Sarcopenia modifies the association between diet quality and cognitive function in community-dwelling older people: a cross-sectional study

Abstract

Background

Sarcopenia and cognitive impairment are common age-related conditions that severely affect older adults’ quality of life. Diet quality, a modifiable factor, has been associated with muscle health and cognitive function, suggesting a potential interaction between sarcopenia, diet, and cognition. However, this relationship remains underexplored.

Methods

This cross-sectional study included 2,572 community-dwelling adults aged 60 + from Shanghai, China. Cognitive function was assessed using the Montreal Cognitive Assessment (MoCA), sarcopenia was diagnosed based on the Asian Working Group for Sarcopenia 2019 guidelines, and diet quality was evaluated with the modified Chinese Healthy Eating Index (CHEI). Multivariate regression models, along with interaction, mediation, and subgroup analyses, were applied to examine these associations.

Results

After adjustments, general sarcopenia and non-sarcopenia were associated with higher MoCA scores (β = 1.25, P = 0.006; β = 2.17, P < 0.001), compared to severe sarcopenia. Diet quality (CHEI scores) was positively associated with MoCA (β = 0.05, P < 0.001). Interaction analysis showed that the positive association between diet quality and MoCA score was attenuated with increasing sarcopenia severity (P for interaction < 0.05). Subgroup analysis revealed that age, body mass index, and education modified the sarcopenia–cognition association, whereas living status and cardiovascular disease history modified the diet–cognition association (all P for interaction < 0.05). Mediation analysis found no significant indirect associations.

Conclusion

This study highlights an interaction between diet quality and sarcopenia in relation to cognitive function. Higher diet quality was associated with better cognitive performance, but this association appeared weaker among individuals with more severe sarcopenia. Given the cross-sectional design, causality and directionality cannot be inferred; longitudinal and interventional studies are warranted to confirm these findings and inform prevention strategies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-026-07176-5.

Introduction

Sarcopenia, an age-related geriatric syndrome, is characterized by a significant decline in muscle mass, muscle strength, and physical performance [1]. It increases the risk of falls and fractures and is closely associated with various chronic conditions, including cognitive impairment, cardiovascular diseases, and depression [1]. The prevalence of sarcopenia is 5%-13% in individuals aged 60–70 years and 11%-50% in those aged 80 and above [2]. Cognitive impairment (CI) is characterized by decreased function in one or more cognitive domains, including memory, language, calculation, comprehension, executive function, and visuospatial skills [3]. CI is further classified into mild cognitive impairment (MCI) and dementia based on severity [3]. Approximately 15–20% of adults aged 65 years and older suffer from MCI, with 11.3% ultimately progressing to dementia [4]. These conditions severely compromise the quality of life and independence of older people, imposing substantial medical and financial burdens on families and society. With the aging of the global population, both sarcopenia and CI have become major public health challenges.

Emerging epidemiological evidence reveals bidirectional links between sarcopenia and CI. A meta-analysis of 77 studies found that individuals with sarcopenia have a significantly higher risk of developing MCI, Alzheimer’s disease (AD), and non-AD dementia compared to those without sarcopenia, with odds ratios of 1.58, 2.97, and 1.68, respectively. Furthermore, of the four cohort studies included in the analysis, three found that sarcopenia significantly predicted the later development of MCI or AD [5]. Conversely, CI may precede sarcopenia, as evidenced by a study by Pacifico et al., which found a threefold higher prevalence of sarcopenia in dementia patients (26.4%) compared to controls (8.4%) [6]. Several mechanisms have been proposed (e.g., inflammation and muscle–brain crosstalk), but the pathways remain incompletely understood [7]. Consequently, early interventions targeting both sarcopenia and CI may delay the progression of disability in aging populations.

Diet, an important modifiable lifestyle factor, plays a critical role in maintaining both muscular [8] and cognitive health [9]. Dietary patterns, rather than single nutrients or foods, may better capture the synergistic and antagonistic interactions within complex diets [10]. Several healthy dietary patterns, including the Mediterranean diet [11], Mediterranean–DASH Intervention for Neurodegenerative Delay diet (MIND) [12], and Dietary Approaches to Stop Hypertension (DASH) [13] have been associated with in delaying cognitive decline. Furthermore, sufficient consumption of critical nutrients, such as protein, vitamin D, omega-3 fatty acids, and antioxidants, is vital for maintaining muscular function and may slow sarcopenia progression [14]. Importantly, the cognitive benefits of diet may differ depending on an individual’s physiological state [15]. Given its close link with cognitive health via the muscle-brain axis, sarcopenia may act as a key modifier of the diet–cognition relationship [16]. These findings suggest that diet and sarcopenia do not independently influence cognitive function but may interact through shared biological pathways.

However, most studies to date have examined diet [17] and sarcopenia [18] as independent factors associated with cognition, with limited research systematically investigating the interrelationships among all three. Moreover, previous studies have primarily focused on Western dietary patterns [11, 12], which may not accurately reflect the dietary habits of Chinese populations. The Chinese Healthy Eating Index (CHEI), developed based on the 2016 Chinese Dietary Guideline (CDG-2016), provides a more culturally appropriate tool for assessing diet quality [19], yet its application in cognitive health research remains scarce.

It remains unknown whether sarcopenia modifies the association between overall diet quality and cognitive function among community-dwelling older adults in China. Using a modified CHEI, we examined the associations among diet quality, sarcopenia, and cognitive function to provide evidence for targeted strategies to to inform risk stratification and future preventive strategies.

Methods

Study design

This cross-sectional study was conducted in Chengqiao and Sanxing towns of Chongming District, Shanghai, China, from March 2025 to June 2025. This study was conducted in accordance with the Declaration of Helsinki. The protocol was approved by the Ethics Committee of Chongming Hospital Affiliated to Shanghai Health University (Approval No. CMEC-2025-KT-45). The study was registered with the China Clinical Trial Registry (Registration No. ChiCTR2500108893) to enhance research transparency, although it was observational in nature. All participants provided written informed consent. The flowchart of this study is shown in Fig. 1. This study was reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (Additional file 1).

Fig. 1.

Flow chart of this study

Sample size calculation

An a priori sample size calculation was performed using G*Power (version 3.1) [20]with the option “linear multiple regression: fixed model, R² deviation from zero.” We assumed a small effect size (Cohen’s f² = 0.02), a two-sided significance level of α = 0.05, power (1 − β) = 0.95, and up to 15 predictors based on the fully adjusted model (including the main exposure, prespecified covariates, and interaction terms) [21]. The minimum required sample size was 1,405 participants.

Study population and sampling method

A stratified cluster sampling approach was used, with towns as the primary sampling units. Based on the age and sex distribution of older residents across the 16 towns in Chongming District, Chengqiao and Sanxing towns were selected as survey clusters.

Inclusion criteria were: (1) aged ≥ 60 years; (2) long-term community residence; and (3) voluntary participation with written informed consent. Exclusion criteria were: (1) severe hearing, visual, or physical impairments that hinder completion of assessments; (2) severe systemic diseases (heart, lung, kidney, or brain); and (3) inability to undergo body composition measurement (e.g., inability to stand, pacemaker, or metal/electronic implants).

A total of 3,110 participants were recruited. After applying the eligibility criteria, 538 participants were excluded and 2,572 were included in the final analysis. Participants were classified into the cognitively impaired group (n = 1,810) and the cognitively normal group (n = 762) (Fig. 1).

Demographic characteristics

Data were collected through questionnaires, physical examinations, and laboratory tests, covering demographic characteristics (age, sex, education level, marital status, living status), anthropometric measurements (height, weight, calf circumference, body mass index [BMI], percentage of body fat [PBF]), medical history (chronic conditions, medications, hypertension, diabetes, hyperlipidemia, stroke, cardiovascular disease, falls, depressive symptoms), lifestyle factors (smoking, alcohol consumption, physical activity, sleep quality assessed by the Pittsburgh Sleep Quality Index [PSQI]), and hematological parameters (plateletcrit, white blood cell count, neutrophil count, lymphocyte count, heart rate).

Assessment of cognitive function

Cognitive function was assessed using the Montreal Cognitive Assessment (MoCA), which includes subdomains of visuospatial skills and executive function, delayed recall memory, attention, naming, language, abstraction, and orientation. The total score ranges from 0 to 30, with higher scores indicating better cognitive function. A cutoff of ≥ 26 was defined as normal cognition, whereas < 26 indicated cognitive impairment. To adjust for educational level, participants with ≤ 12 years of formal education received an additional point [22].

Assessment of sarcopenia

The definition of sarcopenia follows the Asian Working Group for Sarcopenia 2019 (AWGS 2019) guidelines [23]. Sarcopenia was diagnosed as low appendicular skeletal muscle mass index (ASMI) together with either low handgrip strength (HGS) or low physical performance. According to severity, it was further classified as general sarcopenia (low ASMI with either low HGS or low physical performance) and severe sarcopenia (low ASMI with both low HGS and low physical performance). The diagnostic cutoffs were:

1. Low ASMI: <7.0 kg/m² for men and < 5.7 kg/m² for women;

2. Low HGS: <28 kg for men and < 18 kg for women;

3. Low physical performance: 6-meter gait speed (GS) < 1.0 m/s, or Five-times chair stand test (5CST) ≥ 12 s, or Short Physical Performance Battery (SPPB) ≤ 9 points.

Assessment of dietary intake and dietary quality

Dietary intake was assessed using a previously validated 25-item Food Frequency Questionnaire (FFQ-25) [24], which collected information on the frequency and portion size of major food groups to estimate average daily intake. Dietary quality was assessed using the modified CHEI. The standard CHEI comprises 17 components: 12 adequacy components (total grains, whole grains and mixed legumes, tubers, total vegetables, dark-colored vegetables, fruits, dairy products, soy products, fish and seafood, poultry, eggs, and nuts and seeds) and 5 restriction components (red meat, cooking oils, sodium, added sugars, and alcohol) [19].

Because discretionary ingredients such as cooking oils, sodium (salt added during cooking), and added sugars are difficult to recall and quantify accurately using the FFQ-25, including these components could introduce substantial non-differential measurement error, which may attenuate the observed associations between diet quality and health outcomes [25, 26]. Therefore, these three components were excluded from the CHEI, resulting in a 14-component index. The modified CHEI preserves the original framework’s capacity to capture major food groups and overall dietary balance.

Except for fruits, which had a maximum score of 10, all other components had a maximum score of 5. Based on the thresholds of the CDG-2016, scores were divided into four intervals: 0, (0–5), (5–10), and 10 for fruits, and 0, (0–2.5), (2.5–5), and 5 for the remaining 13 components. For adequacy components, a score of 0 is assigned for no intake, and the scores increase proportionately as intakes increase up to the recommended intake. For restriction components, a maximum score is assigned at the recommended intake, and the scores decrease proportionately as intakes exceed the recommended intake. The maximum total score of the modified CHEI is 75 points, and the total modified CHEI score used in analyses was rescaled to a 0–100 range (higher scores indicating better overall diet quality). Therefore, for restriction components, higher scores indicate lower intake.

The mapping of modified CHEI components to FFQ-25 food categories and their recommended intake levels is provided in Supplementary Table 1.

Statistical analysis

All statistical analyses were performed using R software (version 4.3.0). Continuous variables are presented as means ± standard deviation (SD) and were compared using independent-samples t-tests. Categorical variables are presented as frequencies (percentages) and were compared using chi-square tests or Fisher’s exact tests, as appropriate.

Multivariate linear regression models were used to examine the associations of sarcopenia status, modified CHEI (total score and component scores), and CHEI level with MoCA score. Potential confounders were prespecified a priori based on existing literature and theoretical relevance [27], rather than data-driven selection procedures. Three hierarchical models were fitted: Model 1 (unadjusted), Model 2 (adjusted for age, sex, and educational level), and Model 3 (further adjusted for living status, marital status, smoking status, alcohol consumption, number of chronic conditions, number of medications, BMI, and PSQI score).

Interaction was tested by adding a product term (sarcopenia × modified CHEI) to the regression models, and stratified/subgroup analyses were conducted to describe associations within strata. Mediation analysis was performed using a nonparametric bootstrap approach (5,000 resamples); given the cross-sectional design, mediation results were interpreted without assuming temporal ordering.

Missing data were present in a small proportion of variables (0.5%–4.2%), primarily involving smoking status, alcohol consumption, and PSQI score. Missingness was assumed to be missing at random (MAR) and handled using full information maximum likelihood (FIML). Sensitivity analyses included alternative categorizations of sarcopenia status and modified CHEI total score, and repeating the same multivariable regression and interaction models using a proxy-enhanced CHEI constructed from the closest FFQ-25 items.

Primary analyses were prespecified. Interaction tests and subgroup analyses were considered exploratory and hypothesis-generating; therefore, formal adjustment for multiple comparisons was not applied. All tests were two-tailed, and P < 0.05 was considered statistically significant.

Results

Baseline characteristics

Baseline characteristics of the 2,572 participants are shown in Tables 1, 2 and 3, including demographic characteristics (Table 1), sarcopenia status (Table 2), and dietary quality (Table 3).

Table 1.

Baseline characteristics of participants stratified by cognitive status

Variables	Total (n = 2,572)	Cognitively impaired (n = 1,810)	Cognitively normal (n = 762)	P value
Demographics
Age, years, mean ± SD	73.04 ± 5.77	73.75 ± 5.89	71.37 ± 5.10	< 0.001
Female, n (%)	1,503 (58.44)	1,111 (61.38)	392 (51.44)	< 0.001
Education level, n (%)
Illiterate	283 (11.00)	272 (15.03)	11 (1.44)	< 0.001
Primary school	904 (35.15)	787 (43.48)	117 (15.35)
Middle school	938 (36.47)	580 (32.04)	358 (46.98)
High school and above	447 (17.38)	171 (9.45)	276 (36.22)
Marital status, n (%)
Not widowed	2,163 (84.10)	1,483 (81.93)	680 (89.24)	< 0.001
Widowed	409 (15.90)	327 (18.07)	82 (10.76)
Living status, n (%)
Live alone	375 (14.58)	301 (16.63)	74 (9.71)	< 0.001
Live with children	95 (3.69)	64 (3.54)	31 (4.07)
Live with spouse	1,946 (75.66)	1,347 (74.42)	599 (78.61)
Live with spouse and children	156 (6.06)	98 (5.41)	58 (7.61)
Body composition
BMI, kg/m², mean ± SD	24.76 ± 3.34	24.82 ± 3.39	24.61 ± 3.20	0.146
Calf circumference, cm, mean ± SD	34.08 ± 3.06	33.90 ± 3.03	34.52 ± 3.09	< 0.001
Percent body fat, %, mean ± SD	31.13 ± 7.29	31.63 ± 7.38	29.93 ± 6.92	< 0.001
Lifestyle factors
Smoking (ever), n (%)	698 (27.14)	461 (25.47)	237 (31.10)	0.003
Alcohol (ever), n (%)	745 (28.97)	505 (27.90)	240 (31.50)	0.066
Physical activity, n (%)				< 0.001
Low	402 (15.63)	316 (17.46)	86 (11.29)
Moderate	708 (27.53)	498 (27.51)	210 (27.56)
High	1,462 (56.84)	996 (55.03)	466 (61.15)
PSQI total score, mean ± SD	4.70 ± 4.05	4.72 ± 4.07	4.65 ± 4.01	0.704
Medical history, n (%)
Hypertension (yes)	1,659 (64.50)	1,181 (65.25)	478 (62.73)	0.223
Diabetes (yes)	546 (21.23)	385 (21.27)	161 (21.13)	0.936
Hyperlipidemia (yes)	300 (11.66)	187 (10.33)	113 (14.83)	0.001
Stroke (yes)	572 (22.24)	438 (24.20)	134 (17.59)	< 0.001
Cardiovascular disease (yes)	264 (10.26)	174 (9.61)	90 (11.81)	0.094
Falls in the past 1 year (yes)	209 (8.13)	164 (9.06)	45 (5.91)	0.007

P values are from χ² tests for categorical variables and appropriate two-sample tests for continuous variables. Values are mean ± SD or n (%)

Abbreviations: BMI Body mass index, PSQI Pittsburgh Sleep Quality Index, n number

Table 2.

Sarcopenia status of participants stratified by cognitive status

Variables	Total (n = 2,572)	Cognitively impaired (n = 1,810)	Cognitively normal (n = 762)	P value
Sarcopenia status, n (%)				< 0.001
Severe sarcopenia	161 (6.26)	141 (7.79)	20 (2.62)
General sarcopenia	195 (7.58)	145 (8.01)	50 (6.56)
Non-sarcopenia	2,216 (86.16)	1,524 (84.20)	692 (90.81)
Muscle indices
ASMI, kg/m², mean ± SD	6.91 ± 0.93	6.83 ± 0.92	7.09 ± 0.94	< 0.001
HGS, kg, mean ± SD	25.55 ± 8.69	24.89 ± 8.34	27.13 ± 9.28	< 0.001
GS, m/s, mean ± SD	1.10 ± 0.25	1.07 ± 0.26	1.17 ± 0.22	< 0.001
5CST, s, mean ± SD	11.04 ± 3.28	11.37 ± 3.53	10.25 ± 2.43	< 0.001
SPPB score, mean ± SD	10.84 ± 1.26	10.77 ± 1.36	11.02 ± 0.93	< 0.001

P values are from χ² tests for categorical variables and appropriate two-sample tests for continuous variables. Values are mean ± SD or n (%)

Abbreviations: ASMI Appendicular skeletal muscle mass index, HGS Handgrip strength, GS Gait speed, 5CST Five-times chair stand test, SPPB Short Physical Performance Battery, n number

Table 3.

Modified CHEI scores, level and components of participants stratified by cognitive status

Variables	Total (n = 2,572)	Cognitively impaired (n = 1,810)	Cognitively normal (n = 762)	P value
CHEI scores, mean ± SD	51.18 ± 11.59	50.07 ± 11.57	53.82 ± 11.23	< 0.001
CHEI level, n (%)				< 0.001
CHEI-Very low	639 (24.84)	507 (28.01)	132 (17.32)
CHEI-Low	647 (25.16)	463 (25.58)	184 (24.15)
CHEI-Moderate	648 (25.19)	459 (25.36)	189 (24.80)
CHEI-High	638 (24.81)	381 (21.05)	257 (33.73)
CHEI components, mean ± SD
Total grains	4.75 ± 0.67	4.76 ± 0.64	4.72 ± 0.72	0.174
Whole grains and mixed beans	3.12 ± 2.01	3.09 ± 2.03	3.18 ± 1.94	0.268
Tubers	0.87 ± 0.90	0.88 ± 0.92	0.83 ± 0.85	0.159
Vegetables, total	3.28 ± 1.39	3.30 ± 1.40	3.23 ± 1.36	0.277
Vegetables, dark	3.65 ± 1.38	3.65 ± 1.40	3.66 ± 1.33	0.972
Fruits	3.16 ± 2.88	2.86 ± 2.79	3.88 ± 2.95	< 0.001
Dairy	1.97 ± 1.64	1.93 ± 1.65	2.06 ± 1.63	0.069
Soybeans	3.07 ± 1.85	2.99 ± 1.85	3.26 ± 1.82	0.001
Fish & seafood	2.62 ± 1.58	2.53 ± 1.61	2.84 ± 1.51	< 0.001
Poultry	0.71 ± 1.03	0.64 ± 0.92	0.87 ± 1.22	< 0.001
Eggs	3.62 ± 1.86	3.42 ± 1.93	4.09 ± 1.59	< 0.001
Seeds and nuts	2.71 ± 2.10	2.58 ± 2.11	3.03 ± 2.04	< 0.001
Red meat	0.91 ± 1.93	1.01 ± 2.01	0.67 ± 1.70	< 0.001
Alcohol	3.95 ± 2.01	3.91 ± 2.03	4.04 ± 1.94	0.141

P values are from χ² tests for categorical variables and appropriate two-sample tests for continuous variables. Values are mean ± SD or n (%)

Abbreviations: CHEI Chinese Healthy Eating Index, n number

Baseline demographic characteristics

Compared with cognitively normal participants, those with cognitive impairment were older, more often female, had lower educational attainment, were more frequently widowed or living alone, had smaller calf circumference and higher body fat percentage, and had a lower prevalence of ever-smoking and lower physical activity; they also had a higher prevalence of stroke, hyperlipidemia, falls in the past year, and depressive symptoms (all P < 0.05) (Table 1).

Baseline sarcopenia-related characteristics

Sarcopenia status differed between cognitive status groups, with severe sarcopenia more prevalent among cognitively impaired participants (7.79% vs. 2.62%). Cognitively impaired participants had lower ASMI, HGS, GS, and SPPB scores, and longer 5CST times (all P < 0.05) (Table 2).

Baseline dietary quality

Cognitively impaired participants had lower modified CHEI scores and a different distribution of CHEI categories, with more classified as very low and fewer as high (all P < 0.05) (Table 3). At the component level, they scored lower for fruits, soybeans, fish/seafood, poultry, eggs, and seeds/nuts, while the red meat component score was higher in the cognitively impaired group (all P < 0.05) (Table 3).

Associations of sarcopenia with MoCA scores

As shown in Table 4, after full adjustment (Model 3), using severe sarcopenia as the reference group, MoCA scores were significantly higher in the general sarcopenia group (β = 1.25, P = 0.006) and the non-sarcopenia group (β = 2.17, P < 0.001). A change in MoCA score of 1.7 points was considered clinically significant [28]. Interpreted against this threshold, the association for non-sarcopenia versus severe sarcopenia (2.17 points) was clinically meaningful (moderate-to-large), whereas the difference for general sarcopenia (1.25 points) did not reach the threshold.

Table 4.

Associations of sarcopenia status and related indicators with MoCA scores

Variables	Model1		Model2		Model3
	β (95%CI)	P	β (95%CI)	P	β (95%CI)	P
Sarcopenia status
Severe sarcopenia	0.00 (Reference)		0.00 (Reference)		0.00 (Reference)
General sarcopenia	1.97 (0.94–3.00)	< 0.001	1.23 (0.34–2.12)	0.007	1.25 (0.35–2.14)	0.006
Non-sarcopenia	3.28 (2.48–4.07)	< 0.001	2.03 (1.34–2.71)	< 0.001	2.17 (1.46–2.89)	< 0.001
ASMI (kg/m²)	1.12 (0.92–1.33)	< 0.001	0.57 (0.33–0.80)	< 0.001	1.02 (0.68–1.35)	< 0.001
HGS (kg)	0.10 (0.08–0.13)	< 0.001	0.02 (-0.01–0.04)	0.198	0.02 (-0.01–0.04)	0.169
GS (m/s)	5.03 (4.28–5.78)	< 0.001	2.10 (1.41–2.80)	< 0.001	2.37 (1.66–3.09)	< 0.001
5CST (s)	-0.35 (-0.41–-0.29)	< 0.001	-0.17 (-0.22–-0.11)	< 0.001	-0.18 (-0.23–-0.12)	< 0.001
SPPB (score)	0.63 (0.47–0.78)	< 0.001	0.34 (0.21–0.47)	< 0.001	0.36 (0.23–0.49)	< 0.001

β coefficients with 95% confidence intervals (CI) and P values were calculated using multivariable linear regression models

Model 1: Unadjusted

Model 2: Adjusted for age, sex, and education level

Model 3: Model 2 plus living status, marital status, smoking status, alcohol consumption, number of chronic conditions, number of medications, BMI, and PSQI score

Abbreviations: ASMI Appendicular skeletal muscle mass index, HGS Handgrip strength, GS Gait speed, 5CST Five-times chair stand test, SPPB Short Physical Performance Battery, BMI Body mass index, PSQI Pittsburgh Sleep Quality Index

Associations of diet quality with MoCA scores

As shown in Table 5, after full adjustment (Model 3), the total modified CHEI score was positively associated with MoCA scores (β = 0.05, P < 0.001). This corresponds to a 0.50-point higher MoCA score per 10-point increase in modified CHEI, suggesting a small effect size relative to the 1.7-point threshold for clinical significance [28]. Among individual components, higher scores for fruits, soybeans, fish/seafood, poultry, eggs, seeds/nuts, and alcohol were positively associated with MoCA scores, whereas higher red meat scores were inversely associated (all P < 0.05); no significant associations were observed for the remaining components. Sensitivity analyses using a proxy-enhanced CHEI, which incorporated proxies for cooking oils, sodium, and added sugars, showed results consistent with the main analyses (Supplementary Table 3).

Table 5.

Associations between modified CHEI scores, level, and components with MoCA scores

Variables	Model1		Model2		Model3
	β (95% CI)	P	β (95% CI)	P	β (95% CI)	P
CHEI scores	0.09 (0.07–0.11)	< 0.001	0.05 (0.04–0.06)	< 0.001	0.05 (0.03–0.06)	< 0.001
CHEI level (4 level)
CHEI-Very low	0.00 (Reference)		0.00 (Reference)		0.00 (Reference)
CHEI-Low	1.10 (0.56–1.64)	< 0.001	0.52 (0.07–0.97)	0.025	0.52 (0.06–0.97)	0.027
CHEI-Moderate	1.40 (0.86–1.93)	< 0.001	0.75 (0.30–1.20)	0.001	0.72 (0.26–1.18)	0.002
CHEI-High	2.93 (2.39–3.47)	< 0.001	1.58 (1.12–2.04)	< 0.001	1.48 (1.01–1.95)	< 0.001
Total grains (score)	0.04 (-0.25–0.33)	0.806	-0.08 (-0.33–0.17)	0.521	-0.09 (-0.34–0.16)	0.473
Whole grains and mixed beans (score)	0.07 (-0.03–0.16)	0.180	0.05 (-0.03–0.13)	0.266	0.07 (-0.01–0.15)	0.090
Tubers (score)	0.00 (-0.21–0.22)	0.989	-0.06 (-0.24–0.12)	0.518	-0.06 (-0.24–0.12)	0.489
Total vegetables (score)	0.06 (-0.08–0.20)	0.402	-0.02 (-0.14–0.10)	0.729	-0.04 (-0.16–0.08)	0.499
Dark vegetables (score)	0.10 (-0.04–0.24)	0.157	0.01 (-0.11–0.13)	0.878	-0.00 (-0.12–0.11)	0.955
Fruits (score)	0.34 (0.27–0.40)	< 0.001	0.19 (0.14–0.25)	< 0.001	0.18 (0.12–0.24)	< 0.001
Dairy (score)	0.14 (0.02–0.26)	0.019	0.11 (0.01–0.21)	0.028	0.10 (-0.00–0.19)	0.057
Soybeans (score)	0.31 (0.20–0.41)	< 0.001	0.14 (0.05–0.22)	0.002	0.15 (0.06–0.23)	0.001
Fish and seafood (score)	0.42 (0.30–0.55)	< 0.001	0.13 (0.03–0.23)	0.012	0.13 (0.02–0.23)	0.018
Poultry (score)	0.64 (0.45–0.83)	< 0.001	0.40 (0.25–0.56)	< 0.001	0.39 (0.23–0.54)	< 0.001
Eggs (score)	0.54 (0.43–0.64)	< 0.001	0.29 (0.20–0.37)	< 0.001	0.26 (0.17–0.35)	< 0.001
Seeds and nuts (score)	0.32 (0.23–0.41)	< 0.001	0.16 (0.08–0.24)	< 0.001	0.16 (0.09–0.24)	< 0.001
Red meat (score)	-0.21 (-0.31–-0.11)	< 0.001	-0.15 (-0.23–-0.06)	< 0.001	-0.15 (-0.24–-0.07)	< 0.001
Alcohol (score)	0.02 (-0.08–0.11)	0.723	0.16 (0.07–0.25)	< 0.001	0.18 (0.08–0.28)	< 0.001

β coefficients with 95% confidence intervals (CI) and P values were calculated using multivariable linear regression models

Model 1: Unadjusted

Model 2: Adjusted for age, sex, and education level

Model 3: Model 2 plus living status, marital status, smoking status, alcohol consumption, number of chronic conditions, number of medications, BMI, and PSQI score

Abbreviations:CHEI Chinese Healthy Eating Index, BMI Body mass index, PSQI Pittsburgh Sleep Quality Index

Associations of diet quality and sarcopenia with MoCA scores

As shown in Table 6, statistically significant CHEI × sarcopenia interaction terms were observed after full adjustment (Model 3), consistent with differential associations between CHEI and MoCA across sarcopenia status. Δβ represents the reduction in the CHEI–MoCA slope for each sarcopenia group relative to the non-sarcopenia group (Δβ = slope_non-sarcopenia − slope_group), with positive values indicating a weaker association. The estimated Δβ values were 0.03 for overall sarcopenia, 0.03 for general sarcopenia, and 0.04 for severe sarcopenia (all P < 0.001), with the largest attenuation observed in the severe sarcopenia group. Results were robust in sensitivity analyses using the proxy-enhanced CHEI (Supplementary Table 4).

Table 6.

Effect modification of the association between continuous modified CHEI score and MoCA score by sarcopenia status

Interaction term (reference group = non-sarcopenia)	Model 1 Δβ (95% CI)	P	Model 2 Δβ (95% CI)	P	Model 3 Δβ (95% CI)	P
CHEI scores × Overall sarcopenia	0.06 (0.05 ~ 0.06)	< 0.001	0.03 (0.02 ~ 0.04)	< 0.001	0.03 (0.02 ~ 0.04)	< 0.001
CHEI scores × General sarcopenia	0.05 (0.04 ~ 0.06)	< 0.001	0.03 (0.02 ~ 0.04)	< 0.001	0.03 (0.02 ~ 0.04)	< 0.001
CHEI scores × Severe sarcopenia	0.08 (0.06 ~ 0.09)	< 0.001	0.04 (0.03 ~ 0.05)	< 0.001	0.04 (0.03 ~ 0.05)	< 0.001

Δβ denotes the reduction in the CHEI–MoCA slope in the specified sarcopenia group relative to the non-sarcopenia group (Δβ = slope_non-sarcopenia − slope_group). A positive Δβ indicates a weaker CHEI–MoCA association in that sarcopenia group. These interaction terms were estimated from separate models using different sarcopenia definitions (overall, general, and severe vs. non-sarcopenia)

Model 1: unadjusted

Model 2: adjusted for age, sex, and education

Model 3: Model 2 plus living status, marital status, smoking status, alcohol consumption, number of chronic conditions, number of medications, BMI, and PSQI score

Abbreviations:CHEI Chinese Healthy Eating Index, MoCA Montreal Cognitive Assessment, BMI Body mass index, PSQI Pittsburgh Sleep Quality Index

As illustrated in Fig. 2, the adjusted associations between continuous modified CHEI scores and predicted MoCA scores differed by sarcopenia status in the fully adjusted model (Model 3). Among participants without sarcopenia, higher CHEI scores were associated with a clear linear increase in predicted MoCA scores across the observed range. In contrast, the corresponding slopes were progressively attenuated among participants with general sarcopenia and were largely flat among those with severe sarcopenia. These visual patterns are consistent with the significant CHEI × sarcopenia interaction effects observed in Table 6.

Fig. 2.

Adjusted association between continuous modified CHEI score and MoCA score by sarcopenia status in the fully adjusted model (Model 3). Notes. Panels show results from separate subset models consistent with Table 6: A overall sarcopenia vs. non-sarcopenia, (B) general sarcopenia vs. non-sarcopenia, and (C) severe sarcopenia vs. non-sarcopenia. Lines represent adjusted predicted MoCA scores across the range of modified CHEI, and shaded areas indicate 95% confidence intervals. Model 3 adjusted for age, sex, education, living status, marital status, smoking status, alcohol consumption, number of chronic conditions, number of medications, BMI, and PSQI score. Because panels were estimated from different subset models, the absolute levels of predicted MoCA should not be compared across panels; interpretation focuses on within-panel slope differences. Abbreviations: MoCA Montreal Cognitive Assessment, CHEI Chinese Healthy Eating Index, BMI Body mass index, PSQI Pittsburgh Sleep Quality Index

As shown in Table 7, after full adjustment for potential confounders (Model 3), MoCA scores were generally higher among participants with general sarcopenia and non-sarcopenia than among those with severe sarcopenia across CHEI levels. Among participants without sarcopenia, MoCA scores increased progressively with higher CHEI categories, with differences ranging from 1.82 to 3.52 points compared with the reference group (severe sarcopenia with very low CHEI; all P ≤ 0.003). Notably, these differences exceed the 1.7-point threshold commonly considered clinically significant [28], indicating moderate-to-large clinically meaningful differences among participants without sarcopenia.

Table 7.

Joint associations of modified CHEI level and sarcopenia status with MoCA scores (4 × 3 categories)

Variables	Model 1 β (95% CI)	P	Model 2 β (95% CI)	P	Model 3 β (95% CI)	P
CHEI level + Sarcopenia status (4 × 3 level)
CHEI–Very low + Severe sarcopenia	0.00 (Reference)		0.00 (Reference)		0.00 (Reference)
CHEI–Low + Severe sarcopenia	3.07 (0.92–5.22)	0.005	1.40 (− 0.45–3.25)	0.139	1.27 (− 0.57–3.12)	0.177
CHEI–Moderate + Severe sarcopenia	2.58 (0.72–4.44)	0.007	1.44 (− 0.15–3.04)	0.077	1.37 (− 0.22–2.96)	0.091
CHEI–High + Severe sarcopenia	3.21 (0.90–5.52)	0.006	1.14 (− 0.85–3.13)	0.263	0.79 (− 1.23–2.80)	0.443
CHEI–Very low + General sarcopenia	3.72 (1.89–5.55)	< 0.001	1.68 (0.09–3.26)	0.038	1.82 (0.23–3.41)	0.025
CHEI–Low + General sarcopenia	3.75 (1.88–5.63)	< 0.001	2.28 (0.67–3.90)	0.006	2.21 (0.60–3.82)	0.007
CHEI–Moderate + General sarcopenia	3.92 (2.08–5.77)	< 0.001	2.01 (0.41–3.60)	0.014	1.95 (0.36–3.54)	0.017
CHEI–High + General sarcopenia	4.30 (2.26–6.34)	< 0.001	1.85 (0.08–3.61)	0.04	1.37 (− 0.41–3.16)	0.132
CHEI–Very low + Non-sarcopenia	3.88 (2.50–5.25)	< 0.001	1.77 (0.58–2.96)	0.004	1.82 (0.62–3.02)	0.003
CHEI–Low + Non-sarcopenia	4.80 (3.44–6.17)	< 0.001	2.17 (0.98–3.37)	< 0.001	2.22 (1.02–3.43)	< 0.001
CHEI–Moderate + Non-sarcopenia	5.26 (3.89–6.63)	< 0.001	2.60 (1.40–3.79)	< 0.001	2.59 (1.38–3.79)	< 0.001
CHEI–High + Non-sarcopenia	6.75 (5.38–8.11)	< 0.001	3.57 (2.37–4.77)	< 0.001	3.52 (2.31–4.74)	< 0.001

β coefficients (95% confidence intervals [CI]) and P values were estimated using multivariable linear regression models, with CHEI–Very low + Severe sarcopenia as the reference category

Model 1: unadjusted

Model 2: adjusted for age, sex, and education level

Model 3: Model 2 plus living status, marital status, smoking status, alcohol consumption, number of chronic conditions, number of medications, body mass index (BMI), and Pittsburgh Sleep Quality Index (PSQI) score

CHEI levels were categorized into four groups (Very low, Low, Moderate, High) according to the study-defined cutoffs

Abbreviations: CHEI Chinese Healthy Eating Index, MoCA Montreal Cognitive Assessment, BMI Body mass index, PSQI Pittsburgh Sleep Quality Index, CI Confidence interval

Among participants with general sarcopenia, higher MoCA scores were observed at selected CHEI levels, with significantly higher scores in the very low, low, and moderate CHEI categories compared with the reference group (β = 1.82–2.21; all P ≤ 0.025); however, no clear monotonic trend across CHEI levels was evident, and no significant difference was observed in the high CHEI category. In contrast, among participants with severe sarcopenia, no significant differences in MoCA scores were observed across CHEI levels (all P > 0.05). Results were robust in sensitivity analyses using alternative categorizations of sarcopenia status and modified CHEI (Supplementary Table 2).

Associations of sarcopenia and diet quality with MoCA scores in subgroup analyses

Figure 3 presents subgroup analyses for variables with statistically significant interactions (P for interaction < 0.05). As shown in Fig. 3A, participants without sarcopenia had higher MoCA scores than those with sarcopenia in the overall population (β = 1.52, P < 0.001). This association varied across subgroups defined by age, BMI, modified CHEI, and educational level. The overall difference (β = 1.52) was close to the 1.7-point clinical significance threshold [28], indicating a small-to-moderate difference overall, with larger differences observed in specific subgroups. Specifically, the difference in MoCA scores was evident among participants aged ≥ 70 years, those with higher BMI, and those with higher diet quality, whereas no significant difference was observed in younger participants or those with higher educational attainment.

Fig. 3.

Subgroup analyses of associations between (A) sarcopenia status and MoCA scores and (B) modified CHEI level and MoCA scores. Notes:β coefficients with 95% confidence intervals (CI) and P values were estimated using multivariable linear regression models. P for interaction indicates the statistical significance of the interaction between the exposure and subgroup variables. Models were adjusted for age, sex, education level, marital status, living status, smoking status, alcohol consumption, BMI, number of chronic conditions, number of medications, and PSQI score. Abbreviations:MoCA Montreal Cognitive Assessment, CHEI Chinese Healthy Eating Index, BMI Body mass index, PSQI Pittsburgh Sleep Quality Index, SE Standard error

As shown in Fig. 3B, participants with higher modified CHEI scores had higher MoCA scores than those with lower CHEI scores in the overall population (β = 1.22, P < 0.001). The overall association for modified CHEI (β = 1.22) suggests a small difference relative to the 1.7-point threshold [28]. This association differed by living status, cardiovascular disease history, and sarcopenia status. The positive association between CHEI and MoCA was observed among participants living with others, those with or without cardiovascular disease, and those without sarcopenia, but was not evident among participants living alone or those with general or severe sarcopenia. Subgroup analyses without statistically significant interactions are presented in Supplementary Tables 5 and 6.

Mediation analysis of diet quality, sarcopenia, and MoCA

Figure 4 presents the mediation analyses examining the interrelationships among sarcopenia status (overall, general, and severe), diet quality (modified CHEI), and MoCA scores. In all models, no statistically significant indirect effects were observed, either for modified CHEI mediating the association between sarcopenia status and MoCA scores or for sarcopenia status mediating the association between modified CHEI and MoCA scores (all P > 0.05). These findings were consistent across different sarcopenia definitions, with detailed results provided in Supplementary Table 7.

Fig. 4.

Mediation analyses of sarcopenia status, diet quality (modified CHEI), and cognitive function (MoCA). Notes: Numbers on arrows indicate regression coefficients (β). Statistical significance: * P < 0.05; ** P < 0.01; *** P < 0.001. Models were adjusted for age, sex, and education (reference group: non-sarcopenia). Panels A–B: total sarcopenia vs. non-sarcopenia; C–D: general sarcopenia vs. non-sarcopenia; E–F: severe sarcopenia vs. non-sarcopenia. A, C, E: sarcopenia status → modified CHEI scores → MoCA scores; B, D, F: modified CHEI scores → sarcopenia status → MoCA scores.). Abbreviations: MoCA Montreal Cognitive Assessment, CHEI Chinese Healthy Eating Index

Discussion

Principal findings

This study represents the first systematic investigation of the independent, interactive, and mediating effects of diet quality and sarcopenia on cognitive function in Chinese community-dwelling older people. Our findings indicated that sarcopenia was significantly and negatively associated with cognitive function, and this association appeared stronger than the positive association between diet quality and cognition. Sarcopenia status also modified the association between diet quality and cognitive function, indicating that the diet–cognition association differed by physical and functional status. Collectively, these findings add population-based evidence on the joint relationships among sarcopenia, diet quality, and cognition, and may help inform risk stratification and more targeted strategies to support cognitive health in older adults.

Comparative influence of sarcopenia and diet quality on cognition

Our results showed that individuals with severe sarcopenia had MoCA scores approximately 2.17 points lower than those without sarcopenia, whereas a 10-point increase in the modified CHEI was associated with an estimated 0.5-point increase in MoCA scores. Although these estimates are not directly comparable, they suggest that the association between sarcopenia status and cognitive function may be more pronounced than that observed for diet quality.

As a modifiable lifestyle factor, diet is thought to influence cognitive function through nutritional intake and metabolic pathways [29]; however, observed diet–cognition associations may be attenuated by short-term dietary variability and measurement error. In contrast, sarcopenia may reflect a broader phenotype of aging-related physiological decline [1] and may be linked to poorer cognitive function through mechanisms such as chronic inflammation and reduced neurotrophic support [7]. In the Chinese context, differences in body composition and physical activity patterns compared with many Western populations—such as lower BMI and greater reliance on low-intensity aerobic activity rather than structured resistance training—may further increase vulnerability to muscle loss and its cognitive correlates [30, 31].

Taken together, these findings suggest that sarcopenia may serve as a particularly informative indicator of cognitive vulnerability in older adults, while diet quality remains an important, potentially modifiable factor.

Moderating role of sarcopenia in the diet–cognition relationship

Table 6 presents the effect modification analysis, showing that the CHEI × sarcopenia interaction terms remained statistically significant after adjustment for covariates (all P < 0.001). This finding indicates substantial effect modification, suggesting that the association between CHEI and MoCA scores differs across sarcopenia states. Specifically, the interaction terms quantify differences in the CHEI–MoCA slopes between sarcopenia categories, beyond the baseline between-group differences in MoCA at the reference level of CHEI captured by the sarcopenia main-effect terms. Consistent with the pattern of Δβ, the CHEI–MoCA association was weaker among participants with sarcopenia than among those without, with the most pronounced attenuation observed in the severe sarcopenia group.

Figure 2 further illustrates this divergence. A clear positive association between higher CHEI scores and predicted MoCA scores was observed among participants without sarcopenia, whereas this trend was attenuated in those with general sarcopenia and became nearly flat in the severe sarcopenia group. These findings were corroborated by the joint association analysis in Table 7. Within the severe sarcopenia subgroup, MoCA scores across CHEI categories did not differ significantly from the reference group (Very low CHEI + severe sarcopenia). In contrast, among non-sarcopenic participants, β coefficients increased progressively with higher CHEI levels. Together, these results suggest that the positive association between diet quality and cognitive function is primarily evident among individuals without sarcopenia, with those with general sarcopenia showing an intermediate pattern.

From a clinical perspective, these findings imply that for an equivalent increase in CHEI, the magnitude of association with MoCA is likely smaller in participants with sarcopenia—particularly those with severe sarcopenia—than in those without. The null association observed in the severe sarcopenia group should be interpreted with caution, as it may reflect a higher comorbidity burden, reduced physiological reserve, residual confounding, or a potential floor effect due to compressed MoCA variability in this frail subgroup. In addition, the relatively smaller sample size of the severe sarcopenia group may have limited statistical power to detect modest associations. Mechanistically, more severe sarcopenia is often accompanied by heightened systemic inflammation and metabolic dysregulation, which may biologically blunt the neuroprotective benefits typically associated with high-quality dietary patterns [1]. Finally, bidirectional mediation analyses revealed no significant indirect effects. Given the cross-sectional nature of the study, these findings collectively support the interpretation of sarcopenia as an effect modifier rather than a mediator in the association between diet quality and cognitive function.

Associations of muscle indices and cognitive function

This study observed an inverse association between sarcopenia and cognitive function, consistent with previous reports. Importantly, physical performance measures such as GS, SPPB, and 5CST showed stronger associations with cognition than muscle mass or HGS. This discrepancy likely reflects differences in measurement dimensions: muscle mass and HGS represent isolated measures of muscle quantity or strength, whereas performance tests reflect integrated neuromuscular function and engage multiple cognitive domains, including sensorimotor integration, executive function, attention, and spatial perception [32, 33]. These findings suggest that physical performance may serve as more sensitive markers for identifying cognitive risk.

Dietary quality and specific food components related to cognition

This study found that higher modified CHEI scores were positively associated with cognitive function, consistent with existing evidence from healthy dietary patterns such as the Mediterranean diet [11] and DASH diet [13]. This direction of association is also broadly consistent with findings from studies conducted in China and other East Asian populations. For example, analyses of Chinese older adults using the Chinese Healthy Eating Index have reported that higher diet quality is associated with better cognitive performance measured by the Mini-Mental State Examination (MMSE), with additional evidence suggesting potential pathways through psychological well-being and depressive symptoms [34]. Similarly, evidence from the China Health and Nutrition Survey using 24-hour dietary recalls and an instrumental variable approach suggested a positive association of diet quality with cognitive performance assessed by Telephone Interview for Cognitive Status–modified (TICS-m), supporting the robustness of the diet–cognition association in Chinese settings [35].

Greater intake of recommended food groups—including fruits, legumes, fish, poultry, eggs, nuts, and seeds—was closely related to better cognitive performance. From a dietary-pattern perspective, these findings support that adherence to balanced and diverse dietary recommendations is linked to better cognitive health across cultural settings, despite differences in specific food compositions. Various nutritional mechanisms may explain these correlations: protein from poultry and eggs may support neuronal integrity; omega-3 fatty acids from fish may facilitate neurotransmission; antioxidant nutrients like polyphenols and vitamin E from nuts and fruits may help alleviate oxidative stress; and dietary fibre may affect brain function by modulating the gut–brain axis [36]. In addition, limiting alcohol intake was associated with superior cognitive function, in line with the known neurotoxic effects of alcohol [37].

Particularly noteworthy is that red meat intake was positively associated with cognitive performance, in contrast to some international findings [38]. This discrepancy may relate to dietary patterns: in Chinese populations with grain-based diets and relatively low red meat consumption [39], moderate consumption may provide essential nutrients for brain health, such as high-quality protein, iron, zinc, and vitamin B12 [40]. Moreover, individuals with higher cognitive levels frequently possess better socioeconomic conditions, facilitating access to higher-quality red meat. Thus, the observed association may partly reflect socioeconomic and lifestyle factors rather than a direct effect of red meat itself, and may also be subject to residual confounding and reverse causality, highlighting the importance of considering cultural context, intake levels, and food quality in diet–health research.

Population subgroups and effect modifiers

To further elucidate population-specific differences, subgroup analyses were conducted. The results indicated that the inverse association between sarcopenia and cognitive function was more pronounced in certain vulnerable groups. Among older adults (≥ 70 years), the coexistence of sarcopenia with declining physiological reserves may accelerate cognitive decline [41, 42]. In obese individuals (BMI ≥ 24 kg/m²), the coexistence of obesity and sarcopenia, referred to as “sarcopenic obesity,” may exert synergistic detrimental effects on cognition through mechanisms such as chronic inflammation and insulin resistance [43]. In populations with lower educational attainment (middle school or below), limited cognitive reserve diminishes the brain’s compensatory capacity against pathological insults [44]. Notably, the inverse association remained significant and was even more evident among individuals with higher diet quality. This suggests that although better diet quality may be associated with improved cognitive performance, its potential protective role may not be sufficient to fully counteract the adverse impact of sarcopenia. Moreover, individuals with higher diet quality generally have better baseline cognitive levels, which may render the detrimental effects of sarcopenia more apparent, thereby amplifying the between-group differences.

In addition, the positive association between diet quality and cognitive function was also particularly pronounced in certain populations. Among individuals living with their children, household life not only enhances social support and dietary supervision but also promotes mental health and provides additional cognitive stimulation through positive family interactions [45]. In individuals with a history of cardiovascular disease, dietary improvement may indirectly benefit cognition through cardiovascular mechanisms, including optimized lipid metabolism, improved endothelial function, and reduced blood pressure [46]. Collectively, these findings suggest that social support and comorbid conditions may serve as important moderators of the diet–cognition relationship.

Implications for individualized and community-level interventions

The findings of this study provide important evidence for cognitive health management, suggesting the establishment of an integrated framework encompassing individualized interventions and community-level implementation. At the individual level, precision intervention strategies based on sarcopenia stratification may be advisable. For non-sarcopenic individuals, prioritizing improvements in diet quality is recommended, including increased intake of fruits, deep-sea fish, poultry, eggs, nuts, legumes, and whole grains, alongside moderate consumption of high-quality red meat and strict control of alcohol intake. For individuals with general sarcopenia, comprehensive strategies are advised to concurrently manage sarcopenia and diet quality, encompassing dietary optimization, resistance training, and protein supplementation, to maximize potential cognitive benefits. For individuals with severe sarcopenia, where the pathological burden is substantial and the efficacy of dietary interventions alone is limited, priority should be given to targeting sarcopenia directly, with nutritional support serving as an adjunctive measure.

At the community public health level, the establishment of standardized “screening–stratification–intervention–follow-up” pathways is recommended, incorporating assessments of physical function and dietary quality into routine health examinations to enable precise identification and enhanced management of high-risk populations (e.g., older adults, individuals with obesity, low educational attainment, or cardiovascular disease). Given the various needs of older adults across different risk categories, we advocate for multidisciplinary collaborative platforms—such as “nutrition-exercise integrated clinics”—that combine the expertise of dietitians, rehabilitation therapists, and geriatricians for comprehensive evaluations, targeted interventions, and long-term follow-up services.

Strengths and limitations

This study has several strengths. First, it represents one of the early population-based investigations examining the relationships among sarcopenia, diet quality, and cognitive function in Chinese community-dwelling older adults. By stratifying participants according to sarcopenia status, the study highlighted severity-specific heterogeneity in the association between diet quality and cognition, providing population-level evidence that may inform future epidemiological and interventional research. Second, the study employed the modified CHEI, which is tailored to Chinese dietary patterns and food availability, thereby enhancing the cultural relevance and interpretability of the dietary assessment [19]. Finally, the relatively large sample size (n = 2,572) provided adequate statistical power to detect subgroup differences and interaction effects, strengthening the robustness of the findings.

However, several limitations should be acknowledged. First, given the cross-sectional design, temporality between exposures and cognitive outcomes cannot be established; thus, the observed associations should not be interpreted as causal, and reverse causation remains a key concern. Cognitive decline may lead to subsequent changes in dietary behaviors, including reduced appetite, altered food preferences, and difficulty with meal planning and preparation, which may influence the measured diet quality [47]. Cognitive impairment may also contribute to reduced physical activity and functional decline, thereby affecting muscle status [48]. Although multivariable adjustment was applied, residual confounding and reverse causation cannot be fully excluded. Future longitudinal studies are needed to clarify directionality and potential mechanisms, and randomized controlled trials, where feasible, would further strengthen causal inference regarding the effects of dietary interventions on cognitive and muscle health.

Second, dietary intake was assessed using the FFQ-25, which does not fully capture all components of the standard 17-component CHEI, particularly discretionary items such as cooking oils, sodium, and added sugars. Therefore, we constructed a modified 14-component CHEI by excluding these components. This modification may under-represent these diet-quality domains; accordingly, the absolute CHEI scores should be interpreted as reflecting diet quality within the domains covered by the FFQ-25, and their comparability with scores from studies using the full CHEI or other dietary indices may be reduced. In addition, because this modified CHEI has not been formally validated as an independent instrument, cross-study comparisons and the generalizability of findings based on this score may be limited. To examine the robustness of our results, we conducted sensitivity analyses using a proxy-enhanced CHEI based on available information, and the results were consistent with the primary analyses (Supplementary Tables 3–4).

Third, both the FFQ-25 and MoCA serve as screening instruments rather than diagnostic gold standards. The FFQ-25 is susceptible to recall and reporting biases and may not fully capture long-term dietary variability [49], while MoCA scores are sensitive to educational attainment [22]. These measurement errors introduce the potential for misclassification of both exposure and outcome, which might have biased the observed associations toward the null. Future studies incorporating repeated 24-hour dietary recalls, food diaries, or comprehensive neuropsychological assessments may improve exposure and outcome classification [29].

Fourth, the extensive subgroup and interaction analyses increase the risk of Type I error due to multiple testing. These analyses were prespecified as secondary/exploratory, whereas the primary analyses were prespecified; therefore, we did not apply formal multiplicity adjustments (e.g., Bonferroni correction or false discovery rate control) to the exploratory results to avoid overly conservative inference. Accordingly, subgroup- and interaction-specific findings should be interpreted cautiously and warrant confirmation in independent cohorts. In addition, the study population was limited to community-dwelling older adults in China, which may restrict the generalizability of our findings to institutionalized populations or those from different sociocultural backgrounds. Replication in independent cohorts, particularly longitudinal and multicenter studies, is warranted.

Conclusion

In conclusion, this study suggests that sarcopenia was inversely associated with cognitive function and may modify the association between diet quality and cognition. Higher diet quality was associated with better cognitive performance, but this association was attenuated among individuals with more severe sarcopenia. Given the cross-sectional design, these findings should be interpreted as associations rather than evidence of causality, and the temporal direction cannot be determined. Future longitudinal and intervention studies are warranted to further examine these relationships and to assess whether combined approaches targeting diet quality and muscle health may benefit cognitive aging.

Abbreviation

CI

Cognitive impairment

MCI

Mild cognitive impairment

AD

Alzheimer’s disease

MIND

Mediterranean–DASH Intervention for Neurodegenerative Delay diet

DASH

Dietary Approaches to Stop Hypertension

CHEI

Chinese Healthy Eating Index

CDG-2016

2016 Chinese Dietary Guideline

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

MoCA

Montreal Cognitive Assessment

BMI

Body mass index

PBF

Percentage of body fat

PSQI

Pittsburgh Sleep Quality Index

AWGS 2019

Asian Working Group for Sarcopenia 2019

ASMI

Appendicular skeletal muscle mass index

HGS

Handgrip strength

GS

Gait speed

5CST

Five-times chair stand test

SPPB

Short Physical Performance Battery

FFQ-25

25-item Food Frequency Questionnaire

SD

Standard deviation

MAR

Missing at random

FIML

Full information maximum likelihood

MMSE

Mini-Mental State Examination

TICS-m

Telephone Interview for Cognitive Status–modified

ChiCTR

China Clinical Trial Registry

Authors’ contributions

YR, QL and XH contributed equally to this study. All authors contributed to the study conception and design. YR, QL and XH drafted the original manuscript. LM and YS were involved in data analysis. NC supervised the project and provided critical review. All authors approved the final version for publication.

Funding

This work was supported by the National Natural Science Foundation of China “Mechanism of BCAT2-mediated branched-chain amino acid metabolism inhibiting ferroptosis in the progression of sarcopenia alleviated by intestinal P. merdae” (No. 82372575); the Shanghai “Rising Stars of Medical Talent” Youth Development Program for Outstanding Youth Medical Talents (SHWSRS2025-71); the Research Projects of Shanghai Municipal Health Commission “Research on early screening and intervention strategies for cognitive impairment in community-dwelling older people with sarcopenia” (No.20254Z0006); and Chongming Hospital Affiliated to Shanghai University of Medicine and Health Sciences “Community sarcopenia based on spatiotemporal biomechanical analysis of multimodal data” (No.XJJJ2503-1).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Chongming Hospital, Affiliated to Shanghai University of Medicine and Health Sciences (Approval No. CMEC-2025-KT-45). It is registered at the China Clinical Trials Registry (ChiCTR), with registration number ChiCTR2500108893. Detailed registration information is available at https://www.chictr.org.cn/bin/project/edit?pid=286089. All participants provided written informed consent prior to enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.