Tuesday, April 12, 2011

Smoking Gun: Manipulating Definitions

A report just published in Nicotine & Tobacco Research (abstract here) draws an obvious conclusion: How surveys define cigarette smokers and smokeless tobacco users influences the prevalence rate of dual use (consumption of both products).

Robert Klesges and colleagues used information obtained from Air Force recruits to show that if dual users are defined as using both products daily (a narrow definition), prevalence of dual use will be low. However, if dual users are defined as using either product once in the past month (a broader definition), prevalence of dual use will be high.

It is not surprising that Dr. Klesges found that defining tobacco use influences survey results. In 2001, he co-authored a study (with first-author C. Keith Haddock, abstract here; hereafter called Haddock-2001) that perfectly demonstrates how definitions can be manipulated to produce desirable results. Haddock-2001 purported to show that smokeless tobacco use is a gateway to smoking among Air Force recruits; it has been widely cited in American prohibitionist attacks on smokeless tobacco. In fact, its results are based on manipulation of definitions.

Haddock-2001 studied 14,340 men (average age, about 20 years) who had never smoked when they reported to Air Force basic training in 1995 and 1996. After one year, Dr. Haddock followed up with 7,865 subjects, finding that 1,099 of them were smoking. That’s a 14% smoking initiation rate among 20-year old men after one year in the Air Force! At one year, Haddock-2001 also found that, compared with recruits who had never used tobacco before basic training, recruits who were current smokeless users were 2.33 (95% Confidence interval = 1.84 – 2.94) times more likely to be smoking.

First, how did 14% of first-year never-smoking Air Force recruits start smoking? It turns out that this number is grossly inflated. In 1999, Drs. Klesges and Haddock had published a study of these same recruits (abstract here; hereafter called Klesges-1999), in which they reported that only 8% of never smokers had started to smoke after one year in the Air Force.

The discrepancy between Haddock-2001 and Klesges-1999 is due to differences in the category definitions of smoking, summarized in this table.

Different Definitions of Smokers in Klesges-1999 and Haddock-2001
At Enrollment
Never SmokerNever smoked a cigaretteNever Smoked regularly
Experimental SmokerSmoked on one or two occasions, never regularlyNot mentioned
Ex-smokerSmoked regularly, but quitSmoked regularly, but quit
Current SmokerSmoked regularly, at least one cigarette per daySmoked regularly, at least one cigarette per day
At One Year
Current SmokerSmoked even a puff in last 7 daysSmoked even a puff in last 7 days

At enrollment, Klesges-1999 identified a group of “experimental smokers,” 26% of whom became smokers at the one-year follow-up. But Haddock-2001 never mentioned experimental smokers, which means that they were in the “never-smoking” group at enrollment; that significantly accounts for the 14% initiation rate, instead of the 8% rate found in the Klesges-1999 report.

It is likely that many Haddock-2001 smokeless tobacco users were also experimental smokers. The only apparent reason to eliminate the experimental smoker category and effectively shift those subjects to the never-smoking category was to bolster the case for labeling smokeless tobacco use as a gateway to smoking.

Dr. Haddock also defined current smoking differently at enrollment than after one year of follow-up. Good practice in scientific investigation is to establish definitions and stick with them throughout a research project. As the table shows, Haddock-2001 defined a current smoker at enrollment as smoking at least one cigarette per day, but then defined a current smoker at one-year as someone who smoked even a puff in the last 7 days.

The manipulation of smoking definitions in Haddock-2001 casts considerable doubt on its claim that smokeless tobacco use is a gateway to smoking.

No comments: