Introduction

Bone fractures, especially osteoporotic fractures, represent a great burden on the individual and public health care system [1]. They not only have high associated morbidity, hospitalization, permanent disability but also mortality. However, it is estimated that approximately half of the clinical fractures occur in postmenopausal women without osteoporosis based on bone mineral density (BMD) [2]. Hence, it is necessary to identify other risk factors that may help identify people at high risk of fractures.

Sarcopenia is regarded as one of the four major components of muscle wasting disease [3] and an independent condition by an International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10-CM) Code [4]. There are several international published definitions of sarcopenia, including Newman [5], the European Working Group on Sarcopenia in Older People (EWGSOP) [6], Asian Working Group for Sarcopenia (AWGS) [7], International Working Group on sarcopenia (IWGS) [8], and Foundation of the National Institutes of Health (FNIH) [9]. Newman was defined only in terms of low muscle mass; the latter criteria were defined as a loss of muscle mass associated with a loss of muscle function. The cut-off points are the major difference in these definitions, because of the lifestyle, environmental, ethnic differences in body shape or sizes, and cultural backgrounds [7, 10].

In the past several years, sarcopenia has been recognized as an important geriatric condition. There are increasing volumes of articles on the impact of sarcopenia on health outcomes. The decline in muscle mass, function, and strength could contribute to adverse consequences on individual and public health, such as poor quality of life [11, 12], function decline [13], physical disabilities [14], hospitalization [15], and ultimate mortality [16]. In addition, several studies found that sarcopenia was associated with fractures in older adults [17,18,19,20]. However, these consequences were determined from both cross-sectional studies and longitudinal studies, which lead to potential difficulties in attributing causality. Furthermore, the association between sarcopenia and fractures can also differ depending on the definition used for the diagnosis of sarcopenia. Some studies have failed to show that sarcopenia alone significantly associated with fractures [21,22,23,24].

Oliveira and colleagues [25] performed a systematic review describing the association between sarcopenia and osteoporotic hip fracture in 2015. However, almost all studies included in this review were cross-sectional studies, and they did not perform a pooled analysis. In addition, another systematic review [26] that purposed to assess the association sarcopenia and the clinical and socioeconomic consequences (fractures, falls, mortality, functional decline, hospitalization and the length of hospital stay) was published in last year. This review was comprehensive. However, only two studies reported the relationship of sarcopenia and fractures, and the outcomes were simply analyzed and described. Because of the aforementioned reasons, we performed a comprehensive systematic review and meta-analysis of current cohort studies to explore whether sarcopenia is a risk factor for fractures among community-dwelling older adults.

Methods

Search strategy

This study was carried out based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses protocol for reporting systematic reviews and meta-analyses [27]. The following electronic databases were searched: EMBASE (from January 1, 1974, to December 2017), Ovid MEDLINE (from January 1, 1946, to December 2017), Cochrane Central Register of Controlled Trails (from inception to December 2017), and PubMed (from inception to December 2017). Our search terms used for this research include sarcopenia (sarcopenic; muscle mass; muscle strength; muscle atrophy; muscle wasting; myopenia; myopenic; dynapenia; dynapenic) and fracture (broken bone; osteoporotic fractures; fracture*; fractures, bone) and aged (aging; geriatrics; elderly; geriatric*; older adult*; older people). In addition, the reference lists of the eligible articles were also searched to avoid missing any relevant studies.

Inclusion and exclusion criteria

In this systematic search, rigorous inclusion criteria were as follows: (1) participants: older adults (aged 60 years old and older); (2) definition of sarcopenia: studies using the clear definition of sarcopenia; (3) studies reporting hazard ratios (HRs), odds ratios (ORs) or risk ratios(RRs), and their corresponding 95% confidence intervals (CIs) for the relationship between sarcopenia and fractures; (4) study design: cohort studies; and (5) data limitation: from inception to November 2017. Studies excluded were as follows: (1) article type: reviews, conference abstract; (2) language: not English or Chinese.

Study selection and data extraction

The review of potentially eligible studies identified by the searches was completed by two independent reviewers to identify reports for review in full text. Each full-text study was reviewed for eligibility by these reviewers. Data were extracted and summarized independently, according to a standardized data extraction form. The presence or absence of sarcopenia at study baseline was the primary exposure variable. The key outcome variable was the incidence of fracture during follow-up. Any study reported HRs, ORs, or RRs for fracture in individuals with sarcopenia and nonsarcopenia was extracted. Other data from the eligible studies included the following: first author, publication years, sociodemographic (country, type of population, age, the proportion of male), sample size, description of groups, duration of follow-up, frequency of follow-up, methods used for assessing fracture, and statistical adjustment for confounding variables (Table 1). We also extracted sarcopenia criteria, the prevalence of sarcopenia, and the tool and cutoff points of muscle mass, muscle strength, and physical performance (Table 2). If the dataset from the same population had been published in not only one publication, then only the study with longer follow-up period was included in our systematic review and meta-analysis. If studies divided the entire cohorts into male cohort and female cohort separately, then we considered the study as two independent studies. If some studies did not provide required estimates, we calculated those using standard methods. For studies with insufficient information, we contacted the primary authors, when possible, to acquire the data. All extracted data were cross-checked by reviews, and disagreements were resolved by consensus with a third reviewer.

Table 1 Characteristics of studies about the association between sarcopenia and fractures
Full size table
Table 2 Study and sarcopenia criteria
Full size table

Assessment of risk of bias

Two authors independently assessed the risk of bias using the Newcastle-Ottawa Scale (NOS), designed for nonrandomized studies [28], which consists of three factors: patient selection (four items), comparability of the study groups (one item), and assessment of outcome (three items). The overall score ranges from one to nine, nine points represented the highest methodological quality, and NOS scores of ≥ 5 were considered to be of high-quality publications [29]. Disagreements were also settled down by discussion among the third author.

Statistical analysis

HRs, ORs, RRs, and their 95% CIs were assessed for determining the relationship between sarcopenia and fractures. Although published studies presented several estimates of the association between sarcopenia and bone fractures, HRs and ORs were deemed approximate to the RRs [30]. Cochran’ s Q statistic using chi-square and I square (I2) test were performed to assess the impact of study heterogeneity on the results of the meta-analysis, and I2 value was considered to reflect mild heterogeneity (0–25%), moderate heterogeneity (26–75%), and high heterogeneity (76–100%) [31]. Random effect models were chosen if heterogeneity existed. Otherwise, the fixed effect models were used. We performed subgroup analysis to detect probable sources of heterogeneity. The predefined criteria for subgroup analyses were as follows: gender (male, female or both sexes), follow-up duration (< 10 or ≥ 10 years), methods used to assess bone fractures (self-reported or medical records), sarcopenia criteria (EWGSOP, AWGS, and other), and fracture site (all fractures, hip fracture). Medical records included radiology reports and hospital authority electronic database. Because of the limited number of studies, we just perform subgroup analyses based on gender and sarcopenia criteria. Moreover, sensitivity analyses were done evaluating whether the overall estimate depended on the effect size from a single study. We drew funnel plot for outcomes, and Egger’s and Beggar’s tests were done to plot the log RR against its standard error for assessment of potential publication bias [32]. All analyses were analyzed using the statistical software (STATA, version 11.0, Stata Corp, College Station, TX, USA). We used two-tailed P values and P < 0.05 was considered significant.

Results

Search processes

A total of 1468 articles were initially identified through three electronic databases. Among these articles, 675 duplicate articles were removed and 774 articles were excluded by screening the title and abstract. Therefore, only 19 articles remained for full-text review. Of these articles, nine were removed due to being not cohort studies (e.g., cross-sectional study, case-control study), and one [33] was removed due to having no clear definition of sarcopenia. Finally, nine prospective studies [17, 18, 22,23,24, 34,35,36,37] were included in our systematic review. Two studies contained the same population in men (n = 2000), so we choose the outcome which was longer periods of follow-up to avoid double counting data. The flow diagram of article search process is available in Fig. 1.

Fig. 1
figure 1

Flowchart of study selection

Full size image

Included studies

Table 1 presents the characteristic of seven included studies. The overall sample size was 31,513 individuals. All of these studies were quite recent since they were published between 2014 and 2017. Three studies [22, 23, 34] were performed in the USA, two [17, 24] in China, two [35, 37] in Australia, one [18] in Switzerland, and one [36] in Netherlands. All of these studies included community-dwelling older adults. Three studies [17, 22, 35] were conducted on only male subjects, one study [34] was conducted on only female subjects, two [18, 36] on both gender, and three studies [23, 24, 37] divided the entire cohorts into male cohort and female cohort separately. The number of participants ranged from 496 [36] to 10,937 [34], and the follow-up periods varied from 3.4 [18] to 15.9 years [34]. The outcome included incident hip fracture, incident fracture, nonspine fractures, and nonvertebral fractures. Three studies [17, 24, 36] had collected data on fractures using hospital electronic database, four studies [18, 23, 34, 37] used self-reported questionnaires, and two studies [22, 35] used radiology reports. Seven studies directly reported adjusted ORs and HRs and 95% CIs for the association between sarcopenia and fractures. Two studies [35, 37] divided participants into four categories: nonsarcopenic nonobese, nonsarcopenic obese, sarcopenic nonobese, and sarcopenic obese. We merged the first two into nonsarcopenic group, and merged the latter two into the sarcopenic group. Then, we calculated the RR and 95% CI of this study by 2 × 2 contingency tables, which were presented in the tables describing the demographics of study populations. Two studies [23, 34] included more than two groups; we choose the outcome of sarcopenia-alone subjects compared with the nonsarcopenic and normal BMD subjects.

Sarcopenia criteria and prevalence

Five studies [18, 22, 23, 35, 36] adopted European Working Group on Sarcopenia in Older People (EWGSOP), two studies [17, 24] adopted Asian Working Group for Sarcopenia (AWGS), and two studies [34, 37] used the definition of Newman. Table 2 presents the sarcopenia criteria among included studies and shows the different tools and cutoff points of muscle mass, muscle strength, and physical performance. Muscle mass was measured with dual-energy X-ray absorptiometry (DXA) in all of studies. Six studies used both low muscle mass and low muscle function to define sarcopenia, and three studies [18, 34, 37] determined sarcopenia only by low muscle mass. The prevalence of sarcopenia varied from 4.3 [22] to 31.9% [36], 7.3 [24] to 9.4% [17], and 19.9 [34] to 33.1% [37] in EWGSOP, AWGS, and Newman, respectively.

Quality assessment

The quality assessment using NOS is shown in supplemental Table 1. The scores ranged from 6 to 8. Four studies [17, 22, 35, 36] achieved 8 points, three studies [18, 23, 24] achieved 7 points, and two studies [34, 37] achieved 6 points. All included studies had high quality.

The relationship between sarcopenia and risk of fractures

All studies were included in meta-analysis. The association between sarcopenia and risk of fractures is shown in Fig. 2. Sarcopenia was significantly associated with the risk of fractures (combined effect size = 1.34, 95% CI = 1.13–1.58, P = 0.001). Mild heterogeneity was observed in this outcome (I2 = 5.5%, P-heterogeneity = 0.391). We performed subgroup analyses stratified by sarcopenia criteria and sex based on a random effects model. We found a significant positive association when sarcopenia was defined by AWGS (combined effect size = 1.78, 95% CI = 1.25–2.54, P = 0.001). However, the association between sarcopenia and fractures was not significant when EWGSOP (combined effect size = 1.20, 95% CI = 0.96–1.50, P = 0.112) or Newman (combined effect size = 1.25, 95% CI = 0.79–1.98, P = 0.339) was used (Fig. 3). In addition, we found that sarcopenia was significantly associated with the higher incidence of fractures in male (combined effect size = 1.39, 95% CI = 1.13–1.71, P = 0.002), but not in female (combined effect size = 1.09, 95% CI = 0.63–1.87, P = 0.169) and both gender (combined effect size = 1.41, 95% CI = 0.58–3.39, P = 0.447) (Fig. 4).

Fig. 2
figure 2

Forest plot of effect sizes for the association between sarcopenia and fractures. ES, effect size. *The RRs reported in the Scott et al. studies were crude RRs. All other outcomes were adjusted

Full size image
Fig. 3
figure 3

Forest plot of effect sizes for the association between sarcopenia and fractures across different definitions of sarcopenia by using a random effects model. ES, effect size; EWGSOP, the European Working Group on Sarcopenia in Older People; AWGS, Asian Working Group for Sarcopenia. *The RRs reported in the Scott et al. studies were crude RRs. All other outcomes were adjusted

Full size image
Fig. 4
figure 4

Forest plot of effect sizes for the association between sarcopenia and fractures across different gender groups by using a random effects model. ES, effect size. *The RRs reported in the Scott et al. studies were crude RRs. All other outcomes were adjusted

Full size image

Sensitivity analysis

We performed a sensitivity analysis by omitting one single study each time and pooling the others to check which study influenced the main effect. The result of sensitivity analysis for all studies is shown in supplemental Fig. 1. The stability of results had no significant changes, which validated the reliability of our analysis.

Publication bias

We used Egger’s and Beggar’s tests to assess the publication bias in this meta-analysis. Finally, asymmetry was observed by visual inspection of funnel plot (supplemental Fig. 2). However, the results did not show any statistically significant publication bias among the studies using Egger’s test (P = 0.876) and Beggar’s test (P = 0.495).

Discussion

We performed this systematic review and meta-analysis to evaluate whether sarcopenia is a predictive factor for fractures in community-dwelling older adults. We found that associations between sarcopenia and fractures were significant when using the AWGS definition and studies in males. However, sarcopenia was not associated with the risk of fractures when EWGSOP or Newman was used and studies in females. No significant heterogeneity was observed across these studies (I2 = 5.5%, P-heterogeneity = 0.391).

One systematic review [25] described the association between sarcopenia and osteoporotic hip fracture in 2015. However, the authors only included one cohort study in their review, which may have reduced the strength of their conclusion. In addition, this review only examined hip fracture risk but did not assess the risk of all fractures. Another meta-analysis [26] that purposed to assess the association sarcopenia and the clinical and socioeconomic consequences was published last year. This review was comprehensive. However, only two studies reported the relationship of sarcopenia and fractures, and the outcomes were simply described. In comparison with this previous meta-analysis, our meta-analysis included nine cohort studies and focused on all fractures. To the best of our knowledge, this is the most comprehensive review to summarize earlier cohort studies on all definitions of sarcopenia and fractures.

Several plausible mechanisms have been proposed for the associations between sarcopenia and the risk of fractures. On the one hand, low muscle mass changes in muscle related proteins, such as myokines, and also is associated with abnormal glucose metabolism, which has a great impact on bone metabolism [38]. Second, sarcopenic individuals are at high risk of falls, which leads to higher incidence of fractures. At last, the decline in muscle function and strength is associated with low mechanical loading, thus also affecting bone mass directly. Ormsbee et al. also researched that loss of muscle commonly combined with the loss of bone, making these individuals at high risk of fractures [39]. Therefore, sarcopenia is considered an effective predictor of fracture risk in older adults.

A positive result was found when sarcopenia was defined by AWGS. Asia is the most populated region in the world with a wide range of society, lifestyles, culture, ethnic, and religious backgrounds. The population size and the rapid population aging, thus the impact of sarcopenia, may be stronger than other regions [7]. In addition, our finding demonstrated the gender differences in the relationship of sarcopenia and fractures. Sarcopenia was associated with fractures in men, but we did not find an increased risk of fractures with sarcopenia in women, which may be partly explained by the fact that muscle strength decline in men is generally two times faster compared to women [40]. Though higher muscle mass was observed at baseline in men, it has been suggested that men could have a rapid age-related decline in muscle mass compared to women [41]. What is more, epidemiological studies have shown that higher testosterone levels have been associated with less loss of muscle mass and strength in older man [42, 43]. Men lose more testosterone than women with age increasing, which leads to the onset and severity of sarcopenia [44]. Therefore, the impact of sarcopenia on fracture risk prediction in men is more prominent.

Prevalence of sarcopenia is difficult to establish. A meta-analysis found a prevalence ranging from 1 to 29% in community-dwelling older adults [45]. In our research, the prevalence of sarcopenia ranged from 4.3 to 31.9%, 7.3 to 9.4%, and 19.9 to 33.1% in EWGSOP, AWGS, and Newman, respectively. This prevalence could differ depending on geographic regions, study population and the definitions of sarcopenia.

There are multiple highlights in our study. The most important is that all the studies included in this meta-analysis were prospective cohort studies, and thus, they do provide a higher level of evidence. The second advantage is that almost all of our results were based on adjusted estimates, which considered some potential confounding factors and provided results that are more accurate. In addition, the setting of the participants only included community-dwelling older adults, and we performed subgroup analyses to evaluate the pooled effect, which can help to reduce and examine the heterogeneity and analyze the results in detail.

Despite these strengths, some limitations in our study should also be taken into account. First, we choose different diagnostic criteria of sarcopenia, and different cutoff points were used in studies, thus may influence the research outcome and quality. The second limitation is that the number of studies included in this analysis was insufficient, especially in terms of a subgroup analysis. Only results of published studies were included in this meta-analysis. There may be numerous cohort studies currently in progress around the world, and a number of these studies could potentially have measured the relevant constructs, but without publishing these results yet. Moreover, some studies reporting negative results are difficult to publish. Therefore, potential publication bias is likely to exist, in spite of no evidence obtained from our statistical tests. Thirdly, fractures were assessed by self-report in some studies and may be subject to recall bias. However, fractures are major life events and inaccuracy of recall is unlikely. At last, some studies included in this meta-analysis reported nonspine fractures or nonvertebral fracture risk, and we treated these fractures as all fractures. Future studies are needed to focus on the association between sarcopenia and fracture of different sites.

Conclusion

We found that compared to nonsarcopenic, the associations between sarcopenia and fractures among community-dwelling older people were significant only when using the AWGS definition, and only for males in our study. Future studies are needed to establish a possible association between sarcopenia definitions and risk of fracture of different sites.