Relationship Between Weight-Change Patterns and Cognitive Function: A Retrospective Study

Data sources

The NHANES is a nationally representative, continuous cross-sectional study designed to assess health and nutritional status of adults and children in the United States. Briefly, this survey contains information on demographic, socioeconomic, dietary, and health-related factors acquired through interviews; medical, dental, and physiological measurements; as well as laboratory tests administered and acquired by the examiners. We incorporated data on adult participants from four cycles of the NHANES, including the 1999–2000 (n = 1834), 2001–2002 (n = 1872), 2011–2012 (n = 1687), and 2013–2014 (n = 1785) phases, which specifically assessed cognitive performance in individuals aged 60 years and older.

The National Center for Health Statistics Research Ethics Review Board approved the NHANES protocol, and written informed consent was obtained prior to data collection.

Weight-change measures and covariates

Data on weight and height at 25 years of age and 10 years before the baseline survey and the age at maximum weight were recorded at the baseline survey. Baseline weight and height were measured during physical examinations. BMI at age 25 years of age (BMI₂₅), at 10 years before the baseline survey (BMI_10prior), and at the baseline survey (BMI_baseline) were calculated as weight (kg) divided by the square of height (m²). We categorized the participants into two groups: obese (BMI≥30 kg/m²) and non-obese (BMI < 30 kg/m²). Using BMI measurements obtained at three time intervals (BMI₂₅ to BMI_10prior, BMI_10prior to BMI_baseline, and BMI₂₅ to BMI_baseline), we categorized four BMI change patterns (Fig. 2): 1) stable non-obese pattern: BMI  < 30 kg/m² at both time point; 2) obese to non-obese pattern: BMI≥30 kg/m² at the younger age and < 30 kg/m² later; 3) non-obese to obese pattern: BMI < 30 kg/m² at younger age and≥30 kg/m² later; and 4) stable obese pattern: BMI≥30 kg/m² at both time points. This approach has been described extensively elsewhere [18].

All models were adjusted for covariates associated with cognitive decline and dementia, including age, sex, race/ethnicity, education level, annual household income, smoking status, diabetes, hypertension, chronic kidney disease, and cardiovascular diseases (coronary heart disease, myocardial infarction, and stroke).

Assessment of cognitive function

Cognitive function was measured using the baseline NHANES surveys. The cognitive function tests used in the NHANES include the Digit Symbol Substitution Test (DSST), Consortium to Establish a Registry for Alzheimer’s Disease Word Learning test (CERAD-WL) and Delayed Recall test (CERAD-DR), and Animal Fluency (AF) test. In NHANES 1999–2001 and 2002–2003, only one cognitive test, the DSST, was administered to participants aged≥60 years. In NHANES 2011–2012 and 2013–2014, all four tests were administered to participants aged≥60 years.

In the DSST, participants were requested to copy corresponding symbols that were paired to nine numbers in the 133 boxes adjoining the numbers in 2 min. The score represented the total number of correct matches. The maximum score for this test is 100, and the score reflects processing speed, sustained attention, and working memory.

The CERAD-WL consisted of four tests, of which three were consecutive learning trials and one was a delayed recall trail. In the learning part, 10 unrelated words were presented one at a time, and participants were requested to read each one aloud. Immediately after the presentation of the words, the participants were asked to recall as many words as possible. The total score for the learning task was 30. For the delayed recall trail, participants were requested to recall the as many of the words used in the first CERAD-WL trial as they could approximately 8–10 min after the word learning trials. The maximum CERAD-DR score was 10.

In the AF test, participants were requested to name as many animals as possible in 1 min after passing a practice test that required them to name three items of clothing. The maximum score for the AF test was 40, and the score depended on the number of animals named.

For all participants, cutoffs of < 14 for the AF score, < 17 for the CERAD-WL score, and < 5 for the CERAD-DR score were used to distinguish potential cognitive impairment from relative cognitive health in the NHANES; the cutoff values were based on prior literature [19–21]. The DSST cutoff was defined as the dividing line of the lowest quartile (25th percentile) in our study group (≤30 points) to DSST impairment, which is consistent with the methods previously published in the literature [22].

Statistical analysis

All statistical analyses were performed using Stata (version 15.0; StataCorp LP, College Station, USA), and a two-tailed p value of < 0.05 was considered to indicate statistical significance. Demographic variables at baseline were presented as means with standard deviations or proportions.

Linear regression models were used to investigate the differences in the coefficients of each weight-change group with cognitive test scores at each time interval after checking for normality. The stable non-obese weight-change pattern was defined as the reference group. When the cognitive test scores were considered categorical variables, we used logistic regression models to compare the results of different groups by using the stable non-obese group as the reference. Considering the different weight statuses across each individual’s lifespan, participants were asked to recall the age at which they were at their heaviest weight. Linear regression models were used to investigate the potential linearity between the changes in cognitive function and the increase in the year when weight was the heaviest. We then categorized three age_heaviest weight groups: 18–40, 40–60,≥60 years, while the group with the heaviest weight after≥60 years of age served as the referent. Models were adjusted for age, sex, race/ethnicity, education level, and annual household income as well as smoking status, diabetes, hypertension, chronic kidney disease, and cardiovascular diseases (coronary heart disease, myocardial infarction, and stroke).

OPEN IN VIEWER

Fig. 1

Study design for analysis between weight change and cognitive impairment at survey (n = 7,178). It shows how a retrospective cohort of US adults were created. We studied individuals who participated in the NHANES (1999–2002, 2011–2014) cross-sectional survey at ages 60 years and more. It based on participants’ recalled weight history at age 25 (young adulthood) and ten years before the survey (age≥50 years, midlife), which were used to create a measure of weight change between young adulthood and midlife. We then investigated the association between this weight change and risk of cognitive impairment at the survey time.

Characteristics of the participants

Data on cognitive function were available for 7,178 individuals who participated in NHANES 1999–2002 and 2011–2014. After excluding those without weight information ten years before the survey and in young adulthood, 5,809 participants were finally included in the current study (Fig. 1). Of these, 5,152 individuals had undergone the DSST; 2,830, the CERAD-WL; 2,828, the CERAD-DR; and 2,812, the AF test. The results of the DSST, CERAD-WL, CERAD-DR, and AF test indicated cognitive impairment in 12.9%, 21.9%, 21.5%, and 21.3% of the participants, respectively. Approximately 50% of the individuals were considered to have lived past midlife (≥60 years) 10 years before the baseline survey. The baseline characteristics of the enrolled patients are presented in Table 1. In absolute numbers, participants with cognitive impairment were likely to be older, black, diabetic, hypertensive, and have a lower annual household income. Regardless of cognitive impairment, BMI increased significantly from the age of 25 to 10 years before baseline and remained relatively stable thereafter. No discernible differences were observed in BMI at any time point between those who later developed cognitive impairment and those who did not.

OPEN IN VIEWER

Fig. 2

The association between weight change and cognitive impairment at survey. 1) Stable non-obese: BMI < 30 kg/m² at both times; 2) obese to non-obese: BMI≥30 kg/m² at younger age and < 30 kg/m² later; 3) non-obese to obese: BMI < 30 kg/m² at younger age and≥30 kg/m² later; and 4) stable obese: BMI≥30 kg/m² at both times.

Association between weight change patterns and specific cognitive test scores

After adjusting for possible confounders, in comparison with stable non-obese participants, participants who changed from the obese to non-obese category during the 10-year period before baseline tended to have lower DSST scores (β= –3.22; 95% confidence interval [CI]: –5.09 to –1.35), whereas no significant difference was observed in participants belonging to the non-obese to obese group (β= –0.60, 95% CI: –2.29 to 1.08) and the stable obese group (β= –0.32, 95% CI: –1.78 to 1.15). CERAD-WL, CERAD-DR, and AF test scores were similar among the four weight-change pattern groups. In the analysis of BMI changes from 25 years of age to 10 years before the baseline survey, those who changed from the non-obese to obese category tended to have lower DSST scores (β= –1.28; 95% CI: –2.24 to –0.31) than stable non-obese participants. No other significant association was observed for participants belonging to the obese to non-obese (β= –5.43, 95% CI: –11.40 to 0.54) and stable obese groups (β= 1.57, 95% CI: –1.48 to 4.63). CERAD-WL, CERAD-DR, and AF test scores were not significantly different among the four weight-change pattern groups. From the age of 25 years to the baseline survey, no significant difference was found between BMI changes and scores among all four cognitive tests in comparison with the stable non-obese participants (Table 2).

Association between weight-change patterns and specific cognitive impairment

Using a logistic regression method and after adjusting for all potential confounders, we explored the relationship between weight-change patterns and cognitive impairment. Analysis of the data obtained 10 years before baseline showed a higher risk of AF impairment in the obese to non-obese group than in the stable non-obese group (odds ratio [OR] = 1.50; 95% CI: 1.05 to 2.16), which was not found in the non-obese to obese group (OR = 0.89; 95% CI: 0.59 to 1.34) or the stable obese group (OR = 0.97; 95% CI: 0.66 to 1.42). All four weight-change pattern groups showed no difference in the risk of DSST, CERAD-WL, and CERAD-DR impairment during this period. From 25 years of age to 10 years prior to baseline, a significantly higher risk of DSST impairment was found in the non-obese to obese group (OR = 1.40; 95% CI: 1.09 to 1.79), but not in the obese to non-obese group (OR = 2.05; 95% CI: 0.47 to 8.96) or stable obese group (OR = 0.68; 95% CI: 0.35 to 1.31) in comparison with the stable non-obese group. Similar risks of the other three cognitive impairments (CERAD-WL, CERAD-DR, and AF) were found in all four weight-change pattern groups. In the analysis over a longer duration (from 25 years of age to the baseline survey), the results showed that, in comparison with stable non-obese individuals, people in the obese to non-obese group were at higher risk of CERAD-DR impairment (OR = 2.51; 95% CI: 1.17 to 5.40), but not DSST (OR = 0.94; 95% CI: 0.23 to 3.82), CERAD-WL (OR = 1.64; 95% CI: 0.76 to 3.56), or AF impairments (OR = 1.13; 95% CI: 0.30 to 4.32). Participants in the non-obese to obese and stable obese groups were found to have a similar risk of cognitive impairment as those in the stable non-obese group (Table 3).

The effect of age at heaviest weight on different cognitive scores

First, we considered age at the heaviest weight as a continuous variable (Table 4). We observed that the larger age_heaviest, the better the DSST score (β= 0.03; 95% CI: 0.00 to 0.06). In comparison with participants whose age_heaviest was≥60 years, individuals whose age_heaviest was between 18 and 40 years had lower DSST scores (β= –1.46; 95% CI: –2.77 to –1.52; Table 5), but no significant difference existed in those whose age_heaviest was between 40 and 60 years (β= –0.50; 95% CI: –1.56 to 0.56). The age_heaviest showed no significant effects on the CERAD-WL, CERAD-DR, and AF test scores in both continuous and categorical analyses.

Strengths and limitations

Our study included a large and nationally representative sample with a detailed analysis of weight changes from young adulthood through midlife to late adulthood, which allowed us to estimate the associations between weight change across adulthood and cognitive function. Since people are more likely to gain weight until middle age, stabilize, and begin to lose weight near 60 years of age [47], assessment of weight changes in different life periods is important. In addition, weight gain in early adulthood mostly reflects fat accumulation, and weight loss in late life may reflect a loss of lean mass as well as an increase in fat mass [48, 49]. By studying weight-change patterns, we can better understand the dynamic relationship between changes in BMI and cognitive function.

This study also had several limitations. First, we stratified the participants into only two groups, obese, and non-obese, which may have caused data reduction. More specific classifications, including underweight, normal weight, and overweight, may better fit the reality. In our study, the maximum BMI of most participants could be classified as overweight; therefore, if we had further categorized the non-obese pattern into a stable normal pattern and maximum overweight pattern, similar to the approach employed by Pan et al. [50], the group with the stable normal pattern would be relatively small. Second, we only used BMI and not waist circumference or the waist-to-height ratio. Central obesity has been reported to be associated with a higher risk of cognitive decline, even in individuals with a normal BMI [42]. Third, the weight and height were self-reported. Although the self-reported height may be closer to the actual height, the actual weight is likely to be lower than that reported by the participants, which may result in bias. Moreover, recall bias may be large because of the long-time interval. However, since historical weight measures have been proven to show high agreement with the measured weight in previous studies, our results should remain sound [51, 52]. Fourth, the four tests we used could not fully characterize cognitive damage. Fifth, in comparison with their normal-weight counterparts, obese individuals are more likely to die due to diseases such as cardiovascular disease, diabetes, and stroke [8]. Since the participants in our study were approximately 70 years old, a survival bias might exist. Finally, although we controlled the results for several potential confounders, residual confounding cannot be ruled out because of the observational study design. Our study did not include APOE level, exercise, trauma, or medicine, which may have affected the accuracy of the experimental results. In addition, only baseline characteristics and potential confounders were employed in our study; therefore, we could not capture changes in possible confounders over time.

Conclusions

In conclusion, our results suggest that the transition from the non-obese range to the obese range, especially in early adulthood, is positively correlated with worse cognitive function in later life. Our findings have implications for the development of public health strategies to prevent dementia. Regular weight checks can help identify high-risk populations and facilitate prompt interventions. Further research is required to examine the mechanisms underlying the influence of weight changes on cognitive function.