Association Between Weight Status and Rate of Cognitive Decline: China Health and Nutrition Survey 1997–2018

Abstract

Background

There is a close relationship between weight status and cognitive impairment in older adults. This study examined the association between weight status and the trajectory of cognitive decline over time in a population-based cohort of older adults in China.

Methods

We used data from adults aged ≥55 years participating in the China health and nutrition survey (1997–2018). Underweight (body mass index [BMI] ≤ 18.5 kg/m²), normal weight (18.5–23 kg/m²), overweight (23–27.5 kg/m²), and obesity (BMI ≥ 27.5 kg/m²) were defined using the World Health Organization Asian cutpoints. Global cognition was estimated every 2–4 years through a face-to-face interview using a modified telephone interview for cognitive status (scores 0–27). The association between BMI and the rate of global cognitive decline, using a restricted cubic spline for age and age category, was examined with linear mixed-effects models accounting for correlation within communities and individuals.

Results

We included 5 992 adults (53% female participants, mean age of 62 at baseline). We found differences in the adjusted rate of global cognitive decline by weight status (p = .01 in the cubic spline model). Models were adjusted for sex, marital status, current employment status, income, region, urbanization, education status, birth cohort, leisure activity, smoking status, and self-reported diagnosis of hypertension, diabetes, or Myocardial Infarction (MI)/stroke. In addition, significant declines by age in global cognitive function were found for all weight status categories except individuals with obesity.

Conclusions

In a cohort of adults in China, cognitive decline trajectory differed by weight status. A slower rate of change was observed in participants classified as having obesity.

Background

Patterns of cognitive decline in older adults present as a spectrum of cognitive changes that range from normal aging to mild cognitive impairment to dementia (1). Cognitive impairment is a disease of aging associated with many important adverse outcomes, including decreased quality of life (2,3), hospitalization (4), long-term care placement (5,6), and mortality (7,8). The rate of cognitive decline increases with each decade of life (9) and as a result of improved life expectancy and declining birth rates, the prevalence of cognitive impairment has risen steadily (10,11). China’s population is aging rapidly (12) with a recent study estimating the prevalence of mild cognitive impairment and dementia to be 15.5% (38.8 million) and 6% (15 million), respectively (13). The rising prevalence of an aging demographic combined with the associated decline in function has made cognitive impairment a significant public health concern (14,15).

In Western populations, the prevalence of obesity in older adults has risen gradually in recent decades (16). The evidence suggests that the relationship between body mass index (BMI) and adverse outcomes, including mortality, decreases over time in similar populations (12,17). However, the burden of multimorbidity, cognitive impairment, and functional impairment are high among older adults with obesity (18,19). Over the last 3–4 decades, obesity has become considerably more common in China (20). In addition, waist circumference has increased faster than BMI over time in China (21) which is concerning as fat deposition and obesity-related complications develop at lower levels of BMI in this population (22). There has been a clear transition from under- to overnutrition in China (23–26), with a rapid increase in the prevalence of obesity and overweight across all age groups (27), accompanied by obesity-related comorbidities such as hypertension (28) and diabetes (29).

With the prevalence of cognitive impairment rising worldwide, the literature on the relationship between BMI and cognitive decline with aging has been mixed (30–33). A recent systematic review and meta-analysis of prospective studies found that mid-life underweight or obesity, and late-life underweight status, increased the risk of dementia (34). In the same review, late-life overweight and obesity status conferred a 21% and 25% reduction in dementia risk, respectively. However, summarizing the data this way may fail to capture the complex nature of the interaction between weight status and cognitive function over time (30). Specifically, baseline BMI, its trajectory throughout the life course, the differences between diverse population groups, competing causes of death, and measurements of adiposity are essential considerations. This is supported by 2 studies (done in Asian and European populations) that followed patients from mid-life to late life, which reported no association between mid-life measures and later-life dementia (35,36). There is a critical need to examine the relationship between weight status and cognition over an extended period from mid-life through to older age while attempting to account for competing factors at an individual and population level. We used data from a longitudinal, population-based cohort to examine this association in adults residing in China to fill this gap.

Method

Study Population

We used data from the China Health and Nutrition Survey (CHNS), a prospective household-based study conducted over 27 years from 1991 to 2018. It is the only large-scale, longitudinal study of its kind in China. Prior to 2011, the sample only included data from 9 provinces (Heilongjiang, Liaoning, Jiangsu, Shandong, Henan, Hubei, Hunan, Guangxi, Guizhou, Shaanxi, Yunnan, and Zhejiang). Three mega-cities (Beijing, Chonqing, and Shanghai) were added in 2011, and an additional 3 provinces (Shaanxi, Yunnan, and Zhejiang) were added in 2015. A stratified multistage, clustered sampling design was used to select a sample within each province or mega-city that would represent urban, rural, and suburban areas that vary substantially in geography, economic development, public resources, and health indicators. Each participant provided written informed consent and response rates based on the participation in the previous survey range were ranged from 68.7% to 88.1% across years (37). The study was approved by institutional review boards from the University of North Carolina at Chapel Hill and the National Institute for Nutrition and Health, Chinese Center for Disease Control and Prevention. Further information on survey procedures and the rationale of the CHNS can be found elsewhere (38).

Our sample includes 6 rounds of the CHNS survey (1997, 2000, 2004, 2006, 2015, and 2018) when questions were asked to assess cognitive function. Due to substantial differences between provinces and mega-cities, and the fact that they were only present for later waves, we excluded individuals from the 3 mega-cities. Questions on cognitive function were only asked for participants aged 55 years or older at each study visit—our eligible sample included 9 514 adults. As our analysis is focused on changes in cognitive function, we excluded 3 491 participants with only 1 cognitive measure collected. Individuals missing 1 or more covariate values (n = 31) were also excluded from all analyses resulting in a final analysis sample of 5 992 participants.

Primary Outcomes—Cognitive Function

The cognitive screening was conducted in person via face-to-face interview using an adapted version of the telephone interview for cognitive status (39) and included immediate and delayed recall of a list of 10 words, counting backward from 20, serial 7 subtraction, and orientation. A cognitive global score and a verbal memory score were derived from these measures. For the global score, we assigned a maximum of 10 points each to the immediate and delayed 10-word recall (1 point per word). Two points were awarded for counting backward from 20 correctly on the first attempt, and if unsuccessful on the first attempt, 1 point was conveyed for successful completion on the second attempt. Five points were allocated (1 for each correct answer) based on the successful subtraction of serial sevens. Scores ranged from 0 to 27, with higher scores reflecting better cognitive function. For the short-term verbal memory score, we assigned 2 points for each of 2 repetitions of a recall of a series of 10 words that were read to each participant immediately and after a delay of 5 minutes, for a total of 20 points.

Primary Exposure—BMI

Using a standardized protocol, height was measured without shoes to the nearest 0.1 cm using a portable stadiometer, and weight was measured without shoes and in light clothing to the nearest 0.1 kg on a calibrated floor scale to generate BMI (kg/m²). BMI was categorized into underweight (<18.5 kg/m²), normal weight (18.5–23 kg/m²), overweight (23–27.5 kg/m²), or obesity (≥27.5 kg/m²) based on WHO cutpoints for Asian populations (22).

Covariates

Age, sex, marital status, smoking status, highest education level achieved, and a diagnosis of hypertension, diabetes, and myocardial infarction or stroke were self-reported each year using structured survey questionnaires. Age was categorized into 5-year age groups, including 55–60, 60–65, 65–70, 70–75, and ≥75–years. Marital status was categorized as married or not. Whether an individual participated in any active leisure activity was based on a detailed physical activity time-use questionnaire during an average week (40). Due to the low proportion of the sample participating in any active leisure physical activities (8% across all study visits), this variable was categorized as any or no leisure activity. Smoking status was classified as never smoked, former smoker, or current smoker. Employment status was derived from a series of questions about formal and informal occupational employment, which we classified as not currently working or currently working, in either formal or informal employment. Household income was derived from individual and household questionnaires from time-use, asset, and economic activity and inflated to 2018 yuan currency for comparability. To capture income each year, income was categorized into year-specific tertiles. The region was categorized as Northern, Central, and Southern China. Overall urbanization was calculated at the community level each year using a multicomponent and validated continuous index, derived from 12 components to capture the complexity of urbanization, including population density, economic activity, traditional markets, modern markets, transportation infrastructure, sanitation, communications, housing, education, diversity, health infrastructure, and social services (41). Given the dramatic increase in urbanization in China over the study period, urbanization was defined using year-specific tertiles to measure relative urbanization each year. In addition, to account for potential cohort effects, as China has experienced rapid urbanization during this period, we categorized individuals based on birth cohort: born before 1930, between 1930 and 1940, between 1940 and 1950, between 1950 and 1960, or after 1960 (26,38,41). We included an indicator variable for self-reported cardiometabolic diseases: based on whether a doctor had ever told participants if they had hypertension, diabetes, and/or a stroke/MI.

Statistical Analysis

Descriptive statistics were estimated overall and by weight status at the time of the first cognitive measure. In addition, we calculated the average annual change in global cognitive function score and verbal memory score, dividing the difference in each score between 2 subsequent survey years by the years between measures, and calculated the mean (and standard deviation) average annual change by age group (using index age defined as age at first observation for each change period), and weight status. Two separate univariate linear mixed-effects models were used to test whether the unadjusted average annual change differed by age group and weight status.

Linear mixed-effects models were used to test whether the association between age and decline in global cognitive function score differed by weight status. This allowed us to test whether the rate of decline across age differed by weight status. We used a restricted cubic spline model with a cubic spline term for age to allow for the nonlinear association between age and cognitive function. Random intercepts for individual and communities were included to control for repeated measurements on individuals over time and the clustering of individuals within communities. The restricted cubic spline for age had 5 knots at the 5th, 27.5th, 50th, 72.5th, and 95th percentiles, respectively. In a minimally adjusted model, we adjusted for age, sex, marital status, employment status, region, income, education, urbanization level, and birth cohort. A fully adjusted model was also fit, which included all the covariates from the first model along with lifestyle behaviors (leisure activity and smoking status) and medical history (self-reported diagnosis of hypertension, diabetes, or myocardial infarction/stroke). To illustrate the association between the decline in global cognitive function score and weight status, we estimated the adjusted global cognitive function score at each age for each weight status. The same models were fit using verbal memory score as the outcome.

To aid in interpretation, we also fit a model using age group to account for the nonlinear association between age and cognitive decline. As in the cubic spline model, we included random intercepts at individual and community levels and an interaction term for age group by weight status to test whether the rate of cognitive decline differed by weight status, with identical variables in minimally and fully adjusted models. We estimated cognitive function for each age group from the minimally and fully adjusted models. We tested for declines in cognitive function across age groups within each weight status category using Wald tests, with p < .05 considered statistically significant. All analysis was done using SAS version 9.4. Given that individuals with obesity might be more likely to use obesity-related Cardiovascular disease (CVD) medications, we conducted 2 sensitivity tests with alternative control variables: replacing self-reported diagnosis of hypertension with (a) hypertension based on systolic blood pressure ≥130 and/or diastolic blood pressure ≥80 and (b) self-reported use of hypertension medication. Given that diabetes was not measured at each exam, we were unable to include similar diabetes sensitivity tests. In addition, we tested whether associations for global cognitive function and verbal memory score with weight status categories varied by sex in sex-stratified minimally and fully adjusted models.

Results

The 3 522 individuals excluded from the analysis were slightly older, still working, had a higher average income and more education, and lived in more urbanized areas than the participants included in the analysis (Supplementary Table 1). Our final analytic sample consisted of 5 992 adults (3 155 women) who were 55 years or older at 2 or more study visits (Table 1). Sociodemographic, lifestyle, and health factors varied across baseline weight status. We observed lower global cognitive and verbal memory scores on average among individuals who were underweight at the time of their first cognitive measure compared to overweight individuals. In general, individuals with baseline overweight and obesity tended to have higher education, currently be working, be married, have greater income, participate in active leisure, and report a history of hypertension, diabetes, myocardial infarction, or stroke. In addition, we found a higher proportion of underweight in cohorts born before 1950, whereas a higher proportion of obesity in cohorts born after 1950.

Table 1.

Descriptive Statistics for Baseline Visit by Weight Status Category^*

		All N = 5 992	Underweight N = 392	Normal N = 2 378	Overweight N = 2 402	Obese N = 820	p Value for difference by weight status
Age		62.4 (6.4)	64.4 (7.9)	62.5 (6.6)	62.1 (6.1)	62.0 (5.7)	<.0001
Men (%)		47.3	42.3	52.1	45.9	40.4	<.0001
Education (%)	No formal schooling completed	20.2	31.4	21.6	18.0	17.5	<.0001
	Primary school	37.7	42.4	40.4	35.1	34.8
	Primary school+	42.1	26.2	37.9	46.8	47.7
Married (%)		85.5	73.3	84.2	88.1	87.4	<.0001
Currently working		39.6	53.5	45.7	34.5	30.3	<.0001
Total gross HH income (k) inflated to 2018		13.8 (19.1)	9.1 (12.8)	11.9 (19.5)	15.4 (16.9)	16.6 (24.7)	<.0001
Urbanization index—mean (SD)		64.2 (19.4)	54.6 (18.2)	60.3 (19.8)	68.0 (18.5)	68.9 (17.4)	<.0001
Region	North	15.6	12.8	12.3	17.2	22.1	<.0001
	Central	42.5	25.5	37.1	47.5	51.3
	South	41.9	61.7	50.5	35.3	26.6
Birth cohort	Before 1930	9.9	21.4	12.3	6.8	6.8	<.0001
	1930–40	21.1	32.1	22.8	18.6	18.3
	1940–50	33.4	28.1	33.9	34.8	30.4
	After 1950	33.5	17.6	29.8	37.4	40.6
Participating in active leisure (%)		10.1	4.9	7.7	12.5	12.1	<.0001
Smoking status	Never	62.6	57.8	56.4	66.7	70.6	<.0001
	Former	9.9	10.5	10.3	9.2	10.5
	Current	27.5	31.7	33.2	24.1	18.9
Comorbidities
Hypertension (%)		18.0	3.6	11.4	21.9	32.5	<.0001
Diabetes (%)		4.9	1.8	3.1	5.6	9.2	<.0001
MI/stroke (%)		2.5	1.8	1.9	2.7	3.9	.0075
On hypertension medication (%)		14.1	2.1	8.7	17.8	27.1	<.0001
BMI*—mean (SD)		23.6 (3.6)	17.5 (0.9)	21.0 (1.2)	25.0 (1.3)	29.7 (1.9)	<.0001
Global cognitive function score– mean (SD)		14.3 (6.3)	12.3 (6.6)	13.4 (6.3)	14.9 (6.3)	14.5 (6.3)	<.0001
Verbal memory score—mean (SD)		9.7 (4.8)	8.7 (4.5)	9.5 (4.7)	10.0 (4.8)	9.7 (4.8)	<.0001

Notes: BMI = body mass index; SD = standard deviation.

*Weight was categorized according to World Health Organization Asian BMI classification: BMI ≤ 18.5 kg/m² as underweight, between 18.5 and 22.9 kg/m² as normal weight, between 23 and 27.4 kg/m² as overweight and BMI ≥ 27.5 kg/m² as obese.

While we observed declines in global cognitive and verbal memory scores with increasing age, they were not statistically significant (Table 2). We also observed declines in average annual global cognitive and verbal memory scores across weight status categories (except overweight), although they were not statistically significant (Table 2). In minimally adjusted cubic spline models, we found differences in global cognitive score decline across weight status categories (p = .01, Figure 1). To aid in interpretation, we used model parameters (Supplementary Table 2) to estimate adjusted global cognitive declines across age and by weight status categories from the minimally adjusted cubic spline model (Figure 1). In addition, to better understand the differences in these declines within specific age ranges, we estimated the global cognitive score for each age category based on our adjusted model with age categories (Supplementary Table 2). As shown in Figure 2, there are significant declines in global cognitive scores across ages for each weight status category (p < .001), except among individuals with obesity (p = .15). This model allowed us to test for age differences within weight status categories, finding consistent global cognitive decline across all ages for normal-weight individuals but statistically significant declines only in the oldest age categories among underweight individuals. As shown in Supplementary Table 2, results were similar in the fully adjusted models (Supplementary Figures 1 and 2) and in the sensitivity analysis to test whether control for hypertension based on measured blood pressure or medication use was different from the self-reported diagnosis measure used in the primary analyses.

Table 2.

Mean (standard error) Average Annual Change^* in Global Cognitive Function and Verbal Memory Change Within Each Index Age Group^† and Weight Status Category^‡

		Annual Change* in Global Cognitive Function Score	p Value for Differences in Annual Change across Groups§	Annual Change* in Verbal Memory Change	p Value for Differences in Annual Change across Groups§
Age group	55–60	−0.05 (0.04)	.1022	−0.05 (0.04)	.4649
	60–65	−0.07 (0.05)		−0.09 (0.04)
	65–70	−0.19 (0.05)		−0.15 (0.04)
	70–75	−0.18 (0.08)		−0.11 (0.06)
	75+	−0.23 (0.08)		−0.10 (0.06)
Weight status	Underweight	−0.26 (0.09)	.1078	−0.19 (0.08)	.1569
	Normal weight	−0.11 (0.04)		−0.08 (0.03)
	Overweight	0.00 (0.07)		−0.01 (0.06)
	Obesity	−0.14 (0.04)		−0.12 (0.03)

*Annual change in cognitive function was calculated by dividing the difference between 2 subsequent survey years in cognitive function by the years between measures.

^†Age at first observation for each change period.

^‡Weight was categorized according to World Health Organization Asian BMI classification: BMI ≤ 18.5 kg/m² as underweight, between 18.5 and 22.9 kg/m² as normal weight, between 23 and 27.4 kg/m² as overweight and BMI ≥ 27.5 kg/m² as obese*.

^§ p Values are tests for differences in means across groups from 2 separate univariate linear mixed-effects models, 1 for age group and 1 for weight status category, with random effects at the individual and community levels.

Figure 1.

Adjusted global cognitive function scores from 55 to 80 by weight status category^# estimated from minimally adjusted cubic spline model*. *Adjusted cognitive assessment scores are predicted from a model with cubic age spline, weight status, and an interaction term for weight status and age group controlling for sex, education, marital status, income, urbanization index, birth cohort, region, and working status. #Weight was categorized according to World Health Organization Asian BMI classification: BMI ≤ 18.5 kg/m² as underweight, between 18.5 and 22.9 kg/m² as normal weight, between 23 and 27.4 kg/m² as overweight, and BMI ≥ 27.5 kg/m² as obese. ^~p Value for test of differences in rate of cognitive decline by weight status.

Figure 2.

Adjusted global cognitive function scores by age group, by weight status^# estimated from minimally adjusted model with age categories^¥. ^¥Adjusted cognitive assessment scores are predicted from a model with age group, weight status, and an interaction term for weight status and age group controlling for sex, education, marital status, income, urbanization index, birth cohort, region, and working status. ^#Weight was categorized according to World health organization Asian BMI classification: BMI ≤ 18.5 kg/m² as underweight, between 18.5 and 22.9 kg/m² as normal weight, between 23 and 27.4 kg/m² as overweight, and BMI ≥ 27.5 kg/m² as obese. *p Values correspond to Wald tests for any difference in adjusted cognitive function across age groups within each weight status category.

We found nominal differences in the rates of verbal memory decline by weight status only in the minimally adjusted models (p = .08; Supplementary Figure 3A and Supplementary Table 3) with similar patterns to that of adjusted global cognitive declines across age and weight status categories. As with the global cognitive score, adjusted verbal memory declined substantially across age only among underweight, normal weight, and overweight individuals (p < .0001), but not among obese individuals (p = .17) (Supplementary Figure 3B). Results were similar in the fully adjusted models (Supplementary Figure 4A and B, and Supplementary Table 3). In sex-stratified cubic spline models, we found statistically significant differences in rates of global cognitive decline by weight status for men (p < .001) and women (p = .01) in the fully adjusted models (Supplementary Figure 5A and B). We also found statistically significant declines in global cognitive function across ages for underweight, normal-weight, and overweight men and women, but no evidence of such declines for individuals with obesity (Supplementary Figures 5B and 6B). Results were similar in minimally and fully adjusted models.

Discussion

In a cohort of adults (>55 years) participating in a prospective study in China, we demonstrated that trajectories of cognitive decline from mid- to late life differed by weight status. We found that the cognitive decline trajectory in people with obesity was slower than in participants who were underweight, normal weight, or overweight. These findings advance our understanding of the intricate interchange between weight status and cognitive decline in older adults.

Although mid-life obesity in contemporary Western populations has been shown to increase the risk of dementia later in life, underweight status has been associated with cognitive impairment in older adults (34). The etiology underlying this paradox is unclear. However, several pathophysiological processes have been proposed. Higher mid-life weight status in contemporary Western populations is associated with higher rates of other vascular risk factors such as hypertension and hyperlipidemia, which can lead to vascular dementia or contribute to the development of Alzheimer’s disease (42). Conversely, in similar populations of older adults, weight loss and being underweight are markers of frailty and suggest that individuals may have acquired an insurmountable multimorbidity burden (43). This multimorbidity can contribute to cognitive impairment and vice versa, resulting in an inverse association between underweight and cognitive impairment (44–48). We sought to examine these differences in the CHNS, a cohort with considerable variation in weight status given dramatic urbanization and nutrition transition.

Although we found no statistically significant difference in cognitive decline by weight status and age, rates of decline were highest among older adults and in people who were underweight (Table 2). Notably, we found a significant decline in cognitive scores, across age categories, for each weight category, except among people with obesity (Figure 1). These findings highlight the importance of considering the relationship between weight status and cognitive decline trajectory rather than examining associations between cognitive function and weight status at a particular life stage. Studies in Western populations suggest that higher BMI in older adults may be associated with lower mortality, or a protective “obesity paradox” (49). The pathophysiology leading to this paradox is likely complex and multifactorial. Potential mechanisms may include increased availability of energy stored as fat that can be used in times of stress and better overall nutritional status (18). It is also possible that unmeasured confounding or other factors may result in a positive association between BMI and cognitive decline. Two such factors that we considered are (1) over-adjustment for cardiovascular risk factors and (2) increased exposure to medications that improve cardiovascular outcomes among those in the higher weight status cohort. Seeing as results were similar in minimally adjusted and fully adjusted models, we do not believe that over-adjustment explains our findings. Also, in a sensitivity analysis (Supplementary Table 2) findings remained similar after adjusting for hypertension medications. Our findings may support the idea that the relationship between cognitive decline and BMI is complex and is impacted by many individual and environmental factors. The relationship changes over time and thus may not be accurately captured by cross-sectional or short prospective studies (30).

Two important factors must be considered when reflecting on the generalizability of our findings. First, it is difficult to fully account for the major socioeconomic changes in China’s urbanization during the study period (26,41,50). We corrected for a range of lifestyle, sociodemographic, and health measures to account for differences between participant age cohorts. However, there may be some fundamental differences between the younger and older cohorts that we could not capture as a function of exogenous changes occurring across decades of dramatic urbanization in China. Second, as a result of changes in body composition with aging, namely increased adiposity and decreased muscle mass, BMI while easy to use clinically, is less accurate than dual-energy x-ray absorptiometry in older adults (51). Additionally, the use of BMI to measure obesity in Asian populations has long been debated (22,52). In particular, obesity classification using BMI fails to capture individuals with high body fat and normal BMI, which is a well-defined phenotype within individuals of Asian descent (52). Although we used the World Health Organization Asian cutpoints (22), it is possible that we did not capture well-known differences in body composition within the same BMI category. Finally, although weight status is related to diet and nutritional status, to investigate the pathways through which diet is associated with cognitive function both on the BMI pathway and outside of it would require a more complex pathway-based analysis which was beyond the scope of this paper but should be investigated further.

Our findings suggest an association between obesity and slower cognitive decline in older adults. However, obesity is a heterogeneous disease, and in China is ever so complex, as it is associated with higher income, education, and urbanization across time (26,29,41,53–55). This has changed in recent decades with greater access to highly processed foods and sedentary lifestyles and has become increasingly common even in the most rural parts of China (54). Early in the nutritional transition, the first individuals to become obese were individuals of higher income who also had greater access to medical care and infrastructure to support healthy lifestyles, factors that are also associated with healthy cognitive aging. Yet later in the nutrition transition, obesity becomes more common in rural and lower-income areas, with less access to obesity-protective environments and lifestyles (56–61).

It is possible that the changes in the rate of decline in individuals with obesity that we observed add to our understanding of the relationship between biological changes with aging, weight status, and cognitive function. Future research is needed to establish whether this pattern observed in a rapidly modernizing population is observed in longitudinal studies in different contexts.

Funding

This work was supported by the NIH, the National Institute of Diabetes and Digestive and Kidney Diseases (R01DK104371), and the National Heart, Lung, and Blood Institute (R01HL108427). We are grateful to the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01 HD30880), the National Institute on Aging (R01AG065357), and the NIH Fogarty International Center (D43 TW009077) for financial support for the CHNS data collection and analysis files from 1989 to 2015. We are also grateful for funding from the NICHD to Carolina Population Center at the UNC-CH (NIH grant P2CHD050924) and the China–Japan Friendship Hospital, the Chinese Ministry of Health for support for CHNS 2009, the Chinese National Human Genome Center at Shanghai since 2009, and Beijing Municipal Center for Disease Prevention and Control since 2011. We thank the National Institute for Nutrition and Health and China Center for Disease Control and Prevention. Dr. Batsis’ has had his research supported in part by the National Institute on Aging of the National Institutes of Health under Award Number K23AG051681 and R01AG067416. Dr. Batsis also has equity in SynchroHealth LLC, a remote monitoring company.

Conflict of Interest

None declared.

Author Contributions

All authors have read and approved the manuscript.

Ethics Approval

This study was deemed to be non-human subjects research by the University of North Carolina at Chapel Hill, North Carolina (#20-3138). Informed consent was obtained by all participants in this study in a written manner by the China Health and Nutrition Survey.