Associations between physical activity parameters and multidimensional trajectories of muscle health in middle-aged and older adults: a group-based multi-trajectory modeling study

Abstract

Objectives

This study used a group-based multi-trajectory model (GBMTM) to identify distinct muscle health trajectories and examine their associations with physical activity (PA) in middle-aged and older adults.

Methods

Data were obtained from 2818 middle-aged and older adults (aged ≥40 years) in the China Health and Retirement Longitudinal Study (2011–2015). Muscle health was assessed using muscle mass (appendicular skeletal muscle mass index), muscle strength (handgrip strength), and physical performance (5-time chair stand test). PA was assessed using the International Physical Activity Questionnaire Short Form. A GBMTM was applied to jointly identify longitudinal trajectories of muscle mass, muscle strength, and physical performance, and to evaluate their associations with PA.

Results

In this study, four muscle health trajectories were identified: low-function declining, moderate-function declining, moderate-function stable, and high-function stable group. Engaging in ≥150 min/wk of light PA (LPA), moderate PA (MPA), or vigorous PA (VPA) was associated with the moderate-function stable group (LPA: aOR = 3.44, 95% CI: 1.94 - 6.11; MPA: aOR = 2.83, 95% CI: 1.67 - 4.96; VPA: aOR = 2.88, 95% CI: 1.61 - 5.13) and the high-function stable group (LPA: aOR = 5.20, 95% CI: 2.44 - 11.19; MPA: aOR = 4.10, 95% CI: 1.92 - 8.73; VPA: aOR = 3.42, 95% CI: 1.55 - 8.55). In older adults aged ≥70 years, associations persisted for MPA and VPA.

Conclusion

Distinct muscle health trajectories highlight individualized muscle aging and inform personalized PA guidance. Regular PA ≥150 min/wk across intensities was associated with more favorable longitudinal muscle health.

1. Introduction

Skeletal muscle plays a vital role in maintaining human health by supporting posture and movement, while also contributing to energy metabolism, thermoregulation, and nutrient storage.¹ As a key reservoir of amino acids, skeletal muscle helps maintain glucose homeostasis during stress or starvation and supports normal organ function.² Evidence suggests that a decline in muscle mass and function can reduce the body's ability to cope with disease and stress, which affects health and quality of life.³ Therefore, maintaining adequate skeletal muscle mass, strength, and physical performance is essential for healthy aging. In rapidly aging populations, muscle health is increasingly regarded as a key marker of overall health and quality of life.⁴ Sarcopenia is characterized by the progressive, age-related loss of muscle mass and function and is associated with poorer health-related quality of life and increased mortality risk.⁵ Moreover, it has been linked to frailty, disability, falls, and fractures in older adults.⁶ Beyond these individual-level consequences, sarcopenia increases healthcare utilization and long-term care needs, thereby imposing a substantial economic burden on health systems.⁷ Given these consequences, identifying modifiable factors and developing effective strategies to preserve or improve muscle health are critical for promoting healthy aging.⁸

Physical activity (PA) is widely recognized as a key modifiable factor for maintaining muscle health in middle-aged and older adults.⁹ PA denotes bodily movement that results in energy expenditure, whereas exercise is planned and structured PA.¹⁰ Regular engagement in PA contributes to maintaining muscle mass and strength through multiple pathways, including supporting muscle maintenance processes,¹¹ and preserving overall physical function. Through these physiological adaptations, regular PA plays a critical role in preventing or delaying the onset of sarcopenia. In particular, moderate-to-vigorous PA (MVPA), typically defined as the combined time spent in moderate- and vigorous-intensity activities, has been associated with a lower risk of accelerated muscle strength decline and functional impairment in older adults.¹² Higher levels of MVPA have been associated with dose-dependent benefits, including a reduced risk of sarcopenia and other adverse muscle-related outcomes.¹² Conversely, sedentary behavior (i.e., any waking behavior with an energy expenditure ≤1.5 METs while sitting, reclining, or lying) is increasingly recognized as an independent risk factor for adverse muscle health outcomes.¹³ Overall, increasing PA across intensities while reducing sedentary behavior may help maintain muscle health, particularly muscle strength and physical function, and mitigate age-related decline.

The deterioration of muscle health is not abrupt but occurs gradually over an extended period. Evidence suggests that in generally healthy adults, muscle mass begins to decline from midlife (around age 40), with an estimated loss of 3–8% per decade.^14,15 Simultaneously, muscle strength and physical performance also deteriorate, with muscle strength often declining more rapidly than muscle mass.¹⁶ Consistent with this multidimensional decline, the revised European consensus on sarcopenia emphasizes that sarcopenia is evaluated using muscle strength, muscle quantity, and physical performance. Accordingly, monitoring longitudinal changes across these domains warrants a comprehensive approach.⁸ The group-based multi-trajectory model (GBMTM) is a statistical technique for longitudinal data that identifies latent classes sharing similar patterns of change and estimates their trajectories over time.¹⁷ A key strength of GBMTM lies in its capacity to model joint trajectories of multiple indicators, such as muscle mass, strength, and physical performance, thus offering a more comprehensive view of muscle health dynamics. Consequently, GBMTM can capture average trends and reveal distinct patterns across subgroups, thereby facilitating the identification of high-risk populations and supporting the development of targeted intervention strategies.¹⁸

While the multidimensional and dynamic nature of muscle health has received increasing attention, much of the existing evidence has focused on cross-sectional associations between PA and single-domain outcomes, such as physical performance.^19,20 Cross-sectional analyses have also been reported for muscle mass or strength,²¹ or for sarcopenia-related outcomes.²² Although some studies have described joint trajectories of lean mass, body composition, and physical performance, the evidence largely comes from older men and has focused on bone and lean mass alongside physical performance,^23,24 or on PA trajectories in relation to musculoskeletal outcomes.²⁵ Evidence remains relatively limited for jointly modeling muscle strength, muscle mass, and physical performance simultaneously and evaluating their associations with PA, particularly in population-based cohorts in China. Moreover, patterns of muscle health decline may differ substantially across individuals,²⁴ yet such heterogeneity has not been fully captured by single-indicator or cross-sectional approaches.

Muscle health is influenced by a complex interplay of biological aging, metabolic status, chronic conditions, and lifestyle behaviors.^26,27 To address these limitations, the present study employed GBMTM to jointly model changes in muscle mass, strength, and physical performance over time among Chinese adults aged 40 years and older. Unlike prior work, this study incorporated multiple PA dimensions, including frequency, duration, and total weekly volume, to explore dose-response relationships with multidimensional muscle health trajectories. By integrating a multidimensional assessment with a trajectory-based framework, this study adds to the evidence on heterogeneity in muscle aging in a large population-based Chinese cohort. These findings may help inform targeted PA strategies for middle-aged and older adults. Based on previous research, we hypothesized that adults would exhibit distinct joint trajectories of muscle mass, muscle strength, and physical performance. We further hypothesized that greater weekly PA volume, examined across intensity levels, would be associated with membership in more favorable and stable trajectory groups, whereas lower PA volume would be associated with membership in declining trajectory groups.

2. Methods

2.1. Study design and sample

We used data from the China Health and Retirement Longitudinal Study (CHARLS), a large longitudinal survey in China that collects information on demographics, health status, healthcare access, insurance coverage, family structure, and economic conditions. CHARLS includes repeated waves of survey and health assessment data. For this study, we used data from the 2011 (baseline), 2013, and 2015 waves, as they consistently included standardized assessments of muscle health, namely handgrip strength, muscle mass, and physical performance, which are core domains used to define and diagnose sarcopenia and to determine its severity.⁸

In the first wave of the survey in 2011, 6904 community-dwelling middle-aged and older adults with complete PA data were initially included. Of these, 2099 participants were excluded based on the following criteria: (a) aged under 40 years (n = 18); and (b) missing or incomplete data on muscle strength, muscle mass, or physical performance (n = 2081). In the second wave (2013), 1280 additional participants were excluded due to missing or incomplete data on muscle strength, muscle mass, or physical performance. In the third wave (2015), 707 more participants were excluded for the same reason. Consequently, a total of 2818 middle-aged and older adults were included in the final analysis (Supplementary Fig. S1).

2.2. Assessment of PA

To assess PA levels, CHARLS uses the International Physical Activity Questionnaire Short Form (IPAQ-SF) (Supplementary Table S1). PA was assessed using the IPAQ-SF in CHARLS, and IPAQ-based questionnaires have demonstrated acceptable reliability and validity among Chinese adults.²⁸ The questionnaire assesses PA by intensity and captures frequency (days/week, d/wk) and duration (minutes/day, min/d) for three intensity levels: light PA (LPA), moderate PA (MPA), and vigorous PA (VPA). PA intensity was categorized as LPA, MPA, and VPA. VPA referred to activities that made participants breathe much harder than usual (e.g., heavy lifting, digging, plowing, fast bicycling, cycling with a heavy load, or aerobics). MPA referred to activities that made participants breathe somewhat harder than usual (e.g., carrying light loads, bicycling at a regular pace, or mopping the floor). LPA primarily comprised walking, including walking at home or work, walking for transportation, and walking for sport, exercise, recreation, or leisure. In this study, MVPA was used only as an intensity-based umbrella term for MPA and VPA and did not capture resistance-based exercise. In CHARLS, daily PA duration is reported using ordinal categories (<10 min/d, 10–29 min/d, and 30–119 min/d; see Supplementary Table S1) rather than continuous minutes. Because continuous duration data are not available, we used a commonly applied imputation method by assigning representative median values to each category (0, 20, and 75 min, respectively).²⁹ Weekly total minutes of VPA was calculated using the formula: VPA (minutes/week, min/wk) = frequency × duration. The same method was applied to estimate weekly totals for MPA and LPA.

2.3. Measures of muscle strength

Consistent with established frameworks for sarcopenia,⁸ and commonly used functional performance frameworks in aging research,³⁰ muscle health was operationalized as a multidimensional construct encompassing muscle strength, physical performance, and muscle mass. Muscle strength was assessed using handgrip strength, measured following standardized protocols with a Yuejian™ WL-1000 dynamometer (Nantong, China). Each participant performed two trials with each hand. During the test, participants stood upright, held the device with the elbow flexed at 90°, and squeezed maximally for several seconds. The highest value recorded from all trials, in kilograms, was used for analysis.³¹ Within-session reliability across repeated trials was acceptable (ICC = 0.70; CV = 8.71%). ICC was estimated in SPSS using an intraclass correlation coefficient model.

2.4. Measures of physical performance

Physical performance was assessed using the 5-time chair stand test, a validated and widely used measure of lower limb functional strength and functional mobility.³⁰ Participants were instructed to sit on a standard chair (seat height: 42 cm) with their arms crossed over their chest. They were then asked to stand up fully, ensuring complete extension of the hips and knees, and return to the seated position. This movement was repeated five times as quickly as possible without using the arms for support. The test was administered twice, and the shorter completion time (in seconds) was used for analysis. Within-session reliability across the two trials was moderate (ICC = 0.63; CV = 7.53%). ICC was estimated in SPSS using an intraclass correlation coefficient model.

2.5. Measures of muscle mass

We calculated appendicular skeletal muscle mass (ASMM) using an anthropometric equation validated in Chinese adults.^32,33 This method shows good agreement with ASMM values obtained through dual-energy X-ray absorptiometry in Chinese adults.³⁴ The equation used was: ASMM = 0.193 × body weight (kg) + 0.107 × height (cm) − 4.157 × sex (male = 1, female = 2) − 0.037 × age (years) − 2.631. To adjust for body size, we calculated the appendicular skeletal muscle mass index (ASMI, kg/m²) as ASMM/height², consistent with previous studies.³²

2.6. Potential covariates

Covariates were chosen based on prior research^22,35,36 and grouped into three categories: sociodemographic characteristics, lifestyle behaviors, and health-related factors. Covariate consideration and selection were guided by a directed acyclic graph (Supplementary Fig. S2). Sociodemographic variables included age (categorized as 40–49, 50–59, 60–69, and ≥70 years), gender (male or female), education level (illiterate, primary school or below, middle school, and high school or above), and marital status (married and living with a spouse, or other). Lifestyle behaviors included alcohol use (>1/month, ≤1/month, or never) and smoking status (current, former, or never). Health-related factors included baseline diabetes, hypertension, and dyslipidemia, which were treated as cardiometabolic risk factors.

2.7. Statistical analysis

We applied GBMTM to identify distinct joint trajectories of muscle strength, muscle mass, and physical performance across the 2011, 2013, and 2015 waves. GBMTM enables the simultaneous analysis of multiple longitudinal outcomes and classifies individuals into latent subgroups based on shared developmental patterns across these indicators.¹⁷

To determine the optimal number and shape of trajectory groups, we estimated a series of GBMTMs with 2–5 groups. Each model specified quadratic functional forms for all three outcomes and assumed no within-class random slope variance. Model selection was based on a combination of fit and classification criteria: (a) the Bayesian Information Criterion (BIC), with lower values indicating better fit; (b) average posterior probability (AvePP), with values ≥ 0.7 reflecting acceptable classification precision³⁷; (c) odds of correct classification (OCC), with a minimum threshold of 5³⁸; (d) relative entropy (E_k), where values > 0.8 suggest good class separation³⁹; and (e) a minimum class size of 5% of the sample to ensure sufficient representation. The final model was selected based on overall performance across these criteria, following the guidelines proposed by Nagin and colleagues.¹⁸

After selecting the final GBMTM, participants were assigned to one of the four trajectory groups based on the maximum posterior probability. For interpretability, groups were labeled post hoc (with no a priori cut-points) according to (i) the baseline performance level (low/moderate/high) based on the rank order of group-specific predicted levels (better function = higher handgrip strength and ASMI, and shorter 5-time chair stand time), and (ii) the longitudinal pattern between 2011 and 2015 (declining vs stable). Accordingly, the four groups were labeled as low-function declining (Group1), moderate-function declining (Group2), moderate-function stable (Group3), and high-function stable (Group4). Group-specific predicted means (95% CI) at each wave are reported in Supplementary Table S2.

Baseline characteristics across the muscle health trajectory groups were compared using chi-square tests for categorical variables. To assess potential selection bias, baseline characteristics were compared between included participants (complete data) and excluded participants (missing data) using chi-square tests for categorical variables. To explore the relationship between PA and muscle health trajectories, we performed multinomial logistic regression analyses with trajectory group membership as the dependent variable. Results were presented as odds ratios (OR) with corresponding 95% confidence intervals (CI). Sensitivity analyses were performed by excluding participants with hypertension, dyslipidemia, and diabetes to assess the robustness of the findings. In addition, stratified analyses were conducted to account for potential age-related differences. All analyses were performed using SAS 9.4 and SPSS Statistics version 26.0.

3. Results

3.1. Comparison of baseline characteristics across muscle health trajectory groups

Baseline characteristics differed significantly across the muscle health trajectory groups, as shown in Table 1. Significant group differences were observed in age, gender, marital status, education level, alcohol consumption frequency, and the prevalence of chronic conditions such as dyslipidemia and hypertension (P < 0.05). To assess potential selection bias, baseline characteristics were compared between included and excluded participants (Supplementary Table S3); excluded participants differed in age distribution, marital status, education level, diabetes, and hypertension (P < 0.05), whereas gender, alcohol drinking frequency, smoking status, and dyslipidemia did not differ significantly (P > 0.05).

Table 1.

Baseline characteristics of the sample based on the different muscle health trajectory groups.

Characteristic	Low-function declining group (n = 557)	Moderate-function declining group (n = 977)	Moderate-function stable group (n = 888)	High-function stable group (n = 396)	P value
Age
40-49 years	60 (10.8)	252 (25.8)	146 (16.4)	110 (26.2)	<0.001
50-59 years	171 (30.7)	371 (38.0)	324 (36.5)	179 (45.2)
60-69 years	212 (38.1)	259 (26.5)	310 (34.9)	95 (24.0)
≥70 years	114 (20.5)	95 (9.7)	108 (12.2)	12 (3.0)
Gender
Male	21 (3.8)	134 (13.7)	712 (80.2)	393 (99.2)	<0.001
Female	536 (96.2)	843 (86.3)	176 (19.8)	3 (0.8)
Marital status
Married and living with spouse	421 (75.6)	835 (85.5)	775 (87.3)	367 (92.7)	<0.001
Others	136 (24.4)	142 (14.5)	113 (12.7)	29 (7.3)
Alcohol drinking frequency
>1/month	53 (9.5)	114 (11.6)	339 (38.2)	191 (48.2)	<0.001
≤1/month	27 (4.8)	56 (5.7)	92 (10.4)	39 (9.8)
Never drank	477 (85.6)	807 (82.6)	457 (51.5)	166 (41.9)
Smoking status
Current smokers	151 (27.2)	258 (26.5)	284 (32.1)	111 (28.1)	0.136
Former smokers	60 (10.8)	127 (13.1)	99 (11.2)	44 (11.1)
Never smoked	344 (62.0)	586 (60.4)	502 (56.7)	240 (60.8)
Education levels
Illiterate	262 (47.0)	328 (33.6)	149 (16.8)	28 (7.1)	<0.001
≤primary school	227 (40.8)	414 (42.4)	429 (48.3)	147 (37.1)
Middle school	42 (7.5)	168 (17.2)	221 (24.9)	147 (37.1)
≥high school	26 (4.7)	67 (6.8)	89 (10.0)	74 (18.7)
Diabetes
Yes	22 (4.0)	54 (5.6)	40 (4.6)	24 (6.1)	0.360
No	529 (96.0)	918 (94.4)	839 (95.4)	368 (93.9)
Dyslipidemia
Yes	25 (4.5)	90 (9.4)	76 (8.8)	58 (14.9)	<0.001
No	523 (95.4)	869 (90.6)	792 (91.2)	332 (85.1)
Hypertension
Yes	103 (18.6)	212 (21.8)	191 (21.6)	132 (33.5)	<0.001
No	452 (81.4)	761 (78.2)	692 (78.4)	262 (66.5)

Note: Data are presented as n (%).

3.2. Muscle health trajectories

Four distinct muscle health trajectory groups were identified among middle-aged and older adults using GBMTM. Although the 5-group model showed the lowest BIC (−59,606), it demonstrated a slight decrease in E_k (0.869) compared to the 4-group model (0.874), indicating reduced classification certainty. In contrast, the 4-group model provided a better overall balance between model fit and classification quality. All groups had sufficient model-estimated probabilities (P_j>5%), with AvePP values ranging from 0.921 to 0.938 and OCC values between 23.1 and 88.2. Based on these criteria, the 4-group model was selected as the optimal solution. Full model fit statistics are presented in Supplementary Table S4.

Participants were assigned to trajectory groups based on the maximum posterior probability. Based on the group-specific baseline level and longitudinal pattern (as defined in Section 2.7), four latent classes were defined: low-function declining (19.7%), moderate-function declining (34.7%), moderate-function stable (31.6%), and high-function stable (14.0%). These trajectories are illustrated in Fig. 1, and the corresponding group-specific predicted means (95% CI) are provided in Supplementary Table S2.

Fig. 1.

Joint trajectories of muscle health indicators from 2011 to 2015. Group 1, Group 2, Group 3, and Group 4 were labeled as the low-function declining, moderate-function declining, moderate-function stable, and high-function stable groups, respectively. Muscle health indicators included handgrip strength, 5-time chair-stand time, and appendicular skeletal muscle mass index. Shaded areas represent 95% confidence intervals. Group-specific predicted means (95% CI) are presented in Supplementary Table S2.

3.3. Association of PA and muscle health trajectories

As shown in Table 2, compared with the low-function declining group, engaging in LPA on ≥3 d/wk was associated with greater odds of belonging to the moderate-function declining, moderate-function stable, and high-function stable groups. For MPA, participation on 1–2 d/wk and ≥3 d/wk was associated with membership in the moderate-function stable and high-function stable groups, and ≥3 d/wk was also associated with membership in the moderate-function declining group. For VPA, engaging in VPA on 1–2 d/wk was associated with membership in the moderate-function stable and high-function stable groups, while participation on ≥3 d/wk was associated with the moderate-function declining, moderate-function stable, and high-function stable groups.

Table 2.

The association between physical activity frequency and trajectories of muscle health in middle-aged and older adults.

PA frequency	Moderate-function declining group		Moderate-function stable group		High-function stable group
	OR (95% CI)	aOR (95% CI)	OR (95% CI)	aOR (95% CI)	OR (95% CI)	aOR (95% CI)
LPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
1-2 d/wk	1.08 (0.51 - 2.28)	1.59 (0.68 - 3.71)	1.68 (0.80 - 3.53)	1.89 (0.58 - 6.19)	1.10 (0.38 - 3.16)	1.44 (0.30 - 6.93)
≥3 d/wk	1.53 (1.08 - 2.17) ∗	1.79 (1.21 - 2.65) ∗	1.89 (1.30 - 2.73) ∗	3.38 (1.91 - 5.98) ∗	2.06 (1.26 - 3.37) ∗	5.35 (2.50 - 11.46) ∗
MPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
1-2 d/wk	2.80 (0.98 - 5.70)	3.36 (0.98 - 7.17)	3.08 (1.49 - 6.35) ∗	4.78 (1.80 - 12.64) ∗	3.95 (1.69 - 9.23) ∗	9.09 (2.68 - 30.86) ∗
≥3 d/wk	1.37 (1.10 - 1.95) ∗	1.47 (1.00 - 2.19) ∗	1.79 (1.23 - 2.61) ∗	2.92 (1.67 - 5.12) ∗	1.90 (1.16 - 3.13) ∗	4.22 (1.99 - 8.89) ∗
VPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
1-2 d/wk	1.58 (0.82 - 3.05)	1.82 (0.90 - 3.68)	2.01 (1.03 - 3.95) ∗	2.93 (1.13 - 7.55) ∗	3.81 (1.77 - 8.20) ∗	5.96 (1.79 - 19.91) ∗
≥3 d/wk	1.47 (1.01 - 2.14) ∗	1.58 (1.03 - 2.41) ∗	2.51 (1.69 - 3.71) ∗	2.96 (1.66 - 5.26) ∗	2.36 (1.41 - 3.96) ∗	3.58 (1.63 - 7.87) ∗

Abbreviations: 95% CI, 95% confidence intervals; aOR, adjusted odds ratio; d/wk, days/week; LPA, light physical activity; MPA, moderate physical activity; OR, odds ratio; Reference, low-function declining group; VPA, vigorous physical activity.

^†Adjusted for gender, age, marital status, education levels, smoking status, alcohol drinking frequency, hypertension, dyslipidemia, and diabetes.

∗P < 0.05.

According to Table 3, compared with the low-function declining group, higher daily duration of PA was associated with more favorable trajectory membership across intensity levels. Specifically, for LPA, engaging in 10–29 min/d was associated with membership in the moderate-function stable and high-function stable groups, whereas 30–119 min/d was associated with membership in the moderate-function declining group. For MPA and VPA, both 10–29 min/d and 30–119 min/d were associated with membership in the moderate-function stable and high-function stable groups, and higher-duration categories were also associated with membership in the moderate-function declining group.

Table 3.

The association between physical activity duration and trajectories of muscle health in middle-aged and older adults.

PA duration	Moderate-function declining group		Moderate-function stable group		High-function stable group
	OR (95% CI)	aOR (95% CI)	OR (95% CI)	aOR (95% CI)	OR (95% CI)	aOR (95% CI)
LPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
10-29 min/d	1.45 (0.94 - 2.25)	1.74 (0.98 - 2.83)	1.59 (1.00 - 2.53) ∗	2.87 (1.42 - 5.79) ∗	2.13 (1.18 - 3.82) ∗	5.09 (2.03 - 12.74) ∗
30-119 min/d	1.54 (1.08 - 2.18) ∗	1.80 (1.21 - 2.67) ∗	1.94 (1.33 - 2.81) ∗	3.41 (1.92 - 6.06) ∗	2.00 (1.22 - 3.28) ∗	5.15 (2.39 - 11.07) ∗
MPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
10-29 min/d	1.53 (0.85 - 2.74)	1.85 (0.96 - 3.55)	1.84 (1.00 -3.38) ∗	4.24 (1.78 - 10.08) ∗	1.43 (1.33 -3.23) ∗	5.74 (1.75 - 8.81) ∗
30-119 min/d	1.44 (1.01 - 2.05) ∗	1.54 (1.03 - 2.28) ∗	1.86 (1.28 - 2.72) ∗	2.88 (1.63 - 5.05) ∗	2.04 (1.24 -3.23) ∗	4.19 (1.97 - 8.90) ∗
VPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
10-29 min/d	2.58 (0.97 - 6.83)	3.08 (0.99 - 8.82)	3.11 (1.15 - 8.42) ∗	3.98 (1.08 - 14.61) ∗	4.12 (1.32 -12.81) ∗	9.82 (1.93 - 49.83) ∗
30-119 min/d	1.48 (1.02 - 2.14) ∗	1.58 (1.04 - 2.40) ∗	2.43 (1.65 - 3.60) ∗	2.91 (1.64 - 5.17) ∗	2.46 (1.47 - 4.11) ∗	3.72 (1.70 - 8.17) ∗

Abbreviations: 95% CI, 95% confidence intervals; aOR, adjusted odds ratio; LPA, light physical activity; min/d, minutes/day; MPA, moderate physical activity; OR, odds ratio; Reference, low-function declining group; VPA, vigorous physical activity.

^†Adjusted for gender, age, marital status, education levels, smoking status, alcohol drinking frequency, hypertension, dyslipidemia, and diabetes.

∗P < 0.05.

As shown in Table 4, similar patterns were observed for weekly PA volume. For LPA, 10–149 min/wk was associated with membership in the moderate-function stable and high-function stable groups, while ≥150 min/wk was associated with membership in the moderate-function declining group. For MPA and VPA, both 10–149 min/wk and ≥150 min/wk were associated with membership in the moderate-function stable and high-function stable groups, and higher weekly volume was also associated with membership in the moderate-function declining group.

Table 4.

The association between physical activity volume and trajectories of muscle health in middle-aged and older adults.

PA volume	Moderate-function declining group		Moderate-function stable group		High-function stable group
	OR (95% CI)	aOR (95% CI)	OR (95% CI)	aOR (95% CI)	OR (95% CI)	aOR (95% CI)
LPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
10-149 min/wk	1.44 (0.94 - 2.21)	1.80 (0.99 - 2.91)	1.59 (1.01 - 2.50) ∗	2.80 (1.41 - 5.57) ∗	2.02 (1.14 - 3.60) ∗	4.84 (1.96 - 11.90) ∗
≥150 min/wk	1.54 (1.08 - 2.18) ∗	1.78 (1.20 - 2.65) ∗	1.95 (1.34 - 2.83) ∗	3.44 (1.94 - 6.11) ∗	2.03 (1.24 - 3.33) ∗	5.20 (2.44 - 11.19) ∗
MPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
10-149 min/wk	1.37 (0.96 - 1.96)	1.46 (0.98 - 2.17)	2.38 (1.40 - 4.06) ∗	4.63 (2.18 - 9.83) ∗	2.27 (1.15 - 4.49) ∗	7.23 (2.66 - 19.66) ∗
≥150 min/wk	2.04 (1.23 - 3.40) ∗	2.46 (1.40 - 4.32) ∗	1.81 (1.24 - 2.64) ∗	2.83 (1.67 - 4.96) ∗	1.97 (1.20 - 3.25) ∗	4.10 (1.92 - 8.73) ∗
VPA
Sedentary	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
10-149 min/wk	1.44 (0.99 - 2.10)	2.07 (0.99 - 3.96)	2.35 (1.27 - 4.34) ∗	3.48 (1.49 - 8.15) ∗	4.00 (1.95 - 8.18) ∗	7.83 (2.61 - 23.50) ∗
≥150 min/wk	1.81 (1.00 - 3.30) ∗	1.53 (1.00 - 2.35) ∗	2.47 (1.66 - 3.67) ∗	2.88 (1.61 - 5.13) ∗	2.30 (1.37 - 3.87) ∗	3.42 (1.55 - 8.55) ∗

Abbreviations: 95% CI, 95% confidence intervals; aOR, adjusted odds ratio; LPA, light physical activity; min/wk, minutes/week; MPA, moderate physical activity; OR, odds ratio; Reference, low-function declining group; VPA, vigorous physical activity.

^†Adjusted for gender, age, marital status, education levels, smoking status, alcohol drinking frequency, hypertension, dyslipidemia, and diabetes.

∗P < 0.05.

To ensure the robustness of the findings, sensitivity analyses were conducted after excluding participants with hypertension, dyslipidemia, and diabetes. The results (Supplementary Table S5) remained consistent with those of the primary analysis (Table 4), supporting the association between PA and muscle health trajectories.

3.4. PA and trajectories of muscle health: age differences

In age-stratified analyses, significant associations between PA and muscle health trajectories were mainly observed in older age groups. Compared with the low-function declining group, among participants aged 50–59 years, engaging in ≥150 min/wk of VPA was associated with membership in the high-function stable group. Among older adults aged ≥70 years, engaging in ≥150 min/wk of MPA and VPA was associated with membership in the moderate-function stable group. Detailed results are presented in Supplementary Tables S6–S9.

4. Discussion

The aim of this study was to identify distinct joint trajectories of muscle health (muscle mass, muscle strength, and physical performance) among middle-aged and older adults and to examine their associations with PA across intensity levels and PA metrics. The main findings of this study were: (i) using GBMTM, we identified four distinct joint trajectory groups: low-function declining, moderate-function declining, moderate-function stable, and high-function stable; (ii) these groups differed significantly in baseline sociodemographic characteristics and chronic conditions; and (iii) greater engagement in PA across LPA, MPA, and VPA, indexed by higher categories of frequency, duration, and weekly volume, was associated with higher odds of belonging to the moderate-function stable and high-function stable groups compared with the low-function declining group. These associations were more evident in older age strata. Overall, these findings corroborate our initial hypotheses that muscle health follows heterogeneous joint trajectories and that greater PA engagement is associated with more favorable and stable trajectory membership within the PA categories assessed.

Previous studies have demonstrated associations between PA and sarcopenia or its individual components. Scott et al.⁴⁰ found that MVPA was positively associated with greater handgrip strength, higher appendicular lean mass, and better Timed Up and Go (TUG) performance, a measure of functional mobility, whereas LPA was associated only with appendicular lean mass and TUG performance. Similarly, Foong and colleagues found that light, moderate, and vigorous PA, but not sedentary behavior, were positively associated with lower limb strength.²¹ Consistent with these findings, our study showed that engaging in LPA and MVPA, including MPA and VPA, was significantly associated with being in more favorable muscle health trajectories. This integrative, trajectory-based design allowed us to relate multidomain muscle health patterns, including strength, mass, and physical performance, to intensity-specific PA frequency, duration, and weekly volume. In our analyses, participants who performed MVPA at least 1–2 d/wk, with each session lasting 10–29 min/d or accumulating 10–149 min/wk, were significantly more likely to be in the moderate-function stable or high-function stable muscle health groups. Likewise, LPA performed on ≥3 d/wk, with each session lasting ≥30 min/d and totaling ≥150 min/wk, was also significantly associated with being in more favorable muscle health trajectories, although the observed associations were less pronounced than those related to MVPA. Our findings emphasize that not only MVPA, but also LPA, is associated with more favorable muscle health trajectories. This aligns with evidence from systematic reviews and longitudinal studies indicating that both MVPA and LPA contribute to the preservation of muscle mass, strength, and function, and may help reduce the risk of sarcopenia, particularly in older populations.^22,41, 42, 43 LPA may be especially relevant as a feasible approach for individuals unable to engage in higher-intensity activity, partly by displacing sedentary time.

Our age-stratified analysis revealed age-dependent variations in the associations between PA and muscle health trajectories, with stronger associations observed in participants aged 50–59 years and older adults aged ≥70 years. Among older adults aged ≥70 years, individuals engaging in ≥150 min/wk of VPA had significantly higher odds of maintaining a high-function stable muscle health trajectory, underscoring the importance of maintaining PA in later life. These findings are consistent with previous studies suggesting that PA plays a critical role in the prevention of sarcopenia among older adults. For example, a longitudinal study by Mijnarends et al.⁴⁴ found that MVPA may delay the onset of sarcopenia in older adults aged 66–93 years, with lower incidence rates among those with higher activity levels. Similarly, prior research has shown that middle-aged and older adults who maintain or initiate PA are less likely to develop sarcopenia, muscle weakness, and functional decline, whereas those who discontinue PA are more prone to these adverse outcomes.⁴⁵ Importantly, the temporality and persistence of PA may shape these associations, and evidence suggests that sustained or lifelong PA may confer more favorable trajectories or risk factor profiles than PA initiated only later in life.⁴⁶ However, because PA was assessed at baseline, we could not distinguish lifelong maintenance from later-life initiation. In the present study, we focused on baseline PA engagement in mid- and later life and examined its associations with subsequent muscle health trajectories, supporting the relevance of current PA engagement for muscle health during mid- and later life within the limits of our measurement.

Taken together, our findings suggest a dose–response pattern across the weekly PA categories assessed, with the most consistent associations observed among participants achieving at least 150 min/wk. However, because the highest category was open-ended (≥150 min/wk) and PA was self-reported and categorized, we were not able to determine whether benefits plateau at higher volumes or to identify an upper threshold; therefore, ≥150 min/wk should be interpreted as a practical reference rather than an “optimal” upper limit. To translate these findings into practice, the trajectory framework may be implemented in clinical and community contexts through repeated assessment of the routine indicators used in this study, including handgrip strength, the 5-time chair stand test and ASMI, followed by risk stratification consistent with the observed trajectory profiles. Such trajectory-based classification provides an empirically grounded basis for stratified management and individualized exercise prescription. Specifically, individuals assigned to declining trajectories may be prioritized for earlier referral and closer follow-up. Where appropriate, referral to exercise professionals may be arranged. In contrast, those in stable, higher-function trajectories may be supported to maintain favorable activity patterns. From an exercise-prescription perspective, within the intensity categories assessed, accumulating at least 150 min/wk of total PA represents a pragmatic benchmark consistently associated with more favorable muscle health trajectories. Where feasible, prioritizing regular participation in MPA and adding VPA as tolerated may confer additional benefits, while maintaining daily LPA accumulation and minimizing sedentary time. Finally, at the public health level, these findings support integrating multidomain functional screening into routine health checks for middle-aged and older adults and implementing scalable, low-barrier community-based PA promotion programs, with additional resources directed toward higher-risk subgroups.

While numerous studies have examined the relationship between PA and individual components of muscle health, these outcomes are often evaluated in isolation. However, these dimensions are biologically and functionally interrelated, and their trajectories may evolve concurrently rather than independently. To address this limitation, our study employed GBMTM to jointly capture the longitudinal patterns of muscle strength, muscle mass, and physical performance over time. This method allows for a more integrated understanding of muscle health aging by identifying subgroups with distinct, multidimensional aging profiles.³⁹ Trajectory-based approaches have also been applied in large prospective cohorts to characterize long-term patterns of PA and related musculoskeletal outcomes. In particular, evidence from the MrOS study has applied group-based trajectory modeling to summarize longitudinal profiles of PA and body composition, as well as multidomain musculoskeletal phenotypes and their functional implications. In the MrOS study, Laddu et al.²³ used group-based trajectory modeling to describe longitudinal PA patterns alongside concurrent changes in body composition, including lean mass. Using a similar framework, Cawthon et al.²⁴ characterized individual and joint trajectories of key musculoskeletal phenotypes such as appendicular lean mass indexed to height squared, grip strength, and walking speed, supporting the presence of multidomain aging phenotypes. Moreover, higher late-life PA trajectories in MrOS have been linked to smaller declines in objective performance measures, including chair-stand performance.²⁵

The marked heterogeneity across the four trajectory groups indicates that longitudinal muscle health may evolve through distinct pathways. Across the trajectories identified here, a key feature is the dissociation between relatively stable muscle mass and more evident deterioration in muscle strength and physical performance, which is consistent with functional decline being driven, at least partly, by reductions in muscle quality rather than mass loss alone.^14,27 Concurrently, age-related neuromuscular alterations may further compromise force-generating capacity and functional performance, thereby amplifying between-group differences over time.²⁷ The high-function stable group may reflect more favorable behavioral profiles, given consistent evidence that higher PA and lower sedentary time are associated with better muscle function and lower sarcopenia risk.^40,44 The moderate-function stable group may reflect a relatively stable, intermediate activity pattern that supports maintenance of function over time.²⁵ In contrast, the moderate-function declining group may plausibly be characterized by progressively less favorable activity–sedentary patterns, which have been linked to poorer muscle function and higher sarcopenia risk.⁴⁵ Finally, the low-function declining group combines low initial function with continued deterioration, consistent with a higher cumulative burden of adverse factors; accordingly, this pattern may be particularly sensitive to impairments in muscle quality and neuromuscular function even when changes in muscle mass are comparatively modest.^14,27

This study is subject to several limitations. First, PA was measured using a modified version of the IPAQ-SF. As a self-reported instrument, the IPAQ-SF is susceptible to recall and social desirability biases, and respondents may misestimate activity intensity and duration, leading to exposure misclassification. In addition, the categorical PA measures (with the highest weekly-volume category top-coded as ≥150 min/wk) limited our ability to determine a precise upper threshold for diminishing returns beyond this level. Accordingly, the present findings should be interpreted primarily in terms of overall associations across PA categories, while potential non-linear patterns at higher activity volumes may warrant further investigation using more granular PA measurement. Second, muscle mass was not directly measured; instead, it was estimated using an anthropometric formula previously validated in a Chinese population.³⁴ While practical for large-scale studies, this indirect approach may be less precise than imaging-based techniques such as dual-energy X-ray absorptiometry or bioelectrical impedance analysis, thereby introducing measurement error.

5. Conclusion

Distinct muscle health trajectories reveal the dynamic and individualized nature of muscle aging. Regular PA at or above 150 min/wk, across all intensity levels, was associated with more favorable longitudinal muscle health outcomes. Recognizing heterogeneity in muscle aging trajectories provides a basis for developing personalized PA guidelines to preserve muscle function and promote healthy aging in middle-aged and older adults.

Authors’ contributions

DL: Formal analysis; investigation; methodology; writing–original draft preparation. YP: Investigation; methodology. HW: Formal analysis; methodology. XZ: Conceptualization; formal analysis; investigation; methodology; resources; writing – review & editing. All authors have read and approved the final manuscript.

Ethics approval and consent to participate statement

The CHARLS survey was evaluated and approved by Peking University's Ethical Review Committee (license number: IRB00001052-11015). All of the participants in the survey gave their written informed consent.

Funding

This work was supported by the National Social Science Fund of China (Grant Number: 24BTY039). The funding agency did not affect any part of this manuscript.

Declaration of competing interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Xiaoguang Zhao reports financial support was provided by National Social Science Fund of China. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following is the Supplementary data to this article:

Data availability

The data that support the findings of this study are available in Peking University Open Research Data at http://charls.pku.edu.cn/en/index.htm.