Bone Density
Before we talk about bones, it is worth being clear about what is on the table. A well-planned vegan diet is not a lifestyle preference in search of justification. It has some of the strongest evidence in all of nutrition science behind it, and unlike most dietary claims, the best of it comes from randomized trials and genetic causal inference, not just observational studies that can be waved away as “healthy user bias.”
Is a vegan diet actually healthy?
Yes, and the evidence for the biggest wins is unusually strong. Here is the short version, organized by how people actually search for it.
Does a vegan diet lower heart disease risk?
It reduces risk for the disease that matters most, heart disease, by lowering its root cause. A 2023 meta-analysis of 30 randomized controlled trials in the European Heart Journal found that plant-based diets reduced LDL cholesterol by about 10 percent and apolipoprotein B by 14 percent compared with omnivorous diets (Koch 2023). That matters because the European Atherosclerosis Society consensus, synthesizing over 200 cohorts, Mendelian randomization studies, and RCTs across more than two million people, concluded that LDL and apoB causally drive atherosclerotic cardiovascular disease, in a dose- and duration-dependent way (Ference 2017). The 2023 meta-analysis’s senior author put that apoB reduction at about a third of the effect of taking a statin, and estimated it would translate to roughly 7 percent lower cardiovascular risk over five years. In EPIC-Oxford, vegetarians, a group that included the vegans, had 22 percent lower ischaemic heart disease over 18 years (Tong 2019). More on that causal chain here: Cholesterol, From the Ground Up.
Does a plant-based diet work even if heart disease runs in your family?
Stanford’s identical-twin trial randomized 22 pairs of healthy adult identical twins, one twin per diet, to a healthy vegan or a healthy omnivorous diet for 8 weeks, and 21 pairs contributed to the final analysis. The primary outcome was the change in LDL cholesterol. The paper’s own estimate, from a statistical model that accounts for the twin pairing, is that the vegan twins’ LDL fell 13.9 mg/dL more than their omnivorous twins’ over the 8 weeks (standard error 5.8; 95 percent confidence interval 2.4 to 25.3). The individual figures that often get quoted alongside it are raw, unadjusted group means: 110.7 and 118.5 mg/dL at baseline for the vegan and omnivorous groups, and 95.5 and 116.1 mg/dL at week 8. Subtracting one from the other does not reproduce the paper’s estimate, probably because the baseline means include all 44 participants while the week-8 means come from the 42 who completed the trial, so the paper’s estimate is the one to use. The paper’s supplement reports the raw changes directly: the vegan twins’ LDL fell by an average of 16.85 mg/dL while the omnivorous twins’ was essentially unchanged, up 0.75, a difference of 17.6 mg/dL (95 percent confidence interval 6.0 to 29.2), and a paired test gave the same difference. That is larger than the model’s 13.9, although the authors describe the t tests as differing only minimally from the main analysis. The gap was also wider earlier: at week 4, while meals were still being delivered, the model’s difference was 23.1 mg/dL (standard error 5.4), and it had narrowed to 13.9 by week 8, after participants switched to preparing their own food. The vegan twins also lost more body weight, 1.9 kg more than their omnivorous twins in the main model, though both groups lost weight. Fasting insulin fell 2.9 μIU/mL more in the main model, but that result is fragile: it was not significant in the supplement’s paired t test (p = 0.13) or unpaired t test (p = 0.054) (Landry 2023). A secondary analysis of the same trial found significant decreases in several epigenetic age measures among the vegan twins and none among the omnivorous twins, though its authors say it is uncertain whether that reflects the diet itself or mainly the vegan twins’ greater weight loss. Seven of its sixteen authors, including the first author, are employees of TruDiagnostic, the company whose laboratory ran the epigenetic tests (Dwaraka 2024).
Several features of the design limit how much weight it can bear. It was open-label, the sample size was set by available resources rather than a power calculation, and the diets were not matched for calories, so the authors say the LDL change cannot be separated from the weight loss that accompanied it. In self-reported dietary recalls, the vegan twins averaged 1,628 kcal a day during the meal-delivery weeks against 1,815 for the omnivorous twins (Landry 2023). The participants were healthy adults with an average baseline LDL of about 114 mg/dL, and the paper does not describe selecting anyone for a family history of heart disease, so it was not designed to show what happens in people at high genetic risk. The authors note that pairing identical twins removes confounding by age, sex and genetics between the two diet groups (Landry 2023). The trial was funded by the Vogt Foundation, with additional grants from Stanford’s Clinical and Translational Science Award unit and the National Heart, Lung, and Blood Institute, and the paper states that the funders had no role in its design, analysis or reporting. The corresponding author reported receiving funding from Beyond Meat outside the submitted work (Landry 2023).
Does a vegan diet lower blood pressure, weight, and blood sugar?
All three move the right way. Vegetarian diets lowered systolic blood pressure by 4.8 mmHg in controlled trials (Yokoyama 2014). A meta-analysis of 11 RCTs found vegan diets reduced body weight by 4.1 kg, along with HbA1c, total cholesterol, and LDL (Termannsen 2022). A 16-week low-fat vegan RCT in 244 overweight adults produced about 6.4 kg of weight loss and a 34 percent reduction in liver fat (Kahleova 2020), though two things temper it: the control group made no dietary changes at all rather than following an active comparison diet, which inflates the apparent effect, and the trial was run by employees of a plant-based advocacy organization involved in its design, conduct, and analysis. Its direction matches other randomized trials, so it corroborates rather than anchors the point. See also: the best diet for weight loss.
Does a vegan diet lower diabetes risk?
Pooling 9 cohorts and 307,099 people, higher plant-based diet adherence was associated with 23 percent lower type 2 diabetes risk (Qian 2019).
Does a vegan diet lower cancer risk?
The IARC classifies processed meat as a Group 1 carcinogen and red meat as Group 2A, with roughly 18 percent higher colorectal cancer risk per 50 g/day of processed meat (Bouvard 2015). Meta-analysis puts total cancer incidence about 15 percent lower in vegans (Dinu 2017). And the fiber that anchors plant-based eating is associated with 15 to 30 percent lower all-cause and cardiovascular mortality (Reynolds 2019). On the recent colorectal-cancer headline, see no, a 40 percent number in one subgroup didn’t overturn the meat and bowel cancer link.
What else is a vegan diet good for?
There is a long tail of smaller wins. Rheumatoid arthritis symptoms improved significantly in a controlled trial of fasting followed by a vegetarian diet (Kjeldsen-Kragh 1991). For gout, meat and seafood raised risk while purine-rich vegetables did not (Choi 2004). Vegetarians had 31 percent less diverticular disease (Crowe 2011) and lower cataract risk (Appleby 2011). Vegetarian Buddhist priests had reflux esophagitis at 12.2 percent versus 21.6 percent in matched controls (Jung 2013). And vegans have the lowest BMI and diabetes prevalence across all diet groups (Tonstad 2009).
Is a well-planned vegan diet safe and complete?
The professional consensus is settled. The Academy of Nutrition and Dietetics states that “appropriately planned vegetarian, including vegan, diets are healthful, nutritionally adequate, and may provide health benefits for the prevention and treatment of certain diseases,” and are “appropriate for all stages of the life cycle” (Melina 2016). More on that: Plants Over Animals and raising children vegan.
Is a vegan diet better for the environment?
UK vegans’ dietary greenhouse gas emissions are 25 percent of high meat-eaters’, along with 25 percent of the land use, 46 percent of the water, and 34 percent of the biodiversity impact (Scarborough 2023).
What are the honest downsides?
Two honest notes, because this article is going to demand honesty later and cannot start by spending it. Meta-analysis has not found a statistically significant all-cause mortality benefit for vegans specifically (Dinu 2017); the point estimates lean favorable but do not clear significance. And EPIC-Oxford found a higher total stroke rate in vegetarians, mostly haemorrhagic, which its own authors could not explain. Two things soften it: the signal did not replicate in a large Taiwanese cohort, where vegetarians had roughly half the overall stroke rate of non-vegetarians (Chiu 2020), and that study found vitamin B12 intake modified the relationship, which fits this article’s broader point that unsupplemented B12 does a lot of unaccounted-for work in the older, more alarming cohorts, and B12 is a cheap, easily fixed variable, though the study’s own authors called that specific detail unclear and in need of replication. Anyone telling you the evidence is unanimous is selling something.
Do vegans have weaker bones?
Which brings us to bones. Here is the honest version.
Do vegans break more bones?
Vegans in the two best long-term studies did break more hips. This is real, it replicates, and pretending otherwise would be dishonest. The pattern points at causes that can be corrected, and the rest of this article shows how: calcium, vitamin D, body weight and resistance training.
How many vegans actually broke bones?
Far fewer than the headline suggests, and this is worth knowing before anything else. In EPIC-Oxford, the study behind the “more than double the risk” language, the hip-fracture finding rests on 41 vegan hip fractures, out of 945 hip fractures across nearly 55,000 people over 17.6 years. There were only 1,982 vegans in the cohort to start with. The other cohorts are thinner still: the UK Women’s Cohort Study contained 130 vegans, and UK Biobank, for all its 413,914 participants, contained about 400 vegans, among whom 10 hip fractures occurred. The 2025 meta-analysis sounds enormous at 529,672 participants, but only 5,344 of them were vegans, roughly one percent of the total.
Small numbers do not make a finding fake. All four cohorts that reported a vegan estimate leaned the same direction, and 41 events is enough to detect a large effect if one is really there. But small numbers do make it imprecise, and that is why the meta-analysis authors themselves graded the vegan estimate “very low certainty”, with a confidence interval running from 1.17 to 2.63. In plain terms: the real effect could be a 17 percent increase or a 163 percent increase, and the data cannot tell you which. The vegetarian estimate, built on far more people and events, was tighter and graded low rather than very low. In formal terms, “very low certainty” means the true effect is likely to be substantially different from the estimate.
You can see that imprecision directly in the four studies the vegan figure pools. Their individual hazard ratios were 1.04, 1.31, 1.93 and 3.26. Two of the four were not statistically significant on their own, and the spread runs from essentially no effect to more than tripled risk, which is what a heterogeneity statistic of 64 percent is describing. The 1.75 headline is an average across estimates that disagree with one another by a factor of three, and the two most extreme values come from the two smallest vegan samples, 130 and 400 people. The highest of them, 3.26, is a figure the original researchers produced only in a sensitivity analysis and then explicitly disclaimed, writing that there were not enough vegans in their cohort to assess vegan risk independently. The meta-analysis pooled it anyway. That is not a reason to dismiss the finding. It is the reason its own authors gave it the lowest grade available.
This matters most for the subgroup numbers that get quoted with the greatest confidence. When those 41 hip fractures get split by calcium intake, or by protein adequacy, or by supplement use, the resulting groups are tiny. Every stratified figure in this article, including the ones that favor the vegan case, should be read as suggestive rather than settled.
But did those extra fractures actually shorten anyone’s life?
Nobody checked, and that gap says something. The studies that found the extra fractures had exactly the data needed to answer this. They knew who fractured. They had death records linked in. Connecting the two, to ask whether those vegans went on to die sooner or recover worse, would have been a straightforward next analysis. It was never run, not for vegans and not stratified by diet at all. These were incidence studies built to answer “who breaks bones,” and they stopped there. So the alarming number traveled widely while the question of whether it translated into anyone actually being worse off went unasked.
What we can say points the other way. In EPIC-Oxford and the Adventist studies, vegetarians and vegans had lower overall mortality, or equal-to-lower, not higher. Whatever the extra fractures cost them, it was not showing up as more deaths overall. And Adventist vegetarians had lower rates of the very conditions that kill people after a hip fracture, principally heart disease and diabetes. Since a Swedish register analysis attributed roughly 17 to 32 percent of the excess deaths after a hip fracture, meaning deaths above what the general population of the same age would see, to the fracture itself, with the rest attributed to pre-existing illness (Kanis 2003), the group that is healthier going in should, in theory, survive a fracture better.
Two honest caveats. That is inference, not measurement; nobody connected the dots directly. And part of the vegan fracture excess may sit in sites that carry far less mortality than the hip, which would mean the fracture count and the survival picture barely interact. We cannot tell, because the question was never asked. There is a fuller treatment of post-fracture survival further down.
Why do vegans have more fractures?
Three modifiable factors are the leading suspects: low calcium, low vitamin D, and low body weight. In the earlier EPIC-Oxford analysis, vegans getting at least 525 mg of calcium a day had a fracture rate ratio of exactly 1.00. In the Adventist Health Study-2, vegan women taking both calcium and vitamin D had no higher hip-fracture risk than nonvegetarian women, though the estimate is imprecise, versus nearly triple the risk for those taking neither. In UK Biobank, body mass index alone mediated 28 percent of the vegetarian excess. The vitamin D link is the least established of the three: serum vitamin D was not a significant mediator in UK Biobank, and the Adventist analysis could not assess vitamin D supplements alone. The Adventist authors themselves suggest that vegans’ 27 to 51 percent lower dietary vitamin D intake could be an explanation for the threefold hip-fracture risk. All three factors can be addressed with the steps in the action plan below.
Is low protein the reason vegans break more bones?
Probably not, and this is worth correcting because the claim gets repeated a lot, including in an earlier version of this article. Only one study has actually parsed protein out, and the result does not support it. In Tong 2020, adjusting for calcium moved the vegan total-fracture hazard ratio from 1.43 to 1.31, roughly 28 percent of the excess. Adjusting for protein moved it from 1.43 to only 1.39, roughly 9 percent. Calcium did about three times the work. The decisive test came when the authors restricted the analysis to vegans already eating adequate protein, at or above 0.75 g/kg per day. If low protein were driving the excess, those vegans should have looked better. They did not. Their total-fracture hazard ratio was 1.52 and their hip hazard ratio was 2.71, as high or higher than the main estimate.
Protein still belongs in a vegan bone plan, because adequacy matters and protein supports the muscle mass that keeps you from falling. But it rests on the general older-adult and sarcopenia literature, not on the vegan fracture cohorts, and it should not be listed alongside calcium and vitamin D as an equal driver. It is a secondary adequacy factor.
Can vegans have strong bones?
The Nurses’ Health Study, which looked at plant-based diet quality and hip fracture, followed 70,285 postmenopausal women for up to 30 years and found that eating a more plant-based diet over the long term was not linked to hip fracture risk either way. In its primary analysis, the healthiest plant-based diets had a hazard ratio of 0.97, essentially no effect (Sotos-Prieto 2024). A secondary analysis of women’s most recent diets found about 21 percent lower risk with a high-quality plant-based diet, but the authors caution that reverse causation may explain it, since women with health conditions that raised their fracture risk may have changed how they ate. The index also scores diet quality rather than veganism: even the women with the highest healthful scores averaged about four servings of animal foods a day. The reassuring finding is the long-term null: plant-heavy eating, sustained for decades, did not raise fracture risk. So this is not an indictment of veganism. It is a specification. The EPIC-Oxford vegans behind the headline were recruited in the 1990s, before fortified foods were everywhere, and averaged 591 mg of calcium a day, roughly half of what meat-eaters got. That is not a diet philosophy failing. That is a shopping list failing.
How can vegans protect their bones?
This is the 60-second action summary. Everything after it is the receipts.
How much calcium do vegans need, and where do they get it?
- Aim for about 1,000 mg per day. The best plant sources are low-oxalate greens: kale, bok choy and broccoli are absorbed at roughly 49 to 61 percent in tracer studies, better than milk’s roughly 32 percent, and collards and turnip greens belong to the same low-oxalate family (Weaver 1999). Add calcium-set tofu and fortified plant milks.
- Do not count spinach, chard, or beet greens toward calcium. Oxalate blocks about 95 percent of it. They are excellent foods, just not calcium sources.
- Shake fortified plant milk. The calcium settles to the bottom of the carton.
- Spread calcium across the day. Absorption saturates around 500 mg per dose, so two or three modest servings beat one big one.
- If you do not reliably eat those foods, the full calcium playbook below covers supplements, doses, forms, timing, label-reading and preparation.
How much protein do vegans need for bone health?
Treat about 0.8 g/kg of body weight as the floor and aim for 1.0 to 1.2 g/kg, especially in older age, for muscle and fall prevention. This is a real gap rather than a theoretical one: in UK Biobank, 68.2 percent of vegetarians met the 0.75 g/kg minimum against 85.2 percent of regular meat eaters, and only 15.8 percent reached 1.2 g/kg against 33.6 percent. The gap is easy to close with soy, lentils, beans, seitan, and a variety of legumes and grains. Two things worth knowing. First, plant protein is essentially as bone-protective as animal protein at equal intake; a meta-analysis of 7 randomized trials comparing soy against animal protein found BMD differences near zero (Shams-White 2018), a result worth trusting partly because of who paid for it rather than despite it. The grants came from Dairy Management Inc, the Egg Nutrition Center, and the Beef Checkoff, none of whom had a reason to publish a finding that plant protein matched animal protein for bone. Second, the plant protein foods vegans actually eat are associated with lower fracture risk: in the Adventist cohort, eating legumes daily was linked to a 64 percent lower hip fracture rate and meat analogues daily to a 49 percent lower rate (Lousuebsakul-Matthews 2014). The old “animal protein acidifies you and leaches calcium” theory has largely been overturned as an explanation of what protein does, though the broader question of dietary acid load is still argued over, as the next few paragraphs explain. Its mirror image, the idea that plant protein is too weak for bone, has been overturned too.
Here is what that claim actually says in plain terms, since it gets repeated often enough that people accept it without knowing the mechanism behind it. Protein breaks down into acid. The body has to neutralize that acid. The theory holds that bone supplies the alkaline mineral to do it, meaning calcium gets pulled out of your skeleton and shows up in your urine. A urine test alone cannot tell the difference between calcium that came from food and calcium that came from bone; the number on the page looks the same either way. The only way to actually answer the question is to physically track where the calcium started out, which is what isotope tracers do, and it is why they can settle something ordinary calcium-balance studies could not. It is also worth being precise about what is and is not actually disputed. Higher protein intake reliably does raise urinary calcium. Across 26 controlled trials pooled by the same research group, the relationship is clean and nobody contests it (Kerstetter 2007), working out to roughly 0.8 mmol more urinary calcium for every 25 g of added protein, as that pooled analysis is summarized in Kerstetter 2005. What was disputed is where that extra calcium comes from.
In the most direct test, 13 women were fed a moderate-protein diet and a high-protein diet with calcium held constant, and researchers used calcium isotope tracers to trace it. Intestinal absorption rose from 18.5 to 26.2 percent, and that increase alone accounted for about 93 percent of the rise in urinary calcium. The fraction of urinary calcium actually traced back to bone fell slightly rather than rising, from 96.0 to 94.4 percent, the opposite of what the leaching theory predicts (Kerstetter 2005). A separate trial in 16 postmenopausal women, 7 weeks on each diet, that specifically raised the acid load from meat protein, not just the amount of protein, found the same pattern: absorption rose from 22.3 to 26.5 percent, urinary calcium rose from 156 to 203 mg a day, the net amount retained did not differ, and the seven markers of bone metabolism it measured, including markers of bone breakdown and formation, did not change (Cao 2011). That trial was partly funded by the beef checkoff, disclosed the same way the plant-protein funding above is; the checkoff had no role in the analysis.
One older, smaller study complicates a fully clean story, and it belongs here rather than left out. In a study of eight postmenopausal women eating ordinary high-protein foods, not purified protein, at a calcium intake of only 650 mg, five of the eight were in negative calcium balance on each diet, and the deficits on the higher-protein diet were larger, 109 to 197 mg a day against 5 to 103 mg, though the average difference between the diets did not reach statistical significance (Draper 1991). Read together, the honest summary is not that protein is harmless at any calcium intake. It is that the mechanism runs through calcium adequacy: give the body enough calcium and higher protein helps it absorb and retain more of it, rather than pulling it from the skeleton.
What the tracer studies settled is the protein question. The broader acid-load question, whether the overall acid or alkaline balance of a diet affects bone apart from its protein, is more contested, and it is worth being straight about that. In a small Swiss trial, eight healthy young men ate an acid-forming diet and a base-forming diet, the second built on fruit, vegetables and bicarbonate-rich mineral water, with protein held nearly equal at 99 and 92 g a day. Over four days, the acid-forming diet raised urinary calcium by 74 percent and a marker of bone breakdown by 19 percent, which the authors read as calcium coming from bone (Buclin 2001). The study was funded by a grant from Vifor Laboratories. A meta-analysis looking at one supposedly acidic nutrient, phosphate, found the opposite of what the acid-ash idea predicts: more phosphate lowered urinary calcium and slightly improved calcium balance, though the authors note calcium balance is only a stand-in for bone health and most of the underlying studies were of poor quality (Fenton 2009). Longer trials point in different directions. In a two-year trial of 276 postmenopausal women, neither potassium citrate nor an extra 300 g of fruit and vegetables a day slowed bone loss compared with placebo (Macdonald 2008). In another two-year trial, of 201 adults over 65 who were also taking calcium and vitamin D, potassium citrate raised spine density by 1.7 percent compared with placebo (Jehle 2013). And in a short metabolic study, neutralizing the acid from a high-protein diet with potassium citrate did not significantly reduce the extra urinary calcium, suggesting something other than acid load drives it; unlike the tracer studies above, calcium absorption did not measurably change in that study either, so where the extra calcium came from there is not fully settled (Maalouf 2011). A 2018 review that sets out both sides concludes that for people with normal kidneys eating typical Western diets, dietary acid load would not be a readily detectable factor in bone density, but that it could plausibly matter for older people whose kidney function has declined (Frassetto 2018). For a vegan, this debate changes nothing practical: a diet built on fruit, vegetables and legumes already sits on the alkaline side of it.
Which supplements do vegans need for bones?
- Vitamin B12 is mandatory, no exceptions. (Why.)
- Vitamin D if sun exposure is limited: Harvard Health’s guidance is that most people taking a supplement probably need no more than 600 to 800 IU a day, and should avoid more than 4,000 IU unless their medical team recommends it (Harvard Health 2026). EPIC-Oxford vegans averaged 55.8 nmol/L versus 77.0 in meat-eaters. Take it daily rather than as large occasional doses: trials in older adults found more falls with a single yearly dose of 500,000 IU (Sanders 2010) and with 60,000 IU once a month (Bischoff-Ferrari 2016), and the Endocrine Society suggests daily, lower-dose vitamin D for adults 50 and older who need it (Demay 2024).
- Calcium only if food falls short, in small doses of 500 mg or less with meals.
- Algal omega-3s are worth considering (here is who actually needs them).
- Prunes: about 50 g per day (5 or 6 prunes) preserved hip bone density in a 12-month RCT (De Souza 2022).
What is the best exercise for bones, and does it matter as much as food?
It matters as much as the food. Most hip fractures happen because someone fell, not merely because bone was thin. So do both.
- Build bone with progressive resistance training plus impact. The LIFTMOR trial, twice-weekly heavy lifting at 30 minutes a session, raised lumbar spine density and held femoral neck density steady while the control group lost bone, in postmenopausal women with low bone mass, with a single minor adverse event (Watson 2018). Walking alone is not enough, and neither is aerobic exercise on its own.
- For vegans specifically, this may matter more than any single nutrient. In detailed bone-microarchitecture scans of normal-weight vegans, most of whom took supplements, those who lifted at least once a week were nearly indistinguishable from omnivores who lifted, while those who did not lift showed measurably weaker bone structure (Wakolbinger-Habel 2022). In that observational study, lifting went with a much smaller gap, and nutrient intake did not explain it.
- Prevent falls with balance and functional training, which cut fall rates 24 percent; Tai Chi cut them 19 percent; and balance plus strength work together cut them 34 percent (Sherrington 2019).
- Safety: with existing osteoporosis or spinal fractures, avoid loaded forward bending and twisting, and get supervision before heavy or high-impact work.
- The exercise playbook below gives the exact doses: how many jumps, how much rest, how heavy to lift, and how to train the fall reflex itself.
Also watch smoking, heavy alcohol, being significantly underweight, and unnecessary long-term PPI use.
How do you respond to the “vegans break more bones” headline?
It is true but incomplete. It describes 1990s UK vegans low in calcium, protein, and vitamin D, who were also thinner. It does not describe a well-planned vegan diet in 2026 with fortified foods and a barbell.
What does the research say about vegan bone density and fractures?
What did the EPIC-Oxford study find on vegan fractures?
There are two key papers. Appleby, Roddam, Allen and Key 2007 (Eur J Clin Nutr 61:1400) followed 34,696 participants including 1,126 vegans for about 5 years, using self-reported fractures. The vegan incidence rate ratio was 1.30 (95% CI 1.02–1.66). Adjusting for calcium dropped it to 1.15 (0.89–1.49), non-significant. Among those consuming 525 mg per day or more of calcium, the vegan rate ratio was 1.00 (0.69–1.44). Only about 55 percent of vegans hit that threshold, versus roughly 95 percent of everyone else. The authors concluded the excess “appeared to be a consequence of their considerably lower mean calcium intake.” That is a gap that fortified plant milk, calcium-set tofu and low-oxalate greens can close, as the calcium playbook below shows.
Tong et al. 2020 (BMC Medicine 18:353) is the stronger paper, and the one behind the headline. It followed 54,898 participants (1,982 vegans) for a mean of 17.6 years, with fractures identified from hospital and death-certificate linkage rather than self-report. There were 3,941 total fractures, including 945 hip. After adjustment for socioeconomic factors, lifestyle, and BMI:
| Site | Vegans | Vegetarians | Fish-eaters |
|---|---|---|---|
| Hip | 2.31 (1.66–3.22) | 1.25 (1.04–1.50) | 1.26 (1.02–1.54) |
| Total | 1.43 (1.20–1.70) | 1.09 (1.00–1.19) | 0.95 (0.86–1.05) |
| Leg | 2.05 (1.23–3.41) | 1.01 (0.76–1.34) | 1.07 (0.77–1.48) |
| Other main site | 1.59 (1.02–2.50) | 0.99 (0.77–1.27) | 0.98 (0.73–1.33) |
| Arm | 1.56 (0.99–2.44), not significant | 1.25 (1.00–1.56) | 0.92 (0.69–1.22) |
| Wrist | 1.24 (0.84–1.82), not significant | 1.00 (0.83–1.19) | 0.90 (0.73–1.11) |
| Ankle | 1.03 (0.58–1.82), not significant | 1.01 (0.80–1.27) | 0.84 (0.63–1.12) |
The absolute risk difference for vegans was 14.9 more hip fractures per 1,000 over 10 years (7.9 to 24.5). The sequential models are where the story is, and they are worth showing in full because they settle the question of which nutrient is doing the work.
| Vegan model | Total fracture | Hip fracture |
|---|---|---|
| Confounders, no BMI | 1.50 | 2.64 |
| Plus BMI | 1.43 | 2.31 |
| Plus calcium | 1.31 | 1.97 |
| Plus protein | 1.39 | 2.21 |
| Plus both | 1.30 | 1.94 |
Calcium explained roughly 28 percent of the total-fracture excess and 26 percent of the hip excess. Protein explained about 9 percent and 8 percent. Calcium did about three times the work. And the stratified analyses go further: restricting to vegans already eating adequate protein did not lower their risk at all (total 1.52; hip 2.71, from 31 hip fractures), while restricting to vegans getting at least 700 mg of calcium a day also left a substantial excess (total 1.50; hip 2.39, from only 14 hip fractures). So the honest complication is real. Unlike the 2007 paper, calcium adjustment did not fully abolish the signal here, and something beyond the measured nutrients is still unaccounted for. The authors themselves said only that the differences were “possibly” due to lower calcium and protein, and flagged that other factors may matter. That is why the plan in this article covers body weight, vitamin D, B12 and resistance training as well as calcium.
Baseline vegan nutrient intakes were calcium 591 mg per day (versus 1,005 in meat-eaters), protein 13.3 percent of energy (versus 17 percent), and BMI 22.1 (versus 24.5). Plasma vitamin D in vegans was 55.8 nmol/L versus 77.0 (Crowe 2011). Not measured: serum vitamin D at the analysis stage, bone density, falls history, bone-active medications, duration of veganism, dietary quality within the vegan group, fortified-food use, sun exposure, lean mass, and frailty. The authors are explicit about one of these: they could adjust for dietary calcium but had no data at all on calcium supplement use. A vegan who was closing the gap with a supplement is therefore indistinguishable in this dataset from one who was not.
What did the UK Biobank study find?
Webster, Greenwood and Cade 2023 (BMC Med 21:278) followed 413,914 participants with 3,503 hip fractures over a median 12.5 years. In the main analysis, vegans were combined with vegetarians because only 400 vegans and 10 vegan hip fractures were available, and that combined group had a hazard ratio of 1.50 (1.18–1.91), about 3.2 more per 1,000 over 10 years. A sensitivity analysis did split them apart, giving vegetarians 1.38 (1.06–1.79) and vegans 3.26 (1.75–6.08). That 3.26 is the most extreme figure in this entire literature, and it is worth knowing what the authors themselves said about it: that there were not enough vegans in the cohort to assess their risk independently. A confidence interval spanning a 75 percent increase to a sixfold one is what a subgroup of 400 buys you. Worth noting too that the sex pattern here runs opposite to the Adventist finding: vegetarian men had a hazard ratio of 2.04 (1.36–3.08) while vegetarian women were at 1.32 (0.98–1.78), which did not reach significance. The valuable part is the mediation analysis: BMI explained 28 percent of the difference (95% CI 1.1–69.8 percent). Fat-free mass, serum vitamin D, and IGF-1 were not significant mediators, which is worth sitting with for a moment given how often IGF-1 gets raised as the missing piece. Low IGF-1 is a well-established, well-powered predictor of fracture risk in older adults generally, in one Swedish cohort of almost 2,900 men each standard-deviation drop in IGF-1 carried a 45 percent higher hip-fracture risk (Ohlsson 2011), and Tong 2020’s own authors named it as a pathway future research should examine. But when UK Biobank actually tested it as an explanation for the vegetarian-specific excess, it did not hold up. IGF-1 matters for fracture risk in general. It does not appear to be why vegetarians and vegans specifically break more hips. For what it is worth, adequate protein raises IGF-1, by more than 20 percent in one controlled feeding trial (Cao 2011), and protein adequacy is already part of this article’s advice, so nothing here changes what to actually do. The companion UK Women’s Cohort Study found vegetarians at a hazard ratio of about 1.35 (1.04–1.74) (Webster 2022). Its 130 vegans produced a hazard ratio of 1.04, essentially no effect at all, which is the opposite end of the same wide range. Whatever the true size of the vegan hip signal, the response is the same correctable plan laid out above.
What did the Adventist Health Study-2 find?
Thorpe et al. 2021 (Am J Clin Nutr 114:488) followed 34,542 adults with 679 hip fractures. Female vegans had a hazard ratio of 1.55 (1.06–2.26). There was no association in men, and none in the whole-population primary analysis. Then the finding that should have been the headline: vegan women taking neither calcium nor vitamin D had a hazard ratio of 2.99 (1.54–5.82), while vegan women taking both had a hazard ratio of 0.84 (0.42–1.66): no higher risk than nonvegetarians, though the interval is wide enough to leave a modest excess possible, and the authors note that their comparison group, supplemented nonvegetarians, may itself carry slightly higher risk from the supplements. The other arms are less informative: vegan women taking calcium alone had a hazard ratio of 1.62 (0.67–3.93), and vitamin D alone could not be assessed because only 373 women took it without calcium and just 2 of them had hip fractures, so the data cannot say whether calcium, vitamin D or the combination is what matters. Only 32 percent of vegans took both supplements, against about 50 percent in the other diet groups, and the authors conclude that further research is needed to confirm the result. It points the same way as the EPIC-Oxford calcium finding, in a different population, on a different continent, from a different research group, though the two are not the same test: EPIC-Oxford looked at dietary calcium, this analysis at supplement use. A related Adventist analysis, cited above, showed that food choices within the diet matter too: participants eating legumes daily had 64 percent lower hip fracture risk and those eating meat analogues daily 49 percent lower, independent of overall meat intake. Adventists are a homogeneous low-smoking, low-alcohol population, which is a genuine strength for isolating diet. Weaknesses: self-reported fractures, modest vegan event counts, a white-only sample, and case counts in the key supplement table that cannot be right, since it lists 89 hip fractures among the 495 vegan women taking both supplements when all vegan women had 42, so the number of events behind the 0.84 cannot be read from the paper. The study was supported by National Cancer Institute grants, and the Institute had no role in its analysis. More context: what the Adventist Health Studies actually tell us.
What did the Nurses’ Health Study find?
Sotos-Prieto et al. 2024 (JAMA Netw Open 7:e241107) followed 70,285 postmenopausal women for up to 30 years with 2,038 hip fractures. This one compared healthy against unhealthy plant-based diet indices, which score the quality of plant foods without excluding animal foods, so it is not a study of vegans, and every participant was a White woman. In the primary analysis, long-term adherence showed no association with hip fracture: a hazard ratio of 0.97 (0.83–1.14) for the healthful index and 1.02 (0.87–1.20) for the unhealthful one. In a secondary analysis of most recent intake, a high healthful index was linked to 21 percent lower risk (0.79, 0.68–0.92) and a high unhealthful index to 28 percent higher (1.28, 1.09–1.51). The authors write that reverse causation may account for those recent-intake associations, because women with conditions that raised their fracture risk may have changed their diets. So the solid finding here is the null: decades of plant-heavy eating did not raise hip fracture risk. The recent-intake result is suggestive evidence that diet quality matters, not proof of it.
What did the 2025 meta-analysis conclude?
Ballarin et al. in Nutrition Reviews (nuaf241) pooled four cohorts and 529,672 participants. Vegetarians had a hazard ratio of 1.25 (1.11–1.39), with zero heterogeneity and low certainty. Vegans had a hazard ratio of 1.75 (1.17–2.63), with high heterogeneity and very low certainty. In absolute terms that is roughly 2 extra hip fractures per 1,000 for vegetarians, ranging from 1 to 4, and roughly 8 per 1,000 for vegans, ranging from 2 to 17. GRADE downgrades were applied for risk of bias (diet-group changes during follow-up being the leading concern), imprecision, and inconsistency. Note what “very low certainty” means in GRADE: the true effect is likely to be substantially different from the estimate. That is the authors’ own assessment, not mine.
Two further details deserve attention. First, the pooled figures use each cohort’s most fully adjusted model, which for EPIC-Oxford means the estimate that already accounts for calcium and protein intake: 1.94, not the 2.31 that usually gets quoted. So the headline vegan number is not a raw signal. It already has nutrient adjustment built into its largest contributing study. Second, the authors’ own closing recommendation is not that anyone should avoid a plant-based diet. It is that vegans and vegetarians should ensure adequate nutrient intake, supplement where necessary, add resistance exercise, and get bone density screened. That is, almost point for point, the plan this article arrives at, written by the researchers who produced the most alarming number in the field.
Is vegan bone density actually lower?
The density gap is a much smaller story than the fracture gap. A Bayesian meta-analysis found BMD about 4 percent lower in vegetarians and about 6 percent lower at the lumbar spine in vegans, with the authors calling the magnitude “clinically insignificant” (Ho-Pham 2009, meta-analysis). A later analysis of 207 self-identified vegetarians and 9,209 nonvegetarians in a US national survey found the differences became non-significant after adjusting for BMI and waist circumference at the femoral neck and total femur, with only a small lumbar-spine gap under 3 percent remaining; the authors note they could not separate vegans from other vegetarians, so the finding may not apply to vegans (Karavasiloglou 2020). One meta-analysis reported a vegan fracture risk ratio of 1.44 but used crude, unadjusted ratios, which drew a formal critique (Iguacel 2019). So the density gap is small, and in vegetarians it is largely explained by body size; whether that holds for vegans is less certain. The fracture signal is the one worth taking seriously.
What do the other vegan bone studies show?
The four big cohorts are not the whole literature. Roughly twenty more studies exist, and they mostly reinforce the same picture.
In Vietnam, 105 vegan Buddhist nuns were compared with 105 omnivores. Despite lower calcium and lower protein intake, there was no significant bone density difference, and femoral-neck osteoporosis rates were statistically indistinguishable at 17.1 versus 14.3 percent (Ho-Pham 2009, nuns study). A two-year follow-up found no significant fracture difference, though it was underpowered with only about 10 cases (Ho-Pham 2012). A 2024 cross-sectional comparison of vegans, lacto-ovo vegetarians and omnivores also found no significant bone density difference between groups, though it did find higher parathyroid hormone in the vegans, with hyperparathyroidism no more prevalent in any one group (Galchenko 2024). Two of its authors are affiliated with the Scientific Society for Vegetarian Nutrition, which is worth flagging; the funding note near the end explains why that kind of tie weighs less than a commercial one but still deserves naming. And a 2024 Australian study of 240 adults found whole-body bone density statistically indistinguishable from regular meat eaters for every plant-based group except lacto-ovo vegetarians (Austin 2024).
The most recent meta-analyses do find a signal. A 2025 systematic review covering about 243,000 people found vegan diets associated with higher osteoporosis risk, more pronounced in people over 60 and in those following the diet more than 10 years (Zheng 2025). Those are the people for whom the calcium, vitamin D, protein and resistance-training steps below matter most, and each of them is something a long-term or older vegan can do.
One notable gap: despite Taiwan’s large Tzu Chi vegetarian cohorts publishing national-registry analyses on stroke, dementia, diabetes and gout, no equivalent incident-fracture study using vegetarian status appears to have been published. Taiwanese vegetarian bone evidence remains cross-sectional density work only.
Does resistance training erase the vegan bone gap?
This may be the most useful finding in the whole literature, and it has been underreported. An Austrian study used HR-pQCT, which measures bone microarchitecture in far more detail than a standard DXA scan, on 43 long-term vegans and 45 omnivores older than 30 and younger than 50 (Wakolbinger-Habel 2022). What makes it unusually informative is who the vegans were. Recruited through the Austrian Vegan Society, they had been vegan for about ten years on average, none were underweight, most took supplements, their B12 and vitamin D intakes were well above target, and their blood vitamin D levels matched the omnivores’. These were the well-planned vegans that the older fracture cohorts mostly lacked.
Even so, vegans who did not do resistance training had measurably weaker bone structure than omnivores who did not train, mainly in the spongy trabecular bone inside the wrist and shin. Vegans who did train were nearly indistinguishable from omnivores who trained, differing on only one bone-structure measure, trabecular thickness at the shin, though they did differ on some blood and diet measures: the vegan lifters had higher vitamin D and parathyroid hormone levels than the omnivore lifters and ate less protein per kilogram. Here is the detail that makes this more than a general fitness finding: the vegan lifters and the vegan non-lifters ate essentially the same diet, with no significant difference in any nutrient. Lifting helped both groups, but the differences between lifters and non-lifters were stronger among the vegans, and the authors later wrote that resistance training’s benefit might be even stronger on a vegan diet (Wakolbinger-Habel 2023). The bar was low: one session a week of machines, free weights or bodyweight exercises counted. Vegans who did only aerobic exercise looked no better than vegans who did no sport at all, though that comparison rested on very small groups. Lifting, not supplementation or nutrient intake, was what separated the vegans with better bone structure, though in an observational study that is an association, not proof.
The blood work echoed the calcium story running through this article. The vegans averaged about 785 mg of calcium a day against 937 in the omnivores, had slightly lower serum calcium, and showed a trend toward higher parathyroid hormone, which the authors suggest could reflect the body trying to absorb more calcium from the gut, though they also note that parathyroid hormone rises with exercise. Their dietary acid load was less than half the omnivores’, which the old acid-ash theory predicts should protect bone, yet the non-lifting vegans still had weaker structure. Whatever role acid load plays elsewhere, it was clearly not the deciding factor in this group.
The limits are real. This is a small observational study, so it shows association rather than cause, and with 43 vegans it had little power to detect nutrient effects, which the authors acknowledge. Resistance training was self-reported with no data on dose. The scans were of the wrist and shin in adults under 50, not the hip or spine in the older people who actually break hips, and the study measured structure, not fractures. And one confounder cuts directly against the headline: the non-lifting vegans had the heaviest smoking history of any group, about 5.3 pack-years on average against almost none among the vegan lifters, a significant difference, though their gap with the non-lifting omnivores (about 1.8) was not significant and the wide spread suggests a few heavy smokers drive the average. Smoking damages bone on its own. The bone measures were adjusted for body size, but the paper reports only testing whether smoking differed between groups, not adjusting for it, so part of the lifting advantage may be a smoking disadvantage. Even discounted for that, it looks at vegans whose intakes of protein, B12, calcium and vitamin D were adequate, largely through supplements, and asks what else separates strong bone from weak, and what it points to is the lever that costs nothing but effort.
Do vegan children have weaker bones?
This is where the evidence is most concerning, and it deserves to be stated plainly rather than buried. A study of 187 Polish children aged 5 to 10, including 52 vegans, found lower bone mineral content that persisted after adjusting for body size, roughly 3.7 percent lower for total body and 5.6 percent lower at the lumbar spine, and vegan children were about 3 cm shorter, despite having a healthier cardiovascular profile (Desmond 2021). A Finnish study of 40 children found vegan children had lower vitamin D status, lower protein intake, and consistently lower serum essential amino acids (Hovinen 2021). A 2025 Finnish study found higher bone turnover markers in plant-based children and adults even when vitamin D status was good and calcium was roughly adequate (Itkonen 2025). These gaps point to nutrients that can be deliberately supplied, as the next paragraphs lay out.
Why does adjusting for body size close most of the adult gap, at least in the vegetarian data, but not this one? Because a settled adult skeleton and a growing one are not the same problem. An adult vegan who weighs less than an omnivore is often simply smaller, in equilibrium, and in adult vegetarians that equilibrium explains most of the density difference once you account for it. A child is not in equilibrium. They are laying down new bone and new height on a fixed developmental timeline, which means their energy and amino acid needs per kilogram of body weight run well above an adult’s, not as an optional upgrade but as the literal raw material growth is built from. The oldest illustration of that gap comes from century-old rat studies. Osborne and Mendel found that young rats fed wheat gliadin as their only protein could maintain their weight but not grow, and that adding the single amino acid lysine made it adequate for normal growth; their own conclusion was that supplying the missing amino acids should make any digestible protein adequate (Osborne and Mendel 1914). That animal finding is often retold as proof that plant protein is inferior, but what it showed is narrower and still holds: a diet that is entirely adequate for a weight-stable adult can genuinely fall short for a child whose per-kilogram demands are much higher. Plant-based diets are also bulkier and more filling per calorie than the diets they replace, an advantage for an adult managing weight and a real challenge for a small child, who can reach fullness before reaching their calorie or protein target for the day. The 2001 pediatric review’s answer is to build the diet around soy foods, other legumes, nuts and nut butters, and grains, which supply more concentrated energy and can support appropriate growth (Messina 2001). The practical takeaway is not that children should not be vegan; it is that vegan children need deliberate attention to calcium, vitamin D, protein adequacy, and total energy intake. Protein needs run slightly higher for vegan children: a 2001 review estimated them at about 30 to 35 percent above the standard recommendation under age 2, 20 to 30 percent for ages 2 to 6, and 15 to 20 percent over age 6, an extra 2 to 14 g a day (Messina 2001), targets that are easy to reach with generous legumes, tofu, tempeh, soy milk, and nuts. “We adjusted for body size and it went away” does not hold in this age group the way it largely does in adult vegetarians. See raising children vegan.
Which nutrients are vegans low in for bone health?
The RBVD study compared 36 vegans with 36 omnivores (Menzel 2021). Vegans had lower calcaneal ultrasound bone parameters and lower vitamin A, B2, lysine, zinc, selenoprotein P, omega-3s, iodine, and calcium, but higher vitamin K1 and folate. A 12-biomarker pattern tracked with bone status. This is essentially a checklist of what to pay attention to, and note that most of these are correctable through food choice rather than inherent to plant eating.
Lysine deserves a second look, because it is the one item on that list with a plausible mechanism attached to an actual fracture outcome rather than just a biomarker gap. It is the amino acid legumes concentrate and cereal grains lack, and it is one of two amino acids, alongside leucine, that a 2025 digestibility-adjusted analysis of long-term New Zealand vegans found most likely to fall short: while roughly three in four vegans met their overall protein requirement, only about half cleared the bar for lysine and leucine specifically once digestibility was accounted for (Soh 2025). That gap lines up with the finding elsewhere in this article that daily legume eaters had a markedly lower hip-fracture rate than infrequent eaters. It does not prove lysine is the mechanism connecting the two, but it is a coherent enough chain, low legume intake, low lysine, weaker bone measures, that eating legumes daily reads as more than a general protein-adequacy habit. It may be a bone-specific one. More on protein adequacy generally in how much protein you actually need.
Did the vegan bone studies control for confounding variables?
| Confounder | Appleby 2007 | Tong 2020 | Webster 2023 | Thorpe 2021 | Sotos-Prieto 2024 |
|---|---|---|---|---|---|
| Age | yes | yes | yes | yes | yes |
| Sex | yes | yes | yes | yes | women only |
| BMI | partial | yes | yes (28% mediation) | yes | yes |
| Physical activity | crude | crude | yes (MET-minutes) | crude | yes (MET-hours) |
| Smoking | yes | yes | yes | yes | yes |
| Alcohol | yes | yes | yes | yes | yes |
| SES / education | yes | yes | yes | partial | no (all nurses) |
| Menopause / HRT | partial | yes | yes | yes | yes |
| Calcium intake | yes (key) | yes (stratified) | no | yes (supplements) | supplements only |
| Protein intake | yes | yes (stratified) | no | partial | no |
| Serum vitamin D | no | no | yes (null mediator) | supplements only | supplements only |
| Vitamin B12 | no | no | no | no | no |
| Prior fracture / falls | no | no | partial (prior fractures) | no | partial (prior hip fracture excluded) |
| Bone-active meds | no | no | no | no | partial |
| Duration of diet | no | no | no | no | yes |
| Diet change in follow-up | no | partial | yes (little change) | no | yes |
| Diet quality within vegans | no | no | no | no | yes |
| Lean / muscle mass | no | no | yes (null mediator) | no | no |
| Fracture ascertainment | self-report | hospital linkage | hospital linkage | self-report | self-report |
Are vegans just more active, so they break bones falling off bikes?
The fracture-site pattern argues strongly against this. The vegan excess was concentrated in hip, leg and vertebral fractures, the classic frailty fractures from falling at standing height. When the grouped “other main sites” category was broken apart, vertebral fracture was the one site within it that reached significance in vegans. There was no significant excess in wrist or ankle fractures, which are exactly what active people get from sport and from falling on an outstretched hand. If the “healthy active vegan” explanation were right, the signal would show up in the wrists. It does not. It shows up in the hips.
If vegans are healthier, are they more active, and did the studies account for that?
This is a sharper version of the question, and it has two halves that pull in opposite directions.
The first half favors the studies. Every cohort did adjust for physical activity, and if vegans really were more active, that adjustment should have pushed their fracture risk down, not up. More activity means better balance, stronger muscle, and fewer falls, plus a direct bone-loading benefit. So the excess appearing anyway, after that adjustment, means activity is not manufacturing the result out of nothing. If anything, being more active should have masked a real signal rather than created a false one.
The second half favors the objection. The adjustment was crude in most cohorts, which used self-reported activity categories, essentially rough buckets of how much people moved; UK Biobank did better, as the exercise answer below explains. None of them measured the thing that actually builds bone and prevents fractures, which is resistance training and impact loading. A vegan cyclist and a vegan who lifts twice a week land in the same “active” bucket while having very different skeletons. Non-impact activity such as cycling and swimming does close to nothing for bone density. So a group could look equally active on paper and still be doing far less of the kind of loading that matters.
The Austrian bone-microarchitecture study is the clearest evidence that this distinction is doing real work. Vegans who did resistance training had bone structure nearly indistinguishable from omnivores who trained. Vegans who did not train had measurably weaker structure, despite eating the same diet as the vegans who did (Wakolbinger-Habel 2022), with the smoking caveat noted above. That is exactly the variable the big cohorts could not see. So the fair conclusion is that activity was controlled for in the ordinary sense but not in the sense that matters most for bone, and residual confounding by type of activity remains one of the more plausible unmeasured explanations for part of the excess.
Did they control for diet quality, since “vegan” is not one diet?
Mostly no, and this is a real gap. Only the Nurses’ Health Study distinguished healthy from unhealthy plant-based eating, and when it did, long-term diet quality showed no association, while a secondary analysis of recent diet found 21 percent lower risk for high-quality plant-based diets, a result its authors say may reflect reverse causation. Ethical veganism and health-optimized veganism produce very different nutrient profiles, and the fracture cohorts could not tell them apart.
Did they control for exercise?
Better than the usual objection assumes, though still not well enough. EPIC-Oxford and the Adventist study used self-reported activity categories. UK Biobank did more, adjusting for total metabolic-equivalent minutes per week as a continuous measure, and it found physical activity essentially identical across diet groups: about 2,984 MET-minutes a week in regular meat eaters against 2,895 in vegetarians. So in the largest cohort, the “vegans are simply more active” story is not visible in the data at all. What none of them captured is resistance training, impact loading, or fall risk, and since those are the interventions that actually build bone, residual confounding by type of activity remains entirely plausible.
Did they control for how long people had been vegan?
Nobody measured it, and there is good reason to think this matters more than it sounds. In the Adventist cohort, only about 1.1 percent of participants were lifetime vegans, against 31 percent who were lifetime lacto-ovo vegetarians. Most vegans were relatively recent converts (Martins 2017). Bone is built over decades, so a cohort that follows recent converts for six or eight years is substantially measuring the skeleton their previous diet built, then attributing the result to veganism. EPIC assessed diet at baseline with one resurvey around 2010; the Adventist and UK Women’s cohorts never reassessed at all. The 2025 meta-analysis flagged diet-group change during follow-up as its single leading source of bias.
Did they control for vitamin D and other nutrients?
Calcium and protein were measured and adjusted in EPIC-Oxford, and in part in the Adventist study, and they explain a substantial share; the Nurses’ Health Study adjusted only for calcium and vitamin D supplements. Serum vitamin D was tested only in UK Biobank, where it was not a significant mediator, and measured separately in EPIC-Oxford, where vegans ran lower. B12, vitamin K, magnesium, zinc, omega-3, iodine, and selenium were essentially unmeasured.
The B12 gap deserves particular attention, because EPIC-Oxford’s own biomarker sub-study answered the question the fracture paper did not ask. It found that 52 percent of the cohort’s vegans were B12-deficient by a strict cutoff, against roughly one omnivore in 226 (Gilsing 2010). That is a gap a B12 supplement closes, as the next paragraph explains.
Low B12 raises homocysteine, and the link from there to fracture is not a loose association, it is a quantified one from two independent studies published side by side in the same journal issue. In the Framingham cohort, people in the highest quarter of homocysteine levels had roughly four times the hip-fracture risk of the lowest quarter in men and about twice the risk in women (McLean 2004). A separate analysis of 2,406 older adults from two Dutch cohorts found each standard-deviation rise in homocysteine carried roughly 40 percent higher fracture risk, independent of bone density itself (van Meurs 2004). The proposed mechanism is biochemical rather than mineral: elevated homocysteine is thought to interfere with collagen cross-linking, weakening the structural scaffold bone is built on rather than depleting its calcium content. The reassuring part: this specific risk factor is cheap and easy to close. A B12 supplement costing a few dollars a year corrects the deficiency and lowers homocysteine directly, which is a large part of why this article treats B12 as mandatory rather than optional. One honest caveat travels with this, and the researchers who ran both studies raised it themselves: whether homocysteine directly causes fracture or simply rides alongside other B-vitamin problems is still debated, so treat it as a strong, plausible contributor rather than settled causation.
What is not in dispute is that the cohort behind the loudest fracture headline had the deficiency. The single most-cited vegan fracture finding comes from a population in which half the vegans were deficient in a nutrient with a quantified link to the outcome, recruited in the 1990s before routine supplementation, with fracture models that adjusted for neither B12 status nor supplement use. That does not explain away the finding, but it does mean the result describes a specific and largely unsupplemented population rather than veganism as such. The gap itself is closed by a cheap B12 supplement. Nor is EPIC-Oxford alone here. The Adventist cohort, the same one whose calcium-and-vitamin-D finding does much of the reassuring work throughout this article, had a smaller but real version of the same gap: its own calibration study found about 15 percent of vegans fell short of the B12 requirement (Damayanti 2018), and its fracture models did not adjust for B12 status either. More on this in The Great B12 Gamble.
Were the vegans in these studies health-optimizers or ethical vegans?
This is the confounder nobody adjusts for, and it probably runs against the vegans rather than for them. Research on why people avoid animal products finds that moral concern for animals tends to rise with how strict the diet is. A 2018 review summarized studies finding vegans more strongly motivated by ethical concerns than vegetarians, and vegetarians more strongly motivated by concern for animals than flexitarians, who tend to cite health or the environment; the evidence on part-time meat avoiders is mixed, though, and one study it cites found health the most common motive among vegetarians (Rosenfeld 2018). In a separate set of surveys, vegans differed from non-vegans most on environmental and animal-rights motives and least on health; that work was funded by Animal Charity Evaluators (Hopwood 2020). Neither source looked at pescatarians specifically.
Follow that through and the usual objection inverts. The standard complaint about vegetarian cohorts is that health-conscious people self-select into them, inflating the benefits. But if health-motivated people preferentially land on less restrictive diets while ethically motivated people land at veganism, then the health-conscious bias favors the fish-eating and vegetarian groups, and the vegan column fills up with people who went vegan for the animals, in years when fortified foods and supplement guidance were less common, and so had little reason to think about calcium, vitamin D, or B12, all of which are easy to cover. Combined with the finding that half of EPIC-Oxford’s vegans were B12-deficient, the picture is not of an optimized diet underperforming. It is of a largely unoptimized one, being compared against groups with more health-driven members.
Worth stating the limit honestly: this motivation research comes from general populations, not from the fracture cohorts themselves, so the sorting is plausible in those cohorts rather than demonstrated there. It is a reason to hold the finding loosely, not a reason to dismiss it.
Did they control for age?
Adjusted everywhere. Worth noting that vegans were on average younger, which if anything biases toward fewer fractures, so age adjustment strengthens the finding rather than creating it.
Did they control for BMI and body weight?
This one is central. Low BMI is an established independent hip-fracture risk factor via reduced mechanical loading and lower adipose-derived estrogen. UK Biobank’s formal mediation put it at 28 percent of the vegetarian excess. Tong 2020 explicitly noted the associations were stronger without BMI adjustment. So low body weight is a real part of the causal pathway, but it does not account for all of the vegan hip-fracture excess, which survived BMI adjustment. It is also one of the more actionable pieces: keeping energy and protein intake adequate and not drifting underweight is within a vegan’s control.
One honest complication is worth naming, because it cuts the other way on a different question elsewhere. When cohort studies of vegetarian diets and ischaemic heart disease report results both with and without adjustment for body weight, adjusting shrinks the lower risk by roughly a fifth, from 22 to 18 percent (Dybvik 2023), and the fair reading of that is not that the benefit was overstated, but that lower body weight is partly how a plant-based diet protects the heart in the first place, so adjusting it away controls off part of a real effect rather than a pure confounder. The same logic applies here, just pointed the other direction. If lower body weight is genuinely one of the mechanisms by which a vegan diet affects fracture risk rather than an unrelated variable riding alongside it, then reporting only the BMI-adjusted hazard ratios understates the full picture, the same way it would understate a benefit in the heart-disease case. This article generally leads with the BMI-adjusted numbers because they isolate what a person can fix independent of body weight, which is the more useful framing for practical advice. But the unadjusted numbers, which ran higher in both EPIC-Oxford and UK Biobank, are not wrong, and a reader weighing how alarmed to be should know both exist.
So did the studies control for confounders or not?
A common vegan talking point is that these studies “didn’t control for confounders.” That is not accurate, and repeating it will get you correctly fact-checked. They did adjust for age, BMI, activity, smoking, alcohol, socioeconomic status, calcium, and protein. The defensible position is narrower and stronger: unmeasured confounding remains plausible (vitamin D status, B12, diet quality, fall risk, muscle mass, diet duration), and the measured mediators of calcium, protein, BMI, and vitamin D supplementation already explain a substantial share of the excess, and in some analyses all of it, though not in Tong 2020, where an excess remained among vegans eating adequate calcium or protein. That claim survives scrutiny. The other one does not.
Where does the vegan bone evidence rank?
Randomized controlled trials of vegan diets with fracture endpoints do not exist. This is the fundamental limitation of the entire debate. Every diet-group fracture finding is observational.
For RCTs of individual nutrients, the largest and most recent pooled analysis, a 2026 BMJ meta-analysis of 69 trials and 153,902 participants, found that calcium alone and vitamin D alone did not significantly reduce any fracture or fall outcome. Combined calcium and vitamin D lowered any fracture by 9 percent (risk ratio 0.91, 0.84 to 0.99) and hip fracture by 16 percent (0.84, 0.74 to 0.96), which is 1.0 and 0.3 percentage points in absolute terms; the authors judged both too small to matter against thresholds they set in advance, noted that the reductions were largely explained by one trial in very frail women in residential care, and said evidence for high-risk and residential-care groups was limited (Massé 2026). Soy isoflavones are inconsistent. Vitamin K2 and MK-7 are mixed, at lower certainty. Prunes are positive, but industry-funded. Protein tracer studies show it increases calcium absorption rather than pulling calcium from bone, provided calcium intake is adequate.
Prospective cohorts are the core evidence: EPIC-Oxford, AHS-2, UK Biobank, and the Nurses’ Health Study. The 2025 pooled meta-analysis is GRADE-rated low to very low. Cross-sectional and biomarker work, such as RBVD and NHANES BMD comparisons, offers no causal direction. The strongest counter-arguments to the pro-vegan position: the hip signal replicates across EPIC-Oxford and AHS-2; it persisted in Tong 2020’s higher-calcium subgroup; and RBVD shows multiple biomarker deficits beyond calcium alone. Where the risk appears correctable: adequate calcium (700 to 1,000 mg, leaning toward 1,000), adequate protein, vitamin D repletion, adequate body weight, and, per AHS-2, combined calcium and vitamin D supplements, with which vegan women showed no higher hip-fracture risk than nonvegetarians, though the estimate is imprecise.
What actually prevents bone fractures?
Do most fractures come from falling rather than weak bones?
Only about 15 percent of low-trauma fractures in older women are attributable to DXA-defined osteoporosis. Hip fracture risk rises roughly 13-fold between ages 60 and 80, while age-related bone density decline explains only about 2-fold of that (Jarvinen 2015, “Osteoporosis: the emperor has no clothes”). This matters enormously here, because the vegan signal is a hip signal, a fall signal. Which means fall prevention is not a side note. It is half the intervention.
What exercise prevents falls?
| Intervention | Reduction in fall rate |
|---|---|
| Exercise (any) | 23% |
| Balance / functional training | 24% |
| Multiple-component (balance + strength) | 34% |
| Tai Chi | 19% |
The overall exercise estimate was rated high certainty. Exercise also reduced the number of people sustaining a fall-related fracture by 27 percent (Sherrington 2019, Cochrane).
What exercise builds bone density?
The LIFTMOR trial ran 8 months of twice-weekly, 30-minute supervised high-intensity resistance and impact training (5 sets of 5 reps at roughly 80 to 85 percent of one-rep max) in postmenopausal women with osteopenia or osteoporosis. It produced a 0.3 percent gain in femoral neck BMD versus a 2.0 percent loss in controls, a 2.9 percent gain in lumbar spine versus a 1.2 percent loss, and a 13.6 percent gain in femoral neck cortical thickness versus 6.3 percent in controls. There was one minor adverse event, a low-back muscle strain that resolved within a week, and compliance was 92 percent (Watson 2018). Practical translation: heavy compound lifts twice a week, plus impact work, plus balance and agility drills. Walking is good for you but it is not a bone stimulus on its own.
Do vibration plates or Power Plates build bone?
The evidence is weak and mixed. Some small BMD signals exist; fracture-reduction support does not. Reasonable as an adjunct, not a substitute for loading. And the claim that “exercise is twice as effective as medication” has no rigorous head-to-head trial behind it. The direction, that exercise is powerful and underused, is defensible. The specific multiplier is not.
How much protein and B12 for bones?
Protein at 0.8 g/kg is a floor, not a target. For older adults, 1.0 to 1.2 g/kg is the range the evidence on preserving muscle function supports, and muscle matters for falls. For bone specifically, a meta-analysis of four US cohort studies of older adults found higher protein intakes associated with about 11 percent lower hip-fracture risk (hazard ratio 0.89, 0.84–0.94), but its authors rated the evidence low certainty, said it was not strong enough to raise the bone-specific recommendation, and assumed vegans were too rare in those cohorts to matter (Groenendijk 2019). The protein version of the acid-ash hypothesis, that protein acidifies you and dissolves your skeleton, does not hold up against calcium isotope tracer studies, which found the extra urinary calcium on a high-protein diet comes mainly from increased intestinal absorption rather than bone, provided calcium intake itself is adequate (full tracer data above). Whether overall dietary acid load matters apart from protein is still contested, and the short answer for vegans is that their diets already sit on the favorable side of it. B12 is essential regardless, and low B12 with elevated homocysteine is independently linked to fracture risk. Calcium, vitamin D and K2 get their own full section below.
Do prunes help bone density?
Yes, in the best trial so far. 50 g per day prevented total hip BMD loss at 6 months, sustained to 12 (De Souza 2022). The study was funded by the California Prune Board, which is worth flagging.
The calcium playbook for vegans
Here is the honest starting point: plenty of people eating plant-based do not eat bok choy daily, use plant milk only occasionally or an unfortified kind, buy tofu without checking how it was set, and get kale when they can. If that sounds like you, you are in good company, and this section is written for you. It answers, in order: how much you need, whether too much calcium from food can hurt, whether supplements are dangerous, which form and dose if you do supplement, which foods actually deliver, how to prepare and pair them, and who should not supplement without medical advice.
How much calcium do vegans actually need?
The targets disagree more than you would expect. The UK says 700 mg a day for adults. The US RDA says 1,000 mg, rising to 1,200 for women over 50 and everyone over 70. Calcium-balance studies suggest the true requirement may sit lower than the US figure, and the body compensates across a wide range of intakes by absorbing more when intake is low (Hunt and Johnson 2007). What the vegan fracture data add is a floor: risk rose in vegans below roughly 525 mg in the 2007 EPIC-Oxford analysis, and EPIC vegans averaged 591 mg. A pragmatic vegan target is 700 to 1,000 mg a day from food, with the upper end for postmenopausal women and older adults. There is no evidence that pushing past 1,200 mg helps, and some evidence it can hurt.
One more finding pushes toward the upper end of that range rather than the lower. In a controlled trial at about 800 mg of calcium a day, which was the US recommendation for women over 50 until 1997, when it was raised to 1,200 mg (Harvard Health 2026), women stayed in negative bone balance regardless of whether they ate moderate or high protein; more protein raised calcium absorption but did not fully close the gap (Kerstetter 2005). Current US recommendations are higher, 1,000 to 1,200 mg depending on age, though the study’s own authors were careful to say it remained an open question whether the higher target would actually restore balance. That is one more reason 1,000 mg, not 700, is the safer place to land if you are hitting your numbers reliably. Not every authority agrees: Harvard Health’s current guidance is that at least 700 mg from food should be enough, partly because the higher US targets rest on calcium-balance studies only a few weeks long (Harvard Health 2026). Both positions agree on the part that matters most for vegans, which is getting the calcium from food rather than large supplement doses. A 2015 systematic review adds a caution that is not specific to vegans. It found only two small randomized trials of dietary calcium sources that reported fractures, a milk-powder trial that recorded four fractures in total and a trial that did not report fractures separately for its dietary arm, too few to draw conclusions, and most of the cohort studies it reviewed found no association between dietary calcium and fracture (14 of 22 for total fracture, 17 of 21 for hip); it concluded that there is no clinical trial evidence that increasing calcium intake from dietary sources prevents fractures (Bolland 2015). A separate BMJ analysis by two of the same authors states that observational evidence does not suggest that raising dietary calcium intake lowers fracture risk (Grey and Bolland 2015). Two of the three cohort analyses that analysis cites fit that summary: a meta-analysis of seven cohorts found no association between total calcium intake and hip fracture in women, and a risk ratio of 0.92 per extra 300 mg a day in men that fell short of significance (Bischoff-Ferrari 2007), and a US study of 9,704 women aged 65 and older found no important association between dietary calcium and any fracture (Cumming 1997). The third does not fully fit it. In 61,433 Swedish women followed for 19 years, those in the lowest fifth of intake, below about 750 mg a day, had 18 percent more fractures of any kind and 29 percent more hip fractures than women eating 882 to 996 mg; intakes above that brought no further benefit, and the top fifth, above about 1,140 mg, had more hip fractures (Warensjö 2011). The authors say that high-intake hip result should be cautiously interpreted and might reflect reverse causation, with women prone to osteoporosis deliberately eating more calcium, and they found the low-intake excess was more pronounced in women who also had a low vitamin D intake. That pattern, a floor below which risk rises and little benefit above it, is consistent with the floor described above for vegans; it is not evidence that pushing intake higher helps. The case for leaning toward 1,000 mg here rests on the calcium-balance data above and on the vegan cohorts, not on fracture trials of dietary calcium.
Can too much calcium from food weaken your bones?
This claim usually cites hypercalcemia, a condition where blood calcium runs too high, and it gets the direction backwards. Mayo Clinic’s own description says the extra calcium in hypercalcemia most often comes out of the bones, and that the most common cause is overactive parathyroid glands; the other causes it lists include cancer, certain medicines, prolonged immobility, dehydration, and taking too much calcium or vitamin D in supplement form, while calcium from ordinary food is not on the list (Mayo Clinic). In other words, the weakened bone is where the high blood calcium came from, not a consequence of eating calcium. In a healthy person the system runs the other way: when more calcium is absorbed, parathyroid hormone, the signal that releases calcium from bone, goes down rather than up; in the protein trial above it fell as absorption rose, which the authors read as the body compensating (Cao 2011). Harvard Health’s article on calcium draws the same line: the downsides it describes apply to high-dose supplements, not to calcium from a regular diet, and it notes that dietary calcium is thought to protect against kidney stones (Harvard Health 2026). A small tracer study points the same way. In 12 postmenopausal women whose bones had been labeled with a calcium tracer, four servings of dairy a day (about 1,300 mg of calcium) and a 1,200 mg calcium carbonate supplement each reduced the tracer leaving bone within one to two weeks, with no difference between them, and the effect reversed within one to two weeks of stopping (Rogers 2016). That is a six-week measure of bone calcium turnover, not fractures, and its corresponding author has received research funding from the National Dairy Council, as disclosed in another of her papers (Demmer 2016). And for anyone comparing cow’s milk to breast milk on calcium content, fortified soy, almond and rice milks carry the same roughly 300 mg per cup as cow’s milk in the same Harvard table, which Harvard sources to the International Osteoporosis Foundation, so that comparison does not separate them.
Are calcium supplements dangerous?
The concern is real, but it is narrower than the headlines, and it is contested. The alarming figures come from Bolland and Reid: their 2008 trial gave 1,000 mg of calcium citrate a day to healthy older women and saw more heart attacks (Bolland 2008), and their reanalysis of the Women’s Health Initiative, where women took 1,000 mg of carbonate plus vitamin D, found more heart attacks and strokes among those who had not been supplementing before the trial (Bolland 2011). The proposed mechanism is that a large pill spikes blood calcium for several hours in a way food never does, and that repeated spikes may promote clotting and vascular calcification. Two findings support the mechanism: people with genetically higher blood calcium have higher coronary disease risk (Larsson 2017), and in the MESA cohort, supplement users showed more coronary artery calcification while people getting the same calcium from food showed less (Anderson 2016).
The evidence is genuinely split. A 2021 meta-analysis of 13 double-blind placebo-controlled trials came down on the harm side, finding calcium supplements raised cardiovascular risk by about 15 percent in healthy postmenopausal women. Notably, in its subgroup analysis the significant signal sat at a supplementary intake of 1,000 mg a day (Myung 2021). Against that, a large review commissioned by the National Osteoporosis Foundation found no such hazard from calcium, whether from food or supplements, at moderate intakes (Chung 2016). A 2023 pooled analysis of 11 placebo-controlled trials, from an Oxford-led collaboration funded by the UK Medical Research Council and the British Heart Foundation, also found no significant excess risk of heart attack, coronary death or stroke, with calcium alone (heart attack risk ratio 1.15, 0.88 to 1.51) or with vitamin D (1.09, 0.95 to 1.25), and concluded that excess coronary or stroke risks above 0.3 to 0.5 percent a year could be ruled out; none of its trials had heart disease as their main outcome (Huo 2023). What both sides agree on is that supplements buy little fracture benefit in healthy community-dwelling adults. One meta-analysis of the trials found calcium supplements did not reduce nonvertebral fractures (risk ratio 0.92, 0.81–1.05), and in the four trials with separate hip data, hip fractures were more common with calcium than with placebo (1.64, 1.02–2.64), which its authors read as a possible increased risk (Bischoff-Ferrari 2007). The 2026 review found almost the same point estimate for calcium alone, 1.63, but with a wider interval that includes no effect (0.86 to 3.03; 6 trials; very low certainty), against 1.51 (0.93 to 2.48) in the 2015 review; its authors called the trend imprecise, uncertain and biologically implausible, and put calcium alone at 0.91 (0.81 to 1.01) for any fracture, which they read as likely little to no effect (Massé 2026; Bolland 2015). So the fair summary is: whether calcium supplements raise cardiovascular risk is unsettled, with the 2023 pooled analysis finding no significant harm even though its point estimates for heart attack sit above 1, and the benefit supplements buy in the general population is small. That combination is why food-first is the right default.
Is a low-dose supplement safe?
Probably safer than a large one, on the mechanism, but I want to be honest about the gap. Nearly every trial behind the harm signal used 1,000 mg or more, often in one or two doses. Almost no trial has tested 500 mg or less against cardiovascular outcomes, so the safety of low-dose supplementation is inferred from the spike mechanism rather than demonstrated. What the mechanism predicts is that a smaller dose taken with food produces a smaller, slower rise in blood calcium, which is the profile the harm hypothesis says is safe. What the vegan data add is a pattern consistent with supplementing helping: in the Adventist cohort, vegan women who took both calcium and vitamin D showed no higher hip-fracture risk than nonvegetarian women, against nearly triple the risk for those taking neither. That is a fracture outcome, not a marker, but it is an observational comparison in a small subgroup with a wide interval, so it is the best available argument rather than a proven one that a vegan who cannot close the gap with food should close it with a small supplement rather than leave it open. Randomized trials do not supply the missing proof either. The 2026 meta-analysis found no credible difference in effect across its subgroup analyses, which included baseline dietary calcium below versus above 800 mg a day and baseline vitamin D level, though only 19 of its 69 trials enrolled high-risk populations and its authors say evidence for high-risk and residential-care groups was limited (Massé 2026).
Is calcium from fortified plant milk or tofu the same as a pill?
No, and the difference runs in your favor. Three reasons. First, the dose per serving is modest: a cup of fortified soy milk is about 300 mg, a half cup of calcium-set tofu roughly 250 mg or more, nothing like a 1,000 mg tablet. Second, it arrives with food, in a matrix that slows absorption and blunts the blood-calcium spike that the harm hypothesis is built on. Third, the cohort studies that found the harm signal separated dietary from supplemental calcium, and dietary calcium did not show it; in one German cohort, dietary calcium looked protective while supplements looked harmful (Li 2012). In one test, an Australian fortified soy milk was absorbed as well as cow’s milk, though that was a single small dose given to 12 women, and the study was supported by the soy milk’s maker, Sanitarium, which employed one of the authors (Tang 2010). The calcium also has to make it into your glass, which is why the carton needs a hard shake (Heaney 2005). The one honest limit: no trial has tested fortified foods specifically against cardiovascular outcomes, so this is mechanism plus observational data, not a randomized proof.
Carbonate or citrate, and when to take it?
- Calcium carbonate is 40 percent elemental calcium, cheapest, and needs stomach acid to dissolve, so take it with a meal. It is the one most likely to constipate.
- Calcium citrate is about 21 percent elemental calcium, absorbs with or without food, and does not need stomach acid, which matters if you take a PPI or acid-blocker or are over 70 with reduced acid (Recker 1985). Citrate also inhibits kidney-stone formation, a modest bonus.
- Read the elemental calcium figure on the label, not the compound weight. A “1,250 mg calcium carbonate” tablet contains 500 mg of actual calcium.
- Keep each dose at 500 mg or less. Absorption saturates at roughly that amount, so 1,000 mg in one sitting mostly wastes the second half and maximizes the spike. Two small doses with meals beat one large one.
- Separate calcium from iron supplements by about two hours; calcium blocks non-heme iron absorption at the same meal (Hallberg 1991). This matters for menstruating vegan women especially.
- Separate calcium from levothyroxine by four hours; it binds the drug.
- Do not take it at the same time as zinc supplements either, for the same competition reason.
Should calcium be taken with vitamin D?
You need adequate vitamin D for calcium absorption, but you do not need it in the same pill or even at the same time. Vitamin D is fat-soluble and stored; what matters is your blood level, not co-timing. Take vitamin D with a meal that contains some fat, because absorption is meaningfully higher that way (Dawson-Hughes 2015). For vegans with limited sun exposure, a vegan D3 supplement fills the gap; Harvard Health’s guidance is that most people probably need no more than 600 to 800 IU a day and should avoid more than 4,000 IU unless their medical team recommends it (Harvard Health 2026). Take it daily. The clearest harm signal in individual trials comes from large, infrequent doses. In 2,256 women aged 70 and older at high fracture risk, a single 500,000 IU dose each autumn or winter raised the rate of falls by 15 percent and of fractures by 26 percent (Sanders 2010). In 200 adults aged 70 and older who had already fallen, more people fell on 60,000 IU once a month than on 24,000 IU once a month (Bischoff-Ferrari 2016). The Endocrine Society’s evidence review found that vitamin D raised fall risk in trials that spaced doses more than 12 weeks apart (risk ratio 1.08, 1.03 to 1.14) but not in trials dosing every 1 to 12 weeks (0.98, 0.92 to 1.04) (Demay 2024), a figure a later correction clarified refers to falls rather than fractures (correction 2025), and the guideline suggests daily, lower-dose vitamin D for adults 50 and older who need it. Daily doses are not a fall-prevention tool either. In 688 older adults at high fall risk with low-to-normal vitamin D levels, 1,000 to 4,000 IU a day did not prevent falls compared with 200 IU, and the authors flagged safety concerns at 1,000 IU or more, including more serious falls (Appel 2021). The US Preventive Services Task Force’s 2024 draft recommendation, covering trials of 300 to 4,000 IU a day or the daily equivalent in community-dwelling adults 60 and older without known deficiency or osteoporosis, found no benefit for falls or fractures, while stressing that adequate vitamin D and calcium intake still matters (USPSTF 2024 draft). The most recent pooled analysis agrees, with high certainty: in the 2026 BMJ meta-analysis of 69 trials, vitamin D alone had no effect on any fracture (risk ratio 1.00, 0.95 to 1.06), on falling (1.01, 0.98 to 1.04) or on the total number of falls, and its hip-fracture estimate was 1.13 (1.00 to 1.27), a difference of 0.1 percentage points that the authors still graded little to no effect. That review’s subgroup analyses, which compared daily-to-monthly with intermittent or high doses, found no credible subgroup effects, so the bolus-dose signal above comes from those individual trials; the pooled review did not find a credible difference by dosing schedule (Massé 2026). The combination “calcium plus D” pills are fine but carry no special magic.
Should calcium be taken with vitamin K2?
Here the marketing has outrun the evidence, and this is worth saying plainly because K2 is sold specifically on the claim that it “directs calcium into bone and away from arteries” and therefore makes calcium supplements safe. The mechanism is real in principle: K2 activates proteins that inhibit vascular calcification. But the trials do not deliver. A large randomized trial of MK-7 for valve calcification, AVADEC in Denmark, failed to slow aortic valve calcification (Diederichsen 2022). A later analysis of the same trial found no overall effect on coronary calcification, with slower progression only in a high-risk subgroup (Hasific 2023), and a subgroup finding after a null primary result is a hypothesis, not evidence. On bone, one Dutch trial in 244 postmenopausal women found that three years of MK-7 slowed the age-related loss of bone density at the spine and femoral neck, though not at the total hip (Knapen 2013); the trial was funded by the K2 manufacturer NattoPharma, and the broader literature is mixed. There is no trial showing K2 mitigates any cardiovascular risk from calcium supplements. Verdict: K1 from greens is easy and adequate; K2 is a reasonable low-risk option if you want it, from natto or a supplement, but it is not a required companion to calcium and it does not license higher calcium doses.
Which foods actually deliver calcium for a vegan?
Two numbers matter for every food: how much calcium it contains, and what fraction you absorb. Spinach has plenty of calcium and you absorb about 5 percent of it. Kale has less and you absorb about half. So the usable calcium is what counts. The figures below are the measured tracer values from Weaver, Proulx and Heaney’s review, at the serving sizes they used, with cow’s milk included as the benchmark everything else gets compared against (Weaver 1999).
| Food and serving | Calcium (mg) | Absorbed | Usable (mg) |
|---|---|---|---|
| Cow’s milk, 1 cup (the benchmark) | 300 | 32.1% | 96 |
| Fortified plant milk, 1 cup (shaken) | 300 | up to ~32% (as milk) | up to ~96 |
| Juice fortified with calcium citrate malate, 1 cup | 300 | 52.0% | 156 |
| Calcium-set tofu, half cup | 258 | 31.0% | 80 |
| Broccoli, half cup | 35 | 61.3% | 22 |
| Bok choy, half cup | 79 | 53.8% | 43 |
| Kale, half cup | 61 | 49.3% | 30 |
| Chinese mustard greens, half cup | 212 | 40.2% | 85 |
| Chinese cabbage leaves, half cup | 239 | 39.6% | 95 |
| Pinto beans, half cup | 45 | 26.7% | 12 |
| Sweet potato | 44 | 22.2% | 10 |
| White beans, half cup | 113 | 21.8% | 25 |
| Rhubarb | 174 | 8.5% | 10 |
| Spinach, half cup | 115 | 5.1% | 6 |
Three things to notice. First, the low-oxalate greens genuinely beat milk on absorption, broccoli and bok choy by a wide margin, but they carry so little calcium per serving that it takes roughly three servings of kale or four to five of broccoli to match one glass of milk. Absorption and content are different problems, and greens solve only one of them. Second, fortified foods do the heavy lifting: a well-fortified plant milk, shaken, can deliver about as much absorbable calcium as dairy milk, and a juice fortified with calcium citrate malate delivers more than either. Whether drinking milk itself prevents fractures is a separate question, and the cohort evidence on it is in the dairy post. Third, notice which foods are absent. Almonds, tahini, sesame and dried figs have never been run through these tracer studies, so any absorption percentage quoted for them anywhere is an estimate rather than a measurement, and I have left them out rather than manufacture numbers. So the reliable heavy hitters for a vegan are fortified plant milk, calcium-set tofu, fortified juice, and the low-oxalate greens, with nuts, seeds, beans and dried fruit as genuine but minor contributors. You cannot get to 800 mg on almonds and figs.
The authors of that review reach a conclusion worth quoting against interest: the sheer quantity of vegetables needed to reach an adequate intake makes an exclusively plant-food approach impractical for most people unless fortified foods or supplements are included. That is the same conclusion this section arrives at, from researchers with no stake in defending vegan diets, and it is why the fortified-food and small-supplement advice here is not a hedge.
How do I know if my tofu has calcium?
Check the ingredients for the coagulant. “Calcium sulfate” or “calcium chloride” means calcium-set, typically 250 to 860 mg per half cup of firm tofu depending on firmness and how much coagulant was used. “Magnesium chloride” or “nigari” means the tofu was set with magnesium and carries a fraction of that calcium, often under 100 mg. Silken tofu is usually low regardless. This one label check can be the difference between tofu being your best calcium source and being a minor one.
Which plant milks are fortified, and does the settling matter?
It matters a lot. Many organic and “barista” plant milks, and most homemade ones, contain no added calcium at all; check the label for added calcium and a figure around 120 mg per 100 mL or 300 mg per cup. Calcium carbonate has the cleaner record. In a crossover trial in 20 young women, calcium from soy milk fortified with calcium carbonate was absorbed as well as calcium from cow’s milk, while soy milk fortified with tricalcium phosphate was absorbed less well; that trial was supported by White Wave, the maker of Silk, which produced the test soy milk (Zhao 2005). An earlier test of a soy milk fortified with tricalcium phosphate found its calcium absorbed at about 75 percent of the efficiency of cow’s milk calcium (Heaney 2000). The Australian study above is the exception: its proprietary calcium phosphate matched cow’s milk (Tang 2010). All three are small absorption studies, two were funded by the makers of the soy milks tested, and none measured bone or fractures, so carbonate’s edge is suggestive, not proven. Then shake the carton hard before every pour, because the calcium settles to the bottom, and an unshaken glass can deliver a fraction of what the label promises. This is not folklore. When researchers tested 14 calcium-fortified drinks against fat-free milk, 82 to 89 percent of the calcium in the four soy and rice drinks was in a particulate form that could be spun out of the liquid in a centrifuge, and some cartons had a dense sludge at the bottom that took vigorous, prolonged agitation to resuspend; in their words, casual shaking would not have sufficed. One of those four, Silk, was fortified with calcium carbonate, so shaking matters whatever the fortificant (Heaney 2005). The tricalcium phosphate result above came from the same research group, and together these findings say that how well a fortified plant milk performs depends on its fortificant and on getting the calcium out of the carton. Two caveats on the settling study: the drinks were bought in 2003 and 2004, and the work was supported by Creighton University research funds and small grants from Tropicana, General Mills and the National Dairy Council; Tropicana’s calcium citrate malate juices tested best.
How can vegans get 800 to 1,000 mg of calcium a day?
- One cup of fortified soy milk in a smoothie or oatmeal, a half cup of calcium-set tofu at dinner, and one cup of cooked kale or collards: about 800 mg total, roughly 300 mg usable. Add a second cup of fortified milk to reach 1,000. This is the simplest version.
- No greens day: two cups of fortified plant milk across the day plus a serving of calcium-set tofu gets you to about 900 mg.
- No tofu day: one cup fortified milk, one cup fortified orange juice, one cup cooked collards, plus tahini and beans through the day lands around 800 mg.
- If none of those happen, that is the day for a 300 to 500 mg supplement with a meal. The supplement is for the gap, not the baseline.
Does cooking greens change the calcium?
For the low-oxalate greens, cooking concentrates them and you eat more, which is a win; a cup of cooked kale carries far more calcium than a cup of raw. For the high-oxalate greens like spinach and chard, boiling reduces the soluble oxalate substantially, by roughly 30 to 87 percent depending on the vegetable, while steaming removes less, 5 to 53 percent (Chai and Liebman 2005). The authors suggest that may help people prone to kidney stones, but it does not turn spinach into a calcium source. Eat spinach for everything else it offers; count calcium from elsewhere.
Does a little spinach in a meal block the calcium from other foods?
Some, but proportionally, and a small amount is not a concern. Oxalate binds calcium in the gut, so a large spinach salad eaten alongside your tofu will reduce absorption of some of the tofu’s calcium too. A handful of spinach in a mixed dish makes a trivial difference. The practical rule: do not make a high-oxalate green the base of the meal you are counting on for calcium, and do not worry about it otherwise. Rhubarb, Swiss chard, beet greens, parsley and purslane are in the same high-oxalate family as spinach.
What about phytates in beans, grains and nuts?
Phytate binds minerals too, but its effect on calcium is modest, and here is the part that surprises people: in postmenopausal women, higher urinary phytate, a marker of how much phytate someone eats, was associated with less bone loss, not more. In a study of 157 women followed for a year, those with low phytate levels lost about 3.1 percent of their lumbar spine bone density against about 0.4 percent in the high-phytate group (López-González 2013). Three things keep that in proportion. It is observational, and it comes from one research group, which describes it as confirming its own earlier studies and calls for larger intervention trials. Two of its authors hold a pending patent based on aspects of the work, and two are employed by the company Laboratoris Sanifit, one of them also holding stock. And the mechanism the authors propose is physical rather than cellular: phytate adsorbs onto the surface of calcium crystals, the same property that slows kidney-stone crystals from forming, and in bone that is thought to slow the dissolution of existing mineral. The same group has since reported that in a test tube phytate slowed the dissolution of bone mineral about as well as the osteoporosis drug alendronate, and that in about 420 postmenopausal women higher phytate intake tracked with higher spine density, though that was a cross-sectional snapshot with no data on estrogen or lifestyle, and a patent resulted from the work (Sanchis 2021). Evidence that phytate also acts on the bone-resorbing cells themselves comes only from cell-culture studies, and it is mixed: in one, phytate curbed the formation of bone-resorbing cells from human blood cells, but in a mouse cell line it increased resorption by mature cells; its authors also hold a pending patent (Arriero 2012). Soaking, sprouting, fermenting and leavening all reduce phytate anyway. Do not restrict whole plant foods over this. The fuller argument is in plant “toxins” and anti-nutrients.
Is there anything that increases calcium absorption?
- Adequate vitamin D is the big one, and the only one that moves absorption by a large margin.
- Eating calcium with a meal rather than alone improves absorption, mainly via stomach acid and slower transit.
- Prebiotic fibers called inulin-type fructans, found in foods like onions, garlic, leeks, chicory, asparagus and bananas, increased calcium absorption and bone mineralization in a trial of 100 young adolescents who took 8 g a day: absorption ran about 8.5 percentage points higher than placebo at eight weeks and 5.9 points higher after a year, with a modest gain in whole-body bone mineral. The fructan was supplied by its manufacturer, and the lead author consulted for a beverage company (Abrams 2005). That was a measured dose in growing teenagers, so how much ordinary food sources do for adults is less certain. A vegan diet is already rich in these anyway, which is one more reason to keep the onions and garlic in.
- Adequate protein raised calcium absorption in the tracer studies covered earlier, which is part of why the old “protein leaches calcium” story collapsed, though one short metabolic study found no change in absorption.
- Vitamin C and magnesium are often claimed to help; there is no good evidence they meaningfully change calcium absorption. Skip the special pairing.
Is there anything that wastes calcium?
Sodium is the one that matters. Each 2,300 mg of sodium pulls roughly 40 to 60 mg of calcium out through the kidneys, so a high-salt diet raises the calcium you need to replace. Caffeine causes a small loss that is trivial if intake is adequate; it is not a reason to skip coffee. Alcohol in excess and smoking both harm bone by other routes. Tannins in tea have negligible effect on calcium.
Who should not take calcium without talking to a doctor?
- Anyone with a history of kidney stones. Dietary calcium reduces stone risk by binding oxalate in the gut, but supplements taken away from meals can raise it (Curhan 1997).
- Anyone with hypercalcemia, hyperparathyroidism, sarcoidosis, or chronic kidney disease.
- Anyone on thiazide diuretics, digoxin, or levothyroxine.
- Anyone with significant existing coronary calcification, given the unresolved MESA finding.
What is the bottom line on calcium for vegans?
Estimate your intake honestly for a typical week. Fix the gap with fortified plant milk, calcium-set tofu, fortified juice and cooked low-oxalate greens, checking labels and shaking cartons. If you are still short, take 300 to 500 mg of calcium citrate or carbonate with a meal, no more than 500 mg per dose, separated from iron and thyroid medication. Keep vitamin D adequate, take it with fat, and treat K2 as optional. Do not exceed about 1,200 mg total a day from all sources. And remember the outcome that matters: in the one cohort that tested it, vegan women who took both calcium and vitamin D showed no higher hip-fracture risk than nonvegetarian women.
The exercise playbook: loading bone and training the nervous system
The evidence section above established the two facts that drive everything here. First, most hip fractures are fall injuries, so preventing the fall matters as much as strengthening the bone. Second, bone only responds to loads that are heavy, fast, or unusual; walking does little for bone, and cycling and swimming close to nothing, though all three are good for nearly everything else. This section turns that into doses, because “exercise” is not a prescription.
How much jumping does it take to build bone?
Less than you would guess, and the rest between jumps matters more than the count. Bone cells desensitize within a few loading cycles, so twenty jumps done with pauses stimulate more bone than a hundred done in a row; this was worked out in animal studies and it holds in the human trials (Robling 2002). In the cleanest trial, premenopausal women who did just 10 or 20 maximal jumps, with 30 seconds rest between each jump, twice a day for 16 weeks, gained hip bone density while the control group lost it (Tucker 2015). In the hopping trial, 50 hops on one leg every day for six months built hip bone on the hopping side; doing it four days a week or two days a week did not (Bailey and Brooke-Wavell 2010). Daily brief loading beats occasional long loading. A year of similar daily hopping on one leg in men aged 65 to 80 raised femoral neck density in the hopping leg by 0.7 percent while it fell 0.9 percent in the other leg, though it was not risk-free: 3 of the 50 men dropped out with exercise-related knee or back pain, and 5 more needed a few days of rest before continuing (Allison 2013). A high-impact program in postmenopausal women raised femoral neck density without worsening knee osteoarthritis markers (Hartley 2020).
Do jumping jacks count?
Partly, and I will not overstate it. No trial has tested jumping jacks specifically. The bone signal depends on ground reaction force and how fast it arrives; the human studies that produced bone gains used maximal vertical jumps, drop landings, or single-leg hops, which generate high forces on landing. In one study that measured them in children aged 8 to 11, countermovement jumps and the rebound after a drop jump reached about five times body weight, while jumping jacks reached about three and a half (McKay 2005). Adults may land differently, but the gap suggests a jumping jack is a real but weaker stimulus. Doing them continuously for a minute also wastes most of the benefit because the bone cells stop responding after the first handful. If jumping jacks are what you will actually do, do them the way the trials did: 10 to 20 of them, jumping as high as you can rather than as fast as you can, with a full pause between each one, twice a day. That turns a warm-up into a bone stimulus. Better still, once you can, switch to real vertical jumps, or step off a low step and land firmly, a gentler cousin of the drop landings the LIFTMOR trial used.
How do you build up to impact exercise safely?
- Heel drops: rise onto your toes and drop onto your heels, 10 times with a pause between, twice a day. Almost anyone can start here, including people with osteopenia.
- Step-offs: step off a 4 to 6 inch step and land on both feet with a firm, quiet landing. 10 reps, twice a day.
- Two-footed jumps: 10 to 20 maximal jumps, 30 seconds rest between each, twice a day. This is the Tucker protocol.
- Single-leg hops: work up to 50 a day on each leg, with pauses; the Bailey trial used 50 a day on one leg. Build up gradually, since in the older-men version of this several participants needed time off for knee or back pain.
- Drop landings: the LIFTMOR impact exercise was a jumping chin-up, pulling up at the top of a jump and then dropping to land firmly. For people who are already lifting and supervised.
Move up a level only when the current one feels easy and painless. Who should not jump at all without clearance: anyone with an existing vertebral fracture, a DEXA T-score below minus 2.5, a joint replacement, or significant pelvic floor problems. Heel drops and resistance training remain available to nearly everyone.
What resistance training builds bone?
Heavy, compound, progressive, and supervised at first. The LIFTMOR trial used five sets of five repetitions at roughly 80 to 85 percent of one-rep max on the deadlift, back squat and overhead press, plus jumping chin-ups with drop landings, twice a week for 30 minutes, in postmenopausal women with low bone mass, and it built bone at the spine and preserved it at the hip, where the control group lost bone, with one minor low-back strain across more than 2,600 training sessions (Watson 2018). A twelve-month trial in men with low bone mass found both heavy resistance training and a jumping program improved bone density (Hinton 2015). The expert consensus from Exercise and Sports Science Australia lands in the same place: high-intensity resistance plus impact, at least twice a week, is the prescription; light weights for high repetitions do not do it (Beck 2017). If you are new to lifting, hire someone to teach you the deadlift and squat properly; the injury risk in these trials was essentially zero because the lifting was coached.
Does walking build bone density?
Walking alone does not significantly change bone density at the spine or whole body, and only modestly at the hip after six months or more (Ma 2013). Competitive road cyclists, despite being extremely fit, have some of the lowest bone density of any athletes, because cycling loads nothing. Swimming is the same story. Keep doing these for your heart, your mood and your weight; just do not count them as bone training. If walking is what you have, make it count: walk with a weighted vest of 5 to 10 percent of body weight, add stairs and hills, and finish with the heel drops or jumps above.
Can you train your nervous system not to fall?
Yes, and this is the part of the literature that deserves far more attention than it gets. Falling is not mostly a strength failure. It is a failure to react fast enough: to get a foot out sideways when you trip, to catch yourself when you slip. That reaction lives in the nervous system, and the nervous system learns. The evidence that it learns is direct. In one trial, older adults who walked a laboratory walkway with hidden sliding platforms and took 24 unannounced slips in a single session went on to fall less in everyday life over the next year, 15 percent against 34 percent before training, while a comparison group given a single slip did not improve as much (Pai 2014). Two caveats: only about two-thirds of participants completed the year of falls tracking, and a later trial found that a single session of slips on a treadmill did not reduce everyday falls over six months (Wang 2022), so the setup seems to matter. Perturbation-based balance training, in which you are deliberately made to lose balance in a safe setting and practice recovering, roughly halved fall rates in the early trials (Mansfield 2015), and reactive step training likewise cut falls by about half (Okubo 2017). A larger 2025 synthesis of 25 trials was more modest, finding fall rates reduced by 23 percent and injurious falls by 24 percent, rated low certainty (Sharma 2025, preprint). So the honest range is somewhere between a quarter and a half fewer falls from training the reflex itself, which is a large effect for something that takes minutes.
This is what people mean when they talk about neuroplasticity and falls, and stated carefully it is accurate: repeated exposure to controlled loss of balance rewires the reactive stepping response, and the adaptation can persist. Stated carelessly, as if any “brain training” app or crossword protects your hip, it is not. The training has to involve your actual balance being actually disturbed.
What does reactive balance training look like at home?
- Fast lateral steps: stand, then step out wide to one side as fast as you can, push back to center, repeat the other way. Ten each side, daily. This is the Fuhrman drill, and it has a real rationale: sideways falls onto the hip are by far the most dangerous kind, and the recovery is a fast lateral step.
- Reactive stepping with a partner: have someone give you a light, unexpected push at the shoulder or hip while you stand relaxed; your job is to catch yourself with a step. Start gentle, near a wall.
- Stepping targets: place four objects around you and step to whichever one a partner calls out, as quickly as you can. Step-mat and video-game versions of this are part of the step-training evidence that cut falls by about half (Okubo 2017).
- Single-leg stance: stand on one leg while brushing your teeth, then progress to eyes closed, then to standing on a folded towel. Being unable to hold a one-leg stand for 10 seconds in mid-life is associated with markedly higher mortality over the following years, which makes it both a test and a target (Araujo 2022).
- Dual-task practice: walk while counting backwards by sevens, or balance while talking. Falls happen when attention is divided, so practice with attention divided.
Is pickleball good for preventing falls?
Probably, and here is the reasoning plus the caveat. Racquet sports demand exactly the thing that prevents hip fractures: unpredictable, multidirectional, fast stepping under time pressure, thousands of times per session, with your attention on the ball rather than your feet. That is reactive balance training disguised as fun, and it comes with a remarkable observational signal. In the Copenhagen City Heart Study, tennis players lived 9.7 years longer than sedentary people and badminton players 6.2 years, the largest gains of any sport, ahead of cycling, swimming and jogging, and well ahead of gym machines (Schnohr 2018). Another large cohort found racquet sports associated with 47 percent lower all-cause and 56 percent lower cardiovascular mortality (Oja 2017). These are associations, and racquet players are often affluent and social, but the size and consistency are striking, and the mechanism fits. Dance works on the same principle: unpredictable stepping to a cue. A pooled analysis of 29 trials found that “dance-based mind-motor” activities cut the risk of falling by 37 percent and the rate of falls by 31 percent, though that category counted tai chi alongside folk and ballroom dancing, and 16 of the 29 trials were tai chi (Mattle 2020). Tai chi, slower but deliberate weight-shifting, cut falls 19 percent in the Cochrane review and 58 percent versus stretching in a large head-to-head trial (Li 2018).
The caveat is that pickleball is not free of risk. Emergency-department visits for pickleball injuries rose significantly between 2013 and 2022. Adults aged 65 to 80 accounted for 61 percent of them, falls caused most injuries in that group, about a third of all injuries were fractures, and the wrist was the most common fracture site (Yu 2025). So the sport that trains fall reflexes can also cause a fall while you are still learning. The sensible sequence is to build the lateral stepping and single-leg balance first, learn how to fall more safely, wear court shoes rather than running shoes, and then let the sport maintain the reflex you have built. For an older beginner, badminton and table tennis offer the same reactive stepping with less collision risk.
What does a weekly bone-building exercise plan look like?
| Element | Dose | What it does |
|---|---|---|
| Heavy resistance | 2 sessions a week, 30 to 45 minutes; squat, deadlift, press; 5 sets of 5 at a weight you can barely lift 5 times | Builds bone at spine and hip; builds the muscle that catches you |
| Brief impact | Daily: 10 to 20 jumps or 50 hops, with rest between each, ideally twice a day; heel drops if jumping is not appropriate | Direct hip bone stimulus |
| Reactive balance | Daily 5 minutes: lateral steps, one-leg stands, stepping drills; plus 2 sessions a week of a reactive sport, dance, or tai chi | Trains the reflex that prevents the fall |
| Walking, cycling, swimming | As much as you like | Heart, weight, mood; not bone |
The whole bone-specific portion is under 15 minutes a day plus two lifting sessions. That is the honest cost of the intervention that, in the trials, cut falls by a quarter to a half and fall-related fractures by about a quarter.
Are vegans more or less likely to die after breaking a hip?
This is a sharp question, because breaking a hip in old age is not a minor event, and the honest answer starts with an admission: nobody has directly studied it.
How deadly is a broken hip?
Very. One-year mortality after hip-fracture surgery can reach 36 percent, as the 2025 meta-analysis notes, citing earlier work (Ballarin 2025). In a large meta-analysis, the risk of death in the first three months after a hip fracture was about 5.8 times higher than in matched peers for women and about 8 times higher for men, and the excess never fully returns to baseline (Haentjens 2010). Beyond survival, function collapses: only about 40 to 60 percent of survivors regain their pre-fracture mobility, and among people who were independent before the fracture, 20 to 60 percent need help with various daily tasks one and two years later (Dyer 2016). This is exactly why the fall-prevention and bone-building advice above matters so much.
Are most deaths after a hip fracture caused by the fracture itself?
Here is the fact that reframes the whole question. When researchers analyzed 160,000 hip fractures in Sweden, they estimated that 17 to 32 percent, depending on age, of the excess deaths after a hip fracture, meaning the deaths above what the general population of the same age would see, were causally related to the fracture (Kanis 2003). They attributed the rest to the comorbidity the person was already carrying: heart disease, diabetes, frailty, the accumulated damage of a life. People who break hips tend to be sicker to begin with, and much of the excess mortality afterward is that underlying illness playing out. Two cautions from the authors themselves: the split rests on an assumption they say cannot be directly validated, and they think it may understate the fracture’s share. They also stress that the fracture-caused deaths are still substantial, more than 1 percent of all deaths after age 50. That matters here because comorbidity burden is a major driver of who lives and who dies, and it is exactly where a well-planned vegan diet has its clearest, best-established advantage. One disclosure on the source: the Kanis analysis received unrestricted support from the International Osteoporosis Foundation and pharmaceutical and medical-device companies, and one of its authors works for Lilly.
So do vegans survive fractures better or worse?
No study has ever measured post-fracture survival or recovery by diet group. Not EPIC-Oxford, not the Adventist studies, not UK Biobank. So anything here is reasoning from the known predictors of post-fracture death, not direct evidence, and it should be read that way. When you lay those predictors out, vegans land on both sides of the ledger.
On the side that favors vegans, and it is the heavier side, sits comorbidity. Vegans and vegetarians have lower rates of the conditions that kill people after a fracture: less heart disease, less diabetes, less hypertension, lower LDL, lower inflammatory markers such as CRP, and less smoking and drinking. Since pre-existing illness accounts for most of the excess deaths after a fracture, by the Swedish estimate, and lower inflammation aids recovery, this is a real and substantial advantage.
On the other side sit several nutritional predictors, and the important thing about all of them is that they are fixable. The strongest is body weight. In hip fracture there is a well-documented “obesity paradox”: underweight and low-BMI patients die at higher rates, while heavier patients survive better (Modig 2019). Vegans are the leanest diet group: in EPIC-Oxford they had the lowest average BMI of the four diet groups, and across the whole cohort 8 percent of men and 16 percent of women had a BMI under 20 (Spencer 2003), which places part of this population toward the higher-risk end of that curve. Alongside body weight: vegans average lower vitamin D, and if they do not supplement B12 they carry higher homocysteine, both of which are linked to worse outcomes in older adults (Niklewicz 2024). Muscle and protein matter too, since sarcopenia predicts poor recovery, though well-planned protein-adequate vegans who train can match omnivores on muscle.
The honest verdict: for a robust, supplemented, normal-weight vegan, the comorbidity advantage plausibly cancels or outweighs the nutritional disadvantages, landing somewhere around neutral to slightly favorable. For an underweight, unsupplemented, sedentary vegan, the disadvantages stack and the picture likely turns unfavorable. The dividing line is not veganism. It is whether the diet is well planned.
How can vegans recover better after breaking a hip?
Nearly everything on the risk side of that ledger is under your control, which makes this one of the more actionable parts of the whole topic.
- Do not be underweight heading into older age. A BMI comfortably in the low-to-mid 20s addresses the single strongest modifiable nutritional predictor of post-fracture death.
- Keep vitamin D repleted, and especially make sure it is topped up if you are recovering from a fracture or surgery.
- Supplement B12 reliably, which also keeps homocysteine down. Do not rely on diet for this.
- Eat enough protein and enough total energy, particularly around any surgery. Distribute protein across the day, use legumes and soy for lysine and leucine, and consider oral nutritional supplements during recovery, which reduce complications after hip fracture even if they do not clearly reduce death (Avenell 2016, Cochrane). Plant-based versions exist. There is a specific mechanism behind this advice: in a placebo-controlled trial of hip-fracture patients, 20 g of daily protein supplementation raised IGF-1, slowed further bone loss at the hip, and shortened the rehabilitation stay compared with placebo, on top of calcium and vitamin D both groups already received (Schürch 1998). This is the one place in this article where IGF-1 earns its keep, not as an explanation for why vegans fracture more, but as a lever for recovering better once a fracture happens.
- Preserve muscle with resistance training before you ever need it, and resume movement early after a fracture.
- Keep leaning on your strongest asset: the cardiometabolic profile that makes vegans lower-risk on the comorbidities behind most of the excess deaths after a fracture.
One surprising tension is worth naming. The leanness that helps a vegan avoid heart disease and diabetes for decades becomes a modest disadvantage in the acute window after a fracture. That is not a reason to abandon a healthy weight; it is a reason not to drift into being genuinely underweight as you age.
Is the viral “vegans break more bones” post accurate?
Are the claims in the viral post accurate?
The “Recovering Vegan” Facebook post that prompted this piece, checked claim by claim:
| Claim | Verdict |
|---|---|
| About 55,000 people, 17 to 18 years, hospital records | Accurate (54,898; 17.6 years; hospital linkage) |
| Vegans 43% higher total fracture | Accurate (HR 1.43) |
| Hip fracture more than double | Accurate (HR 2.31) |
| Elevated leg and other-main-site fractures | Accurate (2.05; 1.59) |
| About 15 extra hip fractures per 1,000 over 10 years | Accurate (14.9) |
| Vegetarians and pescatarians around 1.25 | Accurate (1.25; 1.26) |
| 2025 meta-analysis, 4 cohorts, ~530,000: +25% / +75%, low-to-very-low certainty | Accurate, exactly |
| AHS-2: female vegans +55%, unless taking calcium and vitamin D | Accurate |
So the facts are right. I am not going to pretend otherwise, and neither should you. What is missing is context: the small absolute risk, the authors’ own very-low certainty rating, the apparent fixability through calcium, vitamin D and body weight, the dated fortified-food environment those vegans lived in, and the fact that the Nurses’ Health Study found that decades of plant-heavy eating did not raise hip fracture risk. One thing the post gets wrong is its editorializing. The author says they would be “extremely skeptical of the Adventist study.” That skepticism is overdone. AHS-2’s low-smoking, low-alcohol homogeneity is a strength for isolating dietary effects, and its supplement-interaction finding is the most actionable result in the literature, though an imprecise and observational one. It is the study that points to how the problem might be fixed.
What do Fuhrman and Greger get right and wrong?
Fuhrman characterized EPIC-Oxford vegans as getting about 45 g protein and about 500 mg calcium. The calcium figure is roughly right (591 mg). The protein figure is understated; actual intake was around 56 to 62 g per day. His targets of 85 g protein and over 1,000 mg calcium are a sensible upgrade over 1990s vegan intakes. His guidance on limiting high-oxalate greens for calcium purposes and on bok choy’s bioavailability is well-supported. His soy estrogen-receptor mechanism is mechanistically valid but clinically inconsistent. Notably, he now concedes that the meat-acid-bone-loss hypothesis he published in Eat to Live did not hold up, which is worth respecting. Greger’s claims mostly check out: lower BMD being largely explained by body size (shown in vegetarians, with vegans not analyzed separately), the 525 mg calcium threshold from Appleby 2007, the persistence of risk in Tong 2020’s higher-calcium subgroup, and the “only 15 percent of fractures attributable to osteoporosis” and “13-fold versus 2-fold” figures from Jarvinen 2015.
Limitations and disclosures
- No RCT of vegan diets with fracture endpoints exists. Causation is inferred throughout, not demonstrated.
- Vegan subgroups are small, producing wide confidence intervals and the very-low GRADE rating. Tong 2020, the strongest single study, rests on 41 vegan hip fractures.
- An earlier version of this article listed low protein as a co-equal driver of the vegan fracture excess. That was an overstatement. Only one cohort has parsed protein out, and it found protein explained roughly 9 percent of the excess against calcium’s 28 percent, with adequate-protein vegans showing no reduction in risk. The correction is noted here rather than quietly edited.
- The residual bone deficit in vegan children persists after body-size adjustment, unlike the adult vegetarian data, where body size explained most of the gap. Extrapolating that adult finding, which comes from vegetarians rather than vegans specifically, to children is not supported.
- No Taiwanese or other Asian incident-fracture cohort using vegetarian status as the exposure appears to exist, despite large vegetarian cohorts in the region publishing on other outcomes. Generalizability beyond white, largely female Western populations is genuinely limited.
- The cohorts are historical. These were 1990s and 2000s vegans with limited access to fortified foods. Generalizability to a modern well-planned vegan diet is uncertain, probably favorable, but that is an inference, not a finding.
- Nobody has studied the specific population that matters most here: well-planned vegan, supplementing, resistance training, followed to fracture endpoints. My assessment that this group’s residual risk is small comes from mediator analyses, not a measured result.
- Funding and affiliation flags, in both directions, though the two are not equivalent and it would be dishonest to present them as such. The stronger conflict is financial, because someone’s revenue depends on the answer: prune research here is funded by the California Prune Board, the vitamin K2 bone trial by the K2 manufacturer NattoPharma, the prebiotic calcium trial with product from its manufacturer and a lead author consulting for a beverage company, the meat-protein calcium trial partly by the beef checkoff, the plant-versus-animal protein review by dairy, egg and beef groups (a review that found against its funders’ interest), the fortified-beverage study partly by small grants from Tropicana, General Mills and the National Dairy Council, the fortified-soy-milk absorption tests by the soy milks’ makers, Sanitarium, which employed one author, and White Wave, the acid-diet calcium trial by Vifor Laboratories, the hip-fracture mortality analysis with unrestricted support from pharmaceutical and medical-device companies including Lilly, which employs one of its authors, the protein and hip-fracture meta-analysis whose two authors have received grants from the dairy company FrieslandCampina for other studies, the calcium-intake and hip-fracture meta-analysis, one of whose authors receives study medications from Wyeth and holds a use patent for calcium as a cancer-preventive agent, the calcium-tracer study whose corresponding author has received National Dairy Council research funding, the monthly vitamin D trial with supplementary funds from Merck Sharp & Dohme, WILD and DSM, which supplied the vitamin D, and a lead author advising DSM, Nestlé and others, the 2015 calcium-and-fracture systematic review, funded by New Zealand’s Health Research Council, one of whose authors has received grants or honoraria from Merck, Amgen, Lilly and Novartis, the vegan resistance-training study, which had no external funding but three of whose authors report honoraria or research support from drug companies including Amgen and Eli Lilly, unrelated to that study by their account, and the phytate bone studies by authors holding patents or pending patents, some of them employed by Laboratoris Sanifit. On the plant-based side, the corresponding author of the identical-twin trial reported receiving funding from Beyond Meat outside that study, and seven of the sixteen authors of its epigenetic follow-up are employees of the company that ran the tests. More broadly, a 2015 BMJ analysis found that 12 of the 22 corporate sponsors of the US National Osteoporosis Foundation and 14 of the 25 sponsors of the International Osteoporosis Foundation were active in nutrition-related businesses, with the food and dairy manufacturers Fonterra, Nestlé, Mengniu Dairy Company and Danone listed among the International Osteoporosis Foundation’s sponsors, and its authors wrote that the foundations’ objectivity may be compromised. Those authors disclose that they have coauthored publications on the efficacy and safety of calcium supplements and vitamin D, and that one of them holds shares in a company that measures bone density (Grey and Bolland 2015). Commercial sponsorship of nutrition research is measurably associated with sponsor-favorable conclusions: in an analysis of studies on soft drinks, juice and milk, articles funded entirely by industry had about seven times the odds of a conclusion favorable to the sponsor, and none of the intervention studies funded entirely by industry reached an unfavorable one (Lesser 2007). That is why it belongs first. The weaker conflict is ideological: some of the studies finding no bone deficit in vegans carry vegetarian-advocacy affiliations, including authors tied to the Scientific Society for Vegetarian Nutrition, and one of the diet-motivation studies cited here was funded by Animal Charity Evaluators. No one’s income turns on those results, so the risk there is motivated interpretation rather than purchased conclusions, and it is a smaller risk. For balance in the other direction, the 2025 meta-analysis that produced the headline vegan figure was funded by a Brazilian public research agency and declared no conflicts of interest, so the most alarming number in this article is not an industry product either. The 2026 BMJ calcium and vitamin D meta-analysis likewise reported no funding and no industry support, though 18 of the 69 trials it pooled reported industry funding and 40 had unclear or potential conflicts. Both kinds of tie are named anyway, because an article that flags only its opponents’ conflicts has told you something about its author rather than about the evidence.
- Other contested points: the all-cause mortality benefit for vegans is not statistically significant in meta-analysis; the EPIC-Oxford haemorrhagic stroke signal did not replicate in a Taiwanese cohort and may share the same unsupplemented-B12 root cause covered above for bones, a cause a cheap B12 supplement addresses; the 2023 MIND-diet RCT was null (Barnes 2023); and Blue Zones data quality is genuinely disputed (our fact-check).
The bottom line
A well-planned vegan diet has RCT- and Mendelian-randomization-grade evidence behind its effects on the causal drivers of the diseases most likely to kill you. That evidence is stronger than what exists for almost any other dietary pattern, and the full accounting across 48 peer-reviewed studies is in what the best evidence actually says about plant-based diets and chronic disease. Bone health is the one place where the observational data flags a genuine, replicated concern, and it points at correctable variables rather than at plant-based eating itself: calcium, vitamin D and body weight on the nutrition side, resistance training on the exercise side, and protein adequacy in support. Notably, an umbrella review of vegan diets rated the evidence for a higher fracture risk, a relative risk of 1.46 (1.03–2.07) from three studies, as low certainty (Selinger 2023). That is close to, but not independent of, the dedicated 2025 meta-analysis, which graded the vegan estimate very low and the vegetarian one low; the umbrella review re-pooled the vegan-only studies from the earlier Iguacel 2019 meta-analysis discussed above. So eat the kale and the bok choy, not just the spinach. Shake the carton. Get enough protein, with legumes daily, and know that plant protein is not the weak link. Take your B12, and your D if the sun is scarce. Pick up something heavy twice a week and practice standing on one leg. Then go be right about everything else.
References
- Tong TYN, et al. BMC Med. 2020;18:353. PubMed
- Appleby P, Roddam A, Allen N, Key T. Eur J Clin Nutr. 2007;61:1400. PubMed
- Crowe FL, et al. Public Health Nutr. 2011;14:340. PubMed
- Webster J, Greenwood DC, Cade JE. BMC Med. 2023;21:278. PubMed
- Webster J, et al. BMC Med. 2022;20:275. PubMed
- Thorpe DL, et al. Am J Clin Nutr. 2021;114:488. PubMed
- Sotos-Prieto M, et al. JAMA Netw Open. 2024;7:e241107. PMC
- Ballarin RS, et al. Nutrition Reviews. 2025;nuaf241. PubMed
- Ho-Pham LT, Nguyen ND, Nguyen TV. Am J Clin Nutr. 2009;90:943. PubMed
- Iguacel I, et al. Nutr Rev. 2019;77:1. PubMed
- Karavasiloglou N, et al. J Nutr. 2020;150:1266. PubMed
- Menzel J, et al. Nutrients. 2021;13:685. PubMed
- Jarvinen TLN, et al. J Intern Med. 2015;277:662. PubMed
- Sherrington C, et al. Cochrane Database Syst Rev. 2019;1:CD012424. PubMed
- Watson SL, et al. J Bone Miner Res. 2018;33:211 (erratum 2019;34:572). PubMed
- De Souza MJ, et al. Am J Clin Nutr. 2022;116:897. PubMed
- Myung SK, et al. Nutrients. 2021;13:368. PubMed
- Weaver CM, Proulx WR, Heaney R. Am J Clin Nutr. 1999;70(3 Suppl):543S. PubMed
- Koch CA, Kjeldsen EW, Frikke-Schmidt R. Eur Heart J. 2023;44:2609. PubMed
- Ference BA, et al. Eur Heart J. 2017;38:2459. PubMed
- Landry MJ, et al. JAMA Netw Open. 2023;6:e2344457 (erratum 2023;6:e2350422); Supplement 2 (eTables 2, 20, 21 and 22) also read. PubMed
- Dwaraka VB, et al. BMC Med. 2024;22:301. PubMed
- Yokoyama Y, et al. JAMA Intern Med. 2014;174:577. PubMed
- Termannsen AD, et al. Obes Rev. 2022;23:e13462. PubMed
- Kahleova H, et al. JAMA Netw Open. 2020;3:e2025454. PubMed
- Qian F, et al. JAMA Intern Med. 2019;179:1335. PubMed
- Tong TYN, et al. BMJ. 2019;366:l4897. PubMed
- Dinu M, et al. Crit Rev Food Sci Nutr. 2017;57:3640. PubMed
- Bouvard V, et al. Lancet Oncol. 2015;16:1599. PubMed
- Reynolds A, et al. Lancet. 2019;393:434. PubMed
- Kjeldsen-Kragh J, et al. Lancet. 1991;338:899. PubMed
- Choi HK, et al. N Engl J Med. 2004;350:1093. PubMed
- Crowe FL, et al. BMJ. 2011;343:d4131. PubMed
- Appleby PN, et al. Am J Clin Nutr. 2011;93:1128. PubMed
- Tonstad S, et al. Diabetes Care. 2009;32:791. PubMed
- Melina V, Craig W, Levin S. J Acad Nutr Diet. 2016;116:1970. PubMed
- Chiu THT, et al. Neurology. 2020;94:e1112. PubMed
- Scarborough P, et al. Nat Food. 2023;4:565. PubMed
- Barnes LL, et al. N Engl J Med. 2023;389:602. PubMed
- Lousuebsakul-Matthews V, et al. Public Health Nutr. 2014;17:2333. PubMed
- Shams-White MM, et al. PLoS One. 2018;13:e0192459. PubMed
- Kerstetter JE, et al. J Clin Endocrinol Metab. 2005;90:26. DOI
- Kerstetter JE, et al. Int Congr Ser. 2007;1297:204. DOI
- Cao JJ, Johnson LK, Hunt JR. J Nutr. 2011;141:391. DOI
- Draper HH, Piché LA, Gibson RS. Nutr Res. 1991;11:273. PubMed search
- Groenendijk I, et al. Comput Struct Biotechnol J. 2019;17:1101. PubMed
- Wakolbinger-Habel R, et al. J Clin Endocrinol Metab. 2022;107:2900. PubMed
- Ho-Pham LT, et al. Osteoporos Int. 2009;20:2087. PubMed
- Ho-Pham LT, et al. Eur J Clin Nutr. 2012;66:75. PubMed
- Desmond MA, et al. Am J Clin Nutr. 2021;113:1565. PubMed
- Hovinen T, et al. EMBO Mol Med. 2021;13:e13492. PubMed
- Itkonen ST, et al. Eur J Nutr. 2025. PubMed
- Zheng Y, et al. Clin Nutr. 2025. PubMed
- Galchenko A, et al. Front Nutr. 2024;11:1390773. PubMed
- Austin G, et al. Front Nutr. 2024;11:1411003. PubMed
- Haentjens P, et al. Ann Intern Med. 2010;152:380. PubMed
- Kanis JA, et al. Bone. 2003;32:468. DOI
- Modig K, et al. J Bone Joint Surg Am. 2019;101:888. PubMed
- Niklewicz A, et al. Nutr Bull. 2024;49:463. PubMed
- Avenell A, et al. Cochrane Database Syst Rev. 2016;11:CD001880. PubMed
- Gilsing AMJ, et al. Eur J Clin Nutr. 2010;64:933. PubMed
- McLean RR, et al. N Engl J Med. 2004;350:2042. PubMed
- van Meurs JBJ, et al. N Engl J Med. 2004;350:2033. PubMed
- Damayanti D, et al. Nutrients. 2018;10:722. PubMed
- Ohlsson C, et al. J Bone Miner Res. 2011;26:865. PubMed
- Schürch MA, et al. Ann Intern Med. 1998;128:801. PubMed
- Soh BXP, et al. PLoS ONE. 2025;20:e0314889. PMC
- Martins MCT, et al. Nutrients. 2017;9:1118. MDPI
- Selinger E, et al. Crit Rev Food Sci Nutr. 2023;63:9926. DOI
- Rosenfeld DL. Appetite. 2018;131:125. PubMed
- Hopwood CJ, et al. PLOS ONE. 2020;15:e0230609. DOI
- Jung HK, et al. Dig Dis Sci. 2013;58:2244. PubMed
- Hunt CD, Johnson LK. Am J Clin Nutr. 2007;86:1054. PubMed
- Bolland MJ, et al. BMJ. 2008;336:262. PubMed
- Bolland MJ, et al. BMJ. 2011;342:d2040. PubMed
- Larsson SC, et al. JAMA. 2017;318:371. PubMed
- Anderson JJB, et al. J Am Heart Assoc. 2016;5:e003815. PubMed
- Chung M, et al. Ann Intern Med. 2016;165:856. PubMed
- Li K, et al. Heart. 2012;98:920. PubMed
- Tang AL, et al. Asia Pac J Clin Nutr. 2010;19:243. PubMed
- Heaney RP, Rafferty K, Bierman J. Nutr Today. 2005;40:39. Article
- Recker RR. N Engl J Med. 1985;313:70. PubMed
- Hallberg L, et al. Am J Clin Nutr. 1991;53:112. PubMed
- Dawson-Hughes B, et al. J Acad Nutr Diet. 2015;115:225. PubMed
- Hasific S, et al. JACC Adv. 2023;2:100643. DOI
- Diederichsen ACP, et al. Circulation. 2022;145:1387. DOI
- Knapen MHJ, et al. Osteoporos Int. 2013;24:2499. PubMed
- López-González AA, et al. Eur J Nutr. 2013;52:717. DOI
- Buclin T, et al. Osteoporos Int. 2001;12:493. PubMed search
- Fenton TR, et al. Nutr J. 2009;8:41. DOI
- Frassetto L, et al. Nutrients. 2018;10:517. DOI
- Mayo Clinic Staff. Hypercalcemia: symptoms and causes. Mayo Clinic, 2024. Mayo Clinic
- Harvard Health Publishing. How much calcium do you really need? 2026. Harvard Health
- Pai YC, et al. J Gerontol A Biol Sci Med Sci. 2014;69:1586. PMC
- Wang Y, et al. Aging Clin Exp Res. 2022;34:1593. DOI
- Mattle M, et al. JAMA Netw Open. 2020;3:e2017688. DOI
- Yu J, et al. Orthop J Sports Med. 2025;13:23259671241305364. DOI
- Dyer SM, et al. BMC Geriatr. 2016;16:158. PubMed
- Lesser LI, et al. PLoS Med. 2007;4:e5. DOI
- Macdonald HM, et al. Am J Clin Nutr. 2008;88:465. DOI
- Jehle S, Hulter HN, Krapf R. J Clin Endocrinol Metab. 2013;98:207. DOI
- Maalouf NM, et al. J Clin Endocrinol Metab. 2011;96:3733. DOI
- Arriero MdM, et al. PLoS One. 2012;7:e43187. DOI
- Sanchis P, et al. Nutrients. 2021;13:2859. DOI
- Messina V, Mangels AR. J Am Diet Assoc. 2001;101:661. PubMed
- Dybvik JS, Svendsen M, Aune D. Eur J Nutr. 2023;62:51. DOI
- Spencer EA, et al. Int J Obes. 2003;27:728. DOI
- McKay H, et al. Br J Sports Med. 2005;39:10. DOI
- Osborne TB, Mendel LB. J Biol Chem. 1914;17:325 (excerpt reprinted in J Am Coll Nutr. 1993;12:484). DOI
- Grey A, Bolland M. BMJ. 2015;351:h3170. DOI
- Bischoff-Ferrari HA, et al. Am J Clin Nutr. 2007;86:1780. PubMed
- Cumming RG, Cummings SR, Nevitt MC, et al. Am J Epidemiol. 1997;145:926. PubMed
- Warensjö E, et al. BMJ. 2011;342:d1473. DOI
- Zhao Y, Martin BR, Weaver CM. J Nutr. 2005;135:2379. DOI
- Heaney RP, et al. Am J Clin Nutr. 2000;71:1166. DOI
- Rogers TS, et al. Bone Rep. 2016;5:117. DOI
- Demmer E, et al. J Nutr Sci. 2016;5:e9. DOI
- Sanders KM, et al. JAMA. 2010;303:1815. PubMed
- Bischoff-Ferrari HA, et al. JAMA Intern Med. 2016;176:175. DOI
- Appel LJ, et al. Ann Intern Med. 2021;174:145. DOI
- Demay MB, et al. J Clin Endocrinol Metab. 2024;109:1907. DOI
- Correction to: Vitamin D for the Prevention of Disease. J Clin Endocrinol Metab. 2025;110:e916. DOI
- US Preventive Services Task Force. Vitamin D, Calcium, or Combined Supplementation for the Primary Prevention of Falls and Fractures in Community-Dwelling Adults. Draft recommendation, 2024. USPSTF
- Huo X, et al. Curr Dev Nutr. 2023;7:100046. DOI
- Bolland MJ, et al. BMJ. 2015;351:h4580. DOI
- Wakolbinger-Habel R, Muschitz C. Nutrients. 2023;15:4709. DOI
- Massé O, et al. BMJ. 2026;393:e088050. DOI
- Chai W, Liebman M. J Agric Food Chem. 2005;53:3027. PubMed
- Abrams SA, et al. Am J Clin Nutr. 2005;82:471. PubMed
- Curhan GC, et al. Ann Intern Med. 1997;126:497. PubMed
- Robling AG, et al. J Bone Miner Res. 2002;17:1545. PubMed
- Tucker LA, et al. Am J Health Promot. 2015;29:158. PubMed
- Bailey CA, Brooke-Wavell K. Bone. 2010;46:1043. PubMed
- Allison SJ, et al. Bone. 2013;53:321. PubMed
- Hartley C, et al. J Bone Miner Res. 2020;35:53. DOI
- Hinton PS, et al. Bone. 2015;79:203. PubMed
- Beck BR, et al. J Sci Med Sport. 2017;20:438. PubMed
- Ma D, et al. Menopause. 2013;20:1216. PubMed
- Mansfield A, et al. Phys Ther. 2015;95:700. PubMed
- Okubo Y, et al. Br J Sports Med. 2017;51:586. PubMed
- Sharma S, et al. medRxiv preprint. 2025. medRxiv
- Araujo CG, et al. Br J Sports Med. 2022;56:975. PubMed
- Schnohr P, et al. Mayo Clin Proc. 2018;93:1775. PubMed
- Oja P, et al. Br J Sports Med. 2017;51:812. PubMed
- Li F, et al. JAMA Intern Med. 2018;178:1301. PubMed
Last updated: October 26, 2026 at 8:15 PM EDT
