2.1. Design and participants
The NHANES is an ongoing survey conducted by the Centers for Disease Control and Prevention (CDC) that uses a representative sample of non-institutionalized civilians in the US; selected by a complex, multistage, stratified, clustered probability design. Information on participants was collected through interviews and physical examinations. The interview includes background information such as socio-demographic, dietary, and health-related questions. The examination component consists of medical and physiological measurements, as well as laboratory tests. The National Center for Health Statistics Ethics Review Board approved all NHANES protocols, and all survey participants completed a consent form. The detailed protocol on NHANES methodology and data collection are available at https://www.cdc.gov/nchs/nhanes/index.htm. For this study, we included adults aged ≥ 20 years enrolled in the NHANES-III between 1988 and 1994, with data on blood lead and urinary cadmium concentrations (n = 16,040). Then we excluded participants with missing data on mortality and covariates (n = 1,729). The final study population included 14,311.
2.2. Measurements of blood lead and urinary cadmium
Blood and urine samples were collected during the medical examination. The laboratory methods for the processing of these samples are described in detail elsewhere (21). Briefly, the blood and urine specimens were frozen (− 30°C and − 20°C; respectively), stored, and shipped for analysis to the Division of Laboratory Sciences, National Center for Environmental Health at the Centers for Disease Control and Prevention in Atlanta, Georgia (USA). Lead (µg/dL) concentration was measured in whole blood using inductively coupled plasma mass spectrometry. Urinary cadmium was measured by graphite furnace atomic absorption with Zeeman background correction using the CDC and Prevention modification (21) of the method proposed by Pruszkowska et al. (22). Specimens were analyzed in duplicate, and the average of the two measurements was reported. The detection limits (LOD) were 1.0 µg/dL (0.048 µmol/L) and 0.03 µg/L for BLLs and UCd; respectively. For study participants (n = 1217; 8.5%) who had BLLs below the LOD, a value of \(\frac{\text{L}\text{O}\text{D}}{\sqrt{2}}\)(0.7 µg/dL) was imputed. Urinary creatinine measured using the Jaffe reaction with a Beckman Synchron AS/ASTRA Clinical Analyzer (Beckman Instruments, Inc., Brea, CA), was used to account for between-participant differences in urine dilution.
2.3. Mortality data
A full description of mortality linkage methods is available from the National Center for Health Statistics (NCHS) (23). Briefly, the de-identified and anonymized data of the NHANES III participants were linked to National Death Index mortality files based on 12 identifiers for each participant (eg, Social Security number, sex, and date of birth) with a probabilistic matching algorithm to determine mortality status. The NCHS public-use linked mortality file provides mortality follow-up data from the date of NHANES III survey participation until December 31, 2015 (1988–2015). Participants with no matched death record at this date were assumed to be alive during the entire follow-up period. In a validation study using mortality-linked data from the first NHANES study (NHANES-I; 1971–75), 96% of deceased participants and 99% of those still alive were classified correctly (24). The underlying cause of death was recorded in the public-use linked mortality files using the following ICD-10 codes: cardiovascular diseases including heart diseases (I00-I09, I11, I13, I20-I51) and cerebrovascular diseases (I60-I69) and malignant neoplasms (C00-C97).
2.4. Dietary inflammatory index and daily fruit and vegetable servings
Dietary intake was collected using 24-hour dietary recalls. The 24-hour recall interview was conducted at a Mobile Examination Center using a standard set of measuring guides to assist in estimating portion sizes. The main fruit and vegetable variable used in these analyses was the seven-digit food coding scheme developed by the United States Department of Agriculture for NHANES III, which categorized items by food group and subgroup. If the main ingredient of a mixed dish was a fruit or vegetable, the item was categorized according to the main fruit or vegetable. Sweets containing fruits where the fruit was not the main ingredient were excluded from the analysis. Serving sizes for each recorded fruit and vegetable were determined using serving size estimations from USDA/U.S. Department of Health and Human Services dietary guidelines. The adapted Dietary Inflammatory Index (DII) was derived from the mean daily intakes of foods/beverages, energy, and nutrients of the 24-hour dietary recalls. We constructed the adapted DII to measure the inflammatory potential of the diet (25). Briefly, we used the DII index proposed by Woudenberg et al (26) in combination with the updated dietary components inflammatory weights designed by Shivappa et al (27) instead of the weights proposed by Cavicchia et al (28). These updated inflammatory weights, based on three additional years of published data (2008–2010, inclusive), resulted in a doubling of the total number of articles scored. The DII was based on a nutritional rationale: first, the inflammatory weights of dietary components are multiplied by the standardized energy-adjusted intake, which acts to reduce between-person variation; second, the intakes of all components are standardized; third, the DII calculation did not include alcoholic beverages such as beer, wine, and liquor, total fat, and energy, to avoid overestimation of the inflammatory effects of ethanol, fat, and energy. A total of 24 of the 35 possible dietary components were used for the DII calculation (carbohydrates, proteins, alcohol, fibers, cholesterol, saturated fatty acids, monounsaturated fatty acids, omega 3, omega 6, niacin, thiamin, riboflavin, vitamin B6, vitamin B12, iron, magnesium, zinc, vitamin A, vitamin C, vitamin D, vitamin E, folic acid, beta carotene, and caffeine). A positive DII score indicates an anti-inflammatory diet and negative values correspond to a pro-inflammatory diet.
2.5. Covariates
Baseline covariates were collected when individuals participated in a household interview and demographic information—including sex (male/female), age (continuous; years), ethnicity (Mexican-American, other Hispanic, not Hispanic), income to poverty ratio (categorized in tertiles), the number of years of education attended and completed (continuous; years), area of residence (metro and nonmetro counties), and smoking status (current, former and never) was obtained. Information on body-mass index ([BMI] continuous; kg/m2), and physical activity (None, 1 to 14 times, 15 or more times; per month) was obtained during the medical examination. Final models were adjusted for age, sex, ethnicity, income to poverty ratio, educational level, area of residence, smoking status, BMI, and physical activity. Metals concentrations were mutually adjusted. The selection of potential confounders was done using a priori knowledge ensuring adjustment only for confounders and avoiding over-adjustment for potential mediators or colliders.
2.6. Hypothetical interventions
For each of the following interventions, we used the parametric g-formula (see details below) to estimate the 27-year mortality risk where all participants follow the prescribed intervention beginning at the start of follow-up.
Lower BLLs and UCd to the 25th percentile for each metal separately.
Lower BLLs and UCd to the 25th percentile for both metals.
Increase the dietary inflammatory index to a level corresponding to a more anti-inflammatory (e.g. 75th percentile), or increase the fruit and vegetable daily servings to the number of 5 servings/day as recommended by many nutritional guidelines.
Best case scenario: Combining interventions 2 and 3; lower BLLs and UCd and increase the dietary inflammatory index or fruit and vegetable daily servings.
Worst case scenario: Increase BLLs and UCd to a level corresponding to the 75th percentile for each metal and lower the DII score to the 25th percentile value of the observed distribution (corresponding to an adherence to a pro-inflammatory diet) or fruit and vegetable daily servings (e.g. number of daily servings < 2).
2.7. Statistical analysis
First, we describe the overall sample BLLs and UCd using geometric means and geometric standard errors (SE) by population characteristics. BLLs and UCd were log2-transformed to limit the influence of outliers. We used the parametric g-computation to estimate risk ratios (RR) and risk differences (RD) of all-cause and specific causes of mortality under the above-described hypothetical interventions. In recent years, there have been substantial advances in the application of these methods, and have been used to evaluate hypothetical interventions on sources of lead exposure on BLLs (29). The parametric g-formula is a generalization of the standardization method, proposed by Robins, (30, 31) and allows for flexibly simulating and estimating the effect of any form of hypothetical intervention. A more detailed discussion of this method is presented elsewhere (32, 33). Briefly, we first fitted a pooled logistic regression model conditional on covariates and follow-up time, after arranging the data into a person-time structure. We included interaction terms between metal concentrations and dietary exposures in the outcome model. The discrete-time hazards of all-cause and specific causes of mortality for each two-fold increase in the baseline dietary exposure and metals concentrations (log2-transformed to reduce skewness) were then estimated. This estimated risk was used to predict the mortality risk under the hypothetical interventions described above. We then compared the estimated risk of 27-year mortality had all participants followed the prescribed interventions with the estimated risk of mortality under 1) no intervention (natural course) and 2) the worst case scenario (Scenario 5).
All analyses were weighted by the provided sample weights to account for the unequal probabilities of inclusion and response rates. We estimated 95% confidence intervals (CIs) of the RR and the RD using non-parametric bootstrap (M = 200) and used the 2.5th and 97.5th percentiles as the lower and upper confidence interval limits, respectively.
Statistical analyses were performed using R version 4.0.4 and the Statistical Analysis System software, version 9.4 (SAS Institute, Cary, NC, USA).