**3. Results**

#### *3.1. Search Outcomes*

We identified 459 records from databases and 6 records from the reference lists. After removing 139 duplicated records, we screened the titles and abstracts of the remaining 326 records, and removed another 168 irrelevant records, leaving 158 studies. Further reviewing the full texts of these studies for eligibility, we excluded another 100 for the reasons specified in Figure 1. We carried out the quantitative synthesis (i.e., meta-analysis and meta-regression) on the remaining 58 records.

The characteristics of the included studies are summarized in Table 1. Fifty-one (87.9%) studies were classified as having a moderate risk of bias and seven (12.1%) were classified as having a low risk of bias. None of the included studies recruited the nationwide representative sample. Twelve studies adopted random sampling [19,20,35–44]. One study was open to the non-response bias [35], Each study collected data of sarcopenia components directly from subjects and had a consistent mode of data collection. All included studies had acceptable case definition and correct calculation of prevalence. Details of the critical appraisal are presented in Supplementary Materials (Table S2).

**Figure 1.** PRISMA flowchart.




**Table 1.** *Cont.*


†: This study provided sarcopenia prevalence for older adults from communities and clinical settings separately.


154

#### *3.2. Prevalence in Older Men*

In the 46 studies that reported the prevalence of sarcopenia in older men, the overall prevalence was 18% (95% CI: 15.7 to 20.4%, I2 = 95.2%). For participants from different settings, the pooled prevalence rates were 12.9% (95% CI: 10.7 to 15.1%, I2 = 93.7%) in community-dwelling older men (*n* = 27), 29.7% (95% CI: 18.4 to 41.1%, I2 = 95.5%) for hospitalized older men (*n* = 9), and 26.3% (95% CI: 19.1 to 33.4%, I2 = 83.7%) in nursinghome residents (*n* = 5). We did not pool the prevalence for outpatients due to an insufficient number of studies (*n* = 2). Substantial heterogeneity in prevalence estimates was found across the studies, with all of the I-square values being greater than 80%.

#### *3.3. Prevalence in Older Women*

In the 44 studies that reported the prevalence of sarcopenia in older women, the overall prevalence was 16.4% (95% CI: 14.1 to 18.8%, I2 = 97.3%). For participants from different settings, the prevalence was 11.2% (95% CI: 8.9 to 13.4%, I2 = 97.1%) in community-dwelling older women, (*n* = 27), 23.0% (95% CI: 17.1 to 28.8%, I2 = 80.9%) for hospitalized older women (*n* = 9), and 33.7% (95% CI: 27.2 to 40.1%, I2 =78.4%) for those from nursing homes (*n* = 4). Because only one study reported the prevalence for female outpatients, we did not perform meta-analysis for outpatients. Similarly, considerable heterogeneity was found across studies.

#### *3.4. Meta-Regression Analysis*

We performed a multivariable meta-regression to explore the potential sources of the considerable heterogeneity in the estimates of sarcopenia prevalence across studies. The results of the regression indicated that, irrespective of gender, the hospitalized patients and nursing-home residents had a higher prevalence of sarcopenia than community-dwelling older adults. For older men, the prevalence rate was lower when muscle mass was assessed using the BIA method rather than the DXA method. Furthermore, those from mainland China appeared to have a higher prevalence of sarcopenia compared with those from Hong Kong or Taiwan. For older women, only the participants from different settings showed a significant association with the prevalence of sarcopenia (Table 2).


**Table 2.** Multivariable meta-regression.

## **4. Discussion**

This meta-analysis provided a comprehensive picture of the prevalence of sarcopenia in Chinese older adults. It filled the knowledge gap by extending the existing reviews of sarcopenia prevalence in Chinese older adults from communities to other settings (clinical settings and nursing homes) as well. We also systematically investigated factors that contributed to the significant heterogeneity in pooled prevalence estimates using multivariable meta-regression. Our findings contribute to the epidemiological literature on sarcopenia in Chinese older adults, and can provide a starting point for future research and efforts in sarcopenia prevention among Chinese populations.

For community-dwelling Chinese older adults, we obtained similar estimates of prevalence (12.9% in men vs. 11.2% in women) compared with previous reports (11% in men vs. 10% in women; 14% in men vs. 9% in women) [24,25]. When compared with the prevalence in Europeans (13% in men vs. 14% in women), our estimates were comparable with them [18]. However, Papadopoulou et al. [18] found non-Asian groups seemed more prone to sarcopenia than Asian groups. The difference in ethnic characteristics, body size, and dietary regimes etc., between Asians and non-Asians might be possible reasons for the potential disparity in the sarcopenia prevalence. As for gender difference, our result was similar to the previous evidence that higher prevalence of sarcopenia was shown in Chinese community-dwelling older males than females [15,22,25]. However, in western countries, the evidence of gender difference was controversial depending on the EWGSOP cut-off values applied [13,87]. For hospitalized older adults, we found the prevalence to be 29.7% in men and 23.0% in women. To our knowledge, this is the first systematic review to pool the prevalence of sarcopenia for Chinese older adults who are hospitalized. Our findings revealed a relatively higher prevalence in this group compared with that from communities. Compared with the pooled estimates of hospitalized older adults mainly from Europe (23% in men and 24% in women), our findings were higher in hospitalized men and comparable in hospitalized women [18]. For outpatients, considering that Cochran's Q is not very informative and tends to be biased when the number of studies is small, we did not perform meta-analysis for this group [88]. For those residing in nursing homes, similarly, no meta-analysis synthesized the prevalence of sarcopenia in this Chinese group before. Shen et al. [28] conducted a systematic review pooling the prevalence of sarcopenia in nursing-home populations from 16 studies. However, they only included one original study that targeted the Chinese population [28]. Compared with Shen et al.'s results (43% in men vs. 46% in women), we obtained lower pooled estimates (26.3% in men vs. 33.7% in women) [28]. Nevertheless, the above evidence suggested the high risk of prevalent sarcopenia in nursing-home residents. Institutionalized populations may be more prone to malnutrition [89,90]. A recent cross-sectional study found that malnutrition and physical frailty were highly prevalent among institutionalized older residents and malnutrition was associated with an increased risk of physical frailty among institutionalized Chinese older adults [89]. Hence, nutrition and exercise intervention should be promoted to the institutionalized population to prevent or reverse muscle function for this population.

Multivariable meta-regression indicated that the populations being from different settings was a significant factor contributing to the variation in the pooled prevalence estimates for both genders. Compared with community-dwelling older residents, those who were hospitalized were at 1.69 to 2.10 times the risk of prevalent sarcopenia, and nursing-home residents had 2.50 to 2.73 times the risk. The much higher prevalence rate of sarcopenia in older adults from hospitals and nursing homes was also reported by Papadopoulou and colleagues [18]. However, while Papadopoulou and colleagues reported variation in the pooled prevalence estimates for older adults from different settings, they did not perform meta-regression to quantify it. Our study provided the evidence regarding how much the risk of prevalence increased for older adults who were hospitalized or residing in a nursing home compared with their community-dwelling counterparts. Our findings sugges<sup>t</sup> that hospitalized older adults and those living in nursing homes are particularly vulnerable to this muscle disorder. Therefore, these populations should be

prioritized in sarcopenia-screening and early-prevention efforts. Furthermore, intervention studies of sarcopenia targeting hospitalized older adults and those in nursing homes are warranted to inform evidence-based prevention and managemen<sup>t</sup> efforts such as exercise and/or nutrition interventions for these susceptible groups. Meta-regression did not show significant difference in the prevalence estimate between the community-dwelling and outpatient groups for both genders. However, there were a small number of studies targeting outpatients, which might reduce the statistical power. Future research could further examine the risk of prevalent sarcopenia in this group of older adults.

We found the prevalence estimate in older men varied depending on the method used to assess muscle mass. Specifically, compared with the DXA method, the risk of prevalence was lower (OR: 0.58, 95% CI: 0.35 to 0.98) when using the BIA method. However, no such difference was detected in older women. There is some evidence that the BIA method tends to yield higher estimates of muscle mass, and therefore lower estimates of the prevalence of sarcopenia than the DXA method [91]. Wu and Li [24] synthesized the prevalence estimates from 16 original studies targeting community-dwelling Chinese older adults and obtained similar prevalence estimates irrespective of muscle mass assessments (DXA: 13%, BIA: 12%). However, they did not perform a meta-regression to further examine whether different muscle mass assessments could result in significantly different prevalence estimates after adjusting other covariates. In fact, the measurement of muscle mass remains a challenge in primary care settings; and, while DXA is the standard and more accurate method, BIA is preferred in primary care settings and research because it is more available, easier to use, and less expensive [92]. There are different types of equipment and frequencies of BIA. Based on previous study, BIA with a multifrequency device showed better agreemen<sup>t</sup> with the DXA method than that with other devices when assessing appendicular skeletal muscle mass [51]. The updated AWGS consensus in 2019 also recommended the use of multifrequency BIA [93]. However, the "model" of BIA measurement was not always stated explicitly in previous studies. Therefore, in future research, if the BIA method is used to assess muscle mass, the multifrequency type should be used and clearly documented.

Our study also found the risk of prevalent sarcopenia in older Chinese men living outside the mainland, specifically, in Hong Kong and Taiwan, to be nearly half that of their counterparts living on the mainland (OR: 0.47, 95% CI: 0.23 to 0.98). However, this difference was not significant in the case of older Chinese women. Regional variation in the prevalence of sarcopenia among community-dwelling Chinese older adults was also reported in a previous systematic review, with 17% in mainland and 6% in Hong Kong and Taiwan [24]. However, that review did not conduct meta-regression to quantify the regional variation in the prevalence of sarcopenia. Possible explanations for the variation between mainland Chinese and those living outside the mainland might be attributed to the different levels of economic development, different healthcare and dietary regimes, etc. [94]. Future studies could further explore the reasons why the prevalence of sarcopenia among older adults living outside the mainland is lower than for those in mainland China. Furthermore, as mainland China has a huge area with diverse geographic and climate variations, the prevalence of sarcopenia might also vary across different geographic locations. Until now, no study has investigated the epidemiological characteristics of sarcopenia among different geographic areas within China. Therefore, we further divided the area into 4 regions (south and east of mainland, west of mainland, north of mainland, and outside of mainland) and ran the univariable mate-regression. Findings showed that significant difference in the prevalence of sarcopenia was only presented between mainland areas and outside of the mainland; there was no difference among different geographic locations within mainland China in both genders (Supplementary Materials, Table S3). Therefore, we only divided the area into two regions (mainland and outside of the mainland) in the main analysis.

Previous observational studies found that older age was significantly associated with sarcopenia and severe sarcopenia in Chinese older adults [46,50]. However, less than half of included studies reported the gender-specific age (Supplementary Materials, Table S4). Therefore, our main results did not consider the age in the multivariable meta-regression. Instead, we conducted a supplementary analysis to fit the univariable meta-regression for age (details are presented in Supplementary Materials, Table S5). Findings indicated that the age group of 80 years and over showed a higher prevalence of sarcopenia compared with that of 60 to 70 years, which is consistent with current knowledge that the advanced age is an important risk factor of sarcopenia [50]. Body mass index (BMI) was also reported to be associated with the prevalence of sarcopenia in some literature, with a higher BMI relating to the lower prevalence of sarcopenia [46,50]. In this systematic review, BMI values reported in articles were all above 23 kg/m<sup>2</sup> for both genders (Supplementary Materials, Table S4). We did not include BMI in the meta-regression, considering that the comparison between overweight and obesity might be less meaningful. For other possible factors which might influence the variance of prevalence reported, such as rural-urban regions, comorbidities, and others (e.g., type of dynamometer, protocol of gait speed test), due to limited information reported in the articles, we could not analyze them through meta-regression.

The present study was subject to several limitations. Firstly, a few of the included studies reported the prevalence by different age groups, so we could not pool the prevalence rate by age subgroup. However, age is a well-established factor of sarcopenia. The lower bound of age for participants in the studies included in this review ranged considerably, from 60 to 80 years. Therefore, age may be a significant source of the substantial heterogeneity that we found across studies. Future epidemiological research of sarcopenia is encouraged to report age-specific prevalence. Secondly, due to the small number of studies which reported the prevalence of sarcopenia for outpatients, we were unable to obtain a pooled prevalence rate for this group. Thirdly, we did not include all of the possible covariates in our meta-regression analysis to investigate the source of considerable variations in sarcopenia prevalence. For example, all of the included studies used dynamometers to assess muscle strength, but we had insufficient information to investigate the potential influence of the type of dynamometer used on the prevalence rate. We were likewise unable to analyze the influence of different measurement protocols used for the gait speed tests due to insufficient information. Finally, our results should be interpreted with caution because most of the included studies had a moderate risk of bias and significant heterogeneity.
