On 2016 May 23, Stanton A Glantz commented:

In June 2015 we published our paper “The smoking population in the USA and EU is softening not hardening” in the journal Tobacco Control. We showed that as smoking prevalence has declined over time, quit attempts increased in the USA and remained stable in Europe, US quit ratios increased (no data for EU), and consumption dropped in the USA and Europe. These results contradict the hardening hypothesis which is often used as part of the tobacco industry’s strategy to avoid meaningful regulation and protect its political agenda and markets, claiming that there is a need for harm reduction among those smokers who “cannot or will not quit.” Indeed, rather than “hardening” the remaining smoking population is “softening.”

In February 2016 we received an email from Robert West, editor of the journal Addiction, informing us that Addiction was about to publish an article by Plurphanswat and Rodu entitled “A Critique of Kulik and Glantz: Is the smoking population in the US really softening?” whose sole purpose was to critique our Tobacco Control paper, and offered to let us respond to the criticism.

The fact that Plurphanswat and Rodu sent their paper to Addiction was unusual because normal scientific procedure would have had them sending a letter to the editor of the journal that originally published the work (Tobacco Control).

As detailed below, we did respond, noting that Plurphanswat and Rodu’s paper followed the well-established pattern of tobacco industry-funded researchers trying to create controversy about research inconsistent with industry interests, the fact that Rodu had understated his financial ties to the industry, and, of course, showing how their criticism was based on statistical error that they made.

Addiction rejected our response because we would not delete the first two points and limit our response only to the statistical issue.

This blog post includes the response that Addiction rejected so that readers of Plurphanswat and Rodu’s critique do not think we did not have a response. We also include a summary of our interactions with the journal and the related email correspondence.

THE REJECTED RESPONSE

Consider the Source

“Harm reduction” is a key part of the tobacco industry’s strategy to avoid meaningful regulation and protect its political agenda and markets.[1] This agenda is premised on the existence of “hard core” smokers who “cannot or will not” quit.[2-4] Our paper, “The smoking population in the USA and EU is softening not hardening”,[5] undermined this agenda because it showed that, contrary to the hardening hypothesis, as smoking prevalence has declined over time, quit attempts increased in the USA and remained stable in Europe, US quit ratios increased (no data for EU), and consumption dropped in the USA and Europe.

There is a longstanding pattern of tobacco industry-funded experts writing letters criticizing work that threatens the industry’s position, first described in 1993 by then-JAMA Deputy Editor Drummond Rennie.[6] Rodu and various co-authors have written several such letters.[7-10] Another similarity to past efforts is industry-linked experts submitting critiques of a paper published in one journal to another,[11-15] which is also the case here, with this critique of our paper published in Tobacco Control being published in Addiction. One would have expected any criticism to have been published as a letter in Tobacco Control.

Addiction requires “full disclosure of potential conflicts of interest, including any fees, expenses, funding or other benefits received from any interested party or organisation connected with that party, whether or not connected with the letter or the article that is the subject of discussion.” As with another investigator supported by the tobacco industry,[16] the conflict of interest statement Plurphanswat and Rodu provide may not truly reflect the extent of Rodu’s involvement with the tobacco industry. For example:

• Rodu’s Endowed Chair in Tobacco Harm Reduction Research at the University of Louisville is funded by the U.S. Smokeless Tobacco Company (US Tobacco) and Swedish Match North America, Inc.[17]

• Rodu is a Senior Fellow at the Heartland Institute, which has received tobacco industry funding.[18-20]

• Rodu is a Member and Contributor to the R Street Institute, which has received tobacco industry funding.[19,21]

• Before moving to Louisville, Dr. Rodu was supported in part by an unrestricted gift from the United States Smokeless Tobacco Company to the Tobacco Research Fund of the University of Alabama at Birmingham.[8]

• Rodu was a keynote speaker at the 2013 Tobacco Plus Expo International, a tobacco industry trade fair to discuss “How has the tobacco retail business evolved; where was it fifteen years ago, where is it today and where is it going”.[22]

• Rodu has worked with RJ Reynolds executives between at least 2000 and 2009 to help promote industry positions on harm reduction, including specific products.[23-26]

The substance of Plurphanswat and Rodu’s criticism is that the statistically significant negative association between smoking prevalence and quit attempts and the positive association between prevalence and cigarettes smoked per day both become non-significant when more tobacco control variables are included in the model (state fixed effects, cigarette excise taxes, workplace smoking bans and home smoking bans). The problem with including all these variables is that it results in a seriously overspecified model, which splits any actual effects between so many variables that all the results become nonsignificant. The regression diagnostic for this multicollinearity is the Variance Inflation Factor (VIF); values of the VIF above 4 indicate serious multicollinearity. For the United States, adding all the other variables increases the VIF for the effect of changes in smoking prevalence from 1.8 in our model for quit attempts to 16.7, and from 1.8 in our model to 17.9 for cigarettes per day, respectively. Plurphanswat and Rodu’s model is a textbook case of why one has to be careful not to put too many variables in a multiple regression.

The Plurphanswat and Rodu criticism misrepresents our conclusions. We did not argue that drops in prevalence caused increased quit attempts and reduced consumption; we simply present the observation that, as prevalence falls, quit attempts increase and consumption fall or remain constant, which is the exact opposite of what the hardening hypothesis predicts.

The references and the full email correspondence with Addiction is available at http://tobacco.ucsf.edu/addiction-refuses-allow-discussion-industry-ties-criticism-our-“softening-paper”

On 2016 May 19, Brad Rodu commented:

We conducted an additional analysis of the American data used in this study, which has been published in Addiction: Plurphanswat N, 2016

The original authors ran a linear regression of state-level measures of quit attempts, quit ratios and cigarette consumption on smoking prevalence using data sets from the Tobacco Use Supplements to the Current Population Surveys from 1992–2011. However, they did not include relevant tobacco control variables such as cigarette excise taxes, workplace and home smoking bans, and other unobserved, time-invariant, state-specific factors that could have accounted for the observed associations.

We extended the original model by including these readily available variables, and we found that the original results were not robust; there was a large drop in the coefficients and they became statistically non-significant. These findings do not support the original authors' claim that the smoking population is "softening."

Since 2005 Dr Rodu has been supported by the Kentucky Research Challenge Trust Fund and by unrestricted grants to the University of Louisville from tobacco manufacturers (Swedish Match AB, U.S. Smokeless Tobacco Company, Reynolds American Inc. Services, Altria Client Services, and British American Tobacco). Dr Plurphanswat has been supported by these grants since 2013.

On 2016 May 19, Brad Rodu commented:

We conducted an additional analysis of the American data used in this study, which has been published in Addiction: Plurphanswat N, 2016

The original authors ran a linear regression of state-level measures of quit attempts, quit ratios and cigarette consumption on smoking prevalence using data sets from the Tobacco Use Supplements to the Current Population Surveys from 1992–2011. However, they did not include relevant tobacco control variables such as cigarette excise taxes, workplace and home smoking bans, and other unobserved, time-invariant, state-specific factors that could have accounted for the observed associations.

We extended the original model by including these readily available variables, and we found that the original results were not robust; there was a large drop in the coefficients and they became statistically non-significant. These findings do not support the original authors' claim that the smoking population is "softening."

Since 2005 Dr Rodu has been supported by the Kentucky Research Challenge Trust Fund and by unrestricted grants to the University of Louisville from tobacco manufacturers (Swedish Match AB, U.S. Smokeless Tobacco Company, Reynolds American Inc. Services, Altria Client Services, and British American Tobacco). Dr Plurphanswat has been supported by these grants since 2013.

On 2016 May 23, Stanton A Glantz commented:

In June 2015 we published our paper “The smoking population in the USA and EU is softening not hardening” in the journal Tobacco Control. We showed that as smoking prevalence has declined over time, quit attempts increased in the USA and remained stable in Europe, US quit ratios increased (no data for EU), and consumption dropped in the USA and Europe. These results contradict the hardening hypothesis which is often used as part of the tobacco industry’s strategy to avoid meaningful regulation and protect its political agenda and markets, claiming that there is a need for harm reduction among those smokers who “cannot or will not quit.” Indeed, rather than “hardening” the remaining smoking population is “softening.”

In February 2016 we received an email from Robert West, editor of the journal Addiction, informing us that Addiction was about to publish an article by Plurphanswat and Rodu entitled “A Critique of Kulik and Glantz: Is the smoking population in the US really softening?” whose sole purpose was to critique our Tobacco Control paper, and offered to let us respond to the criticism.

The fact that Plurphanswat and Rodu sent their paper to Addiction was unusual because normal scientific procedure would have had them sending a letter to the editor of the journal that originally published the work (Tobacco Control).

As detailed below, we did respond, noting that Plurphanswat and Rodu’s paper followed the well-established pattern of tobacco industry-funded researchers trying to create controversy about research inconsistent with industry interests, the fact that Rodu had understated his financial ties to the industry, and, of course, showing how their criticism was based on statistical error that they made.

Addiction rejected our response because we would not delete the first two points and limit our response only to the statistical issue.

This blog post includes the response that Addiction rejected so that readers of Plurphanswat and Rodu’s critique do not think we did not have a response. We also include a summary of our interactions with the journal and the related email correspondence.

THE REJECTED RESPONSE

Consider the Source

“Harm reduction” is a key part of the tobacco industry’s strategy to avoid meaningful regulation and protect its political agenda and markets.[1] This agenda is premised on the existence of “hard core” smokers who “cannot or will not” quit.[2-4] Our paper, “The smoking population in the USA and EU is softening not hardening”,[5] undermined this agenda because it showed that, contrary to the hardening hypothesis, as smoking prevalence has declined over time, quit attempts increased in the USA and remained stable in Europe, US quit ratios increased (no data for EU), and consumption dropped in the USA and Europe.

There is a longstanding pattern of tobacco industry-funded experts writing letters criticizing work that threatens the industry’s position, first described in 1993 by then-JAMA Deputy Editor Drummond Rennie.[6] Rodu and various co-authors have written several such letters.[7-10] Another similarity to past efforts is industry-linked experts submitting critiques of a paper published in one journal to another,[11-15] which is also the case here, with this critique of our paper published in Tobacco Control being published in Addiction. One would have expected any criticism to have been published as a letter in Tobacco Control.

Addiction requires “full disclosure of potential conflicts of interest, including any fees, expenses, funding or other benefits received from any interested party or organisation connected with that party, whether or not connected with the letter or the article that is the subject of discussion.” As with another investigator supported by the tobacco industry,[16] the conflict of interest statement Plurphanswat and Rodu provide may not truly reflect the extent of Rodu’s involvement with the tobacco industry. For example:

• Rodu’s Endowed Chair in Tobacco Harm Reduction Research at the University of Louisville is funded by the U.S. Smokeless Tobacco Company (US Tobacco) and Swedish Match North America, Inc.[17]

• Rodu is a Senior Fellow at the Heartland Institute, which has received tobacco industry funding.[18-20]

• Rodu is a Member and Contributor to the R Street Institute, which has received tobacco industry funding.[19,21]

• Before moving to Louisville, Dr. Rodu was supported in part by an unrestricted gift from the United States Smokeless Tobacco Company to the Tobacco Research Fund of the University of Alabama at Birmingham.[8]

• Rodu was a keynote speaker at the 2013 Tobacco Plus Expo International, a tobacco industry trade fair to discuss “How has the tobacco retail business evolved; where was it fifteen years ago, where is it today and where is it going”.[22]

• Rodu has worked with RJ Reynolds executives between at least 2000 and 2009 to help promote industry positions on harm reduction, including specific products.[23-26]

The substance of Plurphanswat and Rodu’s criticism is that the statistically significant negative association between smoking prevalence and quit attempts and the positive association between prevalence and cigarettes smoked per day both become non-significant when more tobacco control variables are included in the model (state fixed effects, cigarette excise taxes, workplace smoking bans and home smoking bans). The problem with including all these variables is that it results in a seriously overspecified model, which splits any actual effects between so many variables that all the results become nonsignificant. The regression diagnostic for this multicollinearity is the Variance Inflation Factor (VIF); values of the VIF above 4 indicate serious multicollinearity. For the United States, adding all the other variables increases the VIF for the effect of changes in smoking prevalence from 1.8 in our model for quit attempts to 16.7, and from 1.8 in our model to 17.9 for cigarettes per day, respectively. Plurphanswat and Rodu’s model is a textbook case of why one has to be careful not to put too many variables in a multiple regression.

The Plurphanswat and Rodu criticism misrepresents our conclusions. We did not argue that drops in prevalence caused increased quit attempts and reduced consumption; we simply present the observation that, as prevalence falls, quit attempts increase and consumption fall or remain constant, which is the exact opposite of what the hardening hypothesis predicts.

The references and the full email correspondence with Addiction is available at http://tobacco.ucsf.edu/addiction-refuses-allow-discussion-industry-ties-criticism-our-“softening-paper”