This was published in a peer-reviewed journal.
Unfortunately, it is a type of study which remains publishable, even though just about everything about it is crap. Statistics are normally used to determine whether results are valid (if there is more than a 5% chance of the results happening by chance (aka P value), they are rejected). But in this kind of study, statistics are used to find correlations between many different factors. At the usual criteria, where the P value is 5% or less, one out of twenty statistically significant correlations is not real. The other consideration with correlations is that they do not equal causation.
One of the methodical flaws of this study is that it did not propose a testable hypothesis (is consumption of certain food items associated with higher mortality?), but set out to demonstrate an association between consumption of red meat and mortality (to what extent does red meat consumption increase mortality?). Asking the wrong questions, especially questions that assume a predetermined result, leads to the wrong observations being made, and the wrong statistical tests being applied. (There are enough different statistical methods that if one does not produce the desired result, another can be used. One should not choose the statistical test that gets the desired result out of the data--the test should be applied according to the type of data being analyzed, regardless of the outcome.)
Another methodical flaw is the way the data was gathered, through questionaires sent out at intervals asking participants about their diet. Who really remembers what they ate and how much, over the course of a week or longer?
Anyway, sorry for being so wordy here. Bad studies are a pet peeve of mine. You can read the study for yourself here.
great explanation - thanks