The mean age and education of the sample was 61.49 (sd = 12.70; range = 40-96) and 10.76 (sd = 4.37; range = 0-20), respectively. Mean GIS-based selenium concentration was 17.70 μg/L (sd = 10.66; range = 3.96-56.32). Seventy-seven percent of the sample was interviewed in English and 33% in Spanish; 88% of the sample self-reported their race as White, and 38% reported their ethnicity as Mexican American. Mean GDS-30 score of the sample was 10.49 (sd = 5.07; range = 0-26). Demographic characteristics of the sample are presented in Table .
Selenium concentrations were found to be significantly and negatively related to all GDS scores. Specifically, higher selenium levels were significantly related to lower scores in the Total GDS-30 scores (β = -0.34, p < 0.001), explaining 11% of the variance. Higher selenium was also related to a decrease in factor scores of Dysphoria (β = -0.23, p < 0.001), explaining 5% of the variance, Meaninglessness (β = -0.42, p < 0.001), explaining 17% of the variance, and Apathy (β = -0.30, p < 0.001), explaining 9% of the variance. Higher selenium concentration was statistically a significant predictor of lower scores on the subfactor of Cognitive Impairment (β = -0.11, p = 0.01), but only accounted for 1% of the variance. The results of the regression analysis can be found in Table . Higher selenium concentration was also associated with a significantly lower risk of depression, with B = -0.06 (SE = 0.01), Wald (df = 1) = 37.25, p < 0.001, and OR = 0.94 (95% CI = 0.92-0.96).
The link between depression and GPX1 polymorphisms was strongest for CC (homozygous wild type) and TT (homozygous mutant type) carriers, with the greatest impact being for TT individuals as shown in Table . The TT polymorphism was the least frequent in the sample; however, for TT individuals, selenium concentrations accounted for 26% of the variance in GDS Total Score, 13% of the variance in Dysphoria, 37% of the variance in Meaninglessness, and 19% of the variance in Apathy. For the CC polymorphism, selenium concentrations accounted for 15% of the variance in total score, 9% of the variance in Dysphoria, 22% of the variance in Meaninglessness, and 14% of the variance in Apathy. The amount of variance for CT polymorphism was relatively small.
Linear Regression Results by GPX1 Allele
Next, we examined the impact of the GPX1 Pro198Leu polymorphism on the findings. First, we sought to determine if homozygous carriage of the C allele was associated with depression status (GDS ≥ 10 = depressed), compared to T allele carriers (CT plus TT; TT was not analyzed separately due to sample size). Logistic regression was conducted with depressed status as the outcome variable and GPX1 polymorphism (CC versus CT) as a categorical predictor variable. Sixty-six percent of CC homozygotes were depressed compared to 57% for T allele carriers; therefore, homozygous carriage of the C allele was associated with a significantly increased risk of depression (B = 0.50, SE = 0.19, Wald (df = 1) = 6.88, p = 0.009, and OR = 1.65 (95% CI = 1.14-2.40). Next, a linear regression model was created to examine the gene-dose effect of GPX1 C versus T (CC = 2, CT = 1, TT = 0) alleles on depression scores (GDS total score). With age, gender, education, and test language entered as covariates, the increasing presence of the T allele was associated with significantly higher GDS scores (B = 0.94, SE = 0.35, t = 2.69, and p = 0.008). Lastly, we re-ran our regression analyses examining the link between selenium concentrations and GDS scores (total and subfactors) after stratification of the sample on genotype at the GPX1 Pro198Leu polymorphism. The link between selenium levels and depression was strongest among CC and TT, as compared to CT individuals. The TT polymorphism was the least frequent in the sample. Of note, in regression analysis of groundwater selenium levels on depression risk, models that included Pro198Leu genotype along with age, gender, education, and language of test administration accounted for more variance (70%) than those that did not include genotypic status (50%). For CC polymorphism, selenium concentrations alone accounted for 15% of the variance in total score, 9% of the variance in dysphoria, 22% of the variance in meaninglessness, and 14% of the variance in apathy, which was more variance accounted for than found by covariates. The amount of variance for CT polymorphism was relatively small. See Table .
Therefore, these findings point to a gene-environment interaction, which is supported by the gene-dose findings.