Tobacco Documents Online

reviewed" projects from 1989 to 1993, based on information in CLAR's tax forms, which we obtained from the Internal Revenue Service. (These figures do not include awards made in 1990 because these tax forms were not available.) This paper will present our analysis of the content, quality and use of research sponsored by CIAR. The content of CTAR-funded research was assessed by exa_Mining the types of projects and the affiliations of principal investigators funded by CIAR. The quality of CIAR-funded research was assessed by eza*nining articles published by CIA.R-funded principal investigators. Specifically, we determined whether the articles had been peer-reviewed; whether they had been published in symposia proceedings; and what source(s) of funding were acknowledged. The use of CIAR-funded research was assessed by analyzing testimony presented by CIA.R- funded researchers at federal hearings related to environmental tobacco smoke. In all cases, we compared our findings for CIAR's peer-reviewed projects versus special- reviewed projects. We will also present two case studies of special-reviewed projects, which suggest that conflict of interest may have influenced the study design of the projects. Fi.nally, we will discuss the implications of our findings with respect to tobacco industry-funded research in particular, and industry-funded research in general. Content of Research Funded by CIAB. The first phase of our study involved a content analysis of CIAR's peer-reviewed projects and special-reviewed projects. We were particularly interested in determining what percentage of CIAR-funded research is related to environmental A version of this manuscript is to be published in 10 the Journal of Health Politics, Policy and Law.

tobacco smoke (ETS). As noted above, research related to ETS has a high potential for being influenced by conflict of interest, since the tobacco industry could use the results in legal and legislative settings. We obtained i.nformation from CIAR regarding projects funded from January, 1988 to October, 1993 (Center for Indoor Air Research 1992b; Center for Indoor Air Research, written communication, 1993). The information included the title of each project, the name and affliation of each princdpal investzgator, the date the project was originally funded, and a one paragraph description of the purpose of the project. Based on the one paragraph description, we classified the topic of each project as either related to ETS or not related to ETS. A project was considered ETS-related if it studied an aspect of tobacco, tobacco smoke or a tobacco-specific substance such as nicotine. Studies of general indoor air quality that did not speci.fically mention tobacco were considered non-ETS studies. We then sub-classified the topics of the projects into the categories of health effects, exposure measurements, or other. A 'health effects' study evaluated the effects of exposure to a substance; the effects could be either acute or chronic, and they could be measured in cells, animals or humana. For example, a study that measured lung development in rats exposed to ETS was considered a'health effects' . study, as was a study of allergic reactions to ETS in asthmatics. An 'exposure' study measured the amount of exposure to a substance, either directly or indirectly, but did not evaluate the effects of that exposure. For example, studies of substance uptake in the body were considered exposure studies, as were measurements of exposure markers such as cotinine (a nicotine metabolite) in urine. In addition, studies that ~ Ctt ~ ~ A version of this manuscript is to be published in 11 ~ l d I aw. ,,,~ the Journa of Health Pohncs, Pokcy an ~~

measured the levels of various substances, such as the amount of nicotine in indoor air, were considered exposure studies. Studies that did not evaluate either health effects or exposure were classified in the 'other' category. One project, for example, studied whether tobacco smoke is perceived primarily through sight or smell. Our findings on the types of projects funded by CIAR are slTmmarized in Table 1. CIARR funded 40 peer-reviewed and 19 special-reviewed projects from 1988 to 1993. Thirty percent (12/40) of the peer-reviewed projects studied issues related to ETS, compared to 63% (12/19) of the special-reviewed projects (ChiSq=5.869, df=1, P=.0154). In addition, most special-reviewed projects studied exposure to ETS, while peer-reviewed projects were evenly divided between health effects studies and exposure studies. We also Pxa**+ined the affiliations of CIAR-funded principal investigators, in order to determine whether CIAR was funding primarily academics. Based on the information provided by CIAR, we coded the affiliation of each principal investigator (PI) as either academic, corporate or other. PIs were classified as 'academic' if their stated affiliation was a university or college. They were classified as 'corporate' if their stated affiliation was a private sector business, consulting firm or contracting agency. All other a.ffiliations were classified as 'other.' None of the PIs had more than . one affiliation listed • Our findings regarding the affiliations of CIAR-funded principal investigators (PIs) are summarized in Table 2. CIAR funded 52 PIs from 1988 to 1993. Thirty- four PIs ieceived funding for one or more peer-reviewed projects, while 18 PIs received N O L"t 0 ski A version of this manuscript is to be published in 12 Q'y the Journal of Health Politics, Policy and Law. ~ Z <

funding for one or more special-reviewed projects.l There was a significant difference between the affiliations of PIs on peer-reviewed and special-reviewed projects (ChiSq=12.262, df=2, P=.0022). Ninety-four percent (32/34) of PIs on peer-reviewed projects had academic aMations, compared to 56% (10/18) of PIs on special- reviewed projects. Thirty-nine percent (7/18) of PIs on special-reviewed projects were affiliated with private sector companies. We also observed that three of the PIs on special-reviewed projects are currently members of C1AR's Science Advisory Board (two had academic affiliations and one was afl'iliated with the private sector). Our content analysis of projects funded by CIAR shows that there are several significant differences between CIAR's peer-reviewed projects and its special- reviewed projects. First, special-reviewed projects are more likely than peer-reviewed projects to study issues related to ETS. This finding raises concern about conflict of interest, since these projects are awarded based on the recommendations of tobacco industry executives and do not undergo peer review prior to funding. This concern is heightened by the fact that CIAR does not generally disclose that its special-reviewed projects are funded through a non-traditional process. Our second finding was that special-reviewed projects tend to measure exposure to ETS, rather than study its health effects. One of the tobacco industry's . main arguments regarding passive smoking is that people are not generally exposed to high enough levels of ETS to cause any serious daniag-e (Tobacco Institute 1986). Since CIAR's special-reviewed projects tend to study exposure, it is possible that the 1 Z`wo principal investigators (PIs) received funding for both peer-reviewed and special- reviewed ppojeats. They were classified as PIs on special-reviewed projects in our analyses. However, ou'r results would not have changed if we had classified them as PIs on peer-reviewed projects or if we had placed them in a separate category. A version of this manuscript is to be published in 13 the Journal of Health Politics, Policy and Law. 1~

tobacco industry is using them to develop evidence to support its claims that public exposure to ETS is generally low. In addition, from a legal standpoint, a study showing direct health effects caused by ETS would be much more damaging to the industry than a study measuring exposure to ETS; thus, it is "safer" for the industry to sponsor exposure research. Our third finding was that almost 40% of CIAR's special-reviewed projects are awarded to private sector consultants and contractors, while virtually all of its peer- reviewed projects are awarded to academic investigators. It is possible that academic investigators may be wary of accepting funding, particularly from the tobacco industry, for research that has not been peer reviewed. Alternatively, tobacco industry executives may prefer to collaborate with private sector investigators. There is some evidence suggesting that the tobacco industry sometimes develops special relationships with private sector companies, in which it provides virtually all of the funding for an organization in ezchange for consulting services (Levin 1993). Finally, it is interesting to note that 16% (3/19) of CIAR's special-reviewed projects were awarded to principal investigators who are currently members of CIAR's Science Advisory Board. This sort of financiall relationship between the supposedly independent SAB and the tobacco industry could raise concern about the' independence of the SAB. It also suggests that the industry may reward some members of the SAB by providing funding for their projects. o ~ A version of this manuscript is to be published in 14 14 the Journal of Health Politics, Policy and Law. M ~ b

Quality of Research Funded by CIAR The next phase of our study involved an evaluation of the quality of research being funded through CIAR's peer-reviewed and special-reviewed projects. We could not directly evaluate the quality of the projects themselves, because we did not have access to the proposals submitted to CIAR. As a surrogate, we evaluated the quality of publications associated with CIAR-funded research. We first conducted a historical analysis of the body of work associated with CLkR-funded researchers by analyzing any articles they had published from 1989 to 1993. We then analyzed the quality of articles that had specifically resulted from CIAR-funded projects. Quality assessments were based on whether the article had been peer reviewed; whether the article had been published as a symposium; and what source(s) of funding were acknowledged. Mutually exclusive categories were used in all cases. Peer-review status of journal in which publication appeared: peer-reviewed, non peer- reviewed. Peer-reviewed publications have been associated with higher quality (Rochon 1994; Barnes and Bero, unpublished data) and a more balanced presentation of information (Bero et aL 1992) than non-peer-reviewed publications. In addition, the tobacco industry often cites data from non-peer-reviewed publications to support its . position that ETS has not been proven harmfu.I to health (Bero and Glantz 1993). To test the hypothesis that PIs on special-reviewed projects would be more likely than PIs on peer-reviewed projects to publish in non peer-reviewed journals, we deterni~ned the peer-review status of the parent journal for each artide evaluated. A journal w:4s considered peer-reviewed if it stated that it was peer-reviewed, if it ~ .., Clt O A version of this manuscript is to be published in 15 ~ the Journal of Health Politics, Policy and Law. ~y ~ Cs ~

published a list of peer reviewers, or if it required that a manuscript be submitted for review prior to publication. In all other cases, the journ.al was considered non-peer- reviewed. Type of article: symposium /conference proceeding, original researclz,, review, other. Several studies have suggested that some types of articles are associated with higher quality or a more balanced presentation of information than others. For example, original research articles published in symposia proceedings have been associated with lower quality than original research articles published in peer-reviewed journals (Rochon 1994; Barnes and Bero, unpublished data). In addition, review articles in general have been associated with biases based on the affiliation of the reviewer (Chalmers et al. 1990). Also, symposia articles on ETS, regardless of whether they are original research articles or reviews, are more likely than original research articles from peer-reviewed journals to support the tobacco industry position on the health effects of ETS (Bero et al. 1994), and the tobacco industry often cites symposia articles and letters to the editor as if they had been peer reviewed (Bero and Glantz 1993). To test the.hypothesis that PIs on special-reviewed projects would publish a smaller proportion of original research articles compared to PIs on peer-reviewed projects, we classified each article p.z *nined as either a symposium, an original research article, a review or other type of article. Classifications were based on how the article was described in the table of contents of the journal in which it appeared. For ezample, if the table of contents stated that all of the articles in the issue were ~ ~. A version of this manuscript is to be published in 16 the Journal of Health Politics, Policy and Law. . t

being published as part of a symposium or conference proceeding, the article was coded as a symposium. If the table of contents listed an article in the letters-to-the- editor section, it was classified in the 'other' category. All articles published in symposia were classified as symposium articles, regardless of whether they were original research articles or reviews, because of previous studies suggesti.ng that symposia publications in general may be associated with biases (Bero et al. 1994). Source(s) of funding acknor.vledged: governm.ent, tobacco industry, oth.er/ multiple, none. Source of funding has been associated with the quality and outcome of research (Davidson 1986; Swaen and Meijers 1988; Lexchin 1993; Rochon et al. 1994; Cho and Bero, submitted). Therefore, we wanted to determine whether CIAR-funded researchers had previously published research sponsored by the tobacco industry. We categorized the source(s) of funding acknowledged for each article PYa*nined. Government sources included any international, federal, state, or local government agency, such as the National Institutes of Health. Tobacco industry sources of funding included any tobacco company, the Tobacco Institute, CIARR or CTR. Sources of funding classified as 'other/multiple' included non-profit health organizations, such as the American Cancer Society; private-sector corporations, such as Ford Motor Company; universities and academic institutions; etc. , Publications with more than one type of funding acknowledged were coded as 'tobacco industry if the tobacco industry was one of the sources acknowledged; otherwise, publications with multiple types of funding were coded as 'other/multiple.' If no source ' of fundin was acknowled ed it was classified as 'none 2V . g , g C ~ A version of this manuscript is to be published in 17 ~ the Journal of Health Politics, Policy and Law. ~ ~ CA c

Attitude of publication toward tobacco irzdustry position on the health effects of tobacco: pro-ircdustry, con-industry, neutral. There is some evidence suggesting that the tobacco industry awarded CTR special projects to investigators with histories of publishing pro-industry articles (Freedrnan and Cohen 1993). To test the hypothesis that PIs on CIAR special-reviewed projects were more likely than PIs on CLAR peer- reviewed projects to publish pro-industry articles, we categorized each article as pro, con, or neutral to the tobacco industry position on the health effects of exposure to tobacco smoke (either active or passive). An article was considered 'pro-industry' if it concluded there is no evidence that tobacco smoke is associated with adverse health effects; that the evidence is inconclusive; that confounding variables are responsible for health effects attributed to tobacco smoke; that studies on tobacco smoke are statistically flawed; or that the level of exposure to tobacco smoke is not high enough to cause health problems. An article was also considered 'pro-industry' if it discussed exposure to tobacco smoke without mentioning the documented hazards of exposure at any point. For example, an article that concluded smoking could be tolerated indoors, without making any statements about the potential risks of exposure, was considered 'pro-industry.' An article was considered 'neutral' if it did not discuss tobacco or tobacco smoke. An article was considered 'con-industry' if it concluded . that either active or passive exposure to tobacco smoke is associated with adverse health effects. These categories have been used previously (Bero and Glantz 1993; Bero et al. 1994) and are based on the written statements of the tobacco industry (Tobacco Institute 1986). - 7 ~ ~, A version of this manuscript is to be published in 18 CA and Iaw the Tournal Polic f H y lth P litics , o a e o ~ ~ M 9

Qualitv of ardcles published by CIAR-filnded researchers. 19R9-1993 We first conducted a historical analysis of the body of work associated with CIAR- funded principal investigators from 1989 to 1993. We searched two on-line databases of the medical and scientific literature,lVIEDLINE and Current Contents, as well as a database of symposia studies on ETS that had been developed for a previous study (Bero et al. 1994). Searches were conducted using the last name and first initial of each principal investigator. Articles were included if they appeared to have been written or co-authored by the CIAR-funded researcher, based on an Pg ination of the full name and affiliation of the author of the article and the general topic area of research. Publications such as editorials and letters-to-the-editor were included in our analysis. MEDLII~E searches retrieved articles published from January, 1989 to August, 1993; Current Contents searches retrieved articles published from May, 1989 to October, 1993. We identified 617 articles published by CIAR.-funded PIs. Fifty-four (8.6%) of these could not be located in libraries within the University of California system or Stanford University. For the rema~'n~rg 563 articles, we compared the quality of articles published by PIs on peer-reviewed projects to the quality of articles published . by PIs on special-reviewed projects, using the criteria described above.2 Our findings regarding the publication histories of CIAR-funded principal investigators are summarized in Table 3. PIs on peer-reviewed projects published an 2Articles published by the two principal investigators (PIs) who received funding for both peer-reviewed and special-reviewed projects were cla.ssified as articles by PIs on special-reviewed projects. However, our results would not have rhanged if we had classified them as articles by PIs on peer-reviewed projects. A version of this manuscript is to be published in 19 the Journal of Health Politics, Policy and Law.