1. Introduction
Dietary patterns rich in vegetables and fruits may provide benefits to bone health [
1,
2,
3,
4,
5]. A number of studies have explored the relationships of defined dietary patterns with fracture risk. Although the results of these studies are inconsistent [
1,
4,
5,
6,
7], they do suggest that certain components of these diets may contribute to lower fracture risk [
8,
9,
10]. A major contribution may be the high intake of vegetables and fruits which are a key attribute of all healthy dietary patterns. This concept is supported by the results of prospective studies in a variety of populations finding that higher vegetable and fruit intakes are associated with lower risk of fracture [
2,
3,
9,
11]. However, there is little data on the effects of specific vegetable or fruit types on fracture outcomes. It is possible that some vegetables and fruits may be more protective than others due to specific nutrients and bioactive compounds such as phytochemicals.
In this study, we explored the associations of vegetable and fruit intakes, separately, with 14.5 years fracture-related hospitalisations in a prospective cohort of postmenopausal women aged ≥70 years. We then examined the associations of specific types of vegetables and fruits with fracture outcomes.
3. Results
3.1. Baseline Characteristics
Baseline characteristics of participants are presented according to all participants and categories of vegetable (
Table 1) and fruit (
Table 2) intakes. The mean age of participants at baseline was 75.2 years (SD 2.7 years). Significant differences were observed across vegetable intake categories for energy, protein, and calcium intakes (
p < 0.001) (
Table 1). Significant differences were observed across fruit intake categories for energy, protein, calcium, and alcohol intakes, physical activity, and smoking status (
p < 0.01) (
Table 2).
Mean (SD) intake for vegetables was 196.4 (79.1) g/day and 2.6 (1.0) serves per day, and for fruit it was 245.1 (128.6) g/day and 1.6 (0.9) serves per day. Mean (SD) intake (from highest to lowest) of vegetable types were: yellow/orange/red vegetables 51.8 (27.5) g/day; cruciferous vegetables 32.1 (21.9) g/day; legume vegetables 27.5 (18.7) g/day; and leafy green vegetables 18.6 (12.0) g/day. Median (interquartile range (IQR)) intake for allium vegetables was 6.2 (2.9–10.7) g/day. Median (IQR) intake (from highest to lowest) for fruit types were: apple and pears 54.4 (22.2–103.6) g/day; other fruits 49.5 (22.0–95.4) g/day; bananas 44.5 (18.5–72.6) g/day; and orange and other citrus fruits 35.5 (5.6–85.9) g/day.
3.2. Fracture-Related Hospitalisation
Over 14.5 years (16,458 person-years) of follow-up, 415/1468 (28.3%) participants were hospitalised with a fracture, of which 158/1468 (10.8%) were hip fractures.
3.2.1. Vegetable Intake
Vegetable intake (per serve) was inversely associated with all fractures and hip fractures in age-adjusted and multivariable-adjusted models (
p < 0.05) (
Table 3). Compared with low intakes of vegetables (<2 serves/day), intakes of ≥3 serves/day were associated with a 27% lower hazard for all fractures (multivariable-adjusted
ptrend = 0.023) and a 39% lower hazard for hip fractures (multivariable-adjusted
ptrend = 0.037) (
Table 3).
The associations between intakes of vegetable types and fracture-related hospitalisations in multivariable-adjusted models are presented in
Table 4. Intakes of cruciferous and allium vegetables were inversely associated with all fractures (
p < 0.05) (
Table 4). The highest quartiles of cruciferous (multivariable-adjusted
ptrend = 0.030) and allium (multivariable-adjusted
ptrend = 0.003) vegetables in comparison to the lowest quartiles were associated with a 28% and 34%, respectively, lower hazard for all fractures. After additional adjustment for total vegetable intake, the hazard for all fractures was reduced and became non-significant for the highest quartile of cruciferous vegetables (multivariable-adjusted hazard ratio (HR): 0.80, 95% confidence interval (CI) 0.58, 1.10,
ptrend = 0.160) compared to the lowest quartile. Allium vegetables, however, remained statistically significant (multivariable-adjusted HR: 0.70, 95% CI 0.52, 0.95,
ptrend = 0.019).
For hip fractures, the association of allium vegetables was borderline significant (
p = 0.050), but did not reach significance for cruciferous vegetables (
p = 0.157). Intakes of yellow/orange/red, leafy green, and legumes were not associated with all fractures (
p > 0.05 for all) or hip fractures (
p > 0.05 for all) (
Table 4).
3.2.2. Fruit Intake
Fruit intake (per serve) was not associated with all fractures (
p > 0.05) and hip fractures (
p > 0.05) (
Table 5). The associations between intakes of fruit types and fracture-related hospitalisations in multivariable-adjusted models are presented in
Table 6. All fruit types were not associated with all fractures (
p > 0.05 for all) and hip fractures (
p > 0.05 for all).
3.3. Sensitivity Analyses
There was a weak positive correlation between intakes of cruciferous and allium vegetables (Spearman’s rho = 0.11, p < 0.001) and a moderate positive correlation between intakes of allium and total vegetables (Spearman’s rho = 0.43, p < 0.001) and between intakes of cruciferous and total vegetables (Spearman’s rho = 0.52, p < 0.001). In a forward stepwise Cox proportional hazards model, which included all multivariable-adjusted variables and both cruciferous and allium vegetable intakes separately, age (per year increase, HR: 1.10, 95% CI: 1.06, 1.14, p < 0.001), BMI (per kg/m2 increase, HR: 0.98, 95% CI: 0.95, 1.00, p = 0.034), cruciferous vegetables (per 20 g/day increase, HR: 0.89, 95% CI: 0.81, 0.98, p = 0.019), and allium vegetables (per 10 g/day increase, HR: 0.82, 95% CI: 0.70, 0.97, p = 0.018) were associated with all fractures. In a forward stepwise Cox proportional hazards model, which included all multivariable-adjusted variables and total vegetable intakes as well as intakes of cruciferous and allium vegetables, age (per year increase, HR: 1.10, 95% CI: 1.06, 1.14, p < 0.001), BMI (per kg/m2 increase, HR: 0.98, 95% CI: 0.95, 1.00, p = 0.036), and total vegetable intake (per 75 g/day increase, HR: 0.87, 95% CI: 0.79, 0.96, p = 0.007) were associated with all fractures.
Intakes of cruciferous, allium, and total vegetables were compared at baseline (1998), 5 years (2003), and 7 years (2005) in 986 participants. The mean (SD) at each time point for intakes of cruciferous, allium, and total vegetables are presented in
Table S1. One-way repeated measures ANOVA were conducted for intakes of cruciferous, allium, and total vegetables and results confirmed a significant effect for time (Wilks’ Lambda
p < 0.001 for all). Intake of cruciferous vegetables was 2.5 g/day (7.8%) lower at 7 years compared to intake at baseline. Intake of allium vegetables was 1.5 g/day (18.5%) lower at 5 years and 2.2 g/day (27.2%) lower at 7 years compared to intake at baseline. Total vegetable intake was 22.3 g/day (11%) lower at 5 years and 31.5 g/day (16%) lower at 7 years compared to intake at baseline. To account for this change, the average across baseline, 5 years, and 7 years was calculated individually for cruciferous, allium, and total vegetable intakes. The average values for cruciferous, allium, and total vegetable intakes were then entered separately into multivariable-adjusted Cox proportional hazards models for all fractures. This did not substantively alter the hazard ratios observed for baseline values and fracture-related hospitalisations (
Table S2).
Adjustment for the Nutrient-Rich Foods Index in multivariable-adjusted models attenuated the associations of all fractures with intakes of total vegetables and cruciferous vegetables (p > 0.05 for both), but not allium vegetables (p = 0.027). No effect modification by prevalent fracture was observed for intakes of total vegetables (pinteraction = 0.617), cruciferous vegetables (pinteraction = 0.989), or allium vegetables (pinteraction = 0.482). Lastly, there were no differences in calcium (n = 12), calcium plus vitamin D (n = 393), or bisphosphonate (n = 80) use across vegetables serve categories amongst the n = 1007 women with available data (p > 0.05 for all).
4. Discussion
In this prospective cohort study of older women, we identified vegetable intakes, but not fruit intakes, to be associated with a lower hazard of all fractures and hip fractures. We also identified cruciferous and allium vegetable intakes to be individually associated with a lower hazard of all fractures, but not hip fractures. This may be due to the relatively low prevalence of hip fractures in this study and, therefore, the insufficient power to detect an association. The study was only powered to detect relatively large reductions in risk between food intake categories of around 20% for all fractures and 30% for hip fractures. The associations were independent of dietary and lifestyle factors known to be related to fracture risk.
Habitual intakes of vegetables and fruits, combined, have been associated with lower risk of fracture outcomes [
2,
5,
6,
30]. In particular, vegetable and fruit intakes have also been individually associated with lower risk of fracture outcomes [
2,
5,
30]. A meta-analysis has identified vegetables, but not fruits, to be associated with reduced risk of hip fracture [
31]. Our study is consistent with this meta-analysis, having demonstrated vegetable intake, but not fruit intake, to be associated with lower risk of fracture outcomes.
There are two main explanations why higher vegetable intakes could contribute to a lower risk of fracture. These include effects on bone mineral density and risk of falling. A number of studies have explored the link between vegetable and fruit intake and bone mineral density with inconsistent results. Some observational studies have found associations of vegetable and fruit intakes with bone mineral density [
2,
3,
5,
8,
30,
32,
33], whilst others have not observed an association [
34]. In addition, results of randomised controlled trials have not found consistent effects on biomarkers of bone turnover [
35,
36,
37]. Evidence on the relationship of vegetable and fruit intakes with risk of falling are scant. Therefore, although there is strong evidence supporting beneficial effects of vegetables on fracture risk, the pathways and mechanisms responsible remain unclear.
To our knowledge, this is one of the first studies to investigate the associations of different types of vegetables and fruits with fracture outcomes in a population of older postmenopausal women. We demonstrated both cruciferous and allium vegetable intakes to be inversely associated with fracture risk, both of which were independent of each other. Cruciferous and allium vegetables contain an abundance of specific nutrients and phytochemicals that may benefit bone biology and subsequent fracture outcomes. For example, intakes of vitamin K (rich in cruciferous vegetables) have been shown to be inversely associated with hip fractures [
38]. However, other studies have shown conflicting results [
39]. Intakes of allium vegetables, in particular onions, have been shown to be associated with increased bone density in perimenopausal and postmenopausal women [
40].
One mechanism through which phytochemicals may benefit bone biology is by the reductions in oxidative stress. Oxidative stress has been demonstrated to inhibit in vitro osteoblastic differentiation [
41], and in human studies, relationships between excessive reactive oxygen species and bone loss have been observed [
42,
43]. Particular phytochemicals of interest found in both cruciferous and allium vegetables are organosulfur compounds. Sulforaphane, an organosulfur compound found abundantly in cruciferous vegetables, has been shown to inhibit in vitro human osteoclast differentiation [
44], possibly due to the activation of nuclear factor-erythroid 2-related factor 2 (Nrf2) [
45]. Nrf2 is a redox-sensitive transcription factor that regulates the expression of antioxidant proteins protecting against oxidative stress. In addition, sulforaphane has been shown to epigenetically stimulate osteoblast activity and reduce osteoclast bone resorption [
46]. Park et al. [
47] have also shown the alliin-containing vegetable, allium hookeri, to have in vitro and in vivo anabolic effects on bone formation. Allium hookeri is a widely consumed allium vegetable in Southeast Asia and is a rich source of alliin, an organosulfur compound that is also found abundantly in other allium vegetables such as garlic.
Strengths to this current study include the prospective design and population-based setting with ascertainment of verified fracture-related hospitalisations with almost no loss to follow-up. Participants of this study were representative of older women of the Australian population. The average vegetable serves of the women in this study were 2.6 serves which is the same for older Australian women aged ≥75 years [
48]. In addition, there was also relatively detailed information on a number of known confounders including alcohol intake and socioeconomic status. Dietary information was collected at different time points and was collected using a validated and reproducible method of assessment. Limitations, however, need to be acknowledged. Participants of this study may have commenced taking calcium supplements or medications known to affect bone metabolism after the completion of the CAIFOS. However, we have demonstrated the proportion of participants that commenced taking calcium or calcium plus vitamin D supplements at the completion of the CAIFOS were similar across vegetable serve categories. In addition, medications known to affect bone metabolism are most likely prescribed after an osteoporotic fracture. Therefore, it is unlikely this would have influenced the interpretation of our findings. In addition, the dietary information, including habitual intakes of vegetables and fruits, were self-reported which may lead to misclassification of these variables. In addition, even though we adjusted for potential confounders such as dietary and lifestyle factors known to be associated with fracture risk, higher vegetable intakes may be a marker of a healthier lifestyle not completely captured by the lifestyle variables that we included as potential confounders in the multivariable-adjusted analyses. For example, participants consuming ≥3 serves/day of vegetable intake versus participants consuming <2 serves/day reported a 30% higher energy intake despite similar BMI. This suggests that they were more physically active. Even though the relationships of cruciferous, allium, and total vegetable intakes with fracture-related hospitalisations persisted after adjustment for energy intake and physical activity, the reported physical activity is relatively imprecise in comparison with these lifestyle factors. Although physical activity has been associated with geometric indices of bone strength in this cohort [
49], reported physical activity using questionnaires are somewhat unreliable with the likelihood of under adjustment. We attempted to further address the possibility of higher vegetable intakes being a marker of a healthier lifestyle by adjusting for diet quality. This did attenuate the relationship for total and cruciferous vegetables, but not allium vegetables. The attenuation of the relationship for total and cruciferous vegetables and fracture-related hospitalisations indicates other constituents of a healthy diet at least partially explain the observed associations. It should also be noted that moderate correlations did exist between cruciferous and allium vegetables and total vegetable intakes. In addition, the inverse association between intakes of cruciferous vegetables and fracture-related hospitalisations was attenuated when adjusting for total vegetable intake. It is, therefore, possible that some of the effects seen for cruciferous vegetables may be due to their contribution to the overall increase in vegetable intake. Lastly, the observational nature of this study cannot establish a causal relationship, and the results of this study cannot be applied to younger cohorts and cohorts of older men.