if at first you don't succeed

[piranha]

manipulate your data, says paul campos.

The researchers collected data from 527,265 AARP members, who were followed for 10 years. What they found was exactly the same result reported by Flegal and her colleagues: Among both men and women, "overweight" people had the lowest mortality risk. This result, however, was clearly unacceptable. So they began torturing their data.

via firecat.

campos doesn't mention the title nor the researchers in question, but i did a quick search and the study he's referring to must be Overweight, Obesity, and Mortality in a Large Prospective Cohort of Persons 50 to 71 Years Old, by Kenneth F. Adams, Ph.D., Arthur Schatzkin, M.D., Tamara B. Harris, M.D., Victor Kipnis, Ph.D., Traci Mouw, M.P.H., Rachel Ballard-Barbash, M.D., Albert Hollenbeck, Ph.D., and Michael F. Leitzmann, M.D. now to get my hands on the full text.

i notice that in the same issue (what is this, fat week at NEJM? there are 5 items dealing with weight issues) is an article from a baby boomer concerned with his weight that makes reference to this study. anyone here have a subscription to NEJM who can tell me whether that guy noticed the problems with the study?

Comments

[brooksmoses]

The other article you refer to only makes mention of the study in the very brief "there's another article about this in this issue" manner that's quoted in the extract -- there's no mention of the study elsewhere in the article at all. Most of that article is based on older research.

Incidentally, the Adams et. al. study defends basing things on at-age-50 BMI values (and thereby ignoring subsequent weight loss) as "We attempted to correct for potential bias from disease-related weight loss by using participants' recalled weight at the age of 50 years to examine the relation of BMI to the risk of death...."

Also, Campos seems to have missed some additional data-torturing in the study. The authors state, "Among all men (Table 2 and Figure 1A) and women (Table 3 and Figure 2A), including smokers and those with preexisting disease ... Overweight was not associated with an increased risk of death among men but was weakly associated with an increased risk of death among women," and their data figures do seem to bear this out -- but only because they are comparing to a baseline of BMI=25. In both cases, the middle of the "normal weight" range seems to clearly be associated with an increased risk of death compared to the BMI=25 values and to the middle of the "overweight" range.

However, that does seem to be skewed a little bit by smoking; removing smokers gives a pretty symmetrical distribution around a BMI of about 24 for women and 25 for men, with deviations in either direction from that being an increased risk in the sample population. (Whether this applies to people under age 50 is outside the scope of the study, of course.)

[ebourland.livejournal.com]

I never trust any statistics anyone quotes at me, esp anyone who's trying to win an argument or make a point.

I don't even trust scientists, b/c so often their funding or tenure or status depends on the outcome of their experiment.

[lizw.livejournal.com]

Thanks for the link to Paul Campos' column - I've now added it to my syndicated feeds.