**1. Introduction**

Although cadmium is a naturally occurring heavy metal, its increased use in numerous industrial applications has made it one of the most abundant environmental pollutants present in atmospheric, terrestrial, and aquatic systems [1,2]. Cadmium is classified as a probable carcinogen [3] and its lack of degradation in the environment facilitates its persistence and enables sustained human exposure [2,4]. While contaminated air from industrial processes is the most cited source of occupational exposure [2], non-occupational cadmium exposure can occur through the inhalation of tobacco smoke and dust. Cadmium is also present in some commercial fertilizers [5–8] and contamination of agricultural soils results in the ingestion of cadmium through dietary staples [2,4]. In the US and other regulated societies, dietary cadmium intake is estimated at ~1 μg/day [1,9]. The slow excretion of cadmium leads to accumulation in the body over time [4].

Early-life cadmium exposure in children and pregnan<sup>t</sup> women has been associated with low birth weight (either due to growth restriction or shorter gestation) [10–13], childhood disorders including neurodevelopmental disorders, and indicators of metabolic dysfunction such as obesity—which have been recapitulated in zebrafish [14]. Low birth weight and preterm birth are public health concerns, as they are risk factors for early mortality and the onset of later disease and co-morbidities [15]. Approximately 8% of infants in the US are born with a low birth weight (≤2500 g) [16] and approximately 10% of infants are born preterm (prior to 37 weeks gestation) [17] with increased frequencies for African-American infants [17] The managemen<sup>t</sup> of low birth weight and preterm birth and associated co-morbidities poses a large financial burden on families and the healthcare system [18].

Dietary supplementation with essential metals including iron, calcium and selenium, has been recommended by multiple environmental health agencies to mitigate the e ffects of cadmium exposure, in part because cadmium influx occurs with metal transporters for these essential metals [19–21]. However, appropriate doses are unclear, as these essential metals can be toxic at high doses [20,21]. Polyphenols and other antioxidants in the diet have also shown the potential to reduce the negative consequences of cadmium exposure [20,21]. As humans typically consume combinations of nutrients as a part of meals or whole foods, the study of dietary patterns is an important tool to understand how public health recommendations can help reduce the risk from prenatal cadmium exposure. The Mediterranean diet pattern which is characterized by a high intake of iron, selenium, and antioxidants can be easily studied and translated to public health guidelines. Maternal adherence to a Mediterranean diet during pregnancy has been found to be associated with a reduced risk of gestational diabetes [22], normal birth weight [23–26] and longer gestational age [26,27], as well as favorable behavioral patterns [28], and other positive child outcomes [26]. Although there is an abundance of evidence suggesting multiple health benefits from adhering to a Mediterranean diet pattern [29–33], the potential role of this diet in mitigating the e ffects of prenatal cadmium exposure and birth outcomes has not ye<sup>t</sup> been described. These analyses aim to explore the e ffect measure modification of Mediterranean diet pattern adherence during pregnancy in the association between elevated prenatal cadmium exposure and birth outcomes, including birth weight and preterm birth, and exploratory analyses of the association between prenatal cadmium exposure and Apgar scores and infant ponderal index at birth.

#### **2. Materials and Methods**

We used data from participants of the Newborn Epigenetics Study (NEST), a cohort of women–infant dyads from central North Carolina. Enrollment details have been described elsewhere [34]. In brief, 1700 women enrolled during pregnancy between 2009 and 2011 at qualifying prenatal clinics. Women met the following inclusion criteria: 18 years of age or older, plan to deliver in one of two birthing facilities in Durham county, and English or Spanish speaking. We excluded women who planned to give up custody of their child and those who did not carry offspring to term. Of the 1700 enrolled, 1304 remained after additional exclusions (*n* = 115 experienced a fetal death, *n* = 281 refused further participation or an inability to follow-up with the participant). We collected blood and obtained cadmium measures for the first *n* = 310 women. Of the 310, *n* = 298 had non-missing values for birth weight, and *n* = 185 women completed a food frequency questionnaire (FFQ). The median gestational age at enrollment was 11–12 weeks. We have previously shown that the 310 mother–infant pairs in whom cadmium was measured did not vary significantly from the remainder of the cohort [14]. The *n* = 185 on whom FFQ data were available also did not differ from the 310 with respect to sex, race/ethnicity and maternal obesity distribution (*p* > 0.05). Those included were, however, more likely to have a higher educational level and older maternal age at delivery (*p* < 0.05). These factors were adjusted for in the analysis. The women in our cohort were not significantly different with respect to covariates from the women in the overall NEST cohort (*p* > 0.05). This study was approved by the Duke University Institutional Review Board (#Pro00014548) on 19 February 2020.

### *2.1. Cadmium Exposure*

We measured cadmium in whole blood for the first 310 enrolled women at a median gestation age of 12 weeks, using ICP-MS and methods previously described in detail [11,13]. Because cadmium co-occurs with other environmental pollutants [35], we also measured lead and arsenic. Briefly, we measured prenatal cadmium concentrations in whole blood donated at enrollment as nanograms per gram (ng/g; 1000 ng/g = 1035 ng/μL) of blood weight using well-accepted solution-based ICP-MS methods. We homogenized temperature equilibrated whole blood samples (0.2 mL) and pipetted them into a trace metal-clean test tube. We used a calibrated mass balance to confirm the samples gravimetrically to ±0.001mg, and we spiked samples with internal standards consisting of known quantities (10 and 1 ng/g, respectively) of indium (In) and bismuth (Bi) (SCP Science, USA), used to correct for instrument drift. We then diluted the solutions with water purified to 18.2 M Ω/cm resistance, which we will refer to as Milli-Q water (Millipore, Bedford, MA, USA) and acidified the solutions using ultra-pure 12.4 mol/L hydrochloric acid to result in a final concentration of 2% hydrochloric acid (by volume). We prepared all standards, including aliquots of the certified NIST 955c, and procedural blanks using the same process. We measured Cd concentrations using a Perkin Elmer DRC II (Dynamic Reaction Cell) axial field ICP-MS at Duke School of the environment, Durham, NC, USA. Calibration standards used to assess metals in blood included aliquots of Milli-Q water, and NIST 955c SRM spiked with known quantities of each metal in a linear range from 0.025 to 10 ng/g. We prepared standards from 1000 mg/<sup>L</sup> single element standards (SCP Science, USA). We calculated method detection limits (MDLs) consistent with the two-step approach using the t99SLLMV method (USEPA, 1993) at 99% CI (*t* = 3.71). The MDLs generated values of 0.006, 0.005, and 0.071 μg/dL, for cadmium, lead and arsenic, respectively. The thresholds of detection (LODs) were 0.002, 0.002, and 0.022 μg/dL, for Cd, Pb and As, respectively, and limits of quantification (LOQs) (according to Long and Winefordner, 1983) were 0.0007, 0.0006, and 0.0073 μg/dL for Cd, Pb, and As, respectively. The number of samples below the LOD for Cd, Pb, and As were two, two, and one, respectively.

### *2.2. Mediterranean Diet*

We measured overall diet using a modified food frequency questionnaire (FFQ) [36] at enrollment. Women were asked to report their usual intake over the past 3 months, allowing us to capture the periconceptional period. We scored women's diets using the data-driven Mediterranean Diet Score (MDS) [37]. The MDS assesses adherence to a Mediterranean diet pattern based on the reported intake of foods that are deemed to be beneficial: fruit, vegetables, fish, dairy, whole grains, legumes, nuts, and monounsaturated fatty acids, and foods that are deemed detrimental: meat. We excluded alcohol from the diet score, as alcohol is not generally recommended during pregnancy and the reported alcohol intake in our cohort was low. Women who reported an intake of a beneficial foods at or above the median for the study population received a score of 1 and 0 otherwise. Those who reported an intake of detrimental foods below the median received a score of 1 or 0 otherwise. The MDS ranges from 0 to 9, with 0 representing the lowest possible adherence to a Mediterranean diet pattern and 9 representing the highest adherence to a Mediterranean diet pattern. We assessed maternal Mediterranean diet adherence as low (MDS at or below 4) and high (MDS above 4).

### *2.3. Birth Outcomes*

At delivery, we abstracted parturition data, including infant birth weight and gestational age, from medical records. We used standard definitions for low birth weight (≤2500 g) and preterm birth (<37 weeks). We assessed birth weight continuously and preterm birth categorically (>37 weeks gestation/≤37 weeks gestation). We assessed infant Apgar scores continuously (1–10), with a higher score reflecting a greater level of health at birth. We also assessed birth length (cm), and derived the infant ponderal index (PI) at birth. PI is a measure of the proportionality of body growth and is calculated using the formula: weight (g) × (100/length (cm3)).
