Increased intake of vegetables, but not fruits, may be associated with reduced risk of hip fracture: A meta-analysis

Association between dietary intake of vegetables and fruits and risk of hip fracture has been reported for many years. However, the findings remain inconclusive. We conducted a meta-analysis to evaluate the relationship between intake of vegetables and fruits, and risk of hip fracture. Literature search for relevant studies was performed on PubMed and Embase databases. Five observational studies were included in the meta-analysis. Summary hazard ratio (HR) with corresponding 95% confidence interval (CI) was calculated from pooled data using the random-effects model irrespective of heterogeneity. Sensitivity and subgroup analysis were performed to explore possible reasons for heterogeneity. The summary HR for hip fracture in relation to high intake vs. low intake of only vegetables, only fruits, and combined intake of fruits and vegetables, was 0.75 (95% CI, 0.61–0.92), 0.87 (95% CI, 0.74–1.04), and 0.79 (95% CI, 0.61–1.03), respectively. Subgroup analyses based on study design, geographical location, number of cases, and gender showed similar results. Increased intake of vegetables, but not fruits, was found to be associated with a lower risk of hip fracture. Large prospective clinical trials with robust methodology are required to confirm our findings.

Combined intake of fruits and vegetables and the risk of hip fracture. Eight studies from five articles were included in the analysis of combined intake of fruits and vegetables and the risk hip fracture. Pooled results indicated no association between combined intake of fruits and vegetables and the risk of hip fracture (OR: 0.83, 95% CI, 0.70-0.98; P = 0.028). However, significant heterogeneity was observed between the studies (I 2 = 84.7%, heterogeneity P = 0.000) (Fig. 2). Sensitivity analysis showed unstable results, which ranged from 0.76 (95% CI, 0.65-0.90) with significant heterogeneity (I 2 = 85.9%, heterogeneity P = 0.000) (excluding the study by Feart et al. 18 ) to 0.88 (95%CI, 0.78-1.01) with significant heterogeneity (I 2 = 73.5%, heterogeneity P = 0.001) (excluding the study by Xie et al. 13 ). On subgroup analyses stratified by geographic location, study design, gender and number of cases, similar results were revealed ( Table 3). The power of 63.23% was determined to detect an HR of 0.83 for the highest combined intake of fruits and vegetables intake compared with the lowest combined intake of fruits and vegetables intake.
Intake of fruits and the risk of hip fracture. Six studies from five articles investigated the association between intake of fruits and risk hip fracture. The summary HR for highest versus lowest levels was 0.87 (95% CI: 0.74-1.04; P = 0.119) using the random-effects model, with a high heterogeneity (I 2 = 73.0%, P = 0.002, Fig. 3). In sensitivity analysis, the unstable results was observed, which ranged from 0.84 (95%CI: 0.72-0.97) with notable heterogeneity (I 2 = 66.8%, heterogeneity P = 0.017) (excluding the study by Feart et al. 18 ) to 0.91 (0.77-1.07) with significant heterogeneity (I 2 = 73.0%, heterogeneity P = 0.005) (excluding the study by Xie et al. 13 ). Further, the subgroup analysis was also performed ( Table 3). The power of 83.48% was determined to detect an HR of 0.87 for highest versus lowest category of total fruits intake. Intake of vegetables and the risk of hip fracture. Six studies from five articles examined the association between intake of vegetables and the risk of hip fracture. The summary HR for highest versus lowest levels was 0.75 (95% CI: 0.61-0.92; P = 0.005) using a random-effects model, with a high heterogeneity (I 2 = 79.6; P = 0.000, Fig. 4). Sensitivity analysis showed that the overall pooled estimate did not alter substantially with the exclusion of any one study. The overall combined OR after sequential exclusion of one study at a time ranged from 0.70 (95% CI, 0.57-0.86) with significant heterogeneity (I 2 = 80.1, heterogeneity P = 0.000) (excluding the study by Feart et al. 18 ) to 0.80 (95% CI, 0.68-0.96) with significant heterogeneity (I 2 = 71.1, heterogeneity P = 0.008) (excluding the study by Xie et al. 13 ). Further, the subgroup analysis was also performed ( Table 3). The power of 98.34% was determined to detect an HR of 0.75 for highest versus lowest category of total vegetables intake.

Discussion
Clinical practice recommends a daily intake of at least 5 servings of fruits and vegetables 26 . A higher intake of fruits and vegetables has been shown to reduce the risk of type 2 diabetes 27 , prolong life 28 , increase BMD 8,9 , decrease cardiovascular risk 29 , and risk of cancer [30][31][32] . Recently, evidence from multiple observational studies has shown the association between intake of fruits and vegetables and risk of hip fracture [12][13][14][15][16][17][18] . However, these studies have had a modest sample size and the strength of the association was variable among these studies, with HR varying from 0.06 (95% CI: 0.01-0.38) to 1.95 (95% CI, 1.04-3.66), with results across the studies being inconsistent. In case-control studies, higher consumption of fruits and vegetables was shown to be associated with a reduced risk of hip fracture among Chinese men and women 13 , and lower forearm fracture among postmenopausal women 33 . Adherence to a Mediterranean diet, rich in fruits and vegetables was found to be associated with lower risk of hip fracture in a cohort study 18 ; while another cohort study suggested a higher risk of hip fracture 14 . At present, it is still unclear whether higher intake of fruits and vegetables is associated with lower risk of hip fracture. Therefore, we conducted a comprehensive meta-analysis to assess the association between the intake of fruits and vegetables and the risk of hip fracture. The present meta-analysis found that increased intake of vegetables, but not fruits, was inversely associated with the risk of hip fracture. Inferences drawn from pooled analysis tend to be more reliable than that from a single study as the overall HR is based on a large sample size and has an adequate statistical power. The power of the estimate of combined intake of vegetables and fruits, fruits, and vegetables with risk of hip fracture was 63.23%, 83.48% and 98.43%, respectively. Subgroup and sensitivity analyses revealed similar results, which serves to further strengthen the inferences drawn.
Several biological mechanisms may underlie the inverse association found in the meta-analysis. The fruits and vegetables are abundant in alkaline ions (potassium, magnesium and calcium). Calcium is a key element required to maintain bone health. The association of potassium and magnesium with bone health is well-documented 34,35 . The high vitamin K content in fruits and vegetables may help explain their role in maintaining bone health 36 . Further, fruits and vegetables are rich source of antioxidants (e.g., vitamin C, carotene and carotenoids), which by countering age-related oxidative stress and inflammation, may increase osteoclastogenesis and osteoclastic differentiation or suppress osteoblastic differentiation 37 . Furthermore, diets high in fruits and vegetables intake have lower dietary acid load 38 , which is known to inhibit osteoblast function and increase osteoclast activity resulting in reduced bone formation and increased bone resorption 39 . Finally, other nutrients such as polyphenols, tocopherols, tocotrienols and glutathione, may potentially contribute in reducing the risk of hip fracture 40 . However, evidence from animal and experimental studies suggests that reactive oxygen species (ROS), primarily derived from fruits and vegetables, may increase osteoclastogenesis 41 , and osteoclastic differentiation 42 , or inhibit osteoblastic differentiation 43 . Therefore, increased oxidative stress may play an important role in age-related bone loss that leads to osteoporotic fractures 44 . The preventive effect derived from the fruits and vegetables intake may counteract the adverse effect of ROS in turnover.

Study
Year Population To the best of our knowledge this meta-analysis is the first to investigate the association between intake of fruits and vegetables and the risk of hip fracture. The large sample size and a longer time span serve to enhance the statistical power to detect possible associations. The power of HR value of intake of only fruits, and only vegetables with risk of hip fracture was 83.48% and 98.43%. Moreover, the use of multivariable-adjusted risk estimates minimized the confounding factors.
However, the potential limitations of this meta-analysis ought to be taken into account while interpreting the results. First, we cannot exclude the possibility that the observed inverse association between intake of vegetables and hip fracture risk could be owe to unmeasured or residual confounding. Further, though the meta-analysis involved a large number of participants, none of the included studies were interventional studies. Third, although we chose the highest multivariable adjusted effect estimates in our meta-analysis, we cannot rule out the effect of uncontrolled confounding by imprecise quantification of the dietary intake of fruits and vegetables. Fourth, a significant heterogeneity was detected among the studies. We further performed sensitivity and subgroup analyses to determine the sources of heterogeneity. The observed heterogeneity could partly be explained by study design (case-control studies and cohort studies) and the article 18 reported the positive association. The statistical power of this French study (only included 57 incident hip fracture) was considerable lower, the positive association in the small number of cohort study could be observed by chance, which could be explained the high heterogeneity. In addition, the high heterogeneity observed in the case-control studies may be due in part to the high likelihood of recall and selection or memory bias and to the differential misclassification of exposure. Furthermore, unstable results were observed in subgroup and sensitivity analyses, which indicated that more relevant articles are needed to further explore this association. Fifth, the diet assessment methods and exposure ranges were different across the included studies, which may have contributed to the heterogeneity. Sixth, because the number of included studies were less than 10, we could not examine the publication bias, the influence of the results should not be ignored.  (1993)(1994)(1995)(1995)(1996)(1997)(1998), dialect group (Hokkien, Cantonese), body mass index (kg/m2), level of education in categories (no formal education, primary school, secondary school or higher), total energy intake (kcal), smoking status (never smokers, ex-smokers, current smokers), physical activity (none, 2-3 hrs weekly, ≥ 4 hrs weekly), calcium (quartiles, mg/1,000 kcal/day), soy isoflavones (quartiles, mg/1,000 kcal/day), vitamin B6 (quartiles, mg/1,000 kcal/ day), menopausal status (women only; yes, no), use of hormone replacement therapy at recruitment (women only; yes, no), baseline physician-diagnosed history of diabetes (yes, no) and stroke (yes, no).  In summary, our results showed that intake of vegetables, but not fruits, may be associated with a lower risk of hip fracture. However, prospective studies need to be conducted to confirm the findings of this meta-analysis.

Materials and Methods
Literature search. A comprehensive literature search for relevant studies published up to May 2015 was performed in PubMed and EMBASE databases. The following search terms were used without imposing any limitation: "fruit(s)" or "vegetable(s)" or "brassicaceae" or "cruciferae" or "citrous" or "citrus", combined with "hip fracture(s)". The reference lists in the retrieved publications were manually screened to increase the yield of relevant studies. Study selection. Studies were considered eligible for inclusion if they met the following criteria: 1) study design: either cohort, case control or cross-sectional; 2) the exposure variable was fruit and/or vegetable consumption; 3) study outcomes included hip fracture risk; 4) Hazard ratio (HR)/Relative risk (RR) and Odds ratio (OR) (95% Confidence interval (CI) reported (or data to calculate these were available). The exclusion criteria were editorial letters, historical reviews and descriptive studies, such as case reports and case series, or laboratory studies. When multiple publications covered the same study population, only the study with the larger sample size was included. Two authors (JKS and HXY) independently performed the literature search and reviewed all studies for their eligibility for inclusion in the meta-analysis. Any disagreements were resolved by consensus or after consultation with the third author (GLH).
Data extraction and assessment of methodological quality. Data pertaining to the following variables were independently extracted for each study by two authors (JKS and HXY) using a standardized data    45 , which contained 9 items for case-control and cohort studies, with every item having a score of 1 point. Studies with a total score of > 6 were deemed to be of high quality.

Statistical analysis.
Hazard ratio with 95% CI was the common measure across all eligible studies. Since hip fracture is a relatively rare event, the differences among risk estimates (HR, RR and OR) were ignored, and the OR and RR were directly converted into HR. A random-effects model of DerSimonian and Laird was used to calculate summary HR, comparing the highest across all included studies regardless of heterogeneity 46 , which incorporates both intra-and inter-study variability. If a study did not report the total HR for combined intake of vegetables and fruits, but instead reported the HR for only vegetables and only fruits separately in a study, they were considered separately as if obtained from different studies, and used the effect estimate for that study in the meta-analysis of combined vegetables and fruits intake and hip fracture risk. If gender-specific estimates were available, they were also regarded as two different studies. Power analysis was performed using the method described by Cafri et al. 47 .
Sensitivity analysis was performed to evaluate robustness and stability of the results by sequentially omitting one study on each turn. Moreover, subgroup analysis was performed to assess heterogeneity and assess the influence of different inclusion criteria on the overall estimate.