2.6. Statistical Analysis

All analyses in this study included sample weights, clustering, and stratification to account for the complex sampling design of NHANES data, in accordance with the NHANES analysis guidelines. Descriptive analysis was conducted for demographic characteristics between the two groups classified by median total dietary zinc. Selected demographic characteristics and biochemical measures were expressed as mean ± standard deviation of continuous variables and percentage (%) for categorical variables, respectively. Chi-square tests were used to analyze categorical variables. t-tests were used to analyze normal continuous variables.

A multivariate Cox proportional risk regression model was used to calculate risk ratios (HRs) and 95% CIs for dietary zinc and vitamin B6 levels in relation to CVD risk and all-cause mortality. The proportional risk assumption was tested using the KM survival curve, and the results showed that the proportional risk assumption was not violated. Groups were divided according to dietary zinc and median dietary vitamin B6 levels. The first group was set as a reference group that was used to obtain the low and high groups. Two multivariate models were developed. Model 1 was adjusted for age (continuous, years) and gender (male or female). Model 2 was further adjusted for race and ethnicity (Mexican American/non-Hispanic White/non-Hispanic Black/other), education level (<high school, high school or equivalent, or college), household income to poverty ratio (<1.0, 1.0-3.0, or >3.0), BMI (<30.0 or ≥30.0 kg/m2), smoking status (never, ever, or current), drinking status (yes/no), HbA1c (<7% or ≥7%), self-reported history of hypertension (yes or no), and self-reported high cholesterol history of hypertension (yes or no).

Participants were divided into four groups based on median dietary zinc and vitamin B6 intake: low zinc, low vitamin B6 score (LZLV) group; low zinc, high vitamin B6 score (LZHV) group; high zinc, low vitamin B6 score (HZLV) group; and high zinc, high vitamin B6 score (HZHV). Using the HZHV group as a reference, a multi-factor Cox proportional risk regression model was used to assess the risk of the other three groups separately. The p-value of the product term between dietary zinc and vitamin B6 levels was used to estimate the significance of the interaction. The dietary zinc and vitamin B6 ratios were log-transformed prior to analysis. Between the 0.5th and 99.5th percentile of dietary zinc and vitamin B6 ratios, restricted cubic spline analysis in three sections (10th, 50th, 90th percentile) was used to examine the non-linear relationship between dietary zinc levels and vitamin B6 levels and CVD mortality and all-cause mortality. Schoenfeld residuals were used to test for proportional risk hypothesis, and no violations were observed.

The quartiles of the log-transformed dietary zinc and vitamin B6 ratios were divided into four groups. The second and third quartiles were combined into one reference group to obtain the low (quartile 1), medium (quartiles 2 and 3), and high (quartile 4) groups. Cox proportional risk models were used to estimate the risk ratio (HR) and 95% confidence interval (CI) for all-cause mortality and CVD mortality for each standard deviation increase in the ratio of dietary zinc to vitamin B6. Dietary zinc (high/low) and vitamin B6 (high/low) were also added to the adjustment factors in Model 2 to eliminate the effect of dose on the outcome.

A series of sensitivity analyses were carried out in this study. First, participants who died within 2 years of follow-up were excluded. Second, permissible total energy intake limits were set to percentiles, 1 to 99. Finally, additional adjustment was made for a history of diabetes and for dietary fiber intake.

A two-sided p < 0.05 was set as the threshold for statistical significance. All analyses were performed using R version 4.1.3 (R Foundation, https://www.r-project.org/ (accessed on 11 December 2022)).