Principal findings
Vegetarians but not occasional meat-eaters or pescatarians were at a higher risk of hip fracture than regular meat-eaters in this cohort of UK women. There was no clear evidence of effect modification by BMI across diet groups. The risk differences remained after accounting for confounders and were not explained by differences in key nutrient intakes related to bone health between vegetarians and regular meat-eaters, implying the potential importance of other unaccounted factors.
Comparison with previous studies
Prospective evidence of hip fracture risk in individuals on meat-free diets is limited. Our findings largely concur with the results of the only other two cohort studies on this topic [8, 9], strengthening the evidence of a higher risk of hip fracture in UK vegetarian women.
In the EPIC-Oxford cohort, there was evidence of a higher risk of hip fracture in vegetarian women of a similar magnitude (25%) [8]. The slightly higher effect estimate in our study (33%) may be due to our reference group being regular meat-eaters, whereas the reference group in the EPIC-Oxford cohort was meat-eaters of any amount. The AHS-2 also found limited evidence of a 17% higher risk of hip fracture in US vegetarian women [9]. Differences in estimates between the AHS-2 and our results may be due to the different adjustment strategies when accounting for confounders; in the AHS-2, attained age was used as the time frame, and adjustment was made for age and energy, calcium, potassium, and vitamin D intakes at recruitment amongst other factors. This may have resulted in overadjustment and adjustment for factors potentially on the causal pathway, diluting risk estimates. The AHS-2 also relied on self-report for case ascertainment. We identified hip fracture cases using participants’ hospital episode statistics, which incurs less reporting error and selective loss to follow-up. We found no clear evidence of a difference in hip fracture risk in pescatarians or occasional meat-eaters (ate meat < 5 times/week) compared to regular meat-eaters. Similarly, in the AHS-2, there was no clear evidence of a difference in hip fracture risk in semi-vegetarian (ate meat or fish ≤ once/week) or pescatarian women compared to non-vegetarians [9]. In contrast, the EPIC-Oxford cohort study found a 30% increased risk in pescatarian women, potentially due to population differences between EPIC-Oxford and the UKWCS, different intakes of fish or other dietary components, or other sources of residual confounding in either study [8]. Both the EPIC-Oxford and AHS-2 cohort studies reported higher risks of hip fracture in vegans compared to meat-eaters [8, 9]. Due to the low number of vegans in the UKWCS, we could not precisely estimate their risk of hip fracture separate from the vegetarian group. Since vegans may face greater challenges in achieving adequate intake of several nutrients, in particular protein and calcium [6], cohort studies with a high proportion of vegans are needed investigating their risk of hip fracture.
Other epidemiological studies have found that adherence to diets low in meat consumption, such as the Mediterranean diet and Alternative Healthy Eating Index, was protectively associated with hip fracture risk [33, 34], and adherence to Western diets in which meat consumption is high was positively associated with hip fracture risk [35]. Conversely, total meat intake has been inversely associated with hip fracture risk [21]. These results cannot be fairly compared with risks in vegetarians and non-vegetarians, which no other study has directly assessed.
Interpretation and implications
The observed higher risk of hip fracture in vegetarians compared to regular meat-eaters may be partly explained by the differences in body anthropometrics between the diet groups. Whilst there was no clear evidence of BMI modifying associations between diet groups and hip fracture risk, the lower mean BMI in vegetarians partly explained their higher risk. Previous studies have shown BMI and body weight to be lower in vegetarians [26, 36], and inversely associated with hip fracture risk [24, 37]. Possible mechanisms include the protective roles of bone mass, fat mass, and muscle mass, which have each been inversely associated with hip fracture risk independently [38]. Inadequate fat mass may reduce cushioning from impact force at the hip during falls, which account for 90% of hip fractures [39]. Higher fat mass could also increase bone strength through increased mechanical loading and enhanced oestrogen production [38]. Low muscle mass and strength of the hip flexor muscles and spine extensors have also been associated with an increased risk of hip fracture [40], possibly due to reduced balance and mobility. Weight management may therefore be an important consideration in reducing hip fracture risk in vegetarians, but further research is required to explore the roles of BMI and body composition in hip fracture risk in vegetarians and meat-eaters.
A second potential reason for the higher risk of hip fracture in vegetarians is their lower intake of nutrients important to bone health that are abundant in animal products. Previous studies have found lower dietary intakes of protein, calcium, vitamin D, and vitamin B12 in vegetarians [6, 22], and have suggested protective associations of these nutrients with hip fracture risk [6, 41, 42]. In our study, vegetarians had lower dietary intakes of protein, vitamin D, and vitamin B12, but similar dietary calcium intakes to other diet groups. In particular, vegetarians were less likely to meet the UK recommendation for protein intake in adults of 0.75 g/kg body weight/day than regular meat-eaters (88.8% vs 98.3%) [43], but the higher risk of hip fracture in vegetarians was not explained by any dietary nutrient intake. It is likely that measurement error incurred by estimating nutrient intakes from an FFQ precluded accurate estimation of the importance of nutrients from dietary sources to hip fracture risk in vegetarians.
Since the higher risk of hip fracture in vegetarians remained after adjustment for BMI and several dietary nutrient intakes, other factors may be important. Supplemental sources of specific nutrients and circulating vitamin D concentrations could differ between vegetarians and non-vegetarians and may impact the risk of hip fracture [9, 44] but could not be accounted for in this analysis due to a lack of data. Circulating levels of insulin-like growth factor-1 (IGF-1) may also be lower in vegetarians than in non-vegetarians [45] and have been positively associated with BMD and negatively associated with risk of total fracture and hip fracture [46], but could not be considered here. Future studies should investigate the roles of IGF-1 and nutrients abundant in animal products on hip fracture risk in vegetarians to better understand the reasons for their observed higher risk.
Strengths and limitations
This study has three main strengths. Firstly, the large number of pescatarians and vegetarians included gave good statistical power to estimate their risk of hip fracture. Secondly, the identification of hip fractures based on hospital records over a long follow-up period reduced reporting errors and loss to follow-up. Finally, we classified subjects into diet groups based on reported intakes of animal foods using a validated FFQ, which may more accurately allocate participants into diet groups than asking participants to identify their diet group.
On average, UKWCS participants were younger by end of follow-up than the average age at hip fracture in women (83 years) [47], limiting the number of hip fractures observed. Moreover, high-energy trauma may account for more hip fractures in younger adults, whereas fragility hip fractures are more common in older adults [48]. We could not distinguish between traumatic and fragility fractures here since information on the cause of hip fractures was not available. We had insufficient power to detect effect modification by covariates in subgroup analyses. For BMI, the strong correlation with the diet group meant that the number of vegetarians with a high BMI or regular meat-eaters with a low BMI was low. Moreover, BMI was derived from self-reported height and body weight, implying a possible measurement error. Investigation of hip fracture risk in underweight participants by diet group was also not possible but merits further investigation.
Women with missing covariate data (n = 3926) were excluded from the analyses in this study, which introduced a risk of selection bias. However, the magnitude of any selection bias is unlikely to be clinically significant, given that the characteristics of participants included or excluded in analyses here at recruitment were similar (Additional file 1: Table S5). Although we adjusted for likely confounders, residual confounding was possible. For example, we could not adjust for the use of medications that could impact the associations between the diet groups and hip fracture risk due to a lack of data. The risk of hip fracture could differ between moderate and heavy consumers of alcohol [20], but we were unable to differentiate between these groups when adjusting for alcohol consumption. In addition, the exclusion of participants with prior hip fractures was likely an incomplete exclusion, since hospital data of fracture incidences before 1997 was not available, and the questionnaire did not ask about fracture history. The single questionnaire administered at recruitment was the only method of assessing diet and lifestyle information; therefore, we could not account for changes in the diet group or covariates over time. Additionally, food and nutrient intake in vegetarians in recent years could differ from when data were collected at recruitment due to changes over the last two decades in the availability of vegetarian food products, such as increases in the number of available meat substitute products [49]. Consequently, the generalisability of our findings to modern-day vegetarians is reduced. Our findings were also predominantly in white UK women; previous studies have shown that total fracture risk could depend on ethnicity [50]; therefore, more research is needed investigating hip fracture risk in non-European vegetarians and non-vegetarians.