4. Discussion and conclusion

In this study, we assessed the association between chronic exposure to naturally-occurring F in drinking water and cognitive function in school-aged children, as measured using two distinct types of assessments: a simple drawing task of familiar objects, and the CANTAB PAL tests. The sample was recruited from 8 communities primary exposed to chronic F ranging from 0.41 to 15.5 mg/L in the MER. These communities have relatively homogenous populations with similar lifestyles and stable residency, but the residents of different villages use community-based drinking water sources that vary in their F levels. We hypothesized that measures of cognitive performance would decline with exposure to elevated F concentrations. Accordingly, we found adverse associations of F exposures in drinking water with children’s drawing and CANTAB task performance. The strongest and most significant negative impacts were observed for the more challenging drawing task—a donkey (Fig. 1A). It is observed that children struggled more when drawing a donkey than a house or a person, which may be indicative of a greater challenge accessing memory for this task. In contrast, children appeared to have an easier time drawing a person or a house, and associations between drawing performance and F exposures were correspondingly weaker. Consistent with the negative associations between drawing skill and F exposure, children drinking from wells in communities with higher F levels performed worse in CANTAB PAL tasks that are used to test new learning and memory, and especially the PAL total errors adjusted measure. It was also observed that higher F levels were related to higher deficits in the more difficult PALTEA tasks (i.e., increasing number of boxes from 2 to 8) (Fig. 5). The PAL test targets hippocampal function by measuring visual memory and new learning (Barnett et al., 2016; de Rover et al., 2011). A study by Choi et al. (2015) found that measured working memory using the Wechsler Intelligence Scale for Children-Revised (WISC-IV) in children was negatively associated with dental fluorosis (a marker of early life F exposure during critical periods of tooth development, the first 8 years) and Wechsler’s total and backward digit span tests. Goodman et al. (2022) also reported that visual-spatial and perceptual reasoning abilities may be more impacted by F exposure as compared to verbal abilities.

In previous related studies, drawing (e.g., a person) has often been used as a nonverbal screening measure of cognitive ability that may indicate visual sensory input and neuromuscular output (Imuta et al., 2013; Reynolds and Hickman, 2004; Kamphaus and Pleiss, 1991; Kamphaus and Pleiss, 1991; Reynolds and Hickman, 2004; Abell et al., 1996, Abell et al., 2001). A study by Panesi and Morra (2016) assessed dog drawing in relation to executive function and working memory and found that these two parameters jointly accounted for 58.3% of the variance of dog drawing skill. Moreover, working memory individually accounted for the largest variance (15.4%), whereas executive function accounted for 4.4%. The interaction of these two predictors was then responsible for the remaining 38.5% of this joint variance. Evidence of the role of working memory and executive function in drawing flexibility was also reported by Morra, 1994 and Barlow et al., 2003. Moreover, environmental factors can impede these aspects of cognition and drawing. For example, a study by Guillette et al. (1998) observed impairments in memory, social interaction, creativity, drawing ability, and motor skills in a population of Mexican children exposed to pesticides relative to a comparable group living in an unexposed area. Most dramatically, pesticide exposed children scored more poorly in a “draw a person” task, which may indicate lower cognitive ability or poor visuomotor coordination.

In our study, grade level (or age) was positively associated with drawing ability, which is also consistent with prior literature (Panesi and Morra, 2016). Owing to our sampling approach, however, which aimed to balance sex and age within and across communities, the children in each community are similar, such that sex- and age-related effects cannot explain the variation in observed outcomes across communities. Other possible confounders include As and Pb, which are known to be neurotoxic contaminants, but these were found at low levels in drinking water from the sample communities and in samples of children’s urine. The concentrations of As and Pb in drinking water ranged between 0.92 and 21.9 ?g/L (mean:7.3 ± 6.83 ?g/L), and 0.001 to 0.73 ?g/L (mean: 0.23 ± 0.27 ?g/L), respectively. Anemic appearance, as diagnosed from clinical signs of anemia (e.g., pallor on conjunctiva), was observed in 45.5% of the children, and is known to impair motor and mental development in infants, children, and adolescents (Lam and Lawlis, 2017; Burden et al., 2007; Lazoff, 2007). In regression analysis the association of anemic appearance, and As and Pb in water and urine did not significantly correlate with performance measures, however, this exploratory study relied on a relatively small sample and used cross sectional data to proxy for long-term exposure. In addition, the purposeful recruitment of children to obtain representative age and sex distributions around specific community wells limits the representativeness of the sample. As a result, the study may not be viewed as providing a definitive analysis of F’s neurotoxicity in children. Nonetheless, the similar sociodemographic and lifestyles in these communities minimizes the risk of confounding by variables that may be correlated with exposures and cognitive performance measures. An important additional limitation was the small number of sample communities and wells. In particular, when adjusting the standard errors for clustering within wells/communities, the statistical significance of the association between water F and the PALTEA task performance scores was reduced from p = 0.034 to p = 0.09, emphasizing the need to increase the number of wells and study participants to obtain greater statistical power.

Other limitations include a lack of control of parental variables such as maternal age, educational level of parent, socio-economic status, and assessment of chemical mixture models for better exposure and effect characterization other potential neurotoxicants (e.g., As, Pb), and elemental deficiencies such as iodine and iron that may modify cognition (Lam and Lawlis, 2017). For some urine biomarker measures that were collected as spot samples, we accounted for dilution using urine SG, to reflect actual F exposure from drinking water.

5. Conclusion

Our findings suggest that there are cognitive impairments among children exposed to higher F concentrations, evaluated using figure drawing performance and validated CANTAB cognitive tools. This study also successfully demonstrated the use of language and culture neutral CANTAB testing in a rural Ethiopian sample of children for the first time. Thus, CANTAB can feasibly be administered in this and other similar rural African contexts (as also shown by Chetty-Mhlanga et al., 2022, Chetty-Mhlanga et al., 2018; Nkhoma et al., 2013). While this exploratory study adds evidence and concern about the potential neurotoxicity of elevated F exposure, more studies are critically needed to better establish neurodevelopmental impacts of a range of F exposures from gestation to adulthood, using rigorous study designs and advanced methodologies including mixture models for exposure and effect characterization. Such studies would help provide concrete evidence to inform leaders and policy makers on the need for effective approaches to mitigate environmental exposures to F, including in F endemic geographic settings such as the study areas where alternative water sources are limited, or to establish the threshold levels at which such exposures become toxic, and specifically, inform the growing controversy over the safety of water fluoridation.

Funding sources from the PI’s Internal Institutional Start-up Fund, and National Institute of Environmental Health Sciences (NIEHS’s) career development grant (K99/R00 ES023472) that made the research possible.