Methods
This research used data from the China Health and Nutrition Survey (CHNS), a public database. The data and study materials can be found on its official website (http://www.cpc.unc.edu/projects/china). Each CHNS participant provided written informed consent.10
Population and study design
Details of the study design and key results of CHNS have been previously published.11–13 Briefly, CHNS is an ongoing multipurpose longitudinal open cohort study initiated in 1989, with follow-up conducted every 2–4 years. A multistage, random cluster approach was used to draw the sample from nine provinces or autonomous cities. The provinces include Heilongjiang (enrolled in 1997), Liaoning, Shandong, Henan, Jiangsu, Hubei, Hunan, Guizhou and Guangxi, and the three largest autonomous cities were Beijing, Shanghai and Chongqing, with the latter enrolled in 2011. By 2011, the provinces included in CHNS represented 47% of China’s population.11 The CHNS follow-up rounds were completed in 1989, 1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011 and 2015.
In 1997, 2000, 2004 and 2006, cognitive function was assessed repeatedly for participants aged ≥55 years who were capable of completing cognitive function tests, representing the cognitively healthy, older Chinese population. We conducted a prospective cohort based on the four rounds of data. Participants with only one survey wave were excluded. Of the remaining participants with at least two rounds of survey data (n=3119), the first survey round was considered as the baseline. In addition, 13 participants with extreme dietary energy data (male: >4200 or <600 kcal/day; female: >3600 or <500 kcal/day) were also excluded.14 The final analysis included a total of 3106 participants (figure 1).
Figure 1Flowchart of the study.
Dietary nutrient intakes
Dietary measurements in CHNS have been previously published.11 14 Briefly, both individual and household-level dietary data were collected in each survey round. Individual diet data were collected by trained investigators through a 24-hour dietary recall on three consecutive days randomly allocated from Monday to Sunday. The allocated days were almost evenly distributed across the week for each sampling unit. Given that cooking oil and condiments are integral to Chinese cuisine and are added during cooking and preparation, their consumption was assessed by examining changes in household inventory over the same 3 days with a weighting technique. Specifically, on each of the three consecutive days, interviewers conducted a household visit to weigh and record the household food inventories, including food purchased and discarded, as well as individuals’ proportion of food consumption. The cooking oil and condiments consumed at the household level were allocated to each individual according to their consumption proportion. Nutrient intake was calculated using the Chinese food composition tables. The accuracy of the 24-hour dietary recall designed to assess energy and nutrient intake has been validated.15
In our study, 3-day average intakes of dietary macronutrients and micronutrients in each round were calculated. Furthermore, to represent the long-term nutrient status of each participant and minimise within-person variation, the cumulative average intake of each nutrient was calculated for each participant using all results up to the last visit and was used in the final analysis.
Covariate measurements
After the participants had rested for 5 min, seated blood pressure was measured by trained research staff using a mercury manometer following standard procedures. Triplicate measurements on the same arm were taken in a quiet and bright room. The mean systolic blood pressure (SBP) and diastolic blood pressure (DBP) of the three independent measures were used in the analysis.
Demographic and lifestyle information was obtained through questionnaires at each follow-up survey, including age, sex, smoking, alcohol consumption, occupation, education level, urban or rural residency, region, concomitant conditions and drug use. Height and weight were measured following standard procedures with calibrated equipment. Body mass index (BMI) was calculated as weight (kg) divided by height squared (m2). Physical activity was collected by staff-administered questionnaires exploring all occupational, transportation, domestic and leisure activities in adults.
Assessment of cognitive function decline
A subset of items from the Telephone Interview for Cognitive Status–modified (TICS-m)16 was used to measure cognitive performance. The TICS-m can be administered both over the telephone and through face-to-face interviews17 and has been approved as a strong predictor of cognitive decline,18 dementia and mild cognitive impairment,19 even among people with low levels of education or illiteracy.20 The cognitive test consists of three simple tasks, including immediate and delayed recall of a 10-word list (0–10 points for each), counting backward from 20 (0–2 points) and serial seven subtraction five times from 100 (0–5 points), to evaluate verbal memory, attention and calculation, respectively. A higher score in each item indicates better cognitive function, and the global cognitive score ranges from 0 to 27 points. Moreover, to eliminate the influence of the score proportion of the three components on the global score, a composite score in standard units by averaging z scores of verbal memory and other items was also conducted.
During face-to-face interviews, all items used to test cognitive performance were read and/or interpreted in detail by trained and qualified staff. Therefore, as reported in a previous study,17 participants’ vision and literacy skills may not affect their cognitive assessment. However, participants unable to complete the cognitive assessment due to various reasons, such as severe cognitive impairment, were excluded from the study.
The 5-year decline rate in global or composite cognitive scores, which was used to assess the decline of cognitive function, was calculated as the baseline score minus the last survey score, then divided by the follow-up time (in years) and multiplied by 5. A positive value represents a decline in cognitive function, while a negative value represents an increase in cognitive function.
Statistical analysis
For continuous variables with normal and non-normal distributions, the population characteristics were presented as mean (standard deviation (SD)) and median (25th percentile, 75th percentile), respectively. Differences by categories of dietary thiamine intake (<0.40, [0.40, 0.60), [0.60, 0.80), [0.80, 1.00), [1.00, 1.20), [1.20, 1.40), ≥1.40 mg/day) were compared using analysis of variance, or the Kruskal-Wallis test, accordingly. For categorical variables, the population characteristics were presented as proportions; differences by categories of dietary thiamine intake were compared using χ2 tests.
A restricted cubic spline function was applied to display the relationship between dietary thiamine intake and the 5-year decline rate in global or composite cognitive scores. A two-piecewise linear regression was performed to examine the threshold effect of dietary thiamine intake using R package of segmented. The inflection point was determined using the likelihood ratio test and the bootstrap resampling method. The relationship of dietary thiamine intake with the 5-year decline rate in global or composite scores was estimated by linear regression models (β and 95% confidence interval (CI). The models included adjustments for sociodemographic characteristics (age, sex, occupation, region and urban or rural residency), known dementia risk factors (smoking, alcohol consumption, BMI, self-reported diabetes, SBP, DBP, antihypertensive medication, physical activity and education level),21 the dietary intake factors (fibre, sodium, potassium, carbohydrate, protein and fat) associated with cognitive function22–24 and the baseline global cognitive score. Possible modifications of the association between thiamine intake and the 5-year decline rate in global or composite cognitive scores were evaluated by stratified analyses and interaction testing.
Several sensitivity analyses were performed to assess the robustness of the key findings. First, to exclude the confounding effects of total energy intake, the association between energy-adjusted thiamine intake, calculated using the nutrient residual model, and cognitive decline was further examined. Second, to exclude the influence of food consumption, the intake of whole grains, legumes, unprocessed red meat and processed red meat was further included in the adjustments. Third, considering the potential confounding effect of other dietary B vitamins due to similar food sources, we further adjusted the dietary intakes of riboflavin and niacin in the regression models. Lastly, considering that some participants had more than two cognitive measurements, a mixed linear regression model was applied to capture the information from the multiple measurements to assess the relationship between dietary thiamine and cognitive decline. To eliminate false positives caused by multiple tests, the Benjamini-Hochberg method was used to further calculate the multiple testing corrected p values (P-BH) in regression and subgroup analysis.
All analyses were performed using R software (V.3.6.3), and a two-sided p<0.05 was considered statistically significant.