Factors affecting the variability in the observed levels of cadmium in blood and urine among never smoker adults aged ≥ 20–64 years and senior citizens aged ≥ 65 years

Data from National Health and Nutrition Examination Survey for 1999– 2012 were used to evaluate factors that affect observed levels of blood cadmium (BCd) and urine cadmium (UCd) among never smoker adults aged 20–64 years and senior citizens aged ≥ 65 years. Over 1999–2010, adjusted levels of BCd declined for adults aged 20–64 years (β = −0.02457, p < 0.01) and senior citizens aged ≥ 65 years (β = −0.01845, p < 0.01). For UCd, this decline was observed among never smoker adults aged 20–64 years only (β = −0.01605, p < 0.01). Adjusted BCd as well as UCd levels were found to be statistically significantly higher for female never smoker adults as well as senior citizens than male never smoker adults and senior citizens, respectively, (p < 0.01). Non-Hispanic whites had statistically significantly lower levels (p < 0.01) of BCd than non-Hispanic blacks and Mexican Americans for adults as well as senior citizens (p < 0.01). For UCd, this association was not observed. Among adults, levels of both BCd increased with age and decreased with body mass index (p < 0.01). Dietary intake of cadmium was not found to affect the levels of either BCd or UCd. *Corresponding author: Ram B. Jain, Independent Researcher E-mail: jain.ram.b@gmail.com Reviewing editor: Peter Christiaan Speldewinde, The University of Western Australia, Australia Additional information is available at the end of the article

ABOUT THE AUTHOR Ram B. Jain has been engaged in environmental health research since 2002. Since his retirement from Centers for Disease Control in 2010, he practices as a private consultant and an independent researcher from his home and is primarily involved in analyzing data using National Health and Nutrition Examination Survey. He has recently published in Biomarkers, Environmental Science and Pollution Research, and Environmental Toxicology and Pharmacology. His current research interests, among others, include studying concentrations of environmental contaminants among users of e-cigarettes as compared to the smokers of traditional tobacco products.

Introduction
Tobacco leaves contain high amounts of cadmium (Cd) and as such, smoking greatly increases exposure to Cd (https://www.atsdr.cdc.gov/toxprofiles/tp5.pdf, p. 12). Among nonsmokers and children, the largest source of Cd is via consumption of leafy vegetables such as lettuce and spinach; staples such as potatoes and grains; and peanutsoaen, soybeans, and sunflower seeds. Consumption of shell fish and organ meats is also associated with increased exposure to Cd. As quoted by Agency for Toxic Substances and Disease Registry (https://www.atsdr.cdc.gov/toxprofiles/tp5.pdf), half-life of Cd has been estimated to be 6-38 years for the human kidney and 4-19 years for human liver (p. 201). As such, Cd stays in the human body for a very long time. Concentrations of Cd in blood (BCd) and urine (UCd) provide estimates of human exposure to Cd. While, BCd can be used as a measure of the recent exposure to Cd, UCd can be used as a measure of long-term exposure to Cd (Adams & Newcomb, 2014). Since, estimated half-life of cadmium in the kidney is from one to four decades (https://www.atsdr.cdc.gov/toxprofiles/tp.asp?id=48&tid=15), UCd being a measure of long-term exposure to Cd, can be stated to represent exposure to Cd for several decades.
Mortality linked data from National Health and Nutrition Examination Survey (NHANES) have been often used to report on associations between exposure to Cd and all-cause and cause-specific mortality. Some of the authors who have reported elevated levels of BCd and/or UCd to be associated with mortality are Aoki et al. (2016), Cheung, Kang, Ouyang, andYeung (2014), Patel et al. (2013), Lin et al. (2013), , and Wu, Chang, Lin, Caffrey, and Lin, (2011).
Smokers have been reported to have higher levels of Cd than nonsmokers (https://www.cdc.gov/ biomonitoring/Cadmium_BiomonitoringSummary.html). Some of the studies in general population and among several groups of patients reporting higher levels among smokers as compared to nonsmokers are by Ahn, Kim, and Park (2017), Yaprak et al. (2017), Afridi, Kazi, Talpur, and Brabazon (2015), Kazi et al. (2010), Richter, Bishop, Wang, and Swahn (2009), and Sorkun et al. (2007. Adverse health effects associated with occupational and environmental exposure to Cd (and other contaminants) have been extensively studied. Some of the recent studies reporting on these health adverse effects are by Thomas, Hodgson, Nieuwenhuijsen, andJarup (2009), Beveridge, Pintos, Parent, Asselin, andSiemiatycki (2010), Honda et al. (2010), Gil et al. (2011), Haddam et al. (2011), Arrandale, Beach, Cembrowski, and Cherry (2015), De Franciscis et al. (2015, Togawa et al. (2016), andDodd-Butera et al. (2017). https://doi.org/10.1080/23311843.2017.1379258 Adams and Newcomb (2014) used NHANES data from 1999-2010 for those aged ≥ 20 years and reported both BCd and UCd levels to be substantially higher among those aged ≥ 70 years than those aged 20-29 years. Adjusted geometric means of BCd and UCd were reported to be higher in females than males. Riederer, Belova, George, and Anastas (2013) used NHANES data from 1999and reported decrease in UCd levels among children, teens, and adults over 1999 reported declining trends (p < 0.01) in UCd levels over 1988-2008 in overall US population as well as all participant subgroups. Among females of reproductive age, non-smoking Mexican American and non-Hispanic Black females were more likely to have high Cd than non-Hispanic white females (Mijal & Holzman, 2010).
The objectives of the proposed study were to (i) study time trends in unadjusted and adjusted levels of BCd and UCd over 1999-2012 for never smokers adults aged 20-64 years and senior citizens aged ≥ 65 years, and (ii) evaluate factors that affect adjusted levels of BCd and UCd including gender, race/ethnicity, body mass index (BMI), poverty income ratio (PIR), exposure to environmental tobacco smoke at home (ETSH) and work (ETSW), and diet. The proposed study is similar to the study by Adams and Newcomb (2014) in certain aspects but different in other aspects. Adams and Newcomb (2014) did consider using ETSH as a variable but excluded it from the final models because it was not found to be statistically significant. For the purpose of this study, ETSH and intensity of exposure to ETSH will be studied. In addition, ETSW and intensity of exposure to ETSW will also be used as independent variables in the models. The proposed study will fit separate regression models for BCd and UCd for both adults and senior citizens. Adams and Newcomb (2014) did not consider the role of dietary intake on the levels for BCd and UCd, this study will.

Data source and data descriptions
Data from NHANES for 1999-2012 from demographic, body measures, blood metal, urine metal, serum cotinine, home interview smoking questionnaire, recent tobacco use questionnaire, family smoking questionnaire, occupational questionnaire, and first day individual food dietary questionnaire files were downloaded and match merged. NHANES uses a complex, stratified, multistage, probability sampling designed to be representative of the civilian, non-institutionalized US population based on age, sex, and race/ethnicity. Sampling weights are created in NHANES to account for the probabilities of selection and response as well as total US population for the selected combinations of gender, age, and race/ethnicity. All analyses completed for this study used sampling weights as well as sampling design information, i.e. stratification and clustering to compute relevant statistics. Unweighted sample sizes after removing missing data for dependent and independent variables are given in Table 1. Percent observations at or above the limitation of detection (LOD) for BCd were found to be 71.1% for adults and 88.4% for senior citizens. Percent observations at or above LOD for UCd were found to be 92.4% for adults and 97.5% for senior citizens.
Data for exposure to ETS at home (ETSH) were extracted from family smoking questionnaires (https://wwwn.cdc.gov/Nchs/Nhanes/2011-2012/SMQFAM_G.htm) used in NHANES. A total of three variables indicating exposure to ETSH, namely, exposure to ETSH (yes/no), number of cigarette smokers living inside home, and number of cigarettes smoked inside home every day were available. Data for exposure to ETS at work (ETSW) were extracted from occupational questionnaires (https:// wwwn.cdc.gov/Nchs/Nhanes/2009-2010/OCQ_F.htm) used in NHANES. There were two variables indicating exposure to ETSW, namely, exposure to ETSW (yes/no) and number of hours the participants inhaled smoke from other people's cigarettes, cigars, and/or pipes that were available. However, there was a change in NHANES 2011-2012 about how data on exposure to ETSW were collected. Instead of asking if the respondents inhaled tobacco smoke at work as was done for 1999-2010, the question in NHANES 2011-2012 asked the respondents if anyone smoked cigarettes, cigars, or pipes during the last two weeks in the area the respondents worked. Data on number of hours the participants inhaled smoke from other people's cigarettes, cigars, and/or pipes were no longer collected in NHANES 2011-2012. For this reason, adjusted analyses were completed for 1999-2010 only.  Table 1

. Weighted frequency distribution of blood and urine cadmium by gender, race/ethnicity, and exposure to environmental tobacco smoke (ETS) at work and home
Source: Data from National Health and Nutrition Examination Survey 1999-2010. *Observations at or above the limit of detection for blood cadmium was 71.1% and for urine cadmium 92.4%. **Observations at or above the limit of detection for blood cadmium was 88.4% and for urine cadmium 97.5%.

Definition of selected variables
Database for never smokers aged ≥ 20 years was generated by selecting those who self-reported not smoking 100 cigarettes in life, did not self-report using any tobacco products during the last five days and currently had their serum cotinine levels less than 10 ng/mL.
Selection of dietary variables for use in this study was guided by the Canadian study reported by Garner and Levallois (2016). Garner and Levallois (2016) used frequency of the consumption of selected foods in their study during the last 12 months but data on food frequency consumption for NHANES are not made available in public domain on a consistent basis. Instead, data on the consumption of selected foods in grams during the last 24 h (https://wwwn.cdc.gov/Nchs/ Nhanes/2011-2012/DR1IFF_G.htm) were used for this study. Ten food groups based on the United States Department of Agriculture's (USDA) eight-digit food codes were generated. The USDA food codes (https://www.ars.usda.gov/SP2UserFiles/Place/80400530/pdf/fndds/fndds5_doc.pdf) used to generate these food groups are provided by Jain (2016). Preliminarily, 10 food groups selected were: red meats, organ meats, shell fish, nuts and seeds, white bread, cereals, dark green leafy vegetables, white potatoes, fried potatoes, and lettuce based salads. However, percent participants consuming some of these foods were too low to have enough variability in the data. For this reason, an arbitrary decision was made to include only those food groups in the analyses for which there were at least 10% participants consuming them. As such, the food groups included in analyses were: white bread, cereals, white potatoes, fried potatoes, nuts and seeds, and lettuce based salads.

Outcome variables
Since, the distributions of both BCd and UCd were positively skewed, their measurements in ng/L were log10 transformed before being used as dependent variables in the models.

Covariates/independent variables
Categorical independent variables used in the models for adults were: gender (male, female), race/ ethnicity (non-Hispanic white or NHW, non-Hispanic Black or NHB, Mexican American or MA, other unclassified race/ethnicities or OTH), exposure to ETS at home (ETSH) and exposure to ETS at work (ETSW). Since, there were only 32 senior citizens who had exposure to ETS at work, they were deleted from the models and as such, ETSW was not used as an independent variable in the models for senior citizens. Continuous independent variables used in the models were, age, age 2 , log10 transformed values of body mass index (BMI), and poverty income ratio (PIR), survey year (coded as one through seven), number of smokers smoking inside home, number of cigarettes smoked inside home every day, and for adults only, number of hours smoke was inhaled from other people's cigarettes, cigars, and pipes as well as dietary variables as previously specified. Models also tested for significant first-order interactions between gender, race/ethnicity, ETSH, and ETSW but they were included in the final models if they were found to be statistically significant at α = 0.05.

Software and statistical analysis
All analyses conducted for this study used SAS University Edition software (www.sas.com). Proc SURVEYREG was sued to fit multivariate regression models as well as univariate models to test for time trends in unadjusted values of BCd and UCd. Pairwise comparisons were done using t-test.

Univariate analysis
Unadjusted geometric mean levels (UGM) of BCd exhibited a decreasing trends (p < 0.01) over 1999-2012 irrespective of age (Table 2) UGMs of BCd as well as UCd for both adults and senior citizens were found to be higher for females when compared with males (Table 3, p < 0.01). The ratio of females to males UGM were: 1.4 and 1.1 for BCd for adults and seniors, respectively, and 1.3 and 1.1 for UCd for adults and seniors, respectively. The order in which BCd among three major race/ethnic categories were observed was: NHB > MA > NHW and all three pairwise differences were statistically significant for both adults and seniors (Table 3, p < 0.01). Racial/ethnic differences were not found to exist for UCd (Table 3).
Exposure to ETS either at home or at work did not seem to affect UGMs for BCd or UCd among adults as well as seniors (Table 3). The observed difference between exposure to ETSW and no exposure to ETSW as observed for adults was probably due to small sample sizes.
Irrespective of gender, race/ethnicity, and ETS exposure, seniors had higher UGM for both BCd and UCd than adults (p < 0.01, Table 3).

Gender and racial/ethnic differences
Adjusted geometric means (AGM) of BCd as well as UCd for both adults and senior citizens were found to be higher for females when compared with males (Table 4, p < 0.01). The ratios of females to males AGMs were: 1.4 and 1.2 for BCd for adults and seniors, respectively, and 1.2 and 1.5 for UCd for adults and seniors, respectively. The order in which BCd among three major race/ethnic categories were observed was: NHB > MA > NHW and all three pairwise differences were statistically significant for both adults and seniors (Table 4, p < 0.01). Racial/ethnic differences were not found to exist for UCd (Table 3). However, among adults, when statistically significant interaction between race/ ethnicity and ETSH was taken into account, NHW had lower AGM for BCd than NHB and MA only when there was no ETSH exposure (0.228 vs 0.270 and 0.268 μg/L, Figure 1, Panel A).
In the presence of statistically significant interaction between gender and race/ethnicity for BCd for seniors, male-female differences were limited to NHW (0.34 vs. 0.41 μg/L, p < 0.01) and NHB   For adults, there was a statistically significant interaction for UCd, and male-female differences were observed for all four racial/ethnic groups but male-female differences varied by race/ethnicity. For NHW, the male-female difference was 0.066, 0.112 μg/L for NHB, and 0.111 μg/L for MA. Thus, male-female differences were higher for NHB and MA as compared to NHW (Figure 2, Panel B). Between NHW and MA and between NHB and MA, statistically significant differences were observed for both males and females but statistically significant differences between NHW and NHB were observed for females only (Figure 2, Panel B).

Effect of exposure o ETS
AGMs for those who were exposed to ETS and were not exposed to ETS at home and/or work did not differ (Table 4) for either BCd or UCd. However, among adults, when statistically significant interaction between race/ethnicity and ETSH was taken into account, NHB with no ETSH had higher AGM for BCd than NHB without ETSH (0.270 vs 0.211 μg/L, Figure 1, Panel A). Also, when statistically significant interaction between ETSH and ETSW is taken into account for adults, among those with ETSH exposure, exposure to ETS at work was associated with lower BCd levels when compared with no exposure to ETS at work (0.218 vs. 0.274 μg/L, p = 0.04, Figure 1, Panel B); and among those with no ETSH, exposure to ETS at work was associated with higher BCd levels when compared with no exposure to ETS at work (0.274 vs. 0.260 μg/L, p = 0.04, Figure 1, Panel B).

Effect of age, age 2 , body mass index, and poverty income ratio
Among adults (Table 5), levels of both BCd and UCd increased with increase in age (Table 5, p < 0.01). For one-year increase in age, the levels of BCd and UCd increased by 1.3% and 6.1%, respectively. For adults, there was a negative association between the levels of UCd and age 2 . Among seniors, there was a negative association between the levels of UCd and age (β = −0.09013, p < 0.01, Table 6) and a positive association between the levels of UCd and age 2 (β = 0.00063, p = 0.03, Table 6).
For a 10% increase in body mass index, among adults, levels of BCd decreased by 2.5% and the levels of UCd decreased by 2.6% (Table 5). For a 10% increase in body mass index, among seniors, levels of BCd decreased by 1.6% and the levels of UCd decreased by 3.5% (Table 6). Poverty income ratio did not affect the levels of either BCd or UCd among either adults or seniors (Tables 5 and 6).  Table 4. Adjusted geometric means with 95% confidence intervals for blood cadmium in μg/L and urine cadmium in μg/L by gender, race/ethnicity, and exposure to environmental tobacco smoke (ETS) by age for never smokers

Impact of exposure to ETS
Number of cigarette smokers smoking inside home, number of cigarettes being smoked at home every day, or number of hours participants were exposed to ETS at work did not affect the levels of BCd or UCd among either adults or seniors (Tables 5 and 6).

Adjusted time trends
Among adults, for every two-year period, adjusted BCd levels decreased by 5.8% and adjusted UCd levels decreased by 3.8% (Table 5, p < 0.01). Among seniors, for every two-year period, adjusted BCd levels decreased by 4.3% (Table 6, p < 0.01) but adjusted UCd levels did not change over time ( Table 6, p = 0.27).

Impact of diet
Diet did not affect the adjusted levels of BCD or UCd among either adults or seniors. Trends in adjusted levels of BCd paralleled trends for unadjusted levels. For 1999-2010, decline in adjusted levels of BCd was 5.8% for adults and 4.3% for seniors for every two year. Decline in adjusted levels of UCd was 3.8% for adults and 1.8% for seniors for every two year.

Age differences
Unadjusted levels of both UCd and BCd were higher for seniors than adults. While no formal statistical tests were conducted, adjusted levels of both BCd and UCd were in the order: seniors > adults (Table 4). In addition, among never smoker adults, adjusted levels of BCd increased by 1.3% (p < 0.01) but not among seniors. Adjusted levels of UCd increased by 6.1% (p < 0.01) among adults but decreased by 23.1% (p < 0.01) among seniors. However, because of statistically significant negative slope for age 2 (p < 0.01), this increase slowed down with increase in age for BCd as well as UCd for adults.

Gender differences
Higher Adjusted levels of both BCd and UCd for never smoker females than never smoker males are consistent with other studies, for example, Adams and Newcomb (2014). Riederer et al. (2013), however, did not observe these gender differences which according to these authors may be due to them controlling for a wider range of covariates. Females have been stipulated to absorb higher levels of Cd because of relatively lower iron storage than males (Gallagher, Chen, & Kovach, 2011;Julin et al., 2011). Riederer et al. (2013) reported MA and NHB adults born within and outside of U.S. to have higher UCd levels than NHW born within U.S. among adult nonsmokers and MA born outside of U.S. to have higher UCd levels than NHW born within U.S. among adult smokers. Richter et al. (2009) reported unadjusted levels for UCd levels among NHB to be higher than NHW. Consistent with these observations, for this study for never smoker adults and seniors, the order of adjusted levels for BCd and UCd was NHW < NHB as well as NHW < MA and almost all pairwise differences were statistically significant. Riederer et al. (2013) speculated race/ethnic differences to be attributable to where different racial/ethnic groups reside. Those living in urban areas are likely to be associated with higher Cd levels than those living in the rural areas. The speculation of Riederer et al. (2013) may be true to a degree because there are racial/ethnic differences in where they live but the possibility of metabolic differences in how Cd is excreted from the body among different race/ethnic groups cannot be ruled out.

Effect of environmental tobacco smoke
Exposure to ETSH or ETSW did not affect the levels of BCd or UCd in this study. These results are consistent with those reported by Adams and Newcomb (2014) who did not find exposure to ETSH having any effect on the levels of either UCd or BCd.

Effect of body mass index
Negative association between BCd as well as UCd levels and BMI for never smoker adults and seniors is consistent with the results reported by Riederer et al. (2013) and Padilla, Elobeid, Ruden, and Allison (2010). For a 10% change in BMI, the decrease in the levels of BCd was 2.5% for adults and 1.6% for seniors. For a 10% change in BMI, the decrease in the levels of UCd was 2.6% for adults and 3.5% for seniors.

Effect of poverty income ratio
For this study, no association between poverty income ratio and BCd as well as UCd was observed for never smoker adults and seniors. Riederer et al. (2013) did not find any association between UCd and PIR. Garner and Levallois (2016) evaluated the association of the frequency of the yearly consumption of nine different foods with the levels of BCd and UCd among all respondents and never smokers and reported positive associations between BCd levels and (i) the consumption of shellfish, cereals, and potatoes other than french fries for all female respondents, and (ii) cereals and nuts and seeds for all male respondents. They also reported negative associations between UCd levels and potatoes other than french fries among all female respondents. In order to evaluate association between dietary intake and BCd and/or UCd, long-term dietary intake data like the one used by Garner and Levallois (2016) are preferred but due to the non-availability of food frequency data from NHANES on a consistent basis, data on consumption of 10 foods consumed in grams during the last 24 h were used for this study. Since, UCd is a measure of long-term exposure to Cd and dietary data used for this study depicted recent intake of Cd via food, no association between UCd and dietary intake was observed. Also, no association between BCd levels and dietary intake was observed. However, Riederer et al. (2013) also used total consumption of micronutrients like iron, magnesium, selenium, and others during the last 24 h to evaluate association between UCd levels and these micronutrients and found no associations for children and adolescents. But, they did observe a negative association between UCd levels and total calorie intake during the last 24 h and a positive association with total magnesium intake during the last 24 h for nonsmoker adults.

Strengths and limitations
A large nationally representative data over a period of 14 years add strength to the results observed in this study. However, cross-sectional nature of data also dilutes the confidence that can be placed in the observed results. No longitudinal data on the levels of exposure to Cd including those from dietary sources and no data on observed levels of BCd and/or UCd were available. UCd data were based on spot urine samples which further complicate the issue of "true" levels of Cd in these samples.