Follow-Up to: Direct Evidence that Vegans have Lower Cancer Rates
Wow – the post on Direct Evidence that Vegans have Lower Cancer Rates was one of my most popular ever. That could have partly been due to my asking people to share it with others – thank you to everyone who did that! And now, please unshare it. 🙁
I wrote the post on Monday evening. This morning (Wednesday) a reader, Dave, commented that after adjusting for BMI, the finding for vegans was no longer statistically significant AND that the authors had not adjusted for physical activity. I had not paid proper attention to either of these facts, partially due to not reading the charts in the study correctly.
The authors found significantly more physical activity among those who did not get cancer and also among the vegans, so it is quite odd that they did not adjust for physical activity. But since they did not, and while this study still shows vegans to have lower cancer rates, I no longer consider the study to be evidence that a vegan diet leads to lower cancer rates, which is the whole idea. I have corrected the original post to reflect this.
I sincerely apologize to everyone for doing such a poor job analyzing the study and then asking people to share it.
This post has a follow-up:
Take Three: Direct Evidence that Vegans have Lower Cancer Rates
– Jack
November 28th, 2012 at 6:18 pm
Adjustments are the key. Maybe all of us concerned with other peoples’ health should be promoting physical activity.
November 28th, 2012 at 11:34 pm
Jack, I appreciate the manner you took care of this feedback & corrected your post and statements. This shows you care about the “truth”. Thanks.
November 29th, 2012 at 12:44 am
Dear Jack,
I would like to thank you warmly for your honesty. It means a lot to me that you would issue such a warning statement. Yet another example of how animal advocacy should be done.
Best,
T.
November 29th, 2012 at 8:32 am
I second Reijo’s comment: we trust you, Jack, because we know that honesty guides your work. Everyone makes the occasional mistake; it’s inevitable. I see this error and the way you handled it as just another affirmation of your commitment to accuracy, truth, sincerity. It means a lot to me and my family to know that we can trust you in this way, and you have been an invaluable guide to us for years. (and we are thriving for it!) Thank you, and keep up the good work.
November 29th, 2012 at 9:24 am
Thanks for the nice comments, everyone. I will do a better job checking my work the next time I ask people to share a post, that’s for sure! 🙂
November 29th, 2012 at 11:04 am
Thank you for continuing to give clear and unbiased health and nutritional information. Padding the facts about veganism does not help anyone. I have always felt confident that information posted on your site is as accurate as humanly possible.
Thank you for all of your hard work!
November 29th, 2012 at 7:40 pm
Jack,
What about the China Study? Was that just about having a longer life and a reduction of heart cancer (the leading cause of death)? I read it so long ago I can’t remember.
Sincerely,
Ally
November 29th, 2012 at 7:51 pm
Ally,
To make a long story short, the China Study found that counties in China with lower animal product consumption had lower rates of chronic disease. The authors of the study said that they found no threshold that was so low that this effect did not hold true and they extrapolated that a diet with no animal products would be the most healthy of all. But the China Study is limited in the fact that they were measuring counties, not following individuals. It also did not contain vegans or probably even many vegetarians, but rather people who were close to being vegan but ate small amounts of all animal products.
Fast forward to the Adventist Health Study-2, this study contained thousands of vegans, almost 20,000 lacto-ovo vegetarians, and thousands of pesco-vegetarians. Following these people over time gives us much better insight into what sort of disease rates can be expected for people eating these various diets. But as I pointed out in my original post, they only followed them for 4 years (so far) and that is not very long for a study on cancer. 10 years is decent and 15-20 years is ideal. I don’t know how long AHS2 will continue tracking people. I sure hope they track them for a few more years as this will give us a much better idea of the cancer rates of the various diet groups.
I hope that explains it! Let me know if you have more questions.
November 29th, 2012 at 8:52 pm
The way you handled this issue is why you are my “go-to” for vegan nutrition facts. This share just boosts my confidence in your work. Many thanks and don’t be too hard on yourself!
November 30th, 2012 at 6:19 am
Hi Jack. A lot of people trust you, because you are an absolutely reliable nutritionist.
December 6th, 2012 at 9:38 am
Thanks for all the wonderful work you do!
December 6th, 2012 at 10:22 am
I’m not so sure I would throw in the towel on this study yet just because of bmi/statistical significance and the lack of physical activity being controlled for. On the BMI, if a vegan diet makes it easier to obtain an ideal BMI (22 or less) then it wouldn’t matter to me if it became not significant. Dr. Greger has videos to suggest that perhaps this is the case. Second, usually statistical significance is based on a 95th percentile, i.e. less than a likelihood of 1 out of 20 that the result is due to chance). Some sensitivity analysis may be in order here. What if we used the 90th percentile (as EPA did in risk assessments for indoor smoking)? Would it still be statistically insignificant? If it was still a less than 10 percent probability that the result was due to chance, I wouldn’t throw the study out. I’d just use it as evidence with less statistical power.
On physical activity, this one I’m confused about. Exercise could be a confounding factor for some cancers such as colon or breast, but not all cancers. The fact that this wasn’t controlled for by itself doesn’t invalidate the results for me. It throws a wrinkle into what I would expect for colon cancer risk or breast cancer risk but probably not lung or skin or a number of other cancers. True breast and colon play a large part of the total, but the metric being all cancers, I wouldn’t think dismissing the study’s results are necessary, just caveating it and noting where physical activity has data for particular cancers and where it has none.
To me the most important limitation of the Adventist 2 health study though an improvement over prior studies because of numbers of vegans is the lack of an ability to distinguish betweeen long standing whole food vegans and johnny come lately junk vegans. I wouldn’t expect the aggregation of these two groups to demonstrate the power of the former diet or the liablity of the latter. And of course, when you isolate one group, you do lose statistical power. That said, statistics is a science. And it doesn’t mean that non-statistically significant results aren’t proof of anything. It means they are proof of something with a higher than 5 or 10 percent probability of results to due to chance. Repeat enough of those non-statistically significant results and it says something to me.
Paul
December 6th, 2012 at 11:17 am
Paul,
Thanks for your thoughts on this.
The main reason I think that the BMI adjustment ameliorates the results enough for me to reject this study as proof that a vegan diet lowers your risk of cancer is because they didn’t adjust for physical activity which could be a reasonable predictor of BMI (or effect it significantly) *and* the vegans had a higher physical activity level as a group. If they had adjusted for physical activity then I would consider the BMI amelioration to be less problematic because there is good reason to believe that a vegan diet reduces one’s BMI.
As for confidence intervals and nutrition, I actually don’t start to get too excited that we’re seeing a true effect until findings are well out of the 95% CI range, so a 90% CI is really pushing it in my humble opinion (which is, actually, “humble” given that I’m not a biostatistician). For secondhand smoke and lung cancer, I can see a 90% CI as being arguably important.
The question of whole food vegans versus junk food vegans would be a very complicated one to tease out. Is it even fair to compare only whole-food vegans to non-whole-food meat-eaters? And one person’s view of what is “junk food” can be drastically different than another’s. If we cannot find any differences between your average vegan and your average regular meat-eater, then that doesn’t bode well for suggesting that tweaking the average vegan diet is going to produce considerably different results (with the exception of vitamin B12 – if most vegans in a study are B12-deficienct, then fixing this could possibly effect results).
Again, I don’t think this study is the last word on vegans and cancer and I sure hope one of these studies follows vegans longer than 4 years. And if vegans don’t have lower rates of cancer, it’s not the end of the world; I’m happy as long as vegans don’t have *higher* rates of cancer, and this study is evidence that they don’t. 🙂
December 6th, 2012 at 10:24 am
btw Jack, I do appreciate your honesty and transparency. I just see a little different picture… less harsh judgement for you. I think you are great.
December 6th, 2012 at 10:37 am
One additional thought…. what types of cancers were experienced in each of the cohorts? If most were those amenable to risk reduction/prevention through physical activity (e.g. breast, colon), then I would see the lack of control for physical activity as more of a confounding factor. If they were cancers less amenable to prevention through exercise (e.g. skin cancer, pancreas cancer, non-hodgins lymphoma ) I would view it as less of one.
btw, I may have the cancers and their links to exercise wrong. I remember breast and colon being amenable to reduced risk through increased physical activity, not sure about prostate and lung. Am pretty sure exercise hasn’t been shown to reduce the risk for all though.
Paul
December 6th, 2012 at 4:42 pm
Paul,
If you’d like a copy of the paper, let me know.
December 6th, 2012 at 3:03 pm
Wow what a honesty.
I ‘ll trust your word even more then I did until now because of this post.
December 7th, 2012 at 9:12 am
Jack,
Thanks. Actually, I would. And I am planning to talk to Dr. Fraser for his perspective and get a little into the weeds. Given that this is the best we have right now in terms of studies. I think for me personally it’s worth digging a little deeper. There’s a little connect the dots here. I am open to however it plays. Dr. Fuhrman said back in 2003 in the original Eat to Live that a prudent omni who eats a little fish or chicken but lots of greens and fresh fruit will do better than a junk food vegan. So, I’m comfortable with the idea that as a group we might not test as well. In the end, it would be interesting to test long term whole food vegans who eat many greens and fruits against other cohorts. I doubt that is going to be possible this time around. One day maybe.
Paul
p.s. agree to disagree on junk food vegans. If you are eating half or over processed and refined vegan foods, you are a junk food vegan eating no doubt higher levels of isolated protein…. probably isolated soy protein (an IGF trigger), sodium and sugar with lower fiber, and antioxidants and other nutrients. I know in other studies that the length of time one is vegan is a factor also. It would be useful to me to massage the data even if non-statistically significant for things like length of veganism, specific food patterns (fruits and vegetables vs. grains/starches). Just seeing the 15 percent reduction among all vegans is inspiring. I suspect the news just gets better as one is consuming allium and crucifersous veggies with a vegan diet.
December 7th, 2012 at 9:16 am
one last thought… consider the link between eating a vegan diet an enabling greater physical activity. Less saturated fat, lower body fat … it’s going to enable people to be more physically active. Meat is more difficult to digest, is it not? The body expends considerable energy in digesting meat or animal products than fruits and vegetables. Our systems are not designed to have large amounts of animal products. If a vegan diet helps enable phyisical activity… perhaps that accounts for the lower BMIs among vegans. Just a thought.
Paul
December 10th, 2012 at 4:57 pm
I suspect they didn’t adjust for physical activity in the multivariate analyses because it was not significant in the univariate analyses – in the paper they say they did the latter, to determine which variables should be included in the multivariate models. It would have been helpful if they had listed which variables were tested and found not to be significant, to avoid this speculation. But these are experienced researchers and it seems unlikely they would have just left physical activity out of the multivariate models for no good reason.
I do think this study is important, because even though adjustment for obesity made the OR insignificant, the OR itself changed very little (from 0.84 to 0.86), and remained very close to significance – so the difference between omnis and vegans was only slightly mediated by obesity. In another few years there should be a clearer result emerging.
December 10th, 2012 at 5:15 pm
Kate,
Thanks for pointing that out. It would be good to know for sure.
December 10th, 2012 at 8:55 pm
Jack,
Just had to drop in to say how much I respect you for being transparent. No writers have done more to restore my confidence in a balanced vegan diet than you and Ginny Messina, and that’s largely due to your scientific approach and integrity.
Now I can’t wait to read your book, “Vegan for Life.”
December 11th, 2012 at 2:17 am
I wrote to Gary Fraser 10 days ago, and he replied today: “Although not in the end adjusted for exercise, we did test this, and the adjustment made no difference so that variable was not included finally”.
So it is exactly as Kate said, and I don’t see any reason to unshare this study.
December 11th, 2012 at 7:28 pm
Carlo,
I have now also written Dr. Fraser. As I said to him, since the vegans had a higher rate of physical activity, and cases were less physically active, it seems unlikely to me that if physical activity was adjusted for, the result, with an upper CI of .99, would remain significant.
I hope he will respond.
December 12th, 2012 at 12:52 pm
Jack,
in regards to China Study, you missed the take home message. What Campbell & Junshi (1994) actually reported was that elevated serum cholesterol –within a context of a population with significantly lower cholesterol levels than observed in Western socities– was the chief correlate to all diseases of affluence (p=0.01). This was the bottom line, not the diet itself. Obviously they found that even low amounts of animal products was associated with increase in serum cholesterol levels. However, if Western vegans have high cholesterol levels (>3.6mmol/l), which is expected due to high-fat fare, it’s not expected that they enjoy similar protection from chronic disease as the rural Chinese. The authors conclude:
“”These findings suggest that even small intakes of foods of animal origin are associated with significant increases in plasma cholesterol concentrations, which are associated, in turn, with significant increases in chronic degenerative disease mortality rates”.
http://ajcn.nutrition.org/content/59/5/1153S.full.pdf+html
What Campbell & Junshi are saying in essance is that serum cholesterol levels above what is observed in free-ranging mammalians, healthy neonates and people living in cultures where choronic disease is absent is not healthy. The second message is that, plant-based diet based on whole-foods and minimazation of fat intake help to achieve physiological cholesterol levels.
January 3rd, 2013 at 10:04 am
I have posted a follow-up to this discussion:
Take Three: Direct Evidence that Vegans have Lower Cancer Rates