Wednesday, September 26, 2007

Paper Shows No Increase in Marlboro Nicotine Yields from 1997-2006, Yet Concludes There Has Been a Linear Trend of Increasing Yields During this Time

An article published online in this month's issue of the journal Tobacco Control concludes that there has been a linear, non-random trend of increasing nicotine yields in Marlboro brand cigarettes during the period 1997 to 2006 (see: Connolly GN, Alpert AR, Wayne GF, Koh H. Trends in nicotine yield in smoke and its relationship with design characteristics among popular US cigarette brands, 1997-2005. Tobacco Control 2007; 16:e5).

The paper presents a figure (Figure 1) which shows the trend in average nicotine yields of 18 Marlboro family brand styles between 1997 and 2006. The paper concludes: "A multilevel regression analysis of the nicotine yields in smoke of the 18 Marlboro brand styles reported in each year from 1997 through 2006, as reflected by the MDPH data, shows a statistically significant increasing temporal trend of 0.008 mg/year (p,0.001). Average figures for nicotine yield for these Marlboro brands and the regression line showing a linear, nonrandom trend in nicotine yield in smoke are shown in fig 1."

This finding directly contradicts the public statement of Philip Morris regarding Marlboro's nicotine yields. According to the Tobacco Control article: "PM USA issued press releases, observing that the apparent trend in nicotine yield in smoke from 1998 to 2004 was not present when data from 1997, 2005 and 2006 were included. The PM USA website also shows simple linear average nicotine yields in smoke for 18 Marlboro cigarette styles reported in all years from 1997 through 2006, claiming no significant temporal trend and concluding that ‘year-to-year variations in nicotine occur as part of the normal processes of growing tobacco and manufacturing cigarettes.’ The present analysis shows a significant temporal increase in nicotine yield in smoke, contradicting the PM USA claims."

Importantly, the paper also notes that: "All cigarettes are highly addictive and deadly, and relatively minor changes in nicotine yield may not significantly alter the product’s addictive properties." and that "The increase over time in nicotine yield in smoke does not necessarily signify any change in exposure within the population of smokers, particularly as smoking behaviour among humans is compensatory and will adjust for differences in smoke yield."

Finally, the paper notes that "Human studies, including measures of smoking topography and biomarkers of exposure, may be necessary to predict consumer and population effects."

The Rest of the Story

The rest of the story is that the article reports an average nicotine yield of Marlboro cigarettes in 1997 of approximately 1.85 mg/cigarette, and an average yield in 2006 of approximately 1.85 mg/cigarette.

You're not reading that incorrectly.

The average nicotine yield of Marlboro cigarettes in 2006 was, according to this article itself (not relying upon Philip Morris' own data), no higher in 2006 than it was in 1997.

So if the nicotine yield of Marlboro cigarettes was no higher in 2006 than in 1997, then how could there have been a significant, linear trend of increasing nicotine yields of Marlboro cigarettes during this time period?

The answer is: there could not have been such a trend. Such a trend of linearly increasing yield of nicotine yields is not consistent with the study findings, which demonstrate essentially no difference in the yields of Marlboro cigarettes in 2006 vs. 1997.

The curious reader may ask himself or herself: then how does one conclude that there was a trend of linearly increasing nicotine yields?

The answer is that you fit a line to data which clearly do not show a simple linear pattern.

If you examine Figure 1, what you'll see is that the nicotine yield of Marlboro cigarettes increased rather steadily from 1997 through 2003, and then decreased rather steadily from 2003 through 2006. The nicotine level ended up exactly where it had started. But it went up for a few years, and then back down.

In other words, the trend in nicotine yields exhibits an upside-down "V" pattern.

The appropriate way to model this would be to fit two lines: a line with increasing nicotine yield between 1997 and 2003, and a line with decreasing nicotine yield between 2003 and 2006.

In other words, what appears to have actually happened is that Marlboro's nicotine yields increased between 1997 and 2003, but decreased between 2003 and 2006.

Because the period of time during which the increase occurred was longer than the period of time during which the decrease occurred, if you fit a simple line (a single line) to the data, you are going to find a line with a positive slope. But that line does not really indicate what occurred during the study period.

In my opinion, these data clearly demonstrate that Philip Morris was correct and the article is wrong regarding the observed pattern. There was indeed no net increase in nicotine yields of Marlboro cigarettes between 1997 and 2006. The nicotine yields did not increase by 0.008 mg/year. What happened is that the nicotine yields increased by about 0.016 mg/year between 1997 and 2003, and then decreased by about 0.033 mg/year between 2003 and 2006.

What's most interesting to me is the fact that while Philip Morris is apparently being blamed for increasing the nicotine yield of its Marlboro cigarettes between 1997 and 2003, the company is not simultaneously being praised for having decreased the nicotine yields between 2003 and 2006.

In fact, the decreasing nicotine yield of Marlboros between 2003 and 2006 is "impressive." The rate of decline in nicotine yields is twice what the rate of increase was between 1997 and 2003.

Note that I'm not arguing here that Philip Morris did or did not intend to increase or decrease its nicotine yields within Marlboro cigarettes. These data cannot demonstrate whether or not the changes reflect a true attempt to control these levels on the part of the company, or whether the changes merely reflect random variations. While the variation in nicotine does not appear to be random, neither does it appear to represent a linearly increasing trend.

While I would not have thought that I would be taking the side of Philip Morris in a conflict over the interpretation of scientific data, I do find myself in that position based on my review of this research.

Perhaps an even most interesting aspect of the paper, to me, is the fact that the article itself acknowledges that the observed changes in nicotine yield are essentially meaningless, because "relatively minor changes in nicotine yield may not significantly alter the product's addictive properties," and because "the increase over time in nicotine yield in smoke does not necessarily signify any change in exposure within the population of smokers, particularly as smoking behaviour among humans is compensatory and will adjust for differences in smoke yield."

If it is true that relatively minor changes in nicotine do not necessarily alter the product's addictive properties and that the increase (or decrease) in nicotine yields in smoke does not necessarily signify any change in exposure within the population of smokers, then what exactly is the significance of this research?

If it is true, as the authors conclude, that one can only draw inferences about "consumer and population effects" by conducting "human studies, including measures of smoking topography and biomarkers of exposure," then what exactly does this paper tell us about consumer or population implications of the data reported therein?

The answer is: nothing. And the paper acknowledges that, quite appropriately.

But that raises the question of what the value or importance of this research is. It's not clear to me. However, there's no question that the research has been used to support two things: (1) promotion of a specific policy end - the FDA tobacco legislation; and (2) the opportunity for anti-smoking groups to take a crack at the tobacco companies for increasing yields which, on our own admission, have nothing to do with actual human consumer or population effects, addiction potential, or health.

No comments: