A study published in BMC Medicine looks at risk of cancer in regular and low meat-eaters, fish-eaters, and vegetarians.
Prof Gunter Kuhnle, Professor of Nutrition and Food Science, University of Reading, said:
“This is a well conducted observational study in a large British cohort, based on the “Biobank”. The authors investigated the impact of meat and fish-consumption on cancer risk and found that – overall – people who followed a vegetarian or pescatarian diet had an overall slightly lower risk of developing cancer than those consuming meat.
“This confirms many previous studies that show that – overall – a vegetarian lifestyle is associated with a lower risk of chronic diseases. There are several possible explanations: first of all, people who are following such diets are usually more health-conscious and tend to be more physically active and less likely to smoke. Secondly, people on vegetarian or pescatarian diets are often also better educated and live in less deprived conditions – both factors that are known to have a strong link with health.
“It is therefore difficult to distinguish between the actual impact of dietary and other factors, but there is good evidence that supports a strong link between meat intake and colorectal cancer – although the association observed in this study was rather weak. The most noticeable result was a much lower risk of prostate cancer in people following a vegetarian diet, and it will be important to understand the underlying reasons better.
“It will be very interesting to see whether the apparent benefits of a vegetarian diet will persist when there is an overall shift to more plant based meat alternatives and vegetarian diets will become more “mainstream”.”
Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:
“As is usually the case with observational studies of possible causes of disease, we’ve got to be careful with the interpretation. Here, there’s a general issue that single observational studies don’t allow conclusions about cause and effect, and there’s a more specific point that the data on the participants’ diets was collected only once at the start of the study, and may have changed later. Those points are rightly mentioned in the press release. But there are also some points about the statistical uncertainty in the estimates of risk that are relevant to how to interpret the results, and also some issues about the absolute size of the risk differences (many of which are really rather small).
“The issue about cause and effect arises because people who eat different diets will also differ, on average, in terms of many other factors too. So if participants in the study who eat two different diets are also observed to have different rates of getting cancers, that could be caused by the differences in their diets, but it could also be caused, in whole or in part, by some of the other factors on which they differ. Those factors are called ‘potential confounders’ in the jargon. Researchers can make some allowance for this problem by making statistical adjustments for other factors, and the researchers on this study did make adjustments for quite a wide range of factors (age, where people live, their height, the deprivation of the area where they live, alcohol and smoking behaviour, and a lot more) and also allowed for sex by essentially doing different analyses for men and women when looking at all cancers together, and when looking at bowel cancer. That helps – but one can never adjust for every possible potential confounder. You can’t adjust for confounders on which you don’t have data. And the adjustment involves a statistical model of how the confounder might affect the risks, and the model may not be perfect. So it’s just not possible to conclude from this one study whether any of the cancer risk differences that they found was actually caused by differences in diet, or by something else. The study does nevertheless add to the evidence we have from many studies about how diet might be related to risk.
“The issue about only recording diet once is, at one level, very obvious – you can’t take into account any changes of diet after people joined the study. But it does relate to some other points about the particular group of people that were studied here. They are all volunteers for the UK Biobank study, and so are likely to be more health-conscious than the average UK individual. Does that mean they might be more likely to have responded to public health messages about diet and changed what they ate in response, during the study’s average 11 years of follow-up? I can’t say, because there’s no data, but it’s a possibility.
“On the same lines, it’s worth thinking about the diets that people recorded. Yes, there were almost half a million people involved, but for two of the cancers that the researchers looked at (breast cancer in postmenopausal women, and prostate cancer which can only happen in men), around half the participants are irrelevant because they have the wrong sex (or, for breast cancer, they weren’t postmenopausal). Also, for a study like this, the key number is really the number of cancers that occurred, not the total number of people studied. For some of the comparisons, the number of people involved and the number of cancers was large, so that the statistical estimates have narrow margins of error. But that’s not the case for comparisons involving people who eat fish but no meat, or who are vegetarians, because there really weren’t many of them in this study compared to the meat eaters. For example, there were fewer than 3,000 men who said they were vegetarians (no meat or fish eaten) in the comparisons for prostate cancer, and only 64 cases of prostate cancer in that group, so the margin of error in comparing their prostate cancer risk with the risk for men who are regular meat eaters is really very wide.
“For all participants (both sexes), about half (almost 250,000) were classed as ‘regular meat-eaters’, eating meat more than five time a week, with another 205,000, so a bit over two in every five participants, classed as ‘low meat-eaters’, eating meat sometimes but no more than five times a week. (Note that taking out just one meat meal a week could move someone from being a ‘regular’ to a ‘low’ meat-eater, though if that happened during the study, it would not have been recorded.) But only 10,969 participants ate fish but no meat – that’s only a little over 2 in every hundred participants – and the number of vegetarians was even smaller, 8,685 in total, or under 2 in every hundred. The study can’t be very precisely informative about cancer risks in fish-eaters of vegetarians, despite having the best part of 500,000 people involved in all. That’s another general issue with observational studies – people’s diets are observed, not imposed by the researchers, and if there aren’t many vegetarians amongst the participants, one just has to live with that.
“Because of the different sizes of the diet groups, the most statistically precise comparisons are those involving ‘regular’ and ‘low’ meat eaters. The press release correctly reports that the risk of getting a cancer of any kind was 2% lower in ‘low’ meat-eaters than in ‘regular’ meat eaters. In a group of 1000 ‘regular’ meat eaters, like the participants in this study, something like 119 would have a cancer diagnosis (of any cancer) during the length of time that people were followed up in this research (11 years on average). In a group of 1000 people with the same characteristics as the ‘regular’ meat eaters, except that they are classed as ‘low’ meat eaters, eating meat five times or fewer in a week, about 116 would have a cancer diagnosis – yes, fewer, but only about three fewer out of the 1000. I’m being careful here to avoid saying that, if the first lot of 1000 changed their diet to eat less meat, there would be one fewer case, because we just don’t know whether that’s true. The study didn’t look at changing diet, and it can’t tell us whether it’s the diet that is the cause of reduced risk anyway. There’s a bit of statistical uncertainty about that estimate – the data are consistent with there being about five fewer cancers in the low meat-eating 1000, or no difference in the number of cancers. So a pretty small difference in risk, and we can’t know whether it’s caused by the difference in diet anyway. Another issue in relation to this estimate is that it is based on a long list of adjustments for potential confounders, but the researchers also produced an estimate in which they adjusted for body mass index (BMI) as well. On that estimate, the risk difference was only 1%, so the estimated smaller number of cancers in the low-meat-eating 1000 is only about one, and the statistical margin of error means that there could even be very slightly more cancers in the low meat eaters than in the regular meat eaters.
“The position is a bit clearer in comparing regular meat eaters with fish-eaters (who eat no meat) or with vegetarians. In 1000 fish-eaters who are otherwise just like the 1000 regular meat-eaters, the best estimate from this research is that about 107 would get a cancer during the follow-up period, compared to abut 119 regular meat-eaters, so roughly 12 fewer out of 1000, but the greater statistical margin of error means that there could be around 20 fewer out of 1000, or just five fewer. The position is quite similar in comparing vegetarians and regular meat-eaters – the best estimate would be about 17 fewer cancer cases in the thousand vegetarians, but again the margin of error is quite wide. So still arguably not very big differences in risk, in absolute numbers. And again, in these cases, we don’t know how changing diet would change the risk, because the study didn’t look at changes in diet, and because we don’t know whether the difference in diet is what causes the difference in risk.
“For bowel (colorectal) cancer, the difference in risk is clearer, and that does fit in with previous research on meat eating and bowel cancer. However, the statistical estimates are not very precise, because bowel cancer is not very common in people of the age of those in this study. Out of 1,000 regular meat eaters in this study with characteristics like those in this study, only about 13 would have a bowel cancer diagnosis during the follow-up period. In a similar group of 1,000 people, except that they are ‘low’ meat eaters, the number of cancers would change from about 13 to about 12, so just one fewer. Because of the low number of bowel cancer diagnoses in this study, we can’t even be sure that there’s any risk difference at all in fish-eaters or vegetarians compared to regular meat eaters. The data are consistent with there being quite a sizeable risk difference in favour of the non-meat diets, but the very wide margins or statistical error mean that the data are also consistent with there being a very slightly higher risk for the non-meat-eaters. There’s just not enough data to be at all sure what’s going on.
“Similar statistical issues arise with the data on postmenopausal breast cancer. The press release says that postmenopausal women on a vegetarian diet would have an 18% lower risk of breast cancer than women on a regular meat diet (more then five meat meals a week). It’s true that that’s the best estimate, but the data are consistent with a risk difference between a tiny 1% and a large 32%. In absolute numbers, out of 1,000 postmenopausal women eating more than five meat meals a week, like those in the study, about 32 would have a breast cancer diagnosis during the follow-up period. The best estimate is that there would be about six fewer breast cancers in a similar group of women who were similar except that they ate a vegetarian diet, but instead of six fewer, the results are consistent with there being less than one fewer death, or ten fewer, or somewhere between those figures. The press release doesn’t mention anything about risk difference for breast cancer in comparing regular meat eaters with either low meat eaters or fish eaters. That’s an appropriate thing to do, because in both cases the results are compatible with their being no difference in risk at all.
“There is clearer evidence of a difference in prostate cancer risk between regular meat eaters and men on fish or vegetarian diets, though again there is a lot of statistical uncertainty. (There’s no evidence at all of a difference in prostate cancer risk between regular and low meat eaters.) In 1,000 regular meat eating men like those in the study, about 43 would be diagnosed with prostate cancer during the follow-up period. In 1,000 similar men who ate a vegetarian diet, the number of cancers would be considerable less, at about 29, so 14 fewer. The wide statistical margin of error means that that number might plausibly be quite a bit different, somewhere between 5 fewer and 20 fewer – but despite the margin of error, those differences in risk are pretty substantial. The position is similar in comparing men who eat fish but no meat with regular meat eaters, though the differences are a bit smaller. But again we don’t know if the difference in diet is the cause of the risk differences.
“The press release mentions that the reason for the lower breast cancer risk in vegetarian women may be because they tended to have a lower BMI than those who ate meat regularly. That’s true, but the study puts it forward only as a suggestion. The statistical analysis that they used to produce the suggestion, called mediation analysis, does depend on assumptions about potential confounders that might or might not be appropriate here, and (as the researchers point out in their paper) the results are still compatible with there being no real effect of BMI. And, to clarify, it’s of course quite possible that the women’s BMI might be affected by their diet – the idea of the mediation analysis is really to investigate how much of the difference in risk might have arisen by diet having an effect through BMI instead of in some other way.”
Dr Duane Mellor, Registered Dietitian and Senior Teaching Fellow, Aston Medical School, Aston University, said:
“Although this study looked at a large number of people (nearly 1/2 million) to find an association between low-meat and meat-free diets and a lower risk of cancer, it is important that this finding is based on a very short survey of food habits which was completed only once, at the start of the 11 year study.
“It also did not quantify the total amount of meat eaten, it only asked how many days it was, or was not consumed. This means the research was not able to assess if participants were eating more or less than the recommended amount of meat according to healthy eating guidelines.
“Perhaps more importantly it did not consider what people in the study ate with their meat. Previous studies have looked at the interaction between meat intake and fibre (including foods such as vegetables, wholegrain and beans, peas and lentils) which suggest higher fibre intake may reduce the risk of particularly colon cancer from eating meat. This could be as the fibre reduces the risk of potentially cancer causing chemicals in the meat coming into contact with the lining of our bowel.
“A problem with this study (as with many looking at the association of one type of food with cancer) is that it fails to look at the whole diet – it can be as important to look at what is eaten as well as what is not!”
‘Risk of cancer in regular and low meat-eaters, fish-eaters, and vegetarians: a prospective analysis of UK Biobank participants’ by Cody Z. Watling et al. was published in BMC Medicine at 01:00 UK time on Thursday 24 February 2022.
Prof Gunter Kuhnle: “No CoI.”
Prof Kevin McConway: “I am a Trustee of the SMC and a member of its Advisory Committee. My quote above is in my capacity as an independent professional statistician.”
Dr Duane Mellor: “No conflicts of interest.”