**4. Discussion**

Understanding the e ffects of cadmium in early life is important, as this toxic metal is ubiquitous in the environment. With no set upper threshold for children, the accepted tolerable limits based on body weight are likely detrimental to children whose body weight is also smaller. In these analyses, we evaluated the extent to which adherence to a Mediterranean diet modified the association of cadmium and poor birth outcomes. We found that elevated prenatal cadmium exposure was associated with a lower birth weight compared to infants born to mothers with average or low cadmium exposure during pregnancy. These associations persisted after further adjusting for other co-occurring toxic metals that have been previously associated with these poor birth outcomes. Furthermore, after removing gestational age as a confounder, the association between prenatal cadmium exposure and birth weight remained significant. However, the association between prenatal cadmium exposure and preterm birth lost significance after removing birth weight. We found no association between prenatal cadmium exposure and Apgar score or infant PI at birth.

Our findings are consistent with previous data from our group and others that have demonstrated that prenatal exposure to cadmium, at non-occupational levels, is associated with lower birth weight [10–12] and is not associated with preterm birth [10,38]. These findings are, however, not consistent with the hypothesis that dietary patterns rich in iron, selenium, and folate may mitigate exposure. We did not find evidence to support our hypothesis that adherence to a Mediterranean diet prenatally modifies the associations between cadmium exposure and poor birth outcomes, regardless of the cut-off used to define "high Mediterranean adherence". While sample size limits inference, these data sugges<sup>t</sup> that, in this population, at these cadmium levels, adherence to a Mediterranean dietary pattern may not modify the effects of prenatal cadmium exposure on birth outcomes. We were also interested in understanding whether the association between prenatal cadmium exposure and adverse birth outcomes was related to growth restriction or shortened gestation. In our exploratory analysis, the association between cadmium exposure and birth weight remained significant after excluding gestational age from the model; however, the association between cadmium exposure and preterm birth was no longer significant after excluding birth weight. This suggests that cadmium exposure may influence birth weight through growth restriction rather than shortened gestation.

To our knowledge, this is the first attempt to determine the effects of the Mediterranean diet on the association between prenatal cadmium concentrations and documented poor birth outcomes in humans. Previous animal- and cell-based studies have shown that a dietary intake of iron, calcium, selenium, and folate can reduce toxicity from cadmium exposure [20,21]. The Mediterranean diet pattern has been found to be a rich source of iron, folate, and selenium [39]. Thus, our analysis showing that adherence to a Mediterranean diet pattern during pregnancy did not change the association between prenatal cadmium exposure and birth weight or preterm birth was surprising. The average intake of selenium in the US is high at 108.5 mcg/day [40], the intake of iron and calcium from food is lower than recommended at 11.5–13.7 mg/day [40] and 748 to 968 mg/day (females) [41], respectively, and the intake of folate is insufficient for pregnancy at 455 mcg DFE/day (females) [40]. It is possible that usual eating patterns in the US may not provide sufficient amounts of these nutrients to mitigate elevated cadmium exposure; therefore, diet interventions that encourage following a diet pattern with higher levels of iron, calcium, and folate may be warranted. Interventions focused on dietary modifications may hold better prospects for implementation and adherence in exposed populations when compared to interventions focused on costly landscape remediation. Additionally, dietary intervention does not carry the health risks associated with cadmium chelation using agents such as EDTA.

The inability to find associations could be due to one of several possibilities, some related to how cadmium is estimated, and others related to the measurement of diet. For example, maternal circulating levels of cadmium may not reflect cadmium levels that the o ffspring may be exposed to, as there is evidence in support of cadmium being sequestered by the placenta [42]. Secondly, because in the United States the main source of cadmium in the diet is lettuce, milk and cookies [43], it is possible that the additional exposure to cadmium may overwhelm the nutritive benefits of this diet. It is also possible that women's intake in our sample may not represent a true Mediterranean diet pattern, even at high MDS values, as food choices and availability may di ffer by country [44]. Additionally, it may be that adherence to a Mediterranean diet pattern does reduce the e ffects of cadmium exposure on lower birth weight, ye<sup>t</sup> we were underpowered to detect the associations. Although we were unable to establish a modifying e ffect of Mediterranean diet on the association between prenatal cadmium exposure and birth outcomes, the Mediterranean diet has been shown to protect against a number of diseases and inflammatory processes in the body [45], thus providing a rationale for its continued study. Given the ubiquity of this toxic metal in the environment, the e ffects of cadmium on birth outcomes, and the plausibility that a Mediterranean diet may mitigate the adverse e ffects, repeating these analyses in larger data sets is warranted.

Our study findings should be interpreted in the context of the study limitations. In addition to being underpowered to detect a significant e ffect measure modification that may have existed, both the measurement of Mediterranean diet adherence and our inability to "remove" the e ffects of dietary items such as lettuce and milk, which are major sources of cadmium in the US, is a limitation. An analysis of these relationships in di fferent populations may clarify these findings. Furthermore, although implausible values were excluded from analyses, the Mediterranean diet was computed from self-reported diet data, which may have led to misclassified food intake that may be further biased by social desirability. This may have led to the under-reporting of unhealthy foods and the over-reporting of healthy foods. An additional limitation is the use of the MDS to assess diet. Although widely used and associated with a number of health outcomes, the MDS does not assess many foods that are thought to be "detrimental", such as sugar or highly processed convenience foods.

Despite this, our study also exhibits strengths. We used prospectively collected data, therefore we can establish the timing of exposure, modifiers, and outcomes. Another important strength of this study is that it assessed maternal dietary patterns rather than the intake of single nutrients. Humans consume most of their nutrients through foods and combinations of nutrients, therefore it is important to assess the role of dietary patterns in the potential mitigation of negative consequences from toxic exposures.
