*2.4. Covariates*

Information on secondhand tobacco exposure was obtained by questionnaire; mothers reported the minutes per day that children spent with smokers at 4, 6, and 9 years. Children were weighed with the least amount of clothing possible and without footwear using an InBody230, and height was measured with a SECA stadiometer without footwear. The z-score for body mass index (BMI) was estimated according to WHO guidelines [25]: Underweight, normal weight, overweight, and obesity were defined as <sup>&</sup>lt;−2SD, >−2SD to ≤+1SD, >+1SD to +2SD, and >+2SD, respectively. As there were few underweight children, these observations were collapsed to the normal weight category. Physical activity was measured with the International Physical Activity Questionnaire (IPAQ) that was answered by the child's mother at 4, 6, and 9 years. We considered moderate to vigorous activities such as play in the park, run, walk, ride a bike, and dance to estimate minutes of aerobic activities per day. Maternal socioeconomic status was collected at the time of enrollment using a questionnaire according to the Mexican Association of Market Intelligence and Public Opinion Agencies (AMAI, 2007 version). The AMAI classifies Mexican households into seven levels (very low, low, middle-low, middle, middle-high, and high) according to their ability to satisfy the needs of their members. In this study, we collapsed the AMAI levels into three categories: lower, middle, and higher.

### *2.5. Statistical Analysis*

We performed descriptive analyses to identify the food items at each study visit that most contributed to the estimated total dietary Cd intake and reported them as percentages.

We analyzed the distributions and descriptive statistics for each kidney parameter and log-transformed variables for cystatin C, BUN, and eGFR. We used multivariable linear regression models to analyze associations between dietary Cd at 1 and 9 years with kidney parameters at 9 years. In a subset of children with dietary information across all five study visits, we derived an ordinal cumulative Cd score. The ordinal score was derived as follows: A score of 0 indicated no exposure throughout the study visits, a score of 1 indicated a high dietary Cd in one study visit and so forth, and 5 indicated high dietary Cd in all ages (i.e., 1, 2, 4, 6, and 9 years). We incorporated the ordinal score as a discrete and categorical variable in models and examined associations with each kidney function parameter. Final models were adjusted for sex, age in months (to account for the wider age range at the study visit), BMI z-score, physical activity, secondhand tobacco exposure, and socioeconomic status. All statistical analyses were performed in Stata Statistical Software: Release 14. College Station, TX: StataCorp LP.
