Cut red meat? We already have
Eating one less portion of red meat a day can cut your risk of dying by 7-19 per cent, according to a widely-publicised paper (1) published this week.
For many of us this is an impossible recommendation, as we already average less than one portion of red meat a day. So, indeed, did the participants in the study: by 2006, the last time that their diets were recorded, the mean daily intake of red meat in men was 0.63 servings a day, and for women 0.55 servings.
The data came from Harvard School of Public Health, which for at least a decade has been pouring out data on nutrition and health from two long-running cohort studies: the Health Professionals Follow-up Study (nearly 40,000 dentists, vets, optometrists and osteopaths) and the Nurses Health Study (84,000 female nurses). The HPFS started in 1986, the NHS in 1980.
Over the lengthy run of these two studies, the diets consumed by the participants have changed a lot, especially among women. Their average intake of red meat halved between 1980 and 2006, the authors report. The decline among men is less marked – from 0.73 to 0.63 servings a day between 1986 and 2006.
To provide a measure of meat intake over the whole period, the authors averaged consumption from baseline to death. This means that the diets against which the risks are being assessed are not the same as those we are eating today. We have anticipated the conclusions by pre-empting the advice.
So any additional gains from cutting meat consumption are more limited than the paper in Archives of Internal Medicine implies. For example, the lowest quintile of red meat consumption in women – which is the one with the lowest risk of death – is 0.51 portions a day (95 per cent CI 0.37 to 0.61). That is now almost exactly the same as the average woman in the study consumes.
For men, today’s average consumption of 0.63 portions a day is very close to the second quintile (0.61, CI 0.53 to 0.70). At this level of consumption, there is no statistically significant increase in risk of death compared with the lowest quintile. So the average person in the study, it seems, is already at or very close to the minimum level of risk.
But the puzzles with this study do not end there. If the answers given to the dietary questionnaires were accurate, we are expected to believe that women eat more red meat than men, which common experience suggests is unlikely. At every quintile level, the nurses’ taste for red meat exceeded that of the dentists, optometrists, vets and osteopaths (though their average consumption is reported as being less). This could be a consequence of the nurses’ study starting six years earlier, when everybody ate more meat; or it could be caused by uncertainties in accurately ascertaining diet.
There are also very big differences between meat-lovers and meat-avoiders that may confound the findings. For example, the men who ate most meat were also three times as likely to smoke. They also drank considerably more, and were much less likely to take multivitamins, which probably do little good but serve as a marker for the health-conscious.
Among the nurses, the same was true, though to a less marked degree. The final results correct for smoking, drinking, body mass index and family history of various diseases, among other things, but there may be residual confounding that is not corrected for.
Finally, the study does not tell us how many deaths were premature. The men were aged 40-75 at baseline, so by 2008 when the deaths were counted would have been between 62 and 97; the nurses were aged 30-55 at baseline so would have been 62-87 by 2008. By all normal expectations many would have been expected to die. In fact, 8,926 men died (out of 37,698) and 15,000 women (out of 83,644). But how many died before their time is unknown, or unreported.
In both men and women, there were more deaths in the lowest quintile of red meat consumption than in any other but the highest – so the raw data suggest that red meat consumption, except at the highest levels, protects against dying. When adjusted for age, however, the trend is reversed, which tells us that the lowest quintile must have been, on average, older. It is possible they ate less meat because they ate less of everything.
In nutrition science, there are no certainties. This is a good study with a methodology better than most, but it certainly does not prove that we could all benefit from the advice it offers. Only a minority of people eat enough red meat to be able to cut out a portion a day without becoming vegetarians. The rest of us have little to gain by eating less.
- Red Meat Consumption and Mortality, by An Pan et al Arch Intern Med, online March 12 2012.
Anonymous (not verified) wrote,
Thu, 15/03/2012 - 14:28
Why this constant insistence on non-units? "Portions" of red meat, "units" of alcohol - if you mean 10cl or X hundred grammes, why not say so? Rescaling just promotes obscurity.
PJH (not verified) wrote,
Thu, 15/03/2012 - 15:07
"Eating one less portion of red meat a day can cut your risk of dying by 7-19 per cent, according to a widely-publicised paper (1) published this week."
So, instead of the risk of dying generally accepted as being 100%, they're claiming you can reduce it to 79-93%?
Come on Nigel, a little more care please... :)
Colin Fischbacher (not verified) wrote,
Fri, 16/03/2012 - 10:54
As an admirer and regular follower of Straight Statistics I was slightly puzzled by this comment on meat consumption, which concludes that there is little to gain from reducing meat consumption.
Are the results really only relevant to a minority in the UK? A quick look at the UK National Diet and Nutrition Surveys from 2008-10 suggests that British adults aged 19-64 eat an average of 95 grams daily of beef, lamb and pork, liver, burgers and meat pies.(1) This roughly equates to 1.1 portions daily in terms of the measure used in the US study, which found the lowest risk in those eating 0.5 portions of red meat daily. It seems that far more than a minority of Britons could benefit from reducing their meat intake. And of course it is not necessary to reduce intake by a whole portion to benefit.
The point about residual confounding seems very fair, but the criticism of the lack of information on premature deaths was much less convincing. If those who eat more meat die at a faster rate (as they did in this study) then their deaths will be more premature. The feasibility and value of an additional measure of premature deaths is not clear, at least to me.
The statement that in both men and women, there were more deaths in the lowest quintile of red meat consumption than in any other but the highest is not borne out by table 2 in the study. In any case a comparison that doesn’t take account of age differences seems misleading.
I agree that in nutrition science there are no certainties. Indeed, you could extend that statement to all science. It is never a good idea to draw conclusions from a single study, but this study should be set in the context of a body of research pointing to the health risks of high levels of meat consumption.
My own conclusion is that I will continue to enjoy a steak, but would prefer not to eat one every day.
(1) http://www.dh.gov.uk/prod_consum_dh/groups/dh_digitalassets/documents/di...
Nigel Hawkes (not verified) wrote,
Fri, 16/03/2012 - 12:47
I used "risk of dying" because that is more or less what the authors used. They referred to an increased risk of mortality. What they, and I, meant was the risk of dying within a certain period, in this case the period of the study. In the long run, we're all dead.
Jeruza (not verified) wrote,
Sun, 18/03/2012 - 00:22
As Nigel wrote, the lowest and second lowest quintiles are simply not that low for today. I'd like to see the comparison to eating 0.51 portion to zero portions, but maybe they didn't have enough people in that category.
Nigel Hawkes (not verified) wrote,
Tue, 20/03/2012 - 17:03
Colin Fischbacher makes some good points. In response, I’d say:
He quotes diets from the National Diet and Nutrition Survey, but the European Prospective Investigation into Cancer and Nutrition, which used good methods to ascertain diet, found much lower figures. Mean red meat consumption among men in the UK was found to be to be 39.8 grams a day, and among women 24.9 g. That translates into 0.46 of a “serving” (as defined in the recent paper) for men, and 0.29 of a serving for women, lower even than the average for US men and women found by the Harvard team. So I stick by my contention that there is not much room for a reduction of anything like a serving (85g) a day of red meat for the average UK man or woman.
I was indeed wrong to say that all-cause mortality was higher in the lowest quintile of meat consumption than in any other but the highest. That applies to cardiovascular deaths (Table 3) but not to all-cause mortality. I look at uncorrected data because it can often tell you something about the process of adjustment for confounding factors, but the appropriate measure to use is, I agree, one that takes account of age differences.
My basic point is that the quintiles of meat consumption varied in so many ways – smoking behaviour, cholesterol levels, family history of heart attack, vitamin use, alcohol consumption – that to try and tease a weak signal out of all this noise is a very tricky task indeed. Small odds ratios are essentially meaningless, given the uncertainties. But if everyone accepted that, epidemiology would be dead in the water.
Conrad Morton (not verified) wrote,
Fri, 23/03/2012 - 17:38
There are a number of reasons not to accept the comment that this was a "good study". First the method of measuring habitual dietary intake is known to be horribly inaccurate. The Harvard School argue that these inaccuracies even out when they take a very large number of subjects. But this is only true of the mean of a large number if the errors are random. The errors in self-reported dietary intake data are not random. Far more subjects under-report their food intake than over-report it. And the under-reporting is selective. Foods and drinks that the subjects have heard they should avoid for health reasons tend to be under-reported, especially if they are perceived to be "fattenning" and the subjects are overweight.
So the data are highly quetionable. But secondly, the adjutment for known confounders (such a smoking) is subject to substantial errors as well. Recent studies have shown that different (but equally respectable) ways in which these corrections are applied for "attributable risk" from confounders can lead to completely opposite conclusions. This is leaving aside the lack of correction for known (e. g. physical activity) and unknown confounders.
Third, the study was a prospective observational study, this means that a group of people were followed for a number of years and various measurements of their lifestyles were made and the bodies were counted. There was no conscious attempt to change anyone's behaviour and no control group. So the study cannot be considered adequate evidence of causality.
As usual, the authors and the press present certainty when the study was merely an exercise in opinion, as indeed all observational epidemiology is. The result is too heavily influenced by the opinions and assumptions made by the researchers when setting up the study and analysing the results.
Anoop (not verified) wrote,
Sun, 01/04/2012 - 16:16
Hi Nigel,
What did you think about the data that hows that the cholestrol percentage decreased as the meat intake increased n both the studies?
Thanks!
Anoop (not verified) wrote,
Mon, 02/04/2012 - 05:03
Hi Nick,
Do you know what was the HR for fruits and vegetable consumption. I am guessing it must be pretty high since nobody questions about the validity of those studies.
Dora (not verified) wrote,
Wed, 04/04/2012 - 13:22
In response to Anoop, I found it strange that those who ate more meat had a lower proportion with high cholesterol. Offhand I can't explain how that came about. Maybe somebody else can