence that individual differences in WM storage capacity are predictive of performance on many broad measures of complex cognition in healthy people, mainly adults (Cowan et al., 2005; Fukuda et al., 2010; Miyake, 2001).

We found that females outperformed males on many of the DSAT parameters at 18 and 24 months. Other studies have found associations between child sex and WM, with age-matched females generally demonstrating evidence of better WM than males. Female children had higher number of correct consecutive responses on the A-not-B task in a study of 117 preschool aged children (23–66 months) (Espy et al., 1999). The A-not-B task is very similar to the DSAT (Espy et al., 1999). A study of 170 children ages 48–54 months reported that male children had lower inhibit efficiency in the Shape School Task (Canfield et al., 2003), a storybook-based task designed to examine inhibition and switching processes (Espy, 1997). A study of 25 infants age 7.5–12 months old found that female infants tolerated longer delays than male infants in the AB Object Permanence Task (Diamond, 1985) and the Delayed Response task (Diamond and Doar, 1989). Our results are consistent with previous work and suggest that the development of WM may differ by child sex.

Associations with other known predictors of neurodevelopment were less consistent. We found an association between increasing SES and improved DSAT performance at 18 months, but not at 24 months. A cohort of 170 children age 48–54 months reported older female children whose mothers had higher IQ scores completed more phases in the Shape School task (Canfield et al., 2003). Bivariate analyses in the same study population found that maternal education was associated with increased Spatial Span (Span Length) (Froehlich et al., 2007). In our cohort, SES was highly correlated with educational attainment, and was associated with improved DSAT performance at 18 but not 24 months.

To our knowledge this is the first study of prenatal lead exposure and performance on the DSAT. One other study used the A-not-B task, a similar task of working memory (Espy et al., 1999) to assess domain-specific effects of environmental exposures, including prenatal lead (Boucher et al., 2014). Among the cohort of 94 Inuit infants (11 months of age), subjects with higher prenatal lead exposure performed lower on the Fagan Test of Infant Intelligence (FTII), but showed now changes in performance on the A-not-B task (Boucher et al., 2014). Rather than contradicting the well-documented body of scientific literature detailing the neurotoxic effects of prenatal lead exposure cognitive functions in children (Bellinger et al., 1987; Hu et al., 2006; Lidsky and Schneider, 2003; Schnaas et al., 2006; Bellinger et al., 1992), these finding may inform our understanding about the timing of early life outcomes. Prenatal exposure may not impact WM as much as other functional domains, or may have delayed effects that are mechanistic in nature (e.g. epigenetics) and in that regard our results may be able to inform future research that focuses on other domains, other life stages or mechanisms. Furthermore, as this is a longitudinal study, we plan to plot the trajectory of WM over time which will allow more detailed understanding of how lead impacts WM and when effects are expressed. Research in older children on lead and EF include studies using the Shape School task (a preschool age task of executive function (Espy et al., 2006)). Among children aged 4–4.5 years blood lead was negatively associated with focused attention, efficiency, and inhibition of automatic responses (Canfield et al., 2003). A study of 452 six-month old infants from Poland found that children with higher prenatal lead exposure also performed worse on the FTII, although not in the highest-risk group (Jedrychowski et al., 2008).

In a primate study, lead exposure in the first year of life results in permanent behavioral impairment in monkeys (Levin and Bowman, 1983, 1986; Lilienthal et al., 1986; Bushnell and Bowman, 1979). Monkeys with postnatal lead exposure tested at ages 6–7 ha d impaired performance on the DSAT (Rice and Gilbert, 1990). Rats with postnatal lead exposure also performed more poorly in spatial alternation tasks compared to controls, across all time delays (Alber and Strupp, 1996). One study suggested that increased postnatal lead exposure (was associated with improved DSAT performance in young rats, with increased accuracy and lower error frequency (Cory-Slechta et al., 1991). Another study that found monkeys with postnatal lead exposure performed slightly better than controls on the DSAT (Levin and Bowman, 1988). The authors postulated that a slight lead-induced decrease in attentiveness may explain their results, as exposed monkeys were less distracted by extraneous cues and could result in better performance in the task at hand (Levin and Bowman, 1988).

Our study results did not find an association between MBPb and DSAT performance may suggest that pregnancy is not the most biologically relevant window of susceptibility for WM development or that effects are delayed and not expressed until later in life. Alternatively, early childhood may be a more biologically relevant window for lead-induced WM deficits, compared to pregnancy, which has been documented to be associated with global tests of intellectual development (Hu et al., 2006; Schnaas et al., 2000). We previously reported that blood lead at 2 years of age was most predictive of decreased cognitive abilities at age 4 years (Braun et al., 2012b). We postulate that at 24 months, children may be too young for researchers to observe lead-associated decrements in their WM, either due to delayed expression or imprecision in the measures of WM. However, we cannot rule out that a slight lead-induced decrease in attentiveness may render children less susceptible to extraneous stimuli during testing and result in better performance, as seen in studies with monkeys (Levin and Bowman, 1988).

5. Strengths and limitations

This is one of the first studies to use the DSAT in very young children to assess WM in a population of preliterate children 18–24 months old. Our phenotypic data was collected by trained child psychologists and underwent rigorous QA/QC procedures, and in many subjects, was collected at both 18- and 24-month timepoints. To validate the utility of using the DSAT in studies of children’s environmental health, we examine associations with prenatal exposure to a known neurotoxicant, lead. Maternal blood lead was measured in pregnancy years before the DSAT and BSID-III were administered. Despite many strengths, the study has limitations. While the parent cohort is sizeable, the DSAT was only available for a subsample of the cohort potentially limiting our statistical power. Although we found that mother-child pairs who completed the DSAT were similar in demographic characteristics to those who did not, missing data could result in potential bias.

6. Conclusion

We were able to demonstrate the feasibility and validity of the DSAT in infants, and propose that it will be a useful research tool allowing for the assessment of WM developmental trajectories in longitudinal studies. Such studies will be critical to further our understanding of windows of susceptibility to environmental risk factors. This study is one of the first studies to use the DSAT to assess WM in children at 18–24. Our findings suggest this is a feasible test to administer at this age that reflects improvements in working memory with age. As WM continues to develop throughout childhood, our ability to characterize deficits in WM during this transitional period may inform intervention to promote the healthy development of WM.