Osteoporosis is characterized by low bone mass and microarchitectural deterioration of bone tissues, leading to an increased susceptibility to fractures, particularly at the hip, spine, and wrist.1 Dietary intake plays a key role in the maintenance of bone health throughout life, and there are data illustrating the importance of plant-based foods, particularly in relation to fruit and vegetable intake, although the results from published studies have been inconsistent.2 However, the key constituents that may impact on bone mineral density (BMD) are unclear and limited research to date has focused on flavonoids. These occur naturally in plant-based foods and are present in measurable amounts in many commonly consumed fruits, vegetables, grains, herbs, and drinks (including tea, wine, and juices).3, 4 Their structural complexity has led to their subclassification as: flavonols; flavones; flavanones; flavan-3-ols and their oligomeric and polymeric forms (ie, procyanidins); isoflavones; and anthocyanins and other polymeric flavonoids. Important differences in the chemical structure of subclasses influence both their biological efficacy and bioavailability.5, 6
Flavonoids have the potential to affect bone health; certain classes of flavonoid compounds have been shown to prevent bone loss in ovariectomized animal models7–10 and to improve bone quality and strength in orchidectomized rats.11, 12 A number of potential mechanisms have been proposed, including effects on osteoclast differentiation via mechanisms involving NF-κB and AP-1 induced by receptor activator of NF-κB ligand (RANKL) in vitro,13–15 stimulating alkaline phosphatase activity or promoting osteoblast activity and upregulating bone sialoprotein gene promoter.16–18
Most previous research examining flavonoids and bone health, including randomized controlled trials (RCTs), have focused on the isoflavone subclass with equivocal results.19 Two recent meta-analyses examined the effects of soy isoflavones on BMD in women; one suggested that isoflavones at doses greater than 90 mg/d significantly attenuate bone loss at the spine over a 6-month period in menopausal women,20 whereas the other, which only included studies with a minimum intervention of 12 months, suggested that soy supplementation does not have a positive effect on BMD at the lumbar spine and hip.21 However, isoflavones are found almost exclusively in soy foods, and intakes in habitual Western diets are <5 mg per day, below the bioactive range for these compounds,22, 23 and so are unlikely to have any impact on bone density in European populations. Although a number of epidemiological studies have examined the effects of flavonoid-rich foods, such as green tea, on bone metabolism and density, the results to date have also been equivocal.24
Despite the potential for the beneficial effects of flavonoid subclasses on BMD, there has only been one previous study relating a narrow range of flavonoid subclasses to BMD; significant positive associations between total flavonoid intake and BMD were observed.4 Given the potential for flavonoid subclasses to influence BMD in women, coupled with an absence of previous relevant research, the purpose of this study was to investigate the association between BMD and a comprehensive range of flavonoid subclasses in healthy women aged 18 to 79 years.
Subjects and Methods
The TwinsUK adult twin registry is an ongoing study examining a wide range of age-related phenotypes. We examined data from 3160 participants with measured bone density and dietary intake data (2194 were dizygotic [DZ] twins, 966 were monozygotic [MZ] twins). The twins enrolled in the TwinsUK are healthy volunteers, have not been selected on the basis of disease traits, and have been previously shown to be representative of the UK adult singleton population25, 26 with respect to their distribution of BMD and other physical, anthropometric, and lifestyle characteristics. Zygosity was derived by questionnaire and confirmed by multiplex-DNA-fingerprinting (PE Applied Biosystems, Foster City, CA, USA). Ethical approval was obtained from St. Thomas's Hospital Research Ethics committee and informed consent was obtained from all subjects.
The twins included in this study were female, aged 18 to 79 years who had attended dual-energy X-ray-absorptiometry (DXA) (Hologic QDR; Hologic, Inc., Waltham, MA, USA) scans and clinical assessments from 1996 to 2000. Bone density measurements were made at the hip, foreman, femoral neck, and spine. Height and weight were measured, and body mass index (BMI) was calculated. At assessment twins completed a questionnaire detailing their medical history and lifestyle factors. Physical activity (PA), measured by questionnaire, was classified as inactive, moderate, and heavy exercise during leisure time. This previously validated measure of activity correlates well with an in-depth measure of physical activity in the Dunbar Health Survey.27 Menopausal status was established by questionnaire. Smoking and HRT were categorized as never, former, or current.
Assessment of dietary intake and flavonoid intakes
Dietary intake data were collected from participants via a validated food frequency questionnaire (FFQ) administered between 1996 and 2000.28, 29 A database for assessment of intake of the different flavonoid subclasses was constructed using the updated and expanded U.S. Department of Agriculture (USDA) flavonoid content of foods and the proanthocyanidin databases30, 31 as our primary sources. For foods in the FFQ for which no values were available in the USDA database, we searched phenol explorer (www.phenolexplorer.eu) to ensure all available high-quality data on flavonoid values were included in the database. Values for the individual flavonoid compounds were assigned to each of the foods listed in the FFQ, and if values for specific foods were not available we imputed from similar foods if appropriate. For recipes, we assigned values for the specific flavonoids for each ingredient in the mixed dishes. Intakes of individual compounds were calculated as the consumption frequency of each food multiplied by the specific flavonoid content of the food for the specified portion size. We derived intakes of the six main subclasses commonly consumed in the U.S. and UK diet, specifically flavanones (eriodictyol, hesperetin, naringenin), anthocyanins (cyanidin, delphinidin, malvidin, pelargonidin, petunidin, peonidin), flavan-3-ols (catechins, epicatachins), flavonoid polymers (including proanthocyanidins, theaflavins and thearubigins), flavonols (quercetin, kaempferol, myricetin, isohamnetin), and flavones (luteolin, apigenin). Total flavonoid intakes were derived by addition of the component subclasses (flavanones, anthocyanins, flavan-3-ols, polymers, flavonols, and flavones). We did not evaluate isoflavone intakes as these are negligible in the habitual UK diet.22, 23
Statistical analyses were performed with Stata statistical software version 11.0 (Stata Corp, College Station, TX, USA). Quartiles of the different classes of flavonoid intake were calculated. We calculated means and SDs for BMD of the hip and spine according to flavonoid subclasses. Because established influences of BMD are age, menopausal status, BMI, physical activity, smoking, and hormone replacement therapy (HRT) use, we adjusted for these covariates together with energy intake. Two models for adjustment for covariates were used: (1) the data were adjusted by age; and (2) the model included age, menopausal status, BMI, physical activity, smoking, HRT, and energy. As data from members of twin pairs could not be treated as independent we controlled for familial aggregation by treating twin pairs as clusters by using the robust regression cluster option in Stata software. Because there was a large age range in this population, as a further test of the associations and to further account for menopausal status we also performed the analyses stratified for menopausal status. The contribution of each food type to each flavonoid subclass was calculated as the ratio of intake of a specific food type from the intake of all foods. Data from MZ and DZ twin pairs were pooled in the analysis because no important differences were detected in the magnitude of the within-pair or between-pair coefficients for dietary intake or for the associations between flavonoid subclasses and BMD of the hip and spine (data not shown). In order to relate the scale of the differences between top and bottom quintiles of flavonoid subclass intake, the differences were calculated as a percentage of the SD of either hip or spine bone density.
The baseline characteristics of the population of 3160 female participants are presented in Table 1. The mean age of the participants was 48 years (18–79 years); 16% were currently using HRT, 58% were premenopausal, and 18.5% were current smokers. Mean BMD at the spine was 0.99 (0.14) g/cm2. Median total flavonoid intake was 1.1 g/d (interquartile range [IQR] 544–1688 mg/d) and the polymers contributed the highest amount to total intake. Median anthocyanin intake was 13.7 mg/d, while median flavanone intake was 21.2 mg/d. Overall, tea was the main contributor to total flavonoid and to flavan-3-ol, flavonol, and polymer intake (Fig. 1). There were no significant differences in mean flavonoid intakes between MZ and DZ twins. There were five foods that each contributed to more than 10% of anthocyanin intake in this cohort; grapes, pears, wine, berries, and fruit yogurts.
Table 1. Characteristics of the Population of 3160 Female Twins Aged 18 to 79 Years
Values are mean SD unless indicated otherwise.
BMI = body mass index; HRT = hormone replacement therapy; BMD = bone mineral density; IQR = interquartile range.
Never, % (n)
Previous, % (n)
Current, % (n)
Never, % (n)
Previous, % (n)
Current, % (n)
Inactive, % (n)
Moderately active, % (n)
Very active, % (n)
Premenopausal, % (n)
Spine BMD, g/cm2
Total hip BMD, g/cm2
Total flavonoids, mg/d
In multivariate analyses, a higher total flavonoid intake was positively associated with higher BMD at the spine but not the hip (Tables 2 and 3). Overall, the association between flavonoid subclass intake and BMD was larger for the spine than the hip. The magnitude of effect was greatest for anthocyanins with a 0.034 and 0.029 g/cm2 difference in BMD at the spine and hip respectively at highest compared to lowest intakes (Tables 2 and 3). As a percentage of the SD for BMD, these differences with anthocyanin intake were 23.8% for spine and 22.7% for hip bone density. The differences in BMD at the spine associated with anthocyanin intake were greater in postmenopausal than premenopausal women (β coefficient 0.01 g/cm2 per quintile (p = 0.001) versus 0.005 g/cm2 per quintile (p = 0.039), respectively), p for the difference <0.001. Higher intakes of flavones were also associated with a higher BMD at both sites. Interestingly, associations between BMD and flavonols and polymers were only observed at the spine and for flavanones only at the hip.
Table 2. Mean Spine Bone Density (g/cm2) in 3160 Female Twins Aged 18 to 79 Years According to Quintile of Total Flavonoid Intake
In a cohort of female twins, we have shown that a higher intake of total flavonoids is associated with a higher BMD at the spine but not the hip. Our database integrated all flavonoid subclasses, which has enabled us to provide a more complete assessment of the range of flavonoid subclasses and total intake than the only previously published epidemiological study.4 Total flavonoids integrate a range of structurally different flavonoid subclasses, and these differences in chemical structure alter bioavailability and the potential mechanisms of action of these bioactive constituents. It is therefore critical to conduct analyses on the specific subclasses to identify which key constituents may be important in relation to bone health. Specifically, we showed for the first time that a higher intake of anthocyanins was associated with a 3.4% and 3.1% higher BMD at the spine and hip, respectively, compared to those in the lowest quintile representing percentage differences of the SD of BMD of 23.8% and 22.7% for spine and hip, respectively. These associations were significantly greater in postmenopausal than premenopausal women. Plausible biological mechanisms from in vitro experimental data exist that would predict such associations including activation of cell signaling pathways and anti-inflammatory effects.32 However, to date, only limited experimental data are available examining the specific effects of anthocyanins and their metabolites on bone quality and strength in ovariectomized animal models, or on osteoblast and osteoclast activity in vitro, and the main mechanistic focus in relation to bone has been on the flavanone subclass.
Our data also suggest that higher intakes of flavanones are associated with higher BMD at the hip (0.9%), which is in agreement with the previous study that assessed flavanone intake, although only assessment of intake of two of the members of the flavanone subclass (hesperidin and naringenin) were included and intake of eriodictyol was not included.4 These associations are supported by significant experimental and mechanistic evidence of a protective effect of flavanones on bone in animal models. Citrus juice, the main dietary source of flavanones, modulates bone strength and quality, whereas the flavanone hesperedin inhibited bone loss by affecting osteoblast differentiation through bone morphogenetic protein and mitogen-activated protein kinase (MAPK) signaling.7, 10–12, 33, 34 In another study, the flavanone glycoside, naringin, protected against ovariectomy-induced bone loss, potentially mediated through ligand-independent activation of estrogen receptors in osteoblastic cells.10
Mechanistic support also exists for a beneficial effect of higher flavonol intake on BMD; quercetin and kampferol have been found to inhibit bone loss in ovariectomized mice8 via a range of potential mechanisms including effects on osteoclast differentiation involving NF-κB and AP-1 induction by RANKL,13, 15 stimulating alkaline phosphatase activity, promoting osteoblast activity, and upregulating bone sialoprotein gene promoter.16, 18 We found an association between a higher flavonol intake and higher BMD at the spine, similar to findings from a previous study, which had only included intakes of a few flavonols.4
The associations that we found are of a greater scale than the majority of cross-sectional, prospective cohort studies and randomized controlled trials relating fruit and vegetable intake to BMD, which mainly did not find significant positive associations.2, 35–40 Of those which did find a positive association, the magnitude of effect ranged between 0.3% per serving of fruit and vegetables to 2.2% per 100 g fruit and vegetables with the hip and around 4% with the spine.35, 39, 40 Our magnitude of effect, although also relatively small, taken together with other nutritional factors, may play an important role in the prevention of osteoporosis. In comparison with other populations groups, although mean flavanone intake was lower in our population than previous studies, the range was similar compared to previous publications in U.S. women and other EU populations; intakes of all other flavonoid subclasses were very similar.41, 42
To date, the epidemiological findings on the effects of food rich sources of flavonoids and BMD focus on green tea and the findings have been equivocal.24 Green tea is rich in flavan-3-ols, and the previous study showed that BMD was associated with constituents of the flavan-3-ol subclass including catechin.4 We only observed a trend toward higher BMD at the spine when comparing extreme quintiles of flavan-3-ol intake, although in cell culture experiments, tea catechins, particularly epigallocatechin, promoted osteoblast activity and inhibited osteoclast differentiation.17, 24 Our findings with flavonoids may also relate to the flavonoid content of wine because this was one of the main sources of anthocyanins in our study, and a recent study found that alcohol may also be positively associated with BMD.43
The limitations of our study warrant discussion. As with any observational study, no causal associations can be made, and we cannot exclude the possibility of residual confounding; however, we did control for a range of dietary and lifestyle factors known to affect bone health. Dietary flavonoid intakes were calculated from a database developed using the most recent USDA databases,30, 31 with additional input from other sources. These datasets allowed us to quantify a broad range of flavonoid subclass intakes more robustly than previous analyses. However, there is wide variability in flavonoid content of foods depending on geographical origin, growing season, different cultivars, agricultural methods, and processing, and a lack of biomarkers to integrate intake with the extensive metabolism these compounds undergo in vivo. In the future, we need to continue to carefully evaluate all available composition databases, ensure they are robust, be aware of their limitations, and continue to refine and harmonize flavonoid composition for epidemiological research.
We integrated a range of structurally different flavonoid subclasses into total flavonoid intake, and these differences in chemical structure alter bioavailability and the potential mechanisms of action of these bioactive constituents. We did not adjust for calcium intake because our previous analyses in this cohort found no association with calcium intake and BMD.43 Although we used a FFQ to estimate dietary intakes this is based on the validated European Prospective Investigation into Cancer and Nutrition (EPIC)-Norfolk questionnaire,28, 29 which accurately reflects habitual intake44 and was able to distinguish between the main classes of foods contributing to total flavonoids and the subclasses in this cohort.
Although there has been significant interest in the relationship between flavonoids and cardiovascular health and cognitive function,32, 45 to date we are aware of only one other population study investigating the association between flavonoid intake and bone health.4 Our study was able to examine the association with a more comprehensive estimation of flavonoid intake and the range of subclasses than has been used previously. We found associations between BMD of the hip and spine and several flavonoid subclasses that were of greater magnitude than have been found previously in studies of fruits and vegetables and their components.2 These data support a role for flavonoids present in plant-based foods on bone health. Our study therefore provides an important evidence base to further investigate the relative impact of flavonoid subclasses, specifically the anthocyanins and flavonols, on bone health.
All authors state that they have no conflicts of interest.
We are indebted to the participants of the Twins UK Study for their continued dedication and commitment. This study was unfunded but was completed by internal support from the Department of Nutrition, Norwich Medical School, UEA. Twins UK are funded by the Wellcome Trust, Arthritis Research UK, NIHR, and the UK National Osteoporosis Society.
Authors' roles: AC: concept. TS, AMcG: set up and coordinated the collection of all the data from the cohort. AC, AJ: developed the flavonoid database. AW: data analysis and tables. AC, AW: interpreted the data and drafted the paper. AMcG, TS, SFT: provided critical review of the manuscript. All authors contributed to the manuscript and agreed the final version.