Is this news the kind that will motivate a change in your diet?
No.
He found that people who consumed about one serving of red meat (beef, pork or lamb) per day had a 13 percent increased risk of mortality, compared with those who were eating very little meat.
I call bullshit.
Everybody has a 100 percent risk of mortality.
13% increased. As in 13% faster than those not eating it.
Well, the way it's phrased -- "13 percent increased risk of mortality" -- makes it sound like people who eat red meat daily have a 13 percent greater chance of dying.
You can put this down to the journalist's inability to understand statistics.
The study actually reports a Hazard Ratio of 1.13. That doesn't mean that people are 13% more likely to die if they eat red meat, which as
scotpens points out, is a completely nonsensical statement if taken in isolation and without qualification.
The Hazard Ratio of 1.13 means that if you look at the number of people dead vs not dead over a given unit of time during the study, and then looked at whether they were daily unprocessed red meat-eaters or not, there were 113 dead meat eaters for every 100 dead non meat-eaters. In other word the chances of being dead and a daily meat-eater were 13% higher than being dead and not a daily meat-eater. That's VERY different to saying you're 13% more likely to die if you eat unprocessed red meat daily! This eliding of meaning is ridiculously common in many journalists' reporting of the statistics of medical/scientific studies.
I'm also not quite sure why the authors chose to report Hazard Ratios rather than Relative Risks. Relative Risk strikes me as a more useful gauge in this case, as it represents a cumulative risk effect rather than the cross-sectional per unit time approach of a hazard rate (and therefore, hazard ratio). But there are limits to my medical statistical knowhow, so I can't really comment further on the reasons for their choice here.
it's a 13% increased chance of heart disease.
so, if you're chances of heart disease are 1%, you know have a 13% chance of getting heart disease.
that's still an 87% chance of NOT getting it.
Not to be a mathematical pedant (hey, my degree has to be good for something other than developing websites for the marketing arms of dog food and pharmaceutical companies
), but a 13% increase over a 1% chance of dying from heart disease would be 1.13%. A 13% chance would represent a 1200% increase.
This is where language matters more than mathematics.
In
calhoun's heart disease example I would suggest the most precise language would be to say "that the percentage chance of dying from heart disease increased from 1% to 13%. This is a 13-fold increase in the risk of dying from disease, and represents a 1200% increase in the number of people in the population who die from heart disease".
Of course, as already point out upthread, the actual statistical value reported in the study is different again in meaning to all the above!
Interesting you mention pharma sales strategies; they often do indeed use the maximum leeway on the effect imprecise language can have on statistics to quite legally quote the 1200% type of figure when discussing improved efficacy. It, of course, sounds a lot better than a 13% improvement. Not all do this, and regulators occasionally take action if the statistical manipulation is particularly egregious, but language is always going to be more imprecise than mathematics, and marketing departments are superb at playing on this.
The study in question is observational only, with no control, so it's unscientific anyway.
Then there's the issue of the part that red meat plays in the person's overall lifestyle- i.e. people who eat 12-ounce steaks every night are likely to drink and smoke too.
It's a prospective cohort study with good follow-up which is perfectly scientific; they just rank slightly lower than RCTs as a level of evidence. In fact, in many areas of scientific study it's impossible to perform RCTs so either prospective or retrospective cohort studies are the ONLY way to study the topic. Given what this study is attempting to explore, I'd probably suggest it ranks as about 1b evidence rather than an RCT's 1a on the usually accepted scale from 1 to 5.
Also, the authors carried out a perfectly acceptable multivariate correction for other (known) lifestyle risk factors, including, yes, alcohol & smoking.
What we need to do is keep things reasonable.
The art is deciding what is reasonable and what is not. How strong does a correlation have to be before it's worrying; does one always need a definitive physiological pathway before intervening? You quoted smoking as an explicit example of causation rather than correlation; the causative pathways of harm via smoking came a LOT (decades) later than the very strong correlations. Most of the public health campaigns around smoking
preceded the (relatively) definitive identification of causative pathways.
You then went onto saying it should still be legal because people can choose whether to smoke. But you also pointed out that we have legislation keeping a large number of other toxins out of public circulation. Playing Devil's Advocate as a thought experiment, would it be OK to get rid of that legislation, if it were replaced, say, by a warning on packets that "this chicken may contain traces of mercury, which is known to be lethal to humans in the long-term"? People could choose their chicken. If mercury-contaminated one was cheaper, some would undoubtedly buy it over than the clean version and risk the long-term effect, so where's the real difference to tobacco? After all, some people would die of being run over by a bus before the mercury got to them.
Where does one draw the line regarding risk? One man's "reasonable" level of risk is another man's serious public health hazard. Scientific evidence only talks in statistical terms, not moral values. That's up to society to judge, of course. I'm not suggesting I hold any special knowledge of where the line is.
But going full circle to the article, yes, I'll still be eating bacon from time to time!
