The data currently used were obtained from NHANES (https://www.cdc.gov/nchs/nhanes/index.htm). NHANES is a programme that administers continuous 2-year-cycle cross-sectional surveys, which is conducted by the Centers for Disease Control and Prevention (CDC). This programme is designed to assess the health and nutritional status of adults and children in the United States. NHANES 2011–2012 cycle was used in the present study which included 1791 elderly adults aged 60 years or older who were eligible to complete the cognitive function testings. Among them, we further excluded participants who did not complete two 24-h dietary recalls (n=290), whose dietary data were considered unreliable defined as total energy intake less than 500 or greater than 5000 kcal/day for women and less than 500 or greater than 8000 kcal/day for men (n=9), and who had missing data on cognitive testing (n=214). As a result of exclusions, a total of 1278 elderly adults were included in the present study (Figure 1).
Estimation of Dietary quality
The dietary intake data were obtained from NHANES two 24-h recall interviews which were conducted by the trained interviewers based on the automated multiple-pass method. The first interview was arranged face-to-face in the Mobile Examination Center (MEC) and the second was carried out via phone 3-10 days later. Dietary intake was estimated using the mean value of the two 24 recall data. The energy and nutrients for each food or beverage intake were calculated using Food and Nutrient Database for Dietary Studies (FNDDS) and the food groups were determined by Food Patterns Equivalence Database (FPED) from the US Department of Agriculture (USDA).
The dietary quality was estimated using HEI-2015 which was recommended by USDA to assess the adherence to the dietary guidelines of 2015-2020 DGA. The HEI-2015 contains 13 components (food groups or nutrients), including 9 adequacy components (total vegetables, greens and beans, total fruits, whole fruits, whole grains, dairy, total protein foods, seafood and plant proteins and fatty acids) and 4 moderation components (sodium, refined grains, saturated fats and added sugars). These 13 components were expressed as amounts per 1000 kcal except for fatty acids (expressed as a ratio of unsaturated to saturated fats), saturated fats (expressed as % energy) and added sugars (expressed as % energy). For the adequacy components, the intake were encouraged and the higher score means higher intake and better dietary quality; For the moderation components, the intake were limited and the higher score means lower intake and better dietary quality. These components were scored separately and incorporated to a total score with the maximum possible score of 100.
Assessment of cognitive function
Two cognitive function testings, Digit Symbol Substitution Test (DSST) and Animal Fluency Test (AFT), were performed among elderly adults aged 60 years or older in NHANES 2011-2012 and used in the present analysis. In briefly, the DSST is used to assess the abilities of processing speed, sustained attention and working memory, which presents 9 numbers paired with symbols. Participants were asked to match symbols with corresponding numbers in 120 s, and the final scores represent the total number of correct matches. The AFT is used to evaluate the categorical verbal fluency regardless of cultural context. Participants were asked to name as many animals as possible in 60 s. A point is given for each named animal, and the final score represents the total number of correct named animals. Finally, the global cognition score was calculated by summing the z scores of above two individual tests to evaluate global cognitive performance.
The potential confounding factors were collected from the interview, examination and laboratory samples in the household or MEC. As the demographic variables, age, gender (male and female), educational level (less than high school, high school and more than high school) and ethnicity (Hispanic, non-Hispanic White, non-Hispanic Black and other races) were obtained. The family monthly poverty level index was calculated according to the family monthly income and Department of Health and Human Services(HHS)’ poverty guidelines and was divided into three categories (≤1.30, 1.31-1.85, >1.85).
For the lifestyle variables, participants who have smoked at least 100 cigarettes in life were defined as smoker. Participants who drunk at least 12 times in any one year were defined as drinker. In addition, the time spent sitting (sedentary time) except sleeping time in a day was used as a potential physical activity indicator.
Depressive symptoms were assessed using 9-item Patient Health Questionnaire (PHQ-9). The score of each item ranges from 0 (not at all) to 3 (nearly every day), incorporating to a total score with the maximum possible score of 27. Individuals with PHQ-9 total score of 10 or greater were categorized as depressive symptoms.
The blood pressure (BP) were measured in MEC after resting quietly in a seated position for 5 minute. Three consecutive BP determinations were obtained and the average of SBP and DBP were used in this study. Hypertension was defined as SBP≥140mmHg or DBP≥90mmHg or current use of prescribed medicine for hypertension. The body measures data including weight (kg) and standing height (cm) were also collected in MEC by trained health technicians. Body Mass Index (BMI) was calculated as weight in kilograms divided by height in meters squared. Total cholesterol (TC) from serum specimens was detected using enzymatic assay. Hypercholesterolaemia was defined as TC ≥ 240mg/dl or current use of prescribed medicine for hypercholesterolaemia. The quantitative measurement of % hemoglobin A1c (HbA1c) in whole blood specimens was determined using the Tosoh Automated Glycohemoglobin Analyzer HLC-723G8. Diabetes was defined as HbA1c≥6.5% or the current use of insulin or current use of diabetic pills in the present analysis.
Data were presented as mean±standard deviation (SD) for continuous variables and as frequencies and percentages for categorical variables. First, baseline characteristics were summarized according to the tertiles of HEI-2015 scores. The differences between groups were compared using F-tests for continuous variables and chi-square test for categorical variables. Secondly, the multiple linear regression model was used to evaluate the associations of HEI-2015 with cognitive function scores. The first tertile of HEI-2015 was used as reference. The Model 1 was adjusted for age and gender; the Model 2 was additionally adjusted for energy intake, ethnicity, BMI, drinker, smoker, sedentary time, education, family monthly poverty level and depressive symptom; and the Model 3 was further adjusted for hypertension, hypercholesterolaemia and diabetes. Thirdly, the associations of HEI-2015 with cognitive impairment were assessed using binary logistic regression model. For now there has not been gold standard as the cut off score for DSST and AFT to indicate the poor cognitive performance or cognitive impairment. Consistent with methods previously used[27-29], the lowest quartile for different cognitive scores was considered as the potential cognitive impairment in the binary regression analysis. The cut off scores were 33, 12 and -1.26 for DSST, AFT and global cognition, respectively. All the covariates used in the multiple linear regression model were included in the binary logistic regression model. To test for trends, the median of HEI-2015 in each tertile was calculated and used as a continuous variable. Finally, the multiple linear regression model was further conducted to evaluate associations of the individual HEI-2015 components with global cognitive function, which was adjusted for all the covariates in Model3 plus other components. Data management was performed using SAS version 9.4 (SAS Institute, Cary, North Carolina, USA) and all statistical analyses were performed using SPSS 24.0 (IBM, Armonk, NY, USA). The multicollinearity was evaluated by the tolerance greater than 0.1 and the variance inflation factor (VIF) less than 10. The 2-tailed P < 0.05 was considered statistically significant.