Learn more: PMC Disclaimer | PMC Copyright Notice
Abdominal Obesity and Risk of Hip Fracture: A Systematic Review and Meta-Analysis of Prospective Studies
2,4,6Abstract
Data on the association between general obesity and hip fracture were summarized in a 2013 meta-analysis; however, to our knowledge, no study has examined the association between abdominal obesity and the risk of hip fracture. The present systematic review and meta-analysis of prospective studies was undertaken to summarize the association between abdominal obesity and the risk of hip fracture. We searched online databases for relevant publications up to February 2017, using relevant keywords. In total, 14 studies were included in the systematic review and 9 studies, with a total sample size of 295,674 individuals (129,964 men and 165,703 women), were included in the meta-analysis. Participants were apparently healthy and aged ≥40 y. We found that abdominal obesity (defined by various waist-hip ratios) was positively associated with the risk of hip fracture (combined RR: 1.24, 95% CI: 1.05, 1.46, P = 0.01). Combining 8 effect sizes from 6 studies, we noted a marginally significant positive association between abdominal obesity (defined by various waist circumferences) and the risk of hip fracture (combined RR: 1.36; 95% CI: 0.97, 1.89, P = 0.07). This association became significant in a fixed-effects model (combined effect size: 1.40, 95% CI: 1.25, 1.58, P < 0.001). Based on 5 effect sizes, we found that a 0.1-U increase in the waist-hip ratio was associated with a 16% increase in the risk of hip fracture (combined RR: 1.16, 95% CI: 1.04, 1.29, P = 0.007), whereas a 10-cm increase in waist circumference was not significantly associated with a higher risk of hip fracture (combined RR: 1.13, 95% CI: 0.94, 1.36, P = 0.19). This association became significant, however, when we applied a fixed-effects model (combined effect size: 1.21, 95% CI: 1.15, 1.27, P < 0.001). We found that abdominal obesity was associated with a higher risk of hip fracture in 295,674 individuals. Further studies are needed to test whether there are associations between abdominal obesity and fractures at other bone sites.
Introduction
Bone fractures, especially osteoporotic fractures, impose a great burden on the health care system annually (1). Worldwide, the annual incidence of osteoporosis-related fractures is estimated to reach 3 million in 2025 (2). Hip fracture is common among osteoporotic fractures and is associated with high rates of disability, morbidity, and mortality (3–5). Finding appropriate approaches to decrease the risk of hip fracture is urgently required.
The pathogenesis of hip fracture is complex (6). Impaired bone strength or decline in bone mineral density (BMD) and trauma from falling are factors that contribute to hip fracture (7). Physical activity, alcohol consumption, smoking, and dietary intakes also can contribute to hip fracture (8–12). In parallel with the increasing trend in the prevalence of obesity, the prevalence of hip fracture also is rising (13, 14). Although a 2013 meta-analysis suggested an inverse relation between general obesity and the risk of hip fracture in adults (15), findings from prospective studies on the association between abdominal obesity and the risk of hip fracture are conflicting. Some studies have shown a positive association between abdominal obesity and hip fracture (16, 17), whereas others have failed to reach significant associations or have found inverse relations (18, 19). Considering the alarming prevalence of abdominal obesity, particularly among women of postmenopausal age for whom the risk of fractures also is high, finding contributing factors to fractures take priority.
Data on the association between general obesity and hip fracture were summarized in the meta-analysis of Tang et al. (15); however, to our knowledge no meta-analysis has examined the association between abdominal obesity and the risk of hip fracture. We aimed, therefore, to systematically review the current evidence on this association and to perform a meta-analysis summarizing earlier findings in this regard.
Methods
Search strategy
This study was performed based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses protocol for reporting systematic reviews and meta-analyses (20). A systematic search was conducted in the online databases of PubMed, Institute for Scientific Information Web of Science, Scopus, ProQuest, Science Direct, and Embase for relevant publications up to February 2017. The keywords used in our search strategy were [“Obesity” (MAJR) or “obesity/complications” or “obesity, abdominal” or “intra-abdominal fat” or “abdominal fat” or “waist circumference” or “waist-hip ratio” or “waist-height ratio” or “abdominal obesity” or “abdom*” or “abdominal fatness” or “waist” or “WC” or “WHR” or “WHtR” or “anthropometry” or “fatness” or “visceral” or “abdominal obesit*” or “central obesit*” or “viscera”] and (“fractures, bone” or “osteoporotic fractures” or “fractures” or “fracture” or “broken bone”). No restriction was made as to the time of publication and language. In addition, the reference lists of the relevant articles were reviewed to avoid missing any publication. Unpublished studies were excluded. In addition, duplicate citations were removed after the search was completed.
Inclusion criteria
In this meta-analysis, publications with the following criteria were eligible for inclusion: 1) all studies evaluating the association between abdominal obesity and hip fracture or any fracture (including hip fracture); 2) studies that had a prospective design; 3) studies that were conducted on apparently healthy adults (>18 y); and 4) studies that reported ORs, RRs, or HRs along with 95% CIs for the relation between abdominal obesity and hip fracture. If the same dataset had been published in >1 publication, then only the article with more complete findings was included in our systematic review and meta-analysis.
Data extraction
Data extraction and study selection were conducted independently by 2 investigators, and any disagreements were resolved by discussion. The presence or absence of abdominal obesity at study baseline was the key exposure variable. The key outcome variable was incidence of hip fracture during follow-up. Any reported ORs, HRs, or RRs for hip fracture in individuals with abdominal obesity, compared with the reference group, were extracted. Although published studies presented several estimates of the association between abdominal obesity and bone fractures, we considered ORs, HRs, or RRs and their corresponding 95% CIs as a priori analyses in each study. We also extracted the following information from each study: first author, year of publication, cohort name, country of origin, age range at study baseline, sex, sample size, number of participants with incident hip fracture, duration of follow-up, person-year, methods used for assessing abdominal obesity and hip fracture, and statistical adjustment for confounding variables (Tables 1 and and2).2). If some studies did not provide required estimates, then we calculated these estimates using standard methods. If a study presented sex-stratified results, then we considered the study to be 2 separate studies.
TABLE 1
Risk of hip fracture across categories of abdominal obesity indicators in adults aged ≥40 y1
| Authors (ref) | Country | Age range, y | Sample size | n Cases | Duration of follow-up, y | Person-year | Exposure | Exposure assessment | Outcome | Outcome assessment | Comparison | OR, RR, or HR (95% CI) | Adjustment2 |
| Meyer et al. (16) | United States | 40–75 | 35,488 men | 483 | 26 | 568,298 | WC; WHR | Self-reported | Hip fracture | Self-reported | WC: ≥107 vs. <89 cm; WHR: ≥1.00 vs. <0.91 | WC—RR: 0.78 (0.48, 1.27); WHR—RR: 1.04 (0.78, 1.38) | 1, 7, 8, 11–14, 19–23, 25, 30, 32 |
| United States | 40–65 | 61,677 women | 1168 | 26 | 898,925 | WC; | Self-reported | Hip fracture | Self-reported | WC: ≥96 vs. <72 cm; | WC—RR: 1.33 (0.99, 1.79); | 1, 7, 8, 11–14, 19–23, 25, 30, 32 | |
| WHR | WHR: ≥0.9 vs. <0.75 | WHR—RR: 1.29 (1.04, 1.61) | |||||||||||
| Benetou et al. (19) | 5 European countries | 60–86 | 27,982 men and women | 522 | 8 | 232,639 | WHR | Tape measure | Hip fracture | Self-reported and hospital discharge records based on ICD-10 | Men: >1.02 vs. <0.92; women: >0.89 vs. <0.77 | HR: 0.87 (0.54, 1.40) | 1, 8, 28, 30, 32 |
| Parker et al. (21) | United States | 55–69 | 30,652 women | 768 | 18 | 409,601 | WC | Self-measured | Hip fracture | Self-reported | >0.90 vs. <0.76 cm | HR: 0.93 (0.74, 1.17) | 1, 3, 7, 8, 12, 25, 28, 40 |
| Yang et al. (22) | Australia | ≥50 | 360 men | 19 | 5 | — | aFM | X-ray | Any fracture | X-ray | — | — | 1, 5–7, 9, 30 |
| 766 women | 107 | 5 | — | aFM | X-ray | Any fracture | X-ray | ≥2.1 vs. <1.5 Kg | HR: 0.49 (0.30, 0.80) | ||||
| Søgaard et al. (23) | Norway | 60–79 | 23,061 men | 889 | 8.1 | 188,953 | WC; WHR | Tape measure | Hip fracture | Electronic discharge registers based on ICD-9, ICD-10 | WC: >99 vs. <90 cm; WHR: >0.96 vs. <0.90 | WC—HR: 1.94 (1.48, 2.54); WHC—HR: 1.41 (1.16, 1.71) | 31, 32 |
| 19,918 women | 1498 | 8.1 | 173,908 | WC; WHR | Tape measure | Hip fracture | Electronic discharge registers based on ICD-9, ICD-10 | WC: >90 vs. <80 cm; WHR: >0.85 vs. <0.79 | WC—HR: 1.84 (1.49, 2.27); WHR—HR: 1.49 (1.29, 1.72) | 31, 32 | |||
| Lee et al. (24) | Korea | ≥50 | 16,078 men | 158 | 3 | 48,201 | WC | Tape measure | Osteoporotic fracture | Database based on ICD-10 | ≥90 vs. <90 cm | HR: 0.65 (0.40, 1.06) | 1, 5, 7, 8, 15–17, 20, 24, 26, 27, 32 |
| Hansen et al. (25) | United States | 55–69 | 34,703 women | 4378 | 6.5 | 187,035 | WHR | Self-measured | Any fracture | Self-reported | >0.90 vs. <0.76 | RR: 1.06 (0.96, 1.19) | 1, 7, 8, 12, 18, 19, 23, 25, 32 |
| Szulc et al. (26) | French | 50–85 | 762 men | 82 | 10 | — | WC | Tape measure | Any fracture | Self-reported | >102 vs. ≤102 cm | OR: 0.43 (0.17, 1.08) | 1, 3, 5, 9, 32, 41, 42 |
| Owusu et al. (27) | United States | 40–75 | 43,053 men | 38 | 8 | — | WC; WHR | Self-measured | Hip fracture | Self-reported | WC: >107 vs. <84 cm; WHR: >1.01 vs. <0.87 | WC—RR: 3.11 (0.66, 14.6); WHR—RR: 3.92 (1.07, 14.3) | 1, 8, 19, 25, 30–32 |
TABLE 2
Risk of hip fracture based on 1-U increase or decrease in indicators of abdominal obesity in adults aged ≥40 y1
| Authors (ref) | Country | Age range, y | Sample size | n cases | Duration of follow-up, y | Person-Y | Exposure | Exposure assessment | Outcome | Outcome assessment | Per unit increase or decrease | OR, RR, or HR (95% CI) | Adjustment2 |
| Meyer et al. (16) | United States | 40–75 | 35,488 men | 483 | 26 | 568,298 | WC; WHR | Self-reported | Hip fracture | Self-reported | WC: per 10-cm increase; WHR: per 0.1-U increase | WC—RR: 0.98 (0.83, 1.15); WHR—RR: 1.00 (0.86, 1.18) | 1, 7, 8, 11–14, 19–23, 25, 30, 32 |
| United States | 40–65 | 61,677 women | 1168 | 26 | 898,925 | WC; | Self-reported | Hip fracture | Self-reported | WC: per 10-cm increase; | WC—RR: 1.13 (1.04, 1.23); | 1, 7, 8, 11–14, 19–23, 25, 30, 32 | |
| WHR | WHR: per 0.1-U increase | WHR—RR: 1.14 (1.05, 1.23) | |||||||||||
| Machado et al. (17) | Brazil | ≥65 | 433 women | 28 | 4.3 | — | VAT | DXA | Nonspine fracture | Self-reported confirmed by x-ray | Per 100 g/mass increase | OR: 1.42 (1.09, 1.85) | 1, 4–6 |
| Kauppi et al. (18) | Finland | ≥55 | 2300 men and women | 96 | 9.8 | 22,540 | WC | Tape measure | Hip fracture | Electronic discharge registers based on ICD-10 | Per 1-SD increase | HR: 0.71 (0.55, 0.92) | 1, 2, 30, 35–39 |
| Benetou et al. (19) | 5 European countries | 60–86 | 27,982 men and women | 522 | 8 | 232,639 | WHR | Tape measure | Hip fracture | Self-reported and hospital discharge records based on ICD-10 | Per 0.1-U increase | HR: 0.97 (0.79, 1.20) | 1, 8, 28, 30, 32 |
| Sornay-Rendu et al. (28) | France | ≥40 | 595 women | 85 | 13.1 | VFAT | DXA | Osteoporotic fracture | X-ray | Per 1-SD increase | HR: 0.75 (0.55, 1.03) | 1, 5–7, 9, 10 | |
| Yang et al. (22) | Australia | ≥50 | 360 men | 19 | 5 | — | aFM | X-ray | Any fracture | X-ray | Per 1-kg aFM decrease | HR: 0.84 (0.42, 1.72) | 1, 5–7, 9, 30 |
| 766 women | 107 | 5 | — | aFM | X-ray | Any fracture | X-ray | Per 1-kg aFM decrease | HR: 1.50 (1.10, 2.05) | 1, 5–7, 9, 30 | |||
| Søgaard et al. (23) | Norway | 60–79 | 23,061 men | 889 | 8.1 | 188,953 | WC; WHR | Tape measure | Hip fracture | Electronic discharge registers based on ICD-9, ICD-10 | WC: per 10-cm increase; WHR: per 0.1-U increase | WC—HR: 1.56 (1.36, 1.78); WHR—HR: 1.38 (1.22, 1.56) | 31, 32 |
| 19,918 women | 1498 | 8.1 | 173,908 | WC; WHR | Tape measure | Hip fracture | Electronic discharge registers based on ICD-9, ICD-10 | WC: per 10-cm increase; WHR: per 0.1-U increase | WC—HR: 1.32 (1.20, 1.44); WHR—HR: 1.25 (1.14, 1.36) | 31, 32 | |||
| Nguyen et al. (29) | Australia | ≥60 | 78 men | 26 | 14 | — | aFM | DXA | Hip fracture | X-ray | Per 10% aFM decrease | OR: 1.8 (0.80, 4.00) | 29 |
| 189 women | 63 | 14 | — | aFM | DXA | Hip fracture | X-ray | Per 10% aFM decrease | OR: 1.1 (0.70, 1.7) | 29 | |||
| Sheu et al. (30) | United States | ≥65 | 749 men | 252 | 5.2 | — | VAT; SAT | DXA | Nonspine fracture | Validated self-reported measure | VAT: per 1-SD increase; SAT: per 1-SD increase | VAT—HR: 0.88 (0.70, 1.10); SAT—HR: 1.03 (0.78, 1.37) | 1, 4–6, 28, 33, 34, 36, 43 |
Excluded studies
In our meta-analysis we excluded letters, comments, short communication, reviews, meta-analyses, ecologic studies, and animal studies. In our initial search we found 1490 articles. We excluded 1412 studies on the basis of title and abstract. Except for prospective studies, relevant studies with cross-sectional or other designs were excluded in the screening phase based on title and abstract. The other 63 studies were excluded for the following reasons: 1) studies that examined the association between general obesity and risk of fracture (n = 36), 2) studies that assessed the association between abdominal obesity and bone deformities (n = 2) (31, 32), 3) studies that assessed the relation between obesity and bone fracture resulting from accident or severe trauma (n = 21), 4) studies that examined the association between abdominal obesity and non–hip bone fractures (n = 2) (33, 34), and 5) studies that were conducted on children or adolescents (n = 2) (35, 36, 55). Of the 15 articles remaining, 2 reported data from the same population (21, 37). We only included the one with the larger follow-up duration to avoid double counting data (21). Finally, 14 prospective studies [12 cohorts (16–19, 21, 22–28) and 2 nested case-control studies (29, 30)] were included in our systematic review (Figure 1).
Assessment of study quality
A version of the Newcastle Ottawa Scale (NOS), designed for nonrandomized studies, was used to assess the quality of the selected studies (38). The NOS considers a maximum of 9 points to each prospective study: 4 for selection, 2 for comparability, and 3 for assessment of outcomes (9 represents the highest quality). Any discrepancies were resolved by discussion. Those that achieved an NOS score of ≥5 in our study were considered to be high-quality publications (Supplemental Table 1).
Statistical analysis
RRs, ORs, and HRs (and their 95% CIs) for hip fracture were used to calculate log RRs and their SEs. We calculated the overall effect size using random-effects models, which take between-study variation into account. Because a random-effects meta-analysis tends to give disproportionally more weight to small, statistically less robust studies (39), fixed-effects models are more reliable than random-effects models. In addition to random-effects models, we applied fixed-effects models to find possible associations. Cochran’s Q test and I2 were applied to assess between-study heterogeneity. In this study I2 values of >50% were considered between-study heterogeneity (40). To detect probable sources of heterogeneity, we performed subgroup analyses. The predefined criteria for subgroup analyses were as follows: age (<50 or ≥50 y), sex (male, female or both sexes), follow-up duration (<10 or ≥10 y), geographic region (US or non–US countries), methods used to assess bone fractures (x-ray, self-reported or hospital records), and adjustment for BMI (adjusted or nonadjusted; kg/m2). Because of the limited number of studies, we did not perform subgroup analyses based on age group and methods used for assessing bone fractures. In these analyses we used fixed-effects models. In addition to the main analyses, we conducted sensitivity analyses based on main exposure and main outcome to determine whether the overall estimate depended on the effect size from a single study. If this was the case, then the analyses were rerun by excluding the mentioned study. Assessment of publication bias was made by visual inspection of the funnel plots. Statistical analyses were conducted with use of STATA version 11.2 (Stata Corp). P values were considered significant at <0.05.
Results
Findings from systematic review
Our systematic review included 14 studies [12 cohort studies (16–19, 21–28)] and 2 nested case-control studies (29, 30). The characteristics of these studies are shown in Tables 1 and and2.2. All of the included studies presented the required estimates, and we did not require calculating estimates based on formulas. The sample size of these studies varied from 267 to 97,165 (total: 311,436). Participants were apparently healthy and were aged ≥40 y. The included studies had been published between 1998 and 2017. Five studies were conducted in the United States (16, 21, 25, 27, 30), 5 in Europe (18, 19, 23, 26, 28), 2 in Australia (22, 29), 1 in Asia (24), and 1 in Brazil (17). Of the 14 studies, 4 were conducted on male subjects (24, 26, 27, 30), 4 on female subjects (17, 21, 25, 28), and 6 on either sex (16, 18, 19, 22, 23, 29). The mean duration of follow-up ranged from 3 to 26 y.
Abdominal obesity had been assessed by measuring the waist circumference (WC) in 4 studies (18, 21, 24, 26), the waist-hip ratio (WHR) in 2 studies (19, 25), both WC and WHR in 3 studies (16, 23, 27), visceral adipose tissue (VAT) in 3 studies (17, 28, 30), and abdominal fat mass (aFM) in 2 studies (22, 29). In addition to the assessment of abdominal obesity at the study baseline, some studies repeated this assessment during the follow-up period (16, 37). All of the included studies, even those that had repeated the measurement of exposure, had categorized participants based on the baseline values of abdominal obesity indicators. In 9 of 14 studies, the reference group was participants without abdominal obesity or those in the lowest category of abdominal obesity indicators (16, 19, 21–27). Seven studies had assessed abdominal obesity in relation to hip fracture (16, 18, 19, 21, 23, 27, 29), 3 in relation to any fracture (22, 25), 2 in relation to nonspine fractures (17, 30), and 2 other studies in relation to osteoporotic fracture (24, 28). Studies that considered any fracture, nonspine fracture, or osteoporotic fracture also had reported data for hip fracture separately. These studies, therefore, also were included in this systematic review. In the present study we considered nonspine fractures and osteoporotic fractures as any fracture. Seven studies had collected data on fractures using self-reported questionnaires (validated by x-ray in 2 studies) (16, 17, 21, 25–27, 30), 3 studies used hospital databases or electronic discharge registers (based on International Classification of Diseases, ninth or tenth editions) (18, 23, 24), and 1 study had used both self-reported and hospital discharge records based on International Classification of Diseases, tenth edition (19). In addition, bone fractures had been confirmed by use of x-ray in 3 studies (22, 28, 29). All of the included studies reported adjusted ORs, RRs, and HRs and 95% CIs for the association between abdominal obesity and hip fracture. Of the 14 studies, 7 reported BMI-adjusted effect sizes (16, 19, 23–27). Among 8 studies that categorized participants based on WC and WHR (16, 19, 21, 23–27), 4 studies had assessed the risk of bone fractures across quintiles of WC or WHR (19, 21, 25, 27), 1 examined this risk across tertiles of WC and WHR (23), and 3 had considered other cutoffs for WC and WHR (16, 24, 26). Although different cutoffs were used for categorizing participants according to WC and WHR, in 7 (16, 19, 21, 23, 25–27) of 8 studies participants in the highest category of WC and WHR were abdominally obese based on standard cutoffs (men: WC ≥102 cm and WHR ≥0.9, women: WC ≥88 cm and WHR ≥0.85) (41). One study classified participants based on the WC of <90 cm compared with ≥90 cm (24). All 8 studies considered subjects without abdominal obesity (based on standard cutoffs) as a reference group. In addition to such analyses, the risk of bone fractures was presented in 9 studies (Table 2) per 1-U increase or decrease in the indicators of abdominal obesity [WC, WHR, VAT, visceral fat adipose tissue (VFAT), subcutaneous adipose tissue, aFM] (16–19, 23–30). Based on the NOS, all of the included studies were high-quality studies (Supplemental Table 1).
Findings from the meta-analysis
Of the 14 studies included in the systematic review, 9 were included in the meta-analysis (16, 18, 19, 21, 23–27). Five studies were not included in the meta-analysis because they reported effect sizes for a 1-U (including SD, kilograms, 100 g/mass, and 10%) increase or decrease in aFM and VAT (17, 22, 28–30). Given the different units, these studies were excluded from the meta-analysis. Of 9 studies included in the meta-analysis, 3 examined abdominal obesity in relation to any fractures, not simply hip fracture (24–26). Because hip fracture was part of any fractures in the included studies, we examined the association between abdominal obesity and hip fracture with and without considering these 3 studies. The overall sample size of the 9 studies was 295,674 individuals, with 9819 cases of fractures (5201 hip fractures, 4460 any fractures, and 158 osteoporotic fractures). In the present study we considered osteoporotic fracture to be any fracture.
The association between abdominal obesity (based on WC) and risk of hip fracture is shown in Figure 2. Considering 8 effect sizes from 6 studies, abdominal obesity (based on WC) was not significantly associated with the risk of hip fracture (combined effect size: 1.14, 95% CI: 0.81, 1.59, P = 0.45). Between-study heterogeneity was significant (I2 = 84.7%; P-heterogeneity < 0.001). When we conducted fixed-effects analyses, we found a significant positive association (combined effect size: 1.32; 95% CI: 1.18, 1.48; P < 0.001; Supplemental Figure 1). Excluding 2 studies on any fracture and retaining only studies with hip fracture as the main outcome, the relation was marginally significant in a random-effects model (combined effect size: 1.36; 95% CI: 0.97, 1.89; P = 0.07; I2 = 83.9%; P-heterogeneity < 0.001; Supplemental Figure 2). When we performed fixed-effects meta-analyses, the association became significant (combined effect size: 1.40; 95% CI: 1.25, 1.58; P < 0.001; Supplemental Figure 3).
Forest plot for the association between abdominal obesity (based on waist circumference) and risk of fractures (including hip fractures and any fractures) in adults aged ≥40 y by use of a random-effects model. ES, effect size.
To find the source of heterogeneity, we performed subgroup analyses based on a fixed-effects model. Our findings in the subgroup analyses based on sex, geographic region (US compared with non-US countries), duration of follow-up (<10 y compared with ≥10 y), and adjustment for BMI are shown in Table 3. After excluding the studies on any fracture, geographic region and duration of follow-up explained between-study heterogeneity. From these analyses we found a significant positive association between abdominal obesity (based on WC) and the risk of hip fracture in men (combined effect size: 1.59; 95% CI: 1.26, 2.01; P < 0.001) and women (combined effect size: 1.34; 95% CI: 1.17, 1.54; P < 0.001), in studies with <10-y duration of follow-up (combined effect size: 1.89; 95% CI: 1.60, 2.23; P < 0.001), those conducted in noncountries (combined effect size: 1.88; 95% CI: 1.59, 2.22; P < 0.001), and those that considered BMI to be a covariate in the analysis (combined effect size: 1.63; 95% CI: 1.42, 1.87; P < 0.001). Between-study heterogeneity did not disappear, however, when the subgroup analysis was conducted based on sex and adjustment of BMI.
TABLE 3
Subgroup analysis based on fixed-effects models for the association between abdominal obesity and the risk of hip fracture in adults aged ≥40 y1
| Effect sizes, n | I2 | Q test | RR (95% CI) | P-between | |
| Abdominal obesity based on WC and hip fracture | |||||
| Overall | 6 | 83.9 | >0.001 | 1.40 (1.25, 1.58) | |
| Sex | 0.22 | ||||
| Male | 3 | 81.9 | 0.004 | 1.59 (1.26, 2.01) | |
| Female | 3 | 89.2 | >0.001 | 1.34 (1.17, 1.54) | |
| Country | <0.001 | ||||
| United States | 4 | 56.1 | 0.07 | 1.04 (0.88, 1.23) | |
| Non-US countries | 2 | 0 | 0.76 | 1.88 (1.59, 2.22) | |
| Follow-up, y | <0.001 | ||||
| <10 | 3 | 0 | 0.78 | 1.89 (1.60, 2.23) | |
| ≥10 | 3 | 59 | 0.08 | 1.02 (0.86, 1.21) | |
| Adjustment for BMI, kg/m2 | <0.001 | ||||
| Adjusted effect size | 5 | 71.8 | 0.007 | 1.63 (1.42, 1.87) | |
| Nonadjusted effect size | 1 | 0 | 0 | 0.93 (0.74, 1.17) | |
| Abdominal obesity based on WHR and hip fracture | |||||
| Overall | 6 | 73.2 | 0.002 | 1.36 (1.23, 1.49) | |
| Sex | 0.11 | ||||
| Male | 3 | 65.6 | 0.05 | 1.30 (1.11, 1.53) | |
| Female | 2 | 14.2 | 0.28 | 1.43 (1.26, 1.61) | |
| Both | 1 | — | — | 0.87 (0.54, 1.40) | |
| Country | 0.14 | ||||
| United States | 3 | 56.2 | 0.1 | 1.22 (1.02, 1.44) | |
| Non-US countries | 3 | 55.5 | 0.1 | 1.42 (1.27, 1.59) | |
| Follow-up, y | 0.08 | ||||
| <10 | 4 | 56.1 | 0.07 | 1.43 (1.28, 1.60) | |
| ≥10 y | 2 | 27.6 | 0.24 | 1.19 (1.00, 1.42) | |
| Adjustment for BMI, kg/m2 | |||||
| Adjusted effect size | 6 | 55.5 | 0.04 | 1.36 (1.23, 1.49) | |
| Nonadjusted effect size | 0 | 0 | 0 | 0 |
For studies that defined abdominal obesity based on WHR, we found that combining 7 effect sizes from 5 studies showed a significant positive association between abdominal obesity and risk of hip fracture (combined effect size: 1.24; 95% CI: 1.05, 1.46; P = 0.01); however, a significant between-study heterogeneity was found (I2 = 73.5%; P-heterogeneity = 0.001; Figure 3). When we performed fixed-effects analyses, similar findings were obtained (combined effect size: 1.22; 95% CI: 1.14, 1.31; P < 0.001; Supplemental Figure 4). After excluding 1 study on any fracture, the association did not change (combined effect size: 1.30; 95% CI: 1.10, 1.53; P = 0.007; Supplemental Figure 5). Such findings also were seen in a fixed-effects model (Supplemental Figure 6). To find the source of heterogeneity, subgroup analyses were done. Between-study heterogeneity disappeared completely when studies were divided by sex, geographic region, and duration of follow-up. This analysis revealed a significant positive relation between abdominal obesity (based on WHR) and hip fracture in men (combined effect size: 1.30; 95% CI: 1.11, 1.53; P = 0.001) and women (combined effect size: 1.43; 95% CI: 1.26, 1.61; P < 0.001), in studies done in US (combined effect size: 1.22; 95% CI: 1.02, 1.44; P = 0.02) and non-US countries (combined effect size: 1.42; 95% CI 1.27, 1.59; P < 0.001), and in studies with <10 y (combined effect size: 1.43; 95% CI: 1.28, 1.60; P < 0.001) and >10 y of follow-up (combined effect size: 1.19; 95% CI: 1.00, 1.42; P = 0.04).
Forest plot for the association between abdominal obesity (based on waist-hip ratio) and risk of fractures (including hip fractures and any fractures) in adults aged ≥40 y by use of a random-effects model. ES, effect size.
Three studies (16, 18, 23) reported effect sizes for the risk of hip fracture based on a 10-cm increase in WC. We found that a 10-cm increase in WC was not significantly associated with a higher risk of hip fracture (combined effect size: 1.13; 95% CI: 0.94, 1.36; P = 0.19; Figure 4); however, when we conducted fixed-effects meta-analyses, a 10-cm increase in WC was significantly associated with a 21% increase in the risk of hip fracture (combined effect size: 1.21; 95% CI: 1.15, 1.27; P < 0.001; Supplemental Figure 7). With regard to WHR, in combining 5 effect sizes from 3 studies (16, 19, 23), we found that a 0.1-U increase in WHR was associated with a 16% increase in the risk of hip fracture (combined effect size: 1.16; 95% CI: 1.04, 1.29; P = 0.007; Figure 5). This also was the case when we performed fixed-effects meta-analyses (Supplemental Figure 8).
Forest plot for the risk of hip fracture based a 10-cm increase in waist circumference in adults aged ≥40 y by use of a random-effects model. ES, effect size.
Forest plot for the risk of hip fracture based a 0.1-U increase in waist-hip ratio in adults aged ≥40 y by use of a random-effects model. ES, effect size.
On visual inspection of funnel plots we found no evidence of publication bias (Supplemental Figure 9). In addition, sensitivity analysis showed that combined effect sizes, obtained for the association between abdominal obesity and hip fracture, did not depend on a particular study or group of studies.
Discussion
We found that abdominal obesity (based on WHR) was significantly associated with a higher risk of hip fracture (combined effect size: 1.30; 95% CI: 1.10, 1.53; P = 0.007). Such findings also were seen in all of the subgroups, except in studies conducted on both sexes. Furthermore, we found a marginally significant positive association between abdominal obesity (based on WC) and risk of hip fracture (combined effect size: 1.36; 95% CI: 0.97, 1.89; P = 0.07). This relation did become significant in fixed-effects models (combined effect size: 1.40; 95% CI: 1.25, 1.58; P < 0.001), however. In addition, subgroup analyses revealed a significant positive association between abdominal obesity (based on WC) and the risk of hip fracture in men and women, and for studies with <10-y duration of follow-up, those conducted in non-US countries, and those that considered BMI as a covariate in the analysis. To the best of our knowledge, ours is the first study to summarize earlier studies on abdominal obesity and hip fracture.
Abdominal obesity and osteoporosis are increasing at an alarming rate (42–44). Individuals with osteoporosis are more likely to have bone fractures, particularly hip fractures (45). Abdominal obesity and hip fracture both are associated with increased risks of morbidity and mortality (46–50). Several studies have assessed the association between obesity (general and abdominal) and hip fracture (16, 51, 52); however, earlier studies focused mainly on general obesity (defined as BMI ≥30) (52, 53) rather than abdominal obesity. A meta-analysis showed that general obesity or high BMI was protectively associated with hip fracture (15). Little emphasis has been placed on abdominal obesity, however; data in this regard are conflicting.
We found that abdominal obesity was positively associated with the risk of hip fracture. In line with our findings, Folsom et al. (37) reported that WHR was slightly positively associated with the risk of hip fracture. VAT also was positively associated with nonspine fractures (17). Kim et al. (54) found an inverse association between VAT, WHR, and BMD. In opposition to our findings, 2 nested case-control studies showed no significant relation between abdominal obesity, hip fracture, and nonspine fractures (29, 30). In addition, Sornay-Rendu et al. (28) reported that VFAT was not significantly associated with osteoporotic fractures, including hip fractures. In healthy subjects, an inverse association was seen between WC and the risk of hip fracture (18). Furthermore, Berg et al. (55), in a cross-sectional study on a general adult population, reported that BMI, WC, and aFM were positively associated with bone stiffness and inversely associated with bone fractures. It must be kept in mind that studies that found no significant association or found an inverse association between abdominal obesity and the risk of bone fractures had mostly lower sample sizes than those that found a positive relation (18, 26, 28–30). In addition, unlike studies that found a positive association, adjustment for BMI was not often done in studies that found no significant association or found an inverse association (18, 21, 29, 30). Of 9 studies in our systematic review that reached no significant association or found an inverse association between abdominal obesity and the risk of bone fractures (18, 19, 21, 24–26, 28–30), 4 had included <1000 participants (26, 28–30) and 5 had not controlled for BMI in the analysis (18, 21, 28–30). Alternatively, 4 of 5 studies (16, 17, 22, 23, 27) that found a positive association had a sample size of >1000 individuals (16, 22, 23, 27), and 3 studies had controlled for BMI (16, 23, 27). Different cutoffs used in the included studies may be the reason for the weak association found between WC and hip fracture.
Low BMD is the main risk factor contributing to hip fracture (45). It has been shown that individuals with general obesity or higher weight and BMI have higher BMD and lower odds for having bone fracture than those of normal weight (56–59). It seems that heavier weight leads to increased strain on bones and can improve the structural integrity of bones (59). In contrast, we found that abdominal obesity was directly associated with hip fracture. This relation may be explained by the effects of abdominal obesity–related inflammation (57, 60). Earlier studies demonstrated that inflammatory cytokines (including resistin, TNF-α, IL-1, and IL-6) that release by VAT, uncouple bone remodeling by enhancing bone reabsorption and suppressing bone formation (61). Furthermore, adiponectin and leptin increase the outflow of sympathetic impulses on bone by affecting hypothalamic centers that regulate sympathetic tone. These impulses decrease osteoblast differentiation and increase osteoclast recruitment, thereby uncoupling the bone remodeling unit (62). In addition, it has been shown that higher high-sensitivity C-reactive protein levels are associated with a lower trabecular density, lower trabecular number, higher trabecular spacing, and more heterogeneous trabecular distribution (63). Abdominal obesity–relayed inflammation, therefore, can adversely influence trabecular bone score and bone quality index (64). It has been shown that abdominal obesity is associated with higher levels of inflammatory markers than is general obesity; however, heavier weight that leads to increased strain on bones can decrease the effects of inflammation on bones in individuals with general obesity (60, 65). Furthermore, abdominal obesity causes instability and impaired balance and thus increases the risk of falling and consequently bone fractures (66).
Although the present study was the first meta-analysis that, to our knowledge, assessed the association between abdominal obesity and hip fracture, some limitations of our study should be considered. Other studies considered different sites for bone fractures; however, we examined only hip fracture. Because of the limited number of studies with VAT, VFAT, and aFM as exposure, we excluded such studies from the present meta-analysis. Although subgroup analyses were performed based on sex, geographic region, duration of follow-up, and adjustment for BMI, we could not conduct these analyses using other variables, including age group and methods for assessing abdominal obesity and hip fracture, because of the limited number of studies. Among the 14 studies included in the present analysis, only 7 reported BMI-adjusted effect sizes (16, 19, 23–27), and most studies had not reported these estimates. In addition, in some studies, the effect of other influencing factors, particularly diet, were not controlled for; therefore, the observed associations may have been weakened or even disappeared after adjusting for other variables. Studies included in our review also differed in terms of statistical analyses (logistic compared with Cox proportional hazards regression) and methods of assessing bone fractures (self-reported compared with x-ray and electronic discharge registers).
In conclusion, we found that abdominal obesity, the definition of which was based on various WHRs, was adversely associated with the risk of hip fracture in 295,674 individuals studied. In addition, a positive association was seen between abdominal obesity (defined by various WCs) and the risk of hip fracture in a fixed-effects model; however, this relation was marginally significant in a random-effects model.
Acknowledgments
The authors’ responsibilities were as follows—OS, MN, BL, and AE: contributed to the conception, design, and statistical analyses; PS: contributed to the data analysis; and all authors: contributed to the interpretation of the data, drafted the manuscript, and read and approved the final manuscript for submission.
Footnotes
Abbreviations used: aFM, abdominal fat mass; BMD, bone mineral density; NOS, Newcastle Ottawa Scale; VAT, visceral adipose tissue; VFAT, visceral fat adipose tissue; WC, waist circumference; WHR, waist-hip ratio.