Read on Twitter
I received a large number of questions on the vegetarian/vegan diets and fracture paper below. I am not able to respond to all of them individually.
So here’s a thread addressing the common questions/misconceptions, with input also from @anika_knuppel and @KerenPapier 1/ https://twitter.com/tammy_tong/status/1330819443828678656
So here’s a thread addressing the common questions/misconceptions, with input also from @anika_knuppel and @KerenPapier 1/ https://twitter.com/tammy_tong/status/1330819443828678656
I will cover the following common questions/misconceptions:
1) Sample size
2) Are they all junk food vegans?
3) Crude vs adjusted rates
4) Role of confounders/alternative explanations
5) Overall remarks 2/
1) Sample size
2) Are they all junk food vegans?
3) Crude vs adjusted rates
4) Role of confounders/alternative explanations
5) Overall remarks 2/
Because this will likely be a long thread, and my colleagues and I all have full time academic jobs after all, I will be adding to this thread gradually over the coming days/couple of weeks. 3/
Before I go through the list, an important clarification is that our findings could not be used to make the claim that veganism causes fractures. Rather, we are reporting our findings of a higher risk of fractures in vegans in EPIC-Oxford. 4/
Many have written to me about their personal experiences of going vegan and not having fractures. An important point to note when reading any epidemiological study is that our research compares risks across population groups, and cannot inform individual risk. 5/
Indeed, of 1982 vegans in our cohort, 92.6% did not have a fracture. But when we compared differences in proportions across the different diet groups, we found a higher risk in the vegans than the meat-eaters overall. 6/
This could be due to dietary or non-dietary factors associated with their diets, I will elaborate more on this point later. 7/
Similarly, the findings also could not be used to infer the risks of fractures associated with eating meat or dairy – we did not look at whether consuming these individual foods were linked to fracture risks in the paper. 8/
1) Sample size. As much as we would love to see a cohort with 100,000 vegans, such data do not exit. If we recruit people now, we won’t accrue enough cases for another 10-20 yrs, so we need to make use of what’s available. 9/
This is important for designing future studies, but also for understanding risk factors now so we could mitigate risks in the future. EPIC-Ox is one of the two largest studies of veggie/vegans in the world, the other being Adventist Health Study 2 in the US. 10/
Statistical power depends on both the sample size and effect size, that is, how big the difference in outcome is between comparison groups. If the effect size is sufficiently large, a smaller sample size could suffice. 11/
The effect size here was substantial (esp for nutri epi where 10-20% diff are common) and the confidence intervals (CIs) showed clear significant associations. For non-statisticians, this is essentially an assessment of whether the findings might have occurred by chance 12/
The practical interpretation would be to judge statistical significance on whether the CIs cover the reference value of 1, and also how tight the CIs are around the point estimates. 13/
In this case, because the CIs were well away from 1 in the main findings, esp for total and hip fractures. It is highly unlikely that chance explained our findings. 14/
Admittedly, the sample size does drop in subgroup analyses, which limits our interpretation of the reasons for high fracture risks. Again, the judgement is based on CIs, which became much wider here when compared to the main findings 15/
2) Are they all junk food vegans? I don’t believe so, based on what we know about their dietary habits and other health outcomes. I will elaborate on this at a late date. 16/
Multiple papers have been published on nutrient quality across the diet groups in EPIC-Ox. Including these two following papers, the first one based on baseline data:
https://pubmed.ncbi.nlm.nih.gov/12740075/
17/
https://pubmed.ncbi.nlm.nih.gov/12740075/
17/
And the second one using dietary data collected around 2010:
https://pubmed.ncbi.nlm.nih.gov/27101764/
18/
https://pubmed.ncbi.nlm.nih.gov/27101764/
18/
Intake of nutrients mostly showed a gradient effect across diet groups. In both analyses, vegans had much lower intakes of saturated fat and higher intakes of fibre (more on this below); this is not consistent with a junk food diet. 19/
Although fibre intake in EPIC-Ox may appear low on average, this is largely because fibre was estimated as NSP fibre in the study for historic reasons, which gives lower estimates than AOAC fibre used for recommendations in most countries today. 20/
There is no agreed conversion factor since it depends on the source of fibre, but a very rough guide is to multiply NSP by 1.33. So e.g. in women 26g of NSP at baseline=34.6g AOAC fibre, and estimates were slightly higher in 2010, so not as low as it first appeared. 21/
But vegans also had lower intakes of protein (with higher proportion inadequate), calcium, vit D (though dietary intake is poor indicator of vit D status) and vit B12. These may be important for bone health. 22/
Want to know what they ate in terms of foods? Here’s a paper from @KerenPapier which looked at exactly that:
https://pubmed.ncbi.nlm.nih.gov/30979052/
Generally vegans had higher plant foods/lower refined grains.
23/
https://pubmed.ncbi.nlm.nih.gov/30979052/
Generally vegans had higher plant foods/lower refined grains.
23/
For other health characteristics, vegans in EPIC-Ox also showed lower blood pressure and cholesterol. We also published combined analyses of veggie/vegans which showed lower risks of heart disease/diabetes/diverticular disease. 24/
So all in all, the evidence is not consistent with the theory that EPIC-Ox vegans are all junk food vegans. 25/
3) There is some confusion on why the crude rates of fractures were higher in meat-eaters. I have commented on this before:
https://twitter.com/tammy_tong/status/1331979261859782658?s=20
26/
https://twitter.com/tammy_tong/status/1331979261859782658?s=20
26/
Some have misinterpreted that the statistical adjustment made the vegans look artificially worse. This is not true. For outcomes above where they had lower risks for example, the risks remained lower after adjustment. 27/
Aging is also one of very few unavoidable risk factors in life, and there are older vegans and younger meat eaters in the cohort, despite the difference in group average. 28/
4) Confounding/alternative explanations
Apart from age, we also adjusted for a range of other factors here, because we wanted to understand the effect of diet on fracture risk independently of these factors. The full list of adjustment factors can be found in the paper. 29/
Apart from age, we also adjusted for a range of other factors here, because we wanted to understand the effect of diet on fracture risk independently of these factors. The full list of adjustment factors can be found in the paper. 29/
For example, although vegans had higher physical activity levels than meat eaters, because we adjusted for physical activity, it is unlikely that this is why they had higher risks of fractures. 30/
Note: We adjusted for these factors of interest, including BMI and calcium, by fitting them as covariates in a single regression model, which is different from stratification. 31/
Because the associations attenuated when we further adjusted for BMI, calcium and protein, differences in these factors were likely partly the reason why vegans had a higher risk of fractures in this study. 32/
But the significant associations that remained after adjustment suggest other factors, for example, vit D/B12 status, or muscle strength, may also be important. These data were not available for everyone in our study, but should be further investigated. 33/
5) Overall remarks
So, what should we conclude? Given the known associations between BMI, calcium and the other factors above with bone health, paying attention to all these factors will likely help to improve bone heath and reduce fracture risks in most people. 34/
So, what should we conclude? Given the known associations between BMI, calcium and the other factors above with bone health, paying attention to all these factors will likely help to improve bone heath and reduce fracture risks in most people. 34/
When interpreting risk estimates, the absolute differences might be more intuitive. Our estimates suggest the difference was 20 more cases of total fractures and 15 more cases of hip fractures per 1000 ppl over 10 yrs, in vegans than meat eaters (full results in table 3). 35/
EPIC-Ox is a single study, and often thought of as an old cohort. This is true in terms of when people were recruited, but most people in the study are still alive and well today (we used data collected until 2016), so I don’t believe their data could be deemed irrelevant. 36/
But beyond this, we definitely need further studies, both from other long-running veggie/vegan cohorts, such as Adventist Health Study 2, but also to start collecting data on a newer generation of vegans, so we could understand their health too. 37/
Finally, this research needs to be considered in context, and viewed as one small piece of the puzzle in understanding vegetarian and vegan health, as illustrated by our findings of lower risk of some diseases, but higher risk of other outcomes in these populations. 38/
@threadreaderapp unroll please
