**4. Discussion**

In this longitudinal study of 601 Mexican children, we found that the estimated dietary Cd exposure increased with age, from ~4.4 μg/d at 1 year to 8.1 μg/d at 9 years. The main food sources of Cd changed across all study visits; at 1 and 2 years, leafy greens, milk, and carrots were the primary dietary contributors, and beginning at 4 years, we observed a dietary transition to sweets, lettuce, and sandwiches as the primary contributors. This transition most likely reflects maternal control of the child's dietary intake during the first two years of life and then the child's food preferences influencing intake at older ages. TWI was exceeded at all study visits, with 64% at 1 year decreasing to 16% in 9-year children consuming more Cd than is recommended. Finally, we found inverse associations between dietary 9-year high Cd intake, as well as the cumulative score and BUN β = −0.077 (95% CI: (−0.151, −0.003)) and β = −0.037 (−0.072, −0.003), respectively.

Our results are similar to the Cd intake observed among 4–12-year-old Australian children where the Cd intake was 4.0 ± 2.2 (0.98–9.5) μg/d [27] but higher than a study from the United States where the mean Cd intake in 2 to 10-year-old children was 2.96 (2.83, 3.10) μg/d [28]. Regarding the main food contributors, these results are similar to a previous study in the United States where the top Cd

contributors to children's diet at 10–11 years old were milk, lettuce, and cookies [28]. The TWIs are in line with a study in France that observed that diets of 5–6-year-old children exceeded the Cd TWI by 12–15% [12], while diets of 1 to 3-year-old and 6-year-old Finnish children exceeded the TWI by 88% and 64%, respectively [13].

Cd intake was comparable to other populations, suggesting that at these daily intakes, the kidney parameters we studied are not altered in 9-year-old children. There are several possible interpretations of our findings. It is possible that at this early stage of childhood (age 8–12), potential changes in kidney function are not ye<sup>t</sup> apparent. This is likely given that glomerular function, as assessed via eGFR and less specifically with BUN, may not show a marked decrease until substantial dysfunction has occurred [29]. Thus, we note with interest the marginally significant inverse associations with eGFR as early as age 8–12 observed in this study. The earliest indication of kidney damage in humans is typically an increased excretion of low-molecular-weight proteins, such as β2-microglobulin, α1-microglobulin, retinol binding protein, and N-acetyl-β-glucosaminidase, among others; and increased excretion of calcium and metallothionein [30]. We also note that BUN is a nonspecific biomarker that can vary independently of the GFR. BUN can be associated with other factors such as liver dysfunction [31], and low or high BUN may be related to undesirable states like malnutrition, starvation, dehydration, high protein intake, among others [32]. Future studies should assess biomarkers of subclinical injury, including urinary proteins like β2-microglobulin and additional assessment of liver function.

Another interpretation is that there is no kidney damage at these Cd intake concentrations. We would not anticipate 'normal' dietary levels of Cd to cause clinically apparent kidney dysfunction. However, there is evidence that when Cd reaches between 50 and 300 μg/g wet weight in the kidney cortex, the amount of Cd not bound to metallothionein becomes sufficiently high to cause tubular damage [30]. Studies have estimated that to reach these Cd levels, a Cd intake of >200 μg/day or lifetime intake of 1300 mg is necessary [33]. Lastly, it is possible that a compensatory mechanism exists, where kidney function in healthy children can cope with this level of Cd insult, while high levels of Cd are nephrotoxic [34].

An important limitation of our study is the estimation of dietary Cd using FFQs, as these can introduce measurement error from maternal recall. It is unlikely though that this could have led to bias as mothers were unaware of the research question (i.e., did not answer the questionnaire in terms of Cd exposure or kidney function). Nonetheless, FFQs are also subject to measurement error as, beyond maternal recall, children spend time in school or in other contexts where they consume food items most likely not reported in the FFQ. Furthermore, we did not measure the concentrations directly in food items. Except for Cd concentrations for a few food items (processed meat) reported in a study of adults in Mexico City [19], we used data from nutrient composition tables from different countries, where the Cd concentrations reported for food items may be different from Mexico. Ideally, our study would have measured the Cd concentrations in food items reported in the FFQ and purchased samples in local markets; however, this was beyond the scope of this study. By using food content tables from the United States, The European Food Safety Authority, Australia, Hong Kong, and Canada, we aimed at having a higher diversity of food items, as well as a higher variability for the Cd concentrations, also considering the reality of globalized food supply chains and the availability of international food products in Mexican markets.

We therefore cannot rule out that we either under- or over-estimated Cd intake; however, the results are in line with levels reported in previous studies. For example, a study conducted by the United States Food and Drug Administration showed that the top contributors to dietary Cd were grains, prepared foods (e.g., hamburgers, pizza, lasagna, soups), and vegetables [2]; a study of French children showed that the main contributors to dietary Cd were bread and potatoes [35]; and a study of Chinese children found that the three greatest contributors to dietary Cd were rice, leafy vegetables, and wheat flour [36]. Further, Cd intake estimated by FFQ does not reflect its absorption, metabolism, or excretion. Cd absorption differs between individuals based on their nutritional status, particularly levels of zinc and iron. For this study, we lacked data on indicators of children's nutritional status of these

micronutrients. We also lacked urinary Cd measurement, which is an indicator of body burden [37]. PROGRESS has archived urine samples in which Cd concentrations will be analyzed in future studies.

Among the strengths of this study are the repeated measurements of diet at five stages ranging from 1 to 9 years of age. An important observation of this study was the marked increase in children with overweight or obesity; at 4 years, only 18% of the children presented with overweight or obesity, but by age 9 years, this increased to 46.8%. We also observed an important decrease in physical activity between 6 and 9 years. Both obesity and physical activity are important metabolic risk factors that could modify the association between Cd exposure and kidney parameters [33,38]. Future studies will directly assess the role of obesity and physical activity as risk factors for longitudinal childhood kidney function trajectories.

Although we did not observe significant associations between children's Cd intake and kidney parameters, the TWI for Cd was exceeded at all study visits, from 1 to 9 years of age. The TWI for Cd is based on preventing downstream e ffects on kidney, bone, and cardiovascular health in adulthood because it can accumulate throughout life. Follow-ups of this study population at later life stages will elucidate possible Cd nephrotoxicity.

**Author Contributions:** Conceptualization, E.R.-L., M.T.-O., A.C.A. and A.P.S.; data curation, A.P.S.; formal analysis, E.R.-L., M.T.-O., A.C.A., E.O.-P., G.E.-G. and A.P.S.; funding acquisition, M.M.T.-R., R.O.W. and A.P.S.; investigation, E.R.-L. and M.T.-O.; methodology, E.R.-L., M.T.-O., A.C.A., E.O.-P., M.C.T., G.E.-G., S.P.-H., A.E.-N. and A.P.S.; project administration, M.T.-O., I.P., M.M.T.-R., R.O.W. and A.P.S.; resources, I.P. and R.O.W.; supervision, M.T.-O., A.C.A., E.O.-P., M.M.T.-R. and A.P.S.; writing—original draft, E.R.-L., M.T.-O., A.C.A., E.O.-P. and A.P.S.; writing—review and editing, E.R.-L., M.T.-O., A.C.A., E.O.-P., A.L.D., I.P., M.C.T., G.E.-G., S.P.-H., A.E.-N., M.M.T.-R., R.O.W. and A.P.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by funding from the NIH/NIEHS: K99ES027508, R00ES027508, R01ES013744, R01ES020268, and R01ES021357.

**Acknowledgments:** Authors thank the National Institute of Perinatology, Mexico for the support provided by its facilities.

**Conflicts of Interest:** The authors declare no conflict of interest.
