This study used the CHARLS dataset, spanning the 2011, 2013, and 2015 waves. CHARLS implemented a multistage, stratified, probability-proportionate-to-size sampling design to recruit the individuals from 450 villages or urban communities across 28 provinces of China, as described in previous research.^22,26) The present analysis involved participants aged 45 years and above.

This study was conducted in accordance with the principles of the Declaration of Helsinki. Ethical approval for all the CHARLS waves was granted from the Institutional Review Board (IRB) at Peking University (IRB00001052-11015). Furthermore, written informed consents were obtained from all participants who were involved in this study. In addition, this study has been approved by the Institutional Review Board of Beijing Friendship Hospital, Capital Medical University for exemption from the ethical statement for conducting database research (2025-P2-503).

First, we enrolled 17,857 participants at the baseline in 2011. Then, we excluded 751 participants who were under 45 years of age, as well as 2,119 participants who did not participate in the cognitive assessment and 1,776 participants who did not report sarcopenia at baseline. Additionally, a total of 14,529 participants were included in the study, all of whom underwent a comprehensive cognitive function and sarcopenia assessment. In addition, we excluded 607 adults who were diagnosed with sarcopenia at baseline, 155 who were lost to follow-up or had passed away, 5,912 who failed to complete sarcopenia assessment during follow-up, and 3,199 who had missing lifestyle assessment at wave 2 and had missing data on other confounders at baseline. This led to a final sample of 7,548 indi-viduals for the final analysis of the associations between cognitive status and sarcopenia risk (Fig. 1).

As described in previous studies, the CHARLS conducted face-to-face assessments to evaluate participants' cognitive function using the Mini-Mental State Examination (MMSE), which covers four domains: memory, computation, orientation, and drawing.²²⁾ Memory dimensions involve immediate and delayed recalls. Each participant's memory ability was evaluated by counting the number of words recalled immediately (immediate word recall) and after the participant completed the survey of depression scale, computation, and drawing tests (delayed word recall) following the random presentation of 10 unrelated Chinese words. The total memory score was 20 points, with one point assigned for each word that was recalled during either immediate or delayed recall task. The computation task required the participants to subtract 7 from 100 five times in succession, with one point awarded for each correct response. The total score of the computation was five points. Orientation task was evaluated by items identifying year, month, day, and the day of the week. The maximum total score of the orientation dimension is 4 points, with one point given for each item. The interviewer assessed the drawing task by presenting a picture of two intersecting pentagons to determine whether the participant could accurately observe and reproduce the image. One point was awarded if correct. The total cognitive score was calculated by summing the scores from memory (20 points), computation (5 points), orientation (4 points), and drawing (1 point), with a maximum score of 30 points and a higher score indicating a better cognitive function. In our study, MCI was defined as being at least 1 standard deviation (SD) below the age standard according to aging-associated cognitive decline (AACD).²⁷⁾ Participants were stratified into groups according to 5-year age intervals. Within each age group, individuals who met the AACD criteria were classified as having MCI.²²⁾

Sarcopenia was operationalized based on the Asian Working Group for Sarcopenia (AWGS) 2019 diagnostic framework, including evaluations of physical performance, an equation-estimated appendicular skeletal muscle mass (ASM), and muscle strength.²⁸⁾ Handgrip strength was measured in kilograms (kg) using a Yuejian WL-1000 dynamometer (Nantong Yuejian Physical Measurement Instrument Co., Ltd., Nantong, China) for both dominant and non-dominant hands. Two measurements were recorded for each hand, with the dynamometer held at a right angle (90°), and participants were instructed to exert maximal force while squeezing. The mean value of the maximum strength data was used for analysis. If measurement of one hand was impossible for any reason, the highest score from the other hand was recorded. According to the AWGS 2019 criteria, a low handgrip strength is defined as less than 28 kg for men and less than 18 kg for women. Because dual-energy X-ray absorptiometry (DXA)/bioelectrical impedance analysis (BIA) measurements were not available in the CHARLS dataset, ASM was estimated using a previously developed anthropometric equation for Chinese adults:

ASM = 0.193 × weight (kg) + 0.107 × height (cm) − 4.157 × sex − 0.037 × age (years) − 2.631, where sex was coded as 1 for men and 2 for women. This equation has been cross-validated against DXA-measured ASM in Chinese adults and has demonstrated a satisfactory predictive performance.^22,29)

The subjects' weight and height were measured using an Omron HN-286 scale (Omron Co., Kyoto, Japan) and a Seca 213 height meter (Seca GmbH, Hamburg, Germany), respectively. Because the CHARLS does not include the DXA/BIA measurements, ASM was estimated using a validated anthropometric equation, and a sarcopenia diagnosis was constructed by operationalizing the AWGS 2019 diagnostic framework with estimated ASM, consistent with previous CHARLS-based studies.^22,30) Previous validation work had reported a strong correlation between the equation-estimated ASM and the DXA-derived ASM in the Chinese populations.³¹⁾ Nevertheless, because ASM was not directly measured in the CHARLS, some degree of misclassification is possible. To define low muscle mass, the sex-specific lowest 20th percentile of height-adjusted muscle mass (ASM/height²) within the study cohort was used as the threshold. These values are <5.29 kg/m² for women and <6.99 kg/m² for men, according to the findings of previous work.³²⁾

During the physical performance evaluation, the five times sit-to-stand test (FTSST), the Short Physical Performance Battery (SPPB), and gait speed were assessed. To evaluate gait speed, participants were instructed to walk a distance of 2.5 meters at their usual pace, with completion times recorded over two trials. The FTSST measured the time required for participants to rise consecutively from a seated position on a chair with a 47-cm-high seat.²²⁾ The Short Physical Performance Battery (SPPB) consists of three components: a standing balance test, a 2.5-m gait-speed test, and a five-times sit-to-stand test. For the balance test, participants performed side-by-side, semi-tandem, and tandem stances, each held for up to 10 seconds. The balance component was scored from 0 to 4 points according to performance across the three stances. Gait speed was assessed over 2.5 m, with completion times recorded over two trials. Lower-extremity strength was assessed by the time required to rise from a chair five times without using the arms (0–4 points). A cumulative score of 12 was assigned, with up to 4 points awarded for each test. According to the AWGS 2019 guidelines, inadequate physical performance was defined as a gait speed below 1.0 m/s, a FTSST time of 12 seconds or more, or an SPPB score of less than 9.

In sum, sarcopenia is characterized by an accompanying reduction in muscle mass, strength, or physical performance. Consequently, participants were categorized into two groups: non-sarcopenia and sarcopenia.

According to previous studies,^26,33,34) this study assessed five lifestyle factors: physical activity, alcohol consumption, smoking, BMI, and sleep duration. Each factor was assigned a score of 1 for healthy behavior and 0 for unhealthy behavior. Engaging in vigorous (e.g., heavy lifting, cycling with a large load, aerobics) or moderate (e.g., regular cycling, tai chi, brisk jogging) activities for a minimum of 30 minutes on three separate occasions per week was established as physical activity. Scores of 1 were allocated to participants who met this criterion, while scores of 0 were assigned to those who did not. Smoking status was categorized into three categories: current smoker, former smoker, and never smoker. Individuals who had never smoked were allocated a score of 1, while those who had smoked were assigned a score of 0. Participants who had never consumed alcohol received a score of 1, while those who had consumed alcohol received a score of 0. Alcohol consumption was classified as current, former, or never consuming. A BMI of 18.5–24.9 kg/m² was regarded as healthful. Participants were assigned a sleep score of 1 if they reported sleeping for a minimum of 6 hours each night, and 0 otherwise. An overall lifestyle score was determined, with a range of 0 to 5, where a higher score meant a strict adherence to healthful lifestyle practices. Five individual factors were combined to calculate this total score. Participants were divided into three categories based on their overall score: favorable lifestyle (4–5), intermediate lifestyle (2–3), and unfavorable lifestyle (0–1).^35,36) To elucidate the assumed causative framework, we developed a directed acyclic graph demonstrating the associations among baseline cognitive impairment, lifestyle factors, measured covariates and occurrence of sarcopenia (Supplementary Fig. S1).

The covariates were assessed in accordance with the findings of earlier studies.^20,22) Education levels (illiterate, elementary school, middle school, high school, university degree or higher), marital status (married, separated, and unmarried/divorced/widowed), household incomes (<1,000, 1,000–4,999, 5,000–9,999, 10,000–19,999, and ≥20,000 Chinese yuan), smoking status (never, current, former), alcohol consumption (never, current, former), sleep duration, and BMI were all potential confounders. The sleep duration was self-reported by participants with the question “In the past month, how many hours on average did you actually sleep each night? (This may be shorter than the time you spent lying in bed.)” BMI was calculated from height and weight. Height was measured using a Seca 213 stadiometer, and weight was measured using an Omron HN-286 scale. In addition, frailty was considered as a potential confounder of the association between MCI and sarcopenia. We therefore adjusted for baseline frailty using a 29-item frailty index (FI) previously developed and validated in the CHARLS. The FI was constructed by summing the deficits across 29 items and dividing by the total number of items,³⁷⁾ with a higher score indicating greater frailty, and it was included as a continuous covariate in the multivariable models.

Data analyses were conducted using SAS software version 9.4 (SAS Institute, Cary, NC, USA) and R programming language version 4.4.3 (The R Foundation, Vienna, Austria). Graphs were produced using the ggplot2 package. Continuous variables were reported as means with standard deviations (SD) for normally distributed data or as medians with interquartile ranges (IQR) for non-normally distributed data. Categorical variables were presented as counts and percentages. The t-test or Wilcoxon rank sum test was utilized to evaluate the differences between the MCI group and the non-MCI group. The chi-square test was conducted to assess the differences among groups of categorical variables. The Kaplan-Meier method was used to determine the 4-year cumulative incidence of sarcopenia, while a Cox proportional hazards model and hazard ratios (HR) were employed to evaluate the relationship between MCI and sarcopenia. Two models were developed to assess the relationship between the status of MCI and the risk of developing sarcopenia. Model 1 is adjusted for age and sex; Model 2 is further adjusted for residential regions, education levels, marital status, mean household incomes, smoking status, alcohol consumption, sleep duration, and BMI. In sensitivity analyses, the participants who developed sarcopenia within the first 2 years of follow-up and those with possible preclinical sarcopenia at baseline, defined as low handgrip strength or slow gait speed, were excluded to reduce the potential for reverse causality. In addition, to account for selection bias due to participants' exclusion, we performed an inverse probability weight analysis to further assess our findings. A competing risk model was developed, considering loss to follow-up and death as competing events to mitigate potential biases in outcome assessment.

To further estimate the relationship between baseline MMSE scores and future sarcopenia occurrence in 2013 and 2015, we employed a generalized estimating equation (GEE) model with a binomial distribution and a logit link. Repeated measures of sarcopenia within participants were accounted for by specifying an exchangeable correlation structure. The model adjusted for potential confounders, including age, sex, comorbidities, and baseline sarcopenia status. We estimated the odds ratio (OR) with 95% confidence interval (CI) of the association between baseline MMSE scores and sarcopenia. We assessed the interaction between MMSE scores and time in the GEE model, but the interaction was not statistically significant and subsequently excluded from the model.

We performed the latent mixture modeling to identify subgroups with similar underlying trajectories of cognitive function. To achieve this, we evaluated the models with varying numbers of classes and functional forms (including intercept-only, linear, and quadratic slopes). Model fit was assessed using the Bayesian Information Criterion (BIC). Initially, we specified a model with five classes, all in quadratic form, and subsequently compared its BIC values with those of models containing four, three, two, and one class. The model with three classes demonstrated the best fit. We then further refined our analysis by comparing the BIC values of models with different functional specifications. Ultimately, the optimal model consisted of two classes with linear terms and one class with a quadratic term. We calculated the mean posterior probability of final group membership to evaluate the model's discriminative ability. Subgroup analysis was conducted to explore the associations between long-term cognitive function patterns (low, moderate, and normal cognitive function) and the risk of developing sarcopenia. Additionally, we assessed the relationship between a composite healthy lifestyle and the overall risk of sarcopenia within subgroups defined by long-term cognitive function patterns using a logistic regression model. Further, we conducted sensitivity analyses using alternative lifestyle definitions. First, we reconstructed the lifestyle score after excluding BMI to reduce the potential correlation between BMI and equation-estimated ASM. Second, we constructed a regression coefficient–weighted lifestyle score using the non-BMI lifestyle components. The weighted scores were categorized into the tertiles and analyzed using the same covariate adjustments as in the main analysis. Third, we examined the associations between each non-BMI lifestyle component and incident sarcopenia. All statistical tests were two-sided, and a p-value <0.05 was considered statistically significant.

The mean age of 7,548 included participants was 63.5±6.6 years and 55.4% of them (n=4,178) were males. At baseline assessment, 1,142 (15.1%) participants reported MCI. Among those with MCI, 441 (38.6%) were male and 701 (61.4%) were female. Participants who reported MCI were more likely to live in rural area, have a lower education level and a lower average household income (Table 1). The median follow-up time of this study was 4.0 years (IQR, 3.9–4.0) (Supplementary Table S1).

During the follow-up with a total of 30,030.9 person-years, 502 cases of incident sarcopenia were identified. The 4-year cumulative incidence of sarcopenia was 10.5% (95% CI, 8.5–12.4) in the MCI group compared to 6.4% (95% CI, 5.7–7.1) in the no-MCI group (p<0.001). The baseline MCI (HR=1.57; 95% CI, 1.25–1.96) was significantly associated with an increased risk of developing sarcopenia after multivariable adjustments (p<0.001, Table 2). Among middle-aged participants (n=2,465), 54 cases of incident sarcopenia occurred during 9,813.0 person-years of follow-up. The 4-year cumulative incidence of sarcopenia was 10.1% (95% CI, 4.0–16.3) in the MCI group compared to 3.6% (95% CI, 2.2–5.0) in the non-MCI group. Middle-aged participants with baseline MCI had a 2.47 (95% CI, 1.26–4.84) fold higher risk of incident sarcopenia compared to those without baseline MCI (p=0.0085, Table 2). In older participants, 448 cases of incident sarcopenia were observed during 20,217.9 person-years of follow-up. The 4-year cumulative incidence of sarcopenia was 13.8% (95% CI, 11.2–16.6) in the MCI group versus 8.6% (95% CI, 7.6–9.6) in the non-MCI group. For older participants, baseline MCI significantly elevated the risk of sarcopenia (HR=1.47; 95% CI, 1.16–1.87; p=0.0015) (Table 2). Furthermore, the results remained consistent in sensitivity analyses excluding early incident sarcopenia cases, participants with baseline or early-onset frailty, and those with possible preclinical sarcopenia at baseline. In addition, the inverse-probability weighting analysis also supported the robustness of our findings (Supplementary Table S2).

Table 3 summarizes the results of the association between total score of MMSE, including its various dimensions related to cognitive function, and incident sarcopenia using GEE model. As shown in Table 3, lower scores in the total MMSE score, orientation, memory, and computation were all significantly associated with an increased risk of incident sarcopenia. Specifically, the odds ratios for developing sarcopenia were 0.93 (95% CI, 0.90–0.95), 0.83 (95% CI, 0.76–0.91), 0.91 (95% CI, 0.88–0.94), and 0.88 (95% CI, 0.82–0.94) per 1-SD increase in the scores for MMSE, orientation, memory, and computation, respectively (Table 3).

Three discrete trajectories in cognitive scores during the 4-year span were identified (Fig. 2): 1,254 participants (16.6%) maintained a low level of cognitive function throughout 4 years (low cognitive function group; MMSE score median range, 6.0 to 7.0); 3,439 (45.6%) had a stable medium cognitive function (medium cognitive function group; MMSE score median range, 12.0 to 13.0); and 2,855 (37.8%) maintained a high level of cognitive function during 4 years (normal cognitive function group, MMSE score median range, 17.0 to 18.0). Mean±SD probabilities of each each individual belonged to their final group were 0.91±0.13, 0.89±0.13 and 0.91±0.13 across the trajectory groups. The basic characteristics of participants at baseline by cognitive trajectories are shown in Supplementary Table S3.

The results of subgroup analysis by cognitive function trajectories are shown in Table 4. Compared with participants having a normal cognitive function, those with a low cognitive function demonstrated a 2.02-fold (1.45–2.82) higher risk of incident sarcopenia, and participants with a medium cognitive function showed a 1.51-fold (1.16–1.99) higher risk of sarcopenia. Moreover, the same effect was detected in the older participants and those with low cognitive function middle-aged adults.

The combined effects of cognitive impairment and lifestyle categories on sarcopenia incidence are presented in Fig. 3 and Table 5. Using the unfavorable lifestyle group as the reference category, participants with intermediate and favorable lifestyles demonstrated a dose-response reduction in the risk of developing sarcopenia in both the overall population and cognitive-stratified subgroups. In the overall population, those with an intermediate lifestyle had a 55% reduction in the risk of sarcopenia (HR=0.45; 95% CI, 0.34–0.60), while a favorable lifestyle was associated with a 64% reduction in the risk of sarcopenia (HR=0.36; 95% CI, 0.25–0.53). This pattern remained consistent across cognitive-stratified subgroups, with the strongest protective effect observed in the individuals with normal cognitive function. Notably, even among participants with a persistently low cognitive function, intermediate and favorable lifestyles reduced the risk of sarcopenia by 58% (HR=0.42; 95% CI, 0.24–0.75) and 72% (HR=0.28; 95% CI, 0.13–0.62), respectively. In sensitivity analyses, the association between a favorable lifestyle pattern and a lower risk of incident sarcopenia remained generally consistent after BMI was excluded from the lifestyle score, when a regression coefficient–weighted lifestyle score was used, and when lifestyle factors assessed at wave 1 were used (Supplementary Table S4). Significant trends across lifestyle categories were also observed in alternative models.

The primary novelty of this study lies in examining the relationships between both baseline and long-term patterns of cognitive function, and the development of sarcopenia, as well as how favorable lifestyles modify the impact of MCI on the development of sarcopenia. However, several potential limitations should be noted. First, as this was an observational study, it can only establish an association between chronic MCI and the risk of developing sarcopenia, rather than a causal relationship. Given the shortage of detailed data regarding disease management (such as severity, medication adherence and treatment status) in the CHARLS datasets used in this study, residual confounding associated with disease management cannot be entirely ruled out, and the observed association may be partially attributed to inadequate management of comorbidities. Second, muscle mass was not directly measured in the CHARLS datasets; therefore, ASM was estimated using an anthropometric equation rather than the DXA/BIA. Although this approach has been validated against the DXA in external Chinese samples, it still may lead to misclassification of sarcopenia-related outcomes. Besides, ASM was partly derived from body weight, height and age. Because BMI was also included in the healthy lifestyle score, the association between lifestyle score and sarcopenia may partly reflect the structural correlation between BMI and equation-estimated ASM. Although we have additionally performed sensitivity analyses excluding BMI from the lifestyle score, the lifestyle-related findings should be interpreted with caution. Third, the exclusion of participants who fail to offer data on MCI or sarcopenia may have introduced selection bias. Consequently, the findings may not fully reflect the entire population of Chinese community-dwelling adults, and caution should be exercised when generalizing our results. Fourth, cognitive function was assessed at wave 1, whereas lifestyle factors were assessed at wave 2. Thus, the lifestyle behaviors measured after baseline may have partly reflected behavioral changes related to early cognitive impairment. That is, the lifestyle-related findings should be interpreted as the associations with behavioral patterns during follow-up rather than the evidence of a strictly baseline independent modifying effect. Additionally, lifestyle factors may partially contribute to the progression from mild cognitive impairment to the onset of sarcopenia. The correction for lifestyle-related factors in the multivariable models may have led to potentially underestimating the association between MCI and sarcopenia. Thus, future studies are still needed to clarify the causal relationship and explore the potential mechanisms underlying the association between baseline MCI and the risk of developing sarcopenia.

This study demonstrates that baseline MCI, particularly a long-term cognitive decline, is significantly associated with an increased risk of developing sarcopenia-related outcomes among middle-aged and older adults in China. However, healthy lifestyle factors may potentially mitigate the risk of sarcopenia induced by baseline MCI. Our findings highlight the potential of advocating healthy lifestyle practices at the community level as an effective strategy to reduce the incidence of sarcopenia and promote healthy aging.