Risk prediction models for sarcopenia among community-dwelling older adults in China: a systematic review and meta-analysis

Abstract

Background

An increasing number of sarcopenia risk prediction models for older adults in Chinese communities have been developed, but the quality and applicability of these models in clinical practice and future research remain unclear. We conducted a systematic review to evaluate their performance.

Objectives

To systematically review published studies on risk prediction models for sarcopenia among community-dwelling older adults in China.

Methods

We searched the China National Knowledge Infrastructure (CNKI), Wanfang Database, China Science and Technology Journal Database (VIP), SinoMed, PubMed, Web of Science, Cochrane Library and Embase databases up to February 17, 2025, and extracted relevant information from the selected prediction models, including study design, data sources, outcome definitions, sample size, predictors, model development and performance. The risk of bias and applicability were assessed via the Prediction Model Risk of Bias Assessment Tool (PROBAST) checklist.

Results

Initially, we retrieved 2092 studies. After the screening process, 8 development models and 7 validation models were included from 9 studies. The prevalence of sarcopenia ranged from 8.3% to 30.6%, and the most commonly used predictors were BMI and age. All included studies had a high risk of bias, mainly due to inappropriate data sources and insufficient reporting in the analysis area. In the meta-analysis, we observed that the prevalence of sarcopenia was 20% (95% confidence interval: 0.14–0.26), the area under the curve (AUC) value of the development models was 0.89 (95% confidence interval: 0.84–0.95), and the AUC value of the validation models was 0.87 (95% confidence interval: 0.80–0.95).

Conclusion

The overall accuracy of sarcopenia risk prediction models for older adults in Chinese communities is relatively good, but according to the PROBAST checklist, all studies have a high risk of bias. Future research should focus on developing new models with larger sample sizes, rigorous study designs, and multicenter external validation.

Trial registration

The review protocol was registered in PROSPERO (registration ID: CRD420250653096).

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-025-06816-6.

Background

The problem of population aging in China is becoming increasingly severe. According to data from the seventh national census, the population aged 60 years and above in China reached 264 million in 2020, accounting for 18.7% of the total population. It is projected that this number will exceed 400 million by 2035, with the proportion exceeding 30% [1]. Among the various health-related factors leading to disability in older adults, sarcopenia and cognitive impairment have drawn significant attention from both academia and the clinical field [2]. Currently, China’s focus on sarcopenia is relatively recent, and there is insufficient understanding of this condition, which has not yet been classified into a specific domain.

Sarcopenia is an age-related muscle degenerative disease characterized by a progressive decline in skeletal muscle mass, strength, and/or function, and is commonly observed in older adults [3]. The disease progresses slowly and insidiously but poses significant risks. Numerous studies have confirmed that sarcopenia not only leads to falls, fractures, and functional impairments but is also closely associated with increased hospitalization rates, care needs, and mortality risks, posing a serious threat to the quality of life of older adults [3, 4]. Research indicates that over a quarter of older adults in Chinese communities have sarcopenia [5], and the global prevalence rate among those over 60 years of age ranges from 10% to 27% [6]. Although sarcopenia has a wide range of impacts, it is often overlooked because of its nonspecific symptoms, which are frequently mistaken for normal aging. Early identification is particularly challenging in primary care communities [7].

Currently, although various sarcopenia screening tools have been developed, such as the SARC-F questionnaire [8], the SARC-CalF [9], and the Ring Test [10], these tools require additional tests for older adults and demand high cooperation, take a long time and add to the workload of community workers. Therefore, comprehensively and accurately identifying high-risk individuals is difficult. To overcome the limitations of existing tools, researchers have developed multiple sarcopenia risk prediction models based on physical measurements, lifestyle, and basic clinical information in Chinese community-dwelling older adults in recent years, aiming to achieve simple and effective early screening. However, these models vary significantly in terms of research design, indicator selection, sample sources, and validation methods, and their quality and applicability have not been systematically evaluated. This study aims to systematically evaluate the quality and applicability of existing risk prediction models for sarcopenia among community-dwelling older adults in China, with the goal of providing recommendations for model development and clinical application, and supporting the early identification of high-risk individuals.

Methods

This systematic review adheres to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [11, 12]. The study protocol was registered on PROSPERO (registration number: CRD420250653096).

Search strategy

To conduct a comprehensive search, considering the large population size and language universality, we targeted Chinese and English databases. The databases searched included the China National Knowledge Infrastructure (CNKI), Wanfang Database, China Science and Technology Journal Database (VIP), China National Medical Library (SinoMed), PubMed, Web of Science, Cochrane Library, and Embase. The search period was from the establishment of the databases to February 17, 2025. A combination of subject terms and free terms was adopted, and a retrospective analysis of the references of the included literature was conducted to ensure the completeness and accuracy of the retrieved literature. The search keywords used were “Aged”, “old people”, “elderly”, “senior citizens”, “sarcopenia”, “Sarcopenic”, “skeletal muscle reduction”, “muscle loss”, “muscle mass”, “muscle waste”, “muscle wasting”, “myopenia”, “muscle atrophy”, “muscle weakness”, “predictive model”, “prediction model”, “risk prediction”, “risk assessment”, “risk factors”, “risk score*”, “prediction tool*”, “prognostic model”, “nomogram”, “China”, and “Chinese”. The detailed search strategy can be found in the supplementary materials (Supplementary Table 1).

For the systematic review, we used the PICOTS system, which is recommended by the Prediction Model Reporting System for Systematic Reviews (CHARMS) checklist [13]. This system helps to construct the review’s objectives, search strategy, and inclusion and exclusion criteria for studies [14]. The key items of our systematic review are as follows:

• P (Population): Older adults over 60 years of age in Chinese communities.

• I (Intervention model): The developed and published risk prediction model for sarcopenia among older adults in Chinese communities (with ≥ 2 predictive factors).

• C (Comparator): No competing model.

• O (Outcome): The outcome focuses on sarcopenia rather than just muscle mass reduction or muscle strength decline.

• T (Timing): The prediction result is obtained after basic information, clinical scoring scale results, and laboratory indicators are evaluated.

• S (Setting): The purpose of the risk prediction model is to provide individualized predictions of sarcopenia for older adults in Chinese communities, facilitating the implementation of preventive measures and the prevention of the disease.

Inclusion and exclusion criteria

The inclusion criteria for the study were as follows: (1) the study type was observational research; (2) the study participants were community-dwelling older adults aged 60 years and above in China; (3) the study content was the development or validation of sarcopenia risk prediction models for community-dwelling older adults in China; and (4) the outcome of interest was sarcopenia, and the definition conformed to the standards of the Asian Working Group for Sarcopenia (AWGS), the European Working Group on Sarcopenia in Older People (EWGSOP), or the International Working Group on Sarcopenia (IWGS).

The exclusion criteria were as follows: (1) animal or cell-based experiments, reviews, and conference papers; (2) studies with specific populations, such as those limited to patients with diabetes; (3) studies that did not construct prediction models; (4) studies not written in English or Chinese; and (5) studies for which the full text could not be retrieved despite contacting the authors via email.

Study selection and screening

Two authors (LX and XYW) independently screened the literature using EndNote. Duplicate records were removed first. The titles and abstracts of the remaining studies were then reviewed to identify potentially relevant articles. Full texts were obtained for studies that appeared to meet the inclusion criteria or where eligibility was uncertain. The reference lists of included studies were also checked to identify any additional relevant publications. Any disagreements during the selection process were discussed, and a third author (XQN) was consulted when necessary to reach a consensus.

Data extraction

Two reviewers independently reviewed the full-text articles and extracted relevant data based on the eligibility criteria. Disagreements, if any, were resolved through discussion or with input from a third reviewer.

The extracted data were categorized into two main groups. The first included general study characteristics, such as the first author’s name, publication year, study location, study design, participant details, data sources, diagnostic criteria for sarcopenia, and sample size. The second group focused on details related to the predictive models. This included information on variable selection methods, model development approaches, validation strategies, performance metrics, treatment of missing and continuous data, final predictors included in the models, and the way the models were presented.

Initial data extraction was carried out by one reviewer and subsequently cross-checked by another to ensure completeness and accuracy.

Quality assessment

The risk of bias and applicability of each included study were assessed using the Prediction Model Risk of Bias Assessment Tool (PROBAST) [15]. Two reviewers (LX and XYW) independently evaluated each study across the four PROBAST domains: participants, predictors, outcome, and analysis. Any disagreements in judgment were resolved through discussion, and if consensus could not be reached, a third reviewer (XQN) was consulted for arbitration.

PROBAST is specifically designed to appraise studies that develop, validate, or update predictive models. It consists of 20 signaling questions grouped under the four aforementioned domains. Each item is rated as “yes,” “probably yes,” “no,” “probably no,” or “no information.” A domain is considered at high risk of bias if one or more questions within it are rated as “no” or “probably no.” An overall judgment of low risk of bias or applicability is made only when all domains are assessed as low risk.

Data synthesis and statistical analysis

All statistical analyses were conducted using Stata version 18.0. The pooled prevalence of sarcopenia and the area under the curve (AUC) values for the predictive models were synthesized across the included studies. Prevalence estimates were reported as proportions with 95% confidence interval (CI), while the performance of predictive models was summarized using AUC values with corresponding 95% CI.

Between-study heterogeneity was assessed using the I² statistic and the Cochrane Q test. An I² value ≤ 50% and a p-value >0.1 were considered indicative of low heterogeneity, in which case a fixed-effects model was applied [16]. If I² exceeded 50%, a random-effects model was used, reflecting substantial heterogeneity among studies [17].

To evaluate the stability of the pooled results, sensitivity analyses were performed. Potential publication bias was assessed through visual inspection of funnel plots and Egger’s regression test [18], with a p-value >0.05 indicating no significant evidence of bias [19].

Results

Study selection

Overall, 2092 records were identified through the initial literature search. After 436 duplicate records identified in all the databases were eliminated, 1656 titles and abstracts were screened for eligibility. A total of 1595 records were subsequently excluded on the basis of the assessment of titles and abstracts. A total of 61 full-text articles were evaluated for eligibility. Among them, 30 studies were excluded because they were inconsistent with the population of the review. Furthermore, 19 studies were found not to have established predictive models or merely focused on risk factors, 2 studies had diagnostic criteria that did not meet the inclusion requirements, and the full text of 1 study could not be obtained. Finally, this systematic review included a total of 9 studies (Fig. 1).

Fig. 1.

Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) flowchart of literature search and selection

Study characteristics

Table 1 summarizes the design and participant characteristics of the 9 included studies. They were published between 2022 and 2024, among which 3 studies were published in Chinese. Among the included studies, 5 were prospective (including 3 multicenter studies), and 4 were retrospective (including 2 multicenter studies). With respect to data sources, 6 studies originated from community hospitals, 2 focused on physical examination centers, and 1 was derived from the CHARLS database. With respect to the study participants, all studies selected older adults aged 60 years and above, and all employed the Asian Working Group for Sarcopenia 2019 consensus (AWGS 2019) as the diagnostic criterion for sarcopenia. The sample sizes of these studies varied from 180 to 3,454.

Table 1.

Overview of basic data of the included studies

Author (year)	Study site	Data source	Study design	Participants	Diagnostic criteria	Case/sample size (%)
Liu, Beibeia(2022)[26]	Hunan	CH	Prospective Cohort Study	Aged ≥60 years and community-dwelling for ≥6 months	AWGS	357/1442（24.8%）
Mo, Yi-Han(2022)[20]	Hunan	CH	Retrospective Study	Community-dwelling older adults aged 60 and above	AWGS	263/1050 (25.0%)
Zhou, Mengjuana(2023)[28]	Yunnan	CH	Prospective Cohort Study	Community-dwelling older adults aged 60 and above	AWGS	99/626 (15.8%)
Chen, Jiaweia(2023)[27]	Hunan	CH	Prospective Cohort Study	Study participants aged ≥60 years	AWGS	87/556 (15.7%)
Huang, Shuai-Wen(2023)[25]	Hubei	CH	Prospective Cohort Study	Residents aged ≥60 years	AWGS	81/678 (11.9%)
Yang, Yichen(2023)[21]	Zhejiang	HEC	Retrospective Study	Residents aged ≥60 years	AWGS	125/633 (19.7%)
Yin, Guangjiao(2023)[24]	Hubei	HEC	Prospective Cohort Study	Age ≥ 60 years	AWGS	55/180 (30.6%)
Li, Qiugui(2024)[22]	—	CHARLS	Retrospective Study	Age ≥ 60 years	AWGS	997/3454 (28.8%)
Lin, Taiping(2024)[23]	Western of China	CH	Retrospective Study	Age ≥ 60 years	AWGS	87/1042 (8.3%)

CH Community Hospital, HEC Health Examination Center, CHARLS China Health and Retirement Longitudinal Study

^a Study was published in Chinese

Table 2 presents the model information of the included studies. Logistic regression analysis was used to establish predictive models in all included studies. With respect to the handling of continuous variables, 4 studies maintained the continuity of continuous variables, whereas 5 studies transformed continuous variables into categorical variables. In terms of variable selection, 5 studies were based on univariate analysis. For the treatment of missing data, 1 study employed the k-nearest neighbor algorithm to fill in missing values, 3 studies utilized multiple imputation methods to handle missing values, and 1 study chose to directly exclude cases with missing data. The remaining studies did not explicitly report whether there were missing data. The most frequently used predictors in the models were body mass index (BMI) and age, which appeared in 7 and 6 models, respectively. Additionally, sex and calf circumference were included in 5 models each. Grip strength, pain, the serum ALB concentration, and lack of exercise habits were incorporated into 2 models each.

Table 2.

Overview of the information of the included prediction models

Author (year)	Missing data handing	Continuous variable processing method	Variable selection	Model development method	Calibration method	Validation method	Final predictors	Model performances	Model presentation
Liu, Beibei(2022)[26]	—	Categorical variable	Stepwise regression analysis	Multiple linear regression model	—	Temporal validation	Sex, Height, Weight, Grip strength, Calf circumference	B: 0.849  (0.796～0.903)	Fommula of risk score obtained by partial regression coeticient of each factor
Mo, Yi-Han(2022)[20]	—	Categorical variable	Backward stepwise selection	Logistic regression model	Hosmer‒Lemeshow test, Calibration curve, DCA	Sample split	Age, BMI, Marital status, Regular physical activity habit, Uninterrupted sedentary time, Dietary diversity score	A: 0.827 (0.792～0.860) B: 0.755 (0.680～0.837)	Nomogram model
Zhou, Mengjuan(2023) [28]	—	Continuous variable	—	Logistic regression model	Hosmer‒Lemeshow test, Calibration curve	Sample split	Sex, BMI, Frequency of physical exercise, Duration of each exercise, Albumin, Total cholesterol	A: 0.983 (0.971,0.995) B: 0.947 (0.894–0.998.894.998)	Formula of risk score obtained by regression coeffcient of each factor & WeChat Mini Program
Chen, Jiawei(2023)[27]	—	Categorical variable	—	Multivariable logistic regression model	—	—	Age, BMI, Thigh circumference, Calf circumference	A: 0.895 (0.859～0.931)	Nomogram model
Huang, ShuaiWen(2023) [25]	Delete	Categorical variable	Stepwise regression analysis	Logistic regression model	Hosmer‒Lemeshow test, DCA	Bootstrap	Age, BMI, Calf circumference, CHF, COPD	A: 0.930 (0.907～0.952) B: 0.897 (0.858～0.936)	Nomogram model
Yang, Yichen(2023)[21]	KNN	Continuous variable	Forward stepwise selection	Multivariable logistic regression model	Hosmer–Lemeshow test	Sample split	BMI, Age, UA, ALT, Sex	A: 0.974 (0.962–0.987.962.987) B: 0.968 (0.941–0.994.941.994)	Nomogram model
Yin, Guangjiao(2023)[24]	Multiple imputation	Continuous variable	LASSO regression	Multivariable logistic regression model	Hosmer‒Lemeshow test, Calibration curve	Bootstrap	Age, Albumin, BUN, Grip strength, Calf circumference	A: 0.90 (0.85–0.95.85.95) B: 0.92 (0.83–1.00.83.00)	Nomogram model
Li, Qiugui(2024)[22]	Multiple imputation	Continuous variable	LASSO regression	Multivariable logistic regression model	Hosmer‒Lemeshow test, Calibration curve, DCA	Sample split	Sex, BMI, MSBP, MDBP, Pain	A: 0.77 (0.75–0.79.75.79) B: 0.76 (0.73–0.79.73.79)	Nomogram model
Lin, Taiping(2024)[23]	Multiple imputation	Categorical variable	Stepwise regression analysis	Logistic regression model	Calibration curve, DCA	Bootstrap	Age, Sex, BMI, Low physical activity, Malnutrition, Pain, Calf circumference	A: 0.870 (0.83–0.90.83.90) B: 0.85	Nomogram & web calculator

KNN k-nearest neighbor algorithm, DCA decision curve analysis, BMI body mass index, CHF congestive heart failure, COPD chronic obstructive pulmonary disease, UA uric acid, ALT alanine aminotransferase, BUN blood urea nitrogen, MSBP mean systolic blood pressure, MDBP mean diastolic blood pressure

“-”, not reported; A, development cohort; B, validation cohort

The reported AUC or C-statistic values ranged from 0.755 to 0.974. Calibration was conducted for 7 models, with the Hosmer–Lemeshow test being the most commonly adopted method.

Models validation

Among the included studies, the majority of the models underwent either internal or external validation. Among them, only one study carried out external validation, whereas seven studies solely conducted internal validation. The principal methods employed were the bootstrap method and sample splitting. The model of Liu et al. encompassed both internal and external validation, whereas the model of Chen et al. did not undergo any validation after development.

Results of the quality assessment

The risk of bias and concerns regarding applicability for the nine included studies were evaluated using the PROBAST tool (Table 3). and the results indicated that all studies were at high risk of bias, suggesting methodological weaknesses during either the model development or validation phases. In the participants domain, four studies were judged to have a high risk of bias, primarily due to the use of unrepresentative or inappropriate data sources [20–23]. In the predictors domain, one study was rated as having a particularly high risk of bias, as it failed to describe any quality control procedures for predictor measurement, likely due to its retrospective design [21]. In the outcome domain, four studies included predictors as part of the outcome definition [20, 21, 23, 24], and one did not ensure an adequate interval between the measurement of predictors and the determination of outcomes [21]. In the analysis domain, none of the studies were classified as low risk of bias; among them, three had insufficient sample sizes to meet the recommended events per variable (EPV) threshold of 20 [23–25], five converted continuous variables into categorical ones either partially or entirely [20, 25–27], one study excluded some participants from the analysis and did not properly address missing data [25], five studies relied on univariate analysis for predictor selection [20, 23, 25, 27, 28], three failed to provide a comprehensive assessment of model performance [21, 26, 27], and one study developed a model without any form of validation [27]. Additionally, four studies did not adequately consider overfitting, underfitting, or optimism in model performance [20, 21, 27, 28], with three of these using internal validation based solely on a random data split [20, 21, 28]; three studies did not report regression coefficients for the final models [22, 24, 25], and none provided information related to model complexity. With respect to the assessment of applicability, two studies were deemed to have a high risk, while the remaining seven were judged to have a low risk. In the participants domain, the two studies with high applicability concerns were identified due to discrepancies between the target populations or settings and the intended scope of the prediction model [21, 24]. In contrast, all nine studies were rated as having a low risk of concern in both the predictors and outcome domains, indicating that the definitions of predictor variables and outcome measures, as well as their timing and assessment methods, were generally consistent with the study objectives and clinical context, thereby supporting the overall applicability of the prediction models.

Table 3.

PROBAST results of the included studies

Author(year)	Study type	ROB				Applicability			Overall
		Participants	Predictors	Outcome	Analysis	Participants	Predictors	Outcome	ROB	Applicability
Liu, Beibei(2022)[26]	A	+	+	+	-	+	+	+	-	+
Mo, Yi-Han(2022)[20]	B	-	+	-	-	+	+	+	-	+
Zhou, Mengjuan(2023)[28]	A	+	+	+	-	+	+	+	-	+
Chen, Jiawei(2023)[27]	A	+	+	+	-	+	+	+	-	+
Huang, Shuai-Wen(2023)[25]	A	+	+	+	-	+	+	+	-	+
Yang, Yichen(2023)[21]	B	-	?	-	-	-	+	+	-	-
Yin, Guangjiao(2023)[24]	A	+	+	-	-	-	+	+	-	-
Li, Qiugui(2024)[22]	B	-	+	+	?	+	+	+	-	+
Lin, Taiping(2024)[23]	B	-	+	-	-	+	+	+	-	+

PROBAST Prediction model Risk Of Bias Assessment Tool, ROB risk of bias

A indicates “Prospective Cohort Study”; B indicates “Retrospective Study”

+ indicates low ROB/low concern regarding applicability; - indicates high ROB/high concern regarding application;? indicates unclear ROB/unclear concern regarding applicability

Meta-analysis of prevalence rates

A meta-analysis was conducted on the prevalence of sarcopenia in older adults in Chinese communities from 9 studies. The prevalence of sarcopenia in older adults in Chinese communities was 0.20 (95% confidence interval: 0.14–0.26) (Fig. 2). The I² value was 98.0% (P < 0.001), indicating high heterogeneity among the studies. A sensitivity analysis was performed on this prevalence rate, and the results did not change significantly after excluding any one study (Supplementary Fig. 1), so a random effects model was used. The Egger test result was − 0.64 (p = 0.937), suggesting that there was no significant evidence of publication bias.

Fig. 2.

Forest plot of the pooled effect size of sarcopenia prevalence

Meta-analysis of development models

There are differences in the details of these models, and the information provided is incomplete. Only 8 studies met the comprehensive criteria. The random effects model was used to calculate the combined AUC for the development model, with a result of 0.89 (95% confidence interval: 0.84–0.95) (Fig. 3). The sensitivity analysis of individual studies revealed that the combined effect size did not reverse, indicating the robustness of the results (Supplementary Fig. 2). The result of the Egger test was − 9.34 (p = 0.13), suggesting no significant evidence of publication bias.

Fig. 3.

Forest plot of pooled AUC estimates for development models

Meta-analysis of the validation models

Seven studies met the comprehensive criteria for the validation model. A random-effects model was used to calculate the pooled AUC, which was 0.87 (95% confidence interval: 0.80–0.95) (Fig. 4). The I² value was 95.1% (p < 0.001), indicating significant heterogeneity among the studies. Additionally, sensitivity analysis confirmed the robustness of the results, with no single study altering the magnitude of the combined effect (Supplementary Fig. 3). The Egger test result was − 1.47 (p = 0.78), suggesting no significant evidence of publication bias.

Fig. 4.

Forest plot of pooled AUC estimates for validation models

Discussion

In our meta-analysis evaluating sarcopenia risk prediction models for community-dwelling older adults, we analyzed 8 development models and 7 validation models from 9 studies, all of which were based on data from Chinese patients, with AUC values ranging from 0.755 to 0.974. We observed that the AUC value of the development models was 0.89 (95% CI: 0.84–0.95) and that of the validation models was 0.87 (95% CI: 0.79–0.90). Despite their good predictive ability, all studies were considered to have a high risk of bias according to the PROBAST checklist, which might be related to differences in population characteristics, predictors, and methods among the different models. Additionally, during the model evaluation, we found that some prediction models did not fully adhere to the reporting standards of the TRIPOD statement [29]. The lack of reporting transparency not only affects the interpretability and reproducibility of the models but also may increase the risk of bias, thereby limiting their promotion and application in clinical practice. Therefore, future studies should focus on adopting more rigorous research designs, expanding sample sizes, conducting multicenter external validations, and improving reporting transparency.

The 9 studies included in this research initially had a range of 11–65 candidate predictors, and the final models contained 4–7 predictors. The frequent appearance of specific predictors in this model is highly important for clinical guidance. BMI, age, sex, and calf circumference are high-frequency predictive indicators. BMI is a recognized influencing factor for sarcopenia in community-dwelling older adults [30–32], and the lower the BMI value is, the greater the risk of sarcopenia. Studies have shown that for every 1 kg/m² increase in BMI, the risk of sarcopenia decreases by 0.4 times [33]. A lower BMI often indicates insufficient nutritional intake and a reduced protein supply, leading to restricted muscle protein synthesis and decreased muscle mass, thereby increasing the risk of sarcopenia. A relatively high BMI may slow muscle loss due to greater protein reserves and the buffering effect of fat [34]. However, it should be noted that some individuals with high BMI have “obesity-related sarcopenia”, which is characterized by excessive fat and insufficient muscle [35]. In obesity-related sarcopenia, muscle tissue is infiltrated by fat, and excessive fat produces proinflammatory cytokines that indirectly break down skeletal muscle, accelerating muscle loss [36]. Therefore, BMI is recommended as an important indicator for screening and predicting sarcopenia in older adults, and attention should be given to individuals with low BMI and obesity-related sarcopenia to carry out early nutritional intervention and exercise management.

Age is the main determinant of sarcopenia. As people age, muscle mass and strength decline annually [37]. In normal individuals, muscle mass decreases by approximately 1% to 2% annually beginning at the age of 30. After the age of 60, muscle mass decreases by an average of 5% to 13% each year, and by the age of 80, more than 30% of muscle mass has been lost [38]. Currently, sarcopenia is believed to be related to age-related reductions in type II muscle fibers and motor neurons, as well as the deterioration of neuromuscular function [37, 39], decreased levels of sex hormones and growth hormones [40], weakened satellite cell quantity and function [41], mitochondrial dysfunction, and increased chronic inflammation and oxidative stress [42], leading to reduced muscle protein synthesis and decreased regenerative capacity. Older adults, especially those over 80 years old, are recommended to undergo regular screening for sarcopenia.

Sex is an important predictor of sarcopenia, but research results vary. Some studies have shown that the prevalence of sarcopenia in older women is significantly greater than that in older men [43–45], which may be related to the significant decline in estrogen levels after menopause in women, leading to a decrease in skeletal muscle mass and weakened regenerative capacity. On the other hand, some studies suggest that men are more prone to sarcopenia [46, 47], possibly due to the faster decline in leg muscle strength in men than in women [48], as well as genetic and sex chromosome differences [49].Gao et al. [50] included 68 studies with 98,502 cases in a meta-analysis of factors related to sarcopenia in community-dwelling older adults and reported no association between sex and sarcopenia. Thus, there are inconsistent international views on the impact of sex on sarcopenia. When conducting specific analyses, full consideration of regional factors is recommended, and it is expected that future research will include large samples and multicenter and cross-national studies.

Calf circumference has certain applicability in assessing lower-limb muscle mass in older adults. Previous studies have shown that, since the majority of trunk skeletal muscles are distributed in the lower limbs and are less affected by obesity than other body regions, lower-limb skeletal muscles undergo more pronounced changes with aging and disease [51]. Therefore, calf circumference is a commonly used and practical indicator for the clinical assessment of muscle mass [52, 53]. Nevertheless, recent research indicates that this measurement may be influenced by age-related increases in adiposity, as well as changes in skin elasticity, hydration, subcutaneous fat compressibility, and muscle tone, which can lead to measurement bias [54]. The consensus criteria for sarcopenia and malnutrition have not recommended it as a formal or reliable indicator of skeletal muscle mass (SMM) loss [55, 56]. Therefore, in clinical practice, calf circumference should be used cautiously and interpreted together with other assessment measures, especially in older adults.

Furthermore, although grip strength, irregular exercise habits, albumin levels, and chronic pain have relatively low frequencies of occurrence among the predictive factors, they are nonetheless closely associated with the onset of sarcopenia. A decrease in grip strength reflects the deterioration of muscle strength and quality and is one of the crucial indicators for diagnosing sarcopenia; the absence of regular exercise results in reduced muscle activity, readily leading to accelerated reductions in muscle mass and intensity [57]; a lower albumin level indicates suboptimal protein nutritional status, influencing muscle protein synthesis and facilitating muscle loss [58]; and chronic pain, particularly low back pain, can lead to a decline in the strength of the spinal muscle group, affecting trunk stability and reducing exercise tolerance, thereby further accelerating muscle atrophy. Although these factors have received relatively little attention, they can facilitate the occurrence and progression of sarcopenia in terms of mechanism and remain worthy of attention.

Limitations

This review has several potential limitations. Firstly, all the included studies were conducted in mainland China, which may limit the generalizability of the research results to Western populations, and adjustments may be needed when these models are applied in different regions. Therefore, establishing risk prediction models for sarcopenia in community-dwelling older adults in different populations is highly important for future research. Secondly, owing to the transparency and methodological differences in the reports of the included studies, our meta-analysis included only a subset of the developed and validated models from the identified studies. Future research should strictly follow the PROBAST checklist guidelines and provide more comprehensive reports to ensure a more accurate synthesis of evidence. Thirdly, some differences existed in predictor types, data collection methods, and data completeness across studies. These variations may have contributed to a certain degree of heterogeneity, and incomplete handling of missing data could have influenced the stability of the pooled results. Therefore, future research should aim to further standardize predictor selection and data collection procedures, improve data quality, and appropriately handle missing data to enhance the reliability and robustness of sarcopenia risk prediction models. However, these issues do not have an impact on the assessment of the model and partially reflect the methodological and reporting problems we have discovered. In the future, more rigorous methodologies and more transparent reporting are needed. Finally, as this review only included studies published in English and Chinese, problems related to language limitations might exist.

Conclusion

The sarcopenia risk prediction model for older adults in the community generally exhibited good overall predictive accuracy. Nevertheless, there are significant risks of bias during its development and validation phases. Furthermore, in accordance with PROBAST, all the included studies were evaluated as having a high risk of bias. Hence, enhancing the calibration performance of existing models and ensuring their suitability for the general population are highly important. Additionally, researchers should be conversant with the PROBAST checklist and comply with the reporting guidelines outlined in the TRIPOD statement to improve the quality of future studies. Future research should prioritize the development of new models with larger sample sizes, rigorous study designs, and multicenter external validations.

Authors’ contributions

Xin Li: Writing–original draft, Visualization, Methodology, Formal, analysis, Conceptualization. Yiwen Xu: Writing – review & editing, Validation, Data curation.Qinan Xian: Validation, Data curation. Yan Sun: Supervision, Project administration, Methodology. All authors read and approved the final manuscript.

Funding

This research was funded by the Sichuan Science and Technology Department (2023YFS0068).

Data availability

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

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.