Observational Study on "Cancer incidence in British vegetarians"

Key et al. reported in the British Journal of Cancer a large observational study that compared cancer incidence between groups based on diet. This article claims to be free access to the public. The corporate media love to report on research like this because they find it simple to tease good headlines out of them. Unfortunately, the closer you look at observational studies, the less sure of anything you end up being.

The study split 61566 Britons into three groups, 'meat-eating' (N = 32403), 'fish-eating vegetarians' (N = 8562), and 'vegetarian' (N=20601). Unfortunately, the study did not split up lacto-ovo vegetarians from vegans, as that may have been interesting due to the removal of another food group in the form of dairy.

The 'meat eating' group could be better described as the 'standard British diet' group. There are substantial differences in the approach to food between vegetarians and the stereotypical general population. This is particularly evident in the degree of industrial, processed food consumed although when I go to organic food markets I still see plenty of crap in the middle aisles.

There were, of course, differences between the three groups that were not dietary:

The mean age at recruitment was lower in the fish eaters and vegetarians than in the meat eaters. Smoking rates were low overall, with only 14.4% of meat eaters, 11.2% of fish eaters and 11.4% of vegetarians reporting that they were smokers at the time of recruitment. The median BMI was 1.5 kg m^-2 lower in vegetarians than in meat eaters, and the median alcohol consumption was 1.0 g per day lower in vegetarians than in meat eaters. Fish eaters had similar mean BMI to the vegetarians and had similar alcohol consumption to the meat eaters. The proportions of men and women who reported a relatively high level of physical activity were higher among fish eaters and vegetarians than among meat eaters. The proportion of women who were nulliparous at recruitment was higher among fish eaters and vegetarians than among meat eaters, and the proportion of women who had ever used oral contraceptives was lower among fish eaters and vegetarians than among meat eaters.

The authors claim to have corrected for these factors in their results for cancer risk factors, and that an uncorrected analysis had similar results to the corrected one (i.e. that the differences did not impact the results).

Table 1: Relative risks of cancer for various forms, RR = 1.0 for the 'Meat Eaters' group.

Selected Major Cancer Types	Fish Eaters’ Relative Risk to Meat Eaters	Vegetarians’ Relative Risk to Meat Eaters
Upper GI tract	0.44 (0.16–1.25)	0.81 (0.45–1.46)
Stomach	0.29 (0.07–1.20)	0.36 (0.16–0.78)
Colorectum	0.77 (0.53–1.13)	1.12 (0.87–1.44)
Lung	0.59 (0.29–1.23)	1.11 (0.75–1.65)
Melanoma	0.90 (0.55–1.47)	0.89 (0.61–1.29)
Female breast	1.05 (0.86–1.28)	0.91 (0.77–1.08)
Prostate	0.57 (0.33–0.99)	0.87 (0.64–1.18)
Lymphatic	0.85 (0.56–1.29)	0.55 (0.39–0.78)
Cervix	2.05 (0.91–4.63)	2.08 (1.05–4.12)
Overall	0.82 (0.73–0.93)	0.88 (0.81–0.96)

The results were statistically significant only for stomach, lymphatic, and cervical cancers, and the overall results were also significant. The uptick in cervical cancer is interesting, but not particularly important given how easily it can be avoided through vaccination. What's interesting is that the fish-eating vegetarians appear to have lower risk factors than the pure vegetarians.

It's a little hard to make the claim that meat per-say is the causative factor when you take vegetarians and feed them fish and their cancer incidence drops. In the discussion, the authors' implication focuses on nitrates and other work has shown that heterocyclic amines produced by charing/burning food (not just meat) can be hazardous. Alternatively it could be something like deep frying food in vegetable oil, since its polyunsaturated content is so easily oxidized in such an environment. Could vegetarians be buying more organic food and ingesting lower levels of pesticides? Etc. To paraphrase Simon Pegg, identifying the causative factor is like trying to hit the bull-eye on a dart-board that is mounted on a F1 car, while blind-folded, and riding a horse.

This is the drawback of observational studies: you can't make any definitive conclusions from it. At best, it can act as a guide, but it's sort of like hiking in the mountains: do you follow that old blaze when the trail forks or do you try and orient yourself? The drops in cancer rates (-18 % for fish-eaters, -12 % for vegetarians) are no smoking gun. The drop in stomach cancer is impressive. By way of comparison, the rate of lung cancer in non-smokers is roughly 3 % (i.e. RR = 0.03) compared to heavy smokers.