*2.2. Variables*

In this study, the independent variables were daily total dietary nutrient intakes, containing protein, carbohydrate, total fat, dietary fiber, vitamin B1, vitamin B2, vitamin

B6, total folate, vitamin B12, vitamin E, calcium, magnesium, iron, zinc, and copper. It is worth mentioning that the dietary energy value different from the category of dietary nutrients was also included in the follow-up analysis as an independent important parameter. The dependent variable was RGCS (HbA1c < 6.5% represents good RGCS, and HbA1c ≥ 6.5% represents poor RGCS). All variables involved in this study were divided into continuous variables and categorical variables. Continuous variables included energy, protein, carbohydrate, dietary fiber, total fat, total folate, vitamin B1, vitamin B2, vitamin B6, vitamin B12, vitamin E, calcium, magnesium, iron, zinc, copper, poverty income ratio (PIR), insulin, glucose, and hemoglobin. Categorical variables included gender, age, race, education level, body mass index (BMI), moderate or severe physical activity, hypertension, the doctor informing them they had diabetes, having at least 12 cups of alcoholic drink per year, consuming over 100 cigarettes in their lifetime, and adult food security. Details of all variable acquisition procedures can be found at http://www.cdc.gov/nchs/nhanes/.

### *2.3. Statistical Analysis*

The results of normality test showed that it could not be considered that all continuous variables obeyed normal distribution. Therefore, in the stages of statistical description and single variable analysis, all continuous variables and categorical variables were expressed as median (25% percentile–75% percentile) and percentage (proportion), respectively. We used a nonparametric test (Mann–Whitney *U* test) for all continuous variables that did not obey normal distribution, as well as Pearson's chi-squared test for all categorical variables. Then, in the multivariate analysis stage, we controlled different confounders and established four binary logistic regression models with the RGCS as the dependent variable to adjust the potential bias. Eventually, we performed model fitting with the HbA1c index as the dependent variable, and receiver operator characteristic (ROC) analysis was performed to calculate the area under the curve (AUC). The result of the collinearity diagnosis showed that there was no collinearity (variance inflation factor, VIF < 10) among the independent variables studied. Statistical significance was considered when *p*-value was below 0.05 (two-tailed). Data processing, statistical analysis, and graphic drawing were carried out with Stata version 13.1, IBM SPSS version 26.0, GraphPad Prism version 7.00, R version 4.0.2 (http://www.R-project.org, The R Foundation), and EmpowerStats software version 2.1 (http://www.empowerstats.com, X&Y Solutions, Inc., Boston, MA, USA).

#### **3. Result**

## *3.1. Baseline Characteristics*

The description of demographic and medical characteristics is shown in Table 1. Among the participants, 49.5% (*n* = 20,458) were male, 44.6% (*n* = 18,404) were non-Hispanic White, 20.5% (*n* = 8458) were non-Hispanic Black, and 17.3% (*n* = 7153) were Mexican American. In addition, the statistical description of daily dietary nutrient intakes in our study showed that their distribution fluctuated over time (Figure 1). Therefore, the time effect was often a potential confusion factor, which should be placed in subsequent analysis.

**Table 1.** Characteristics by RGCS of non-pregnant adults 20+ years old from NHANES 1999–2018 (except for 2003–2004).


