This cross-sectional study utilized data from NHANES, a national survey conducted by the National Center for Health Statistics (NCHS) to assess Americans’ health and nutritional conditions. NHANES employed a sophisticated, multistage probability sampling design to obtain a nationally representative sample of the non-institutionalized US population. Participants provided information about demographics, socioeconomic status, and health status through a household interview, whereas mobile examination centers (MECs) handled laboratory and physical assessments.

The NCHS Research Ethics Review Board authorized all NHANES study methods, and all participants provided informed consent. Detailed information about the study design and data can be obtained at www.cdc.gov/nchs/NHANES/.

The study population was recruited from NHANES 2011–2014, and all participants with complete CERAD test, AFT, DSST, BMI, and sleep duration data were included in this study. According to the World Health Organization (WHO), overweight is defined as a BMI between 25 and 29.9 kg/m², and obesity is defined as a BMI of 30 kg/m² or higher. Therefore, in this study, we defined overweight and obesity in the study population as a BMI ≥ 25 kg/m². The World Health Organization also classifies those 60 years of age and older as older adults (Beard et al., 2016). Consequently, individuals aged 60 and above were classified as older adults in our research. A total of 19,931 participants were enrolled. After excluding participants with age < 60 years (n = 16,299), missing cognitive assessment (n = 484), BMI (n = 61), and sleep duration (n = 7) data, and excluding BMI < 25 kg/m² (n = 837), a final total of 2,243 eligible participants were enrolled in this study (Figure 1).

Participants’ sleep duration was assessed with the question, “How much sleep (do you/does SP) usually get at night on weekdays or workdays?” Sleep duration ranged from 2 to 11 h, and if participants reported more than 12 h, it was recorded as 12 h, representing total nighttime sleep. Nighttime sleep plays a vital role in brain neurorepair, memory consolidation, metabolic waste clearance, and cognitive function maintenance. This includes synchronization with the circadian rhythm, which regulates different sleep stages like slow-wave and rapid-eye-movement sleep, crucial for brain health. Nighttime sleep also supports cognitive functions by influencing neurotransmitter balance (e.g., dopamine, norepinephrine) and hormones (e.g., growth hormone, cortisol). Furthermore, it minimizes external disruptions, ensuring higher sleep quality (Bubbico et al., 2019; You et al., 2019; Gabelle et al., 2017). In this study, sleep duration was considered a continuous variable.

The CERAD test evaluates immediate and delayed recall of newly learned verbal information (Fillenbaum et al., 2008). The CERAD test comprises three consecutive learning trials followed by a delayed recall. Participants are asked to read aloud a list of 10 unrelated words in each learning trial. Afterward, they immediately recall as many words as possible. The delayed recall occurs approximately 10 min after the learning trials. Each trial has a maximum score of 10 points, with a total possible score of 40 points, combining the results of the three learning trials and the delayed recall.

The AFT assesses verbal category fluency, which is a measure of executive function, along with other cognitive abilities such as semantic memory and processing speed (Clark et al., 2009). In this task, participants are required to generate as many animal names as possible within a one-minute timeframe, with one point awarded for each correct response.

The DSST is a comprehensive assessment of cognitive functioning involving processing speed, visual scanning, sustained attention, and short-term memory (Casagrande et al., 2021). The test is administered on paper, with a key at the top showing nine numbers paired with unique symbols. Participants are given 2 min to match and copy the corresponding symbols into the 133 boxes that are aligned with the numbers.

According to factors identified in previous research that are associated with sleep duration or cognitive function, this study controlled for several covariates of interest, including sex (male/female), age (years), marriage (married/divorced/widowed/living alone), race (Mexican American/other Hispanic/non-Hispanic white/non-Hispanic black/other race), the income-poverty ratio of family (<1/≥1), education level (less than middle school/less than high school/high school or GED/ college or AA degree/college or higher), smoking status (≥ 100-lifetime cigarettes/ <100-lifetime cigarettes), alcohol intake (≥12 drinks per year/ <12 drinks per year), hypertension (yes/no), hyperlipidemia (yes/no), and diabetes mellitus (yes/no/borderline).

In descriptive analyses, continuous variables are summarized by mean and standard error (SE). To examine the association between sleep duration and cognitive test scores in overweight and obese older adults, multivariable regression models were used, accounting for the NHANES complex sampling design (sampling weights). Three different model analyses were performed. The crude model was unadjusted for any covariates, Model 1 adjusted for sex, age, and race, and Model 2 adjusted for sex, age, race, marital status, education level, income-poverty ratio, hypertension, hyperlipidemia, diabetes, smoke status, and alcohol intake status. Generalized additive models (GAM) and smoothed curves were also used to address potential non-linear relationships between sleep duration and cognitive test scores. If a non-linear association was observed, a piecewise linear regression model (segmented regression) was used to fit each interval and calculate the threshold effect. A likelihood ratio test comparing a linear model (non-segmented) with the piecewise linear regression model was conducted to determine the presence of a threshold effect. The inflection point (K) connecting the two segments was determined using the maximum likelihood model and a two-step recursive method. Subgroup analysis was performed using stratified multivariable logistic regression models, with stratification factors including sex (male/female), age (≤69/70–79/>79 years), smoke status (≥100 lifetime cigarettes/<100 lifetime cigarettes), alcohol intake (≥12 drinks per year/<12 drinks per year), hypertension (yes/no), hyperlipidemia (yes/no), and diabetes mellitus (yes/no/borderline). These stratification factors were also treated as potential effect modifiers, and interaction terms were added using likelihood ratio tests to assess heterogeneity in associations across different subgroups. All statistical analyses were performed using the R software (version 4.2) and the EmpowerStats add-in (version 4.2). The threshold for statistical significance was set at a two-tailed p-value of 0.05.

The study population comprised 2,243 individuals aged 60 years or older (including 60 years) with a BMI above the normal range, drawn from NHANES (2011–2014). Of these, 1,095 (48.82%) were male and 1,148 (51.18%) were female. The mean age of the 2,243 participants was 69.35 years, the mean sleep duration was 7.04 h, and the mean values of the three tests related to cognitive function (CERAD test, AFT, and DDST) were 24.38, 16.51, and 45.88, respectively.

Analysis of sleep duration as a continuous variable across different demographic, lifestyle, and health categories revealed significant differences in sleep duration based on age, race, hypertension status, overweight or obesity status, and different CERAD test score ranges. These findings were statistically significant (p < 0.05) (Table 1).

This study found that in the demographic model (Model 1), each additional hour of sleep was associated with a decrease of 0.22 in CERAD test scores (β = −0.22, 95% CI: −0.40, −0.04), a decrease of 0.16 in AFT scores (β = −0.16, 95% CI: −0.30, −0.01), and a decrease of 0.60 in DSST scores (β = −0.60, 95% CI: −1.04, −0.17).

In the fully adjusted model (Model 2), sleep duration remained negatively associated with CERAD test, AFT, and DSST scores. Each additional hour of sleep was associated with a decrease of 0.25 in CERAD test scores (β = −0.25, 95% CI: −0.43, −0.07), a decrease of 0.17 in AFT scores (β = −0.17, 95% CI: −0.32, −0.02), and a decrease of 0.78 in DSST scores (β = −0.78, 95% CI: −1.16, −0.40). All results were statistically significant (p < 0.05) (Table 2).

Our study indicates a negative correlation between cognitive function and sleep duration. Based on previously identified confounding factors, we further assessed the relationship between sleep duration and cognitive function in both predefined and exploratory subgroups. Stratified analyses and interaction tests were performed as shown in Table 3 and Figure 2 (Association between sleep duration and different cognitive function scores stratified by age, sex, alcohol intake, smoke status, diabetes mellitus. Adjusted for all presented covariates. (A) CERAD test; (B) AFT; (C) DSST.* p < 0.05).

The results demonstrated that there were no significant interactions between any of the stratifying variables and the relationship between sleep duration and cognitive function. Specifically, the negative association between sleep duration and cognitive test scores in overweight and obese older adults was consistent across all stratified groups (all p-values for interaction >0.05).

The solid red line represents the smooth curve fit between variables. The dotted line represents the 95% confidence interval for the fit. The dose–response relationship between sleep duration with the score of the CERAD test (A), the score of the Animal Fluency test (B), and the score of the Digit Symbol Substitution test (C) in obese elders (Figure 3).

After fully adjusting for covariates, we applied a threshold effect model and identified the inflection points (K) for the CERAD test, AFT, and DSST at 5, 6, and 6 h of sleep, respectively. To the left of the inflection point for CERAD test, a positive association between sleep duration and CERAD test was detected (β = 2.11, 95% CI: 1.17, 3.05). However, to the right of this point, sleep duration was negatively associated with CERAD test (β = −0.51, 95% CI: −0.71, −0.30), with a log-likelihood ratio (LLR) of <0.001.

With regard to AFT, no statistically significant association was observed between sleep duration and the inflection point (β = 0.25, 95% CI: −0.17, 0.67). However, a robust inverse correlation was observed on the right side of the inflection point (β = −0.32, 95% CI: −0.52, −0.12), with an LLR of 0.0019.

Similarly, no significant association was identified between DSST and sleep duration on the left side of the inflection point (β = 0.49, 95% CI: −0.57, 1.56). In contrast, a clear negative association was observed on the right side (β = −1.21, 95% CI: −1.71, −0.70), with an LLR of 0.012 (Table 4).