Publication Bias and Public Health Policy on Environmental
Tobacco Smoke
(JAMA. 1994;272:133-136)
Lisa A. Bero, PhD; Stanton A. Glantz, PhD; Drummond Rennie, MD
Objective.--To examine the tobacco industry's claim
that publication bias against negative studies invalidates the risk
assessment of environmental tobacco smoke (ETS) conducted by the US
Environmental Protection Agency and other reviews of the health effects
of ETS.
Design.--Determination of the number of published original
research articles that tested the hypothesis that ETS exposure is
associated with adverse health effects and that reported statistically
significant ("positive") or nonsignificant ("negative")
results; the number of articles that concluded that ETS is a health
risk; and unpublished studies on the effects of ETS on health.
Participants.--Articles identified by a computerized search
of the medical literature supplemented with material obtained from the
tobacco industry and hand searching. Articles were classified as
peer-reviewed journal articles or articles from sponsored symposia.
Main Outcome Measure.--The statistical significance of
results reported in the article and whether or not the article
concluded that ETS exposure is a health risk.
Results.--More symposium articles than journal articles were
reviews (46% vs 6%; P =.0001). More original journal
articles than original symposium articles reported the use of
statistical tests (96% vs 54%; P =.0001). Of articles
with statistical analyses, similar proportions of journal articles and
symposium articles reported statistically significant results (57% vs
47%; P =.329). The conclusions of 80% of the original
journal articles were positive, compared with 51% of the original
symposium articles (P =.006).
Conclusions.--There is no publication bias against
statistically nonsignificant results on ETS in the peer-reviewed
literature. The high proportion of articles in symposia that reach the
conclusion that ETS is not harmful primarily results from the inclusion
of review articles.
(JAMA. 1994;272:133-136)
Publication bias exists against negative studies: studies that do not present
statistically significant results.[1]
[2]
[3]
[4]
However, little is known about the effect of
publication bias on clinical practice and public health policy.
Meta-analysis, the combining of numerous small studies into one study
for analysis, is a means of answering questions regarding the
usefulness of clinical procedures.[5] If negative
studies are not published, they might not be included in meta-analyses or widely
disseminated.[6] Inclusion of unpublished
studies in meta-analysis could invalidate or weaken the meta-analysis.
One example of how publication bias has been featured in a health
policy debate has been the tobacco industry's criticism of scientific
consensus, documents, and review articles that conclude that exposure
to environmental tobacco smoke (ETS) is
harmful.[7]
[8]
[9]
[10]
[11]
For many years some researchers have contended that conclusions regarding
the adverse health effects of ETS are invalid because publication bias
exists.[12]
[13]
[14]
[15]
[16]
[17]
The industry's publication bias argument has
been disseminated to the lay community through tobacco industry press
releases and the lay press.[18] [19]
In December 1992, the US Environmental Protection Agency (EPA)
published a risk assessment of ETS[9] that concluded that ETS
causes lung cancer in adults and respiratory problems in children.
The risk assessment has significant policy implications, since its
approval classified ETS as a known human carcinogen (group A). In June
1993, several tobacco companies filed a lawsuit against the EPA to
require the EPA to withdraw both its classification of ETS as a group A
carcinogen and the ETS risk assessment. The tobacco industry's
pleading states that "various sources of bias, including
publication bias ... could explain any association claimed by
EPA between ETS and lung cancer"[20] (emphasis added) and
asserts that the EPA "cherry-picked scientific data, ignored recent
studies that contradicted its conclusions."[20]
This article tests the validity of the tobacco industry's argument
that publication bias invalidates the EPA's risk assessment of ETS and
other reviews of the health effects of ETS. Our previous work shows
that tobacco industry-sponsored symposium proceedings are cited by the
tobacco industry to support their position that ETS exposure is not
harmful.[21] [22] Since symposium proceedings on ETS comprise a
substantial proportion of the literature on ETS,[22] we
assessed the extent of publication bias among articles on the effects
of ETS on health that had been published as either peer-reviewed
journal articles or proceedings of symposia.
MATERIALS AND METHODS
Published Studies on the Effects of ETS on Health
Symposium articles (n=297) and a random sample of peer-reviewed
articles (n=100) on ETS were identified as described
previously.[22] The MEDLINE, DIALOG, CATALOG, CONFERENCE
PAPERS INDEX, TOXLINE, and INTERNATIONAL GUIDE TO PERIODICALS databases were searched electronically by using the key words command with
"environmental tobacco smoke," "tobacco smoke pollution,"
"passive smoking," "involuntary smoking," and "symposia,"
"proceeding," and "conference paper" to identify 297 symposium
articles published between January 1, 1965, and March 31, 1993. Two
symposia were identified from tobacco industry publications and press
releases. A MEDLINE search identified 583 nonsymposium articles on ETS
published in medical or scientific journals between January
1, 1965, and March 31, 1993. Random numbers were generated by computer
and used to select 100 articles. Both original research and review
articles were included in the random selection. Editorials, articles
from controlled-circulation journals, and articles from supplements
were excluded.
From the 297 symposium articles and random sample of 100 journal
articles, we selected all articles that studied at least one health
effect of ETS (eg, epidemiologic or animal studies on the effect of ETS
on lung cancer, pulmonary function, cardiac disease; in vitro studies
of ETS carcinogenicity). We identified 114 articles (65 symposium and
49 peer-reviewed articles).
We have found previously that symposia on ETS contain a larger
proportion of review articles (41% reviews) than do the peer-reviewed
literature (10% reviews).[22] Literature reviews provide an
opportunity for negative conclusions, whether or not they reflect the
conclusions of the literature as a whole.[23] Therefore, we
hypothesized that review articles would contain a larger proportion of
negative conclusions than would original research articles. One of us
(L.A.B.) classified each selected article as original or review
according to definitions used by the National Library of Medicine.
Original articles were those that presented new data or new
meta-analyses. Review articles were defined as articles published after
examination of published material on a subject that provide an
examination of recent or current literature but not extensive critical
analyses or statistical syntheses of the literature (Lois Ann
Colaianni, written communication, November 26, 1993). Our
classification was checked against the electronic coding for articles
that were listed in MEDLINE, and agreement was 100%.
Our first measure of publication bias was to determine the number of
articles that reported statistically nonsignificant results. For each
article, two reviewers determined independently whether statistical
tests were reported for original data or meta-analyses. The quoting of
statistical tests from other articles in review articles was not
counted. If the results of statistical tests were reported, we
classified the article as "positive" if it reported
P<.05 (two tailed), confidence intervals excluding 1 for
odds ratios, or other statistics that are typically characterized as
"significant." The article was classified as "negative" if it
reported P >.05 (two tailed), confidence intervals
including 1 for odds ratios, etc. If multiple health outcomes were
measured and some of the reported statistics were significant and some
were not, then the article was classified as showing "mixed
significance." There was 100% agreement between reviewers.
Our second measure of publication bias was to determine the number of
articles that concluded that ETS is not a health risk. We tested the
hypothesis that the authors' conclusions regarding the health effects
of ETS might not correspond with the statistical results reported. For
each article, two reviewers determined independently whether the
authors' conclusion regarding the effects of ETS on health was a
"positive conclusion" or a "negative conclusion." One reviewer
was masked to the authors, source of the article, and funding of the
article. A "positive conclusion" was defined as one that stated
that there was an association between ETS exposure and some adverse
health outcome. A "negative conclusion" was defined as one that
stated that there was no association between ETS exposure and any
adverse health outcome. Agreement between reviewers was 92%, and
discrepancies were adjudicated by discussion of the masked article.
Identification of Unpublished Studies
We attempted to identify studies on the health effects of ETS
that have not been published by (1) contacting the US Tobacco Institute
(Washington, DC), (2) examining publications that cited unpublished
studies, and (3) examining the citations made by tobacco
industry-affiliated reviewers in response to the EPA draft risk
assessment. This identification process was intentionally biased in
favor of the interests of the tobacco industry.
The Tobacco Institute publishes several pamphlets that review the
scientific literature on ETS and conclude that ETS is not
harmful.[24]
[25]
[26]
[27]
[28]
Therefore, we contacted the Tobacco Institute
by telephone to see if they could provide for us information on
negative studies that have not been published. The Tobacco Institute
did not respond to our request for information.
We identified publications claiming that negative studies have
been excluded from evaluations of the health effects of ETS by
searching on the key phrases "publication bias" and "tobacco
smoke pollution" in the MEDLINE database and by reading tobacco
industry literature on ETS.[12-17] [29]
[30]
[31]
[32]
[33]
We examined these
publications to identify which unpublished studies, if any, they cited
to support the argument regarding publication bias.
During preparation of the EPA risk assessment of ETS, the draft
was open to technical comment from June 25 to October 1, 1990, and was
also reviewed by the EPA's Science Advisory Board.[34]
During this time, the EPA received 49 comments from tobacco
industry-affiliated individuals criticizing the conclusions of the
risk assessment.[21] One argument used by tobacco
industry-affiliated reviewers to criticize the draft risk assessment
was that negative studies were excluded from EPA's risk
assessment.[21] Both the EPA and tobacco industry-affiliated
commentors cited unpublished studies to support their conclusions
regarding the health effects of ETS.[21] Since the public
comment period offered the tobacco industry the opportunity to produce
unpublished studies that did not demonstrate an association between ETS
and health effects, we hypothesized that the tobacco
industry-affiliated reviewers would cite unpublished negative studies.
We counted the number of negative unpublished studies identified by
either the EPA or tobacco industry-affiliated reviewers.
Analyses
Proportions of statistically significant results and positive and
negative conclusions were compared by means of contingency table
analysis of symposium vs peer-reviewed articles.
RESULTS
Published Studies on the Effects of ETS on Health
Table 1 shows that symposium
articles on the health effects of ETS were more likely to be reviews than were
peer-reviewed articles.
None of the review articles described statistical tests used to pool
data and derive conclusions. A higher proportion of original,
peer-reviewed articles (96% [44/46]) than original, symposium
articles (54% [19/35]) on ETS reported the use of statistical tests
(chi2=30.5, df=1, P<.0001).
Table 2 shows that, of the articles
that did report
statistical tests, comparable proportions of peer-reviewed articles and
symposium articles reported statistically significant results.
Table 3 shows that a larger proportion of symposium
articles than journal articles had a negative conclusion, ie, they
concluded that exposure to ETS is not harmful. When only the
conclusions of original studies, regardless of whether they reported
statistical results, are considered, the conclusions of 20% (9/46) of
the peer-reviewed articles were negative, compared with 49% (17/35) of
the symposium articles (chi2=7.67, df=1,
P=.006). The proportion of review articles in symposia (27%
[8/30]) that concluded that ETS is a health hazard (a positive
conclusion) was lower than the proportion of peer-reviewed review
articles (100% [3/3]) that had a positive conclusion (Fisher's
Exact Test, P=.03).
Identification of Unpublished Studies
Of the 11 publications identified that discussed publication bias
and the literature on ETS, four cited unpublished
studies.[14,31-33] The remaining publications
discussed theoretical reasons for the existence of
publication bias but did not cite any unpublished
studies.[12,13,15-17,29,30] Woodward and
McMichael[32] conducted the most comprehensive search for
unpublished data on ETS because they contacted the tobacco industry and
surveyed investigators listed in the Directory of Ongoing Research in
Cancer Epidemiology regarding unpublished studies. They identified one
unpublished, negative study of ETS and bladder cancer and one
dissertation. Wells[31] cited unpublished data from two
studies--the unpublished data on men from the Humble et
al[35] and Brownson et al[36] studies. When Wells
added these two unpublished studies to a previously published analysis,
the relative risk of lung cancer increased from 1.5 to 1.7, suggesting
that inclusion of the unpublished studies did not invalidate the
previous conclusion that ETS increased the risk of lung cancer.
Kilpatrick,[14] in an article published in a symposium
proceeding,[14] added one unpublished study to a previously
published meta-analysis of ETS and lung cancer and found that the
relative risk decreased to less than one. The unpublished study
identified by Kilpatrick was the same dissertation identified by
Woodward and McMichael.[32,33] presented an
unsubstantiated claim that the American Cancer Society has never
published the results of its large study on ETS and heart disease
because the results are negative.
The EPA draft risk assessment contained 391 citations[34] of
which three were unpublished studies, two of these being
dissertations.[37]
[38]
[39]
Tobacco industry-affiliated reviewers
cited 1620 references.[21] Three references were the same
unpublished studies cited by the EPA.[37-39] In addition,
tobacco industry-affiliated reviewers cited six additional unpublished
studies on the health effects of ETS (T. Sterling, D. Sterling, J.
Weinkam, unpublished data).[40]
[41]
[42]
[43]
[44]
[45]
Four of the six negative unpublished studies were submitted for
publication in peer-reviewed journals.[40-43] By the time
this article was written, one of these four had been published in a
peer-reviewed journal and one had been published as a letter to the
editor.[46] [47] Two of the unpublished studies cited by the
tobacco industry-affiliated reviewers were
dissertations.[44,45]
In summary, we identified five unpublished negative studies, two of
which were dissertations[44,45] and none of which was cited
among the 501 references in the final draft of the EPA risk assessment
of ETS.[9]
COMMENT
Our results suggest that there is no publication bias against original
research that does not report statistically significant results on the
effects of ETS on health. Almost half of the original research
published in the peer-reviewed literature reported statistically
nonsignificant results or results of mixed significance. The lack of
publication bias against negative studies on ETS may not correspond to
the occurrence of publication bias as a whole because the great public
interest in ETS might favor publication of negative results on ETS.
Although similar proportions of both peer-reviewed journal
articles and symposium articles on the health effects of ETS report
statistically significant results, a larger proportion of journal
articles than symposium articles have negative conclusions, ie,
conclusions that ETS exposure is not harmful. Although 80% of the
journal articles had positive conclusions, only 57% of the studies
actually reported statistically significant results. This discrepancy
is primarily caused by the reporting of results of mixed significance
as being positive. In contrast, the symposia report results of mixed
significance as being negative. When health outcomes are studied,
drawing positive conclusions from results of mixed significance is
justified because an intervention (such as ETS exposure) can have
adverse health consequences when it affects one physiologic measure,
even if it does not affect every possible physiologic measure
associated with the health outcome.
In the symposia, the large proportion of articles with negative
conclusions primarily results from the inclusion of 22 review articles
and 19 original articles that did not report statistical tests.
Generating review articles might be one means of publishing negative
data on ETS, and such reviews may not be based on a thorough evaluation
of the peer-reviewed literature.[22] In addition, the
conclusions of reviews have been shown to be influenced by the biases
of the authors. [48] As none of the review articles identified
in this study reported the methods used for pooling data or performing
statistical tests, it was impossible to evaluate the selection criteria
for articles or the validity of the conclusions.
The finding that the symposia contain more review articles and
articles without statistical analyses concluding that ETS is not
harmful than do peer-reviewed journals supports the supposition that
sponsorship can influence the presentation of results as being positive
or negative. We found previously that six of 11 symposia on ETS were
sponsored by the tobacco industry.[22] An association has
been demonstrated between pharmaceutical-industry sponsorship for
research and positive conclusions that support the use of new
therapies.[49] Publication bias has been
defined as a "systematic tendency to publish any one type of result, be it
positive, negative, or null."[50] In the ETS symposia, there appears to be a bias toward the publication of negative
conclusions that are not supported by the available data.
It is hypothetically possible that statistically nonsignificant studies
on ETS have been submitted to peer-reviewed journals, rejected for
publication, and remain unpublished. However, exhaustive searches for
unpublished data by ourselves and others have produced only a few
dissertations and articles. Although industries strive to keep their
data proprietary,[51] they should make available negative
data if they are going to claim that publication bias invalidates
conclusions based on the medical literature. The failure of researchers
with negative findings to submit their papers to peer-reviewed journals
has been shown.[52] It has also been argued that relying on
statistical cutoffs, such as P<.05 or 95% confidence
intervals, is inappropriate and discourages authors from submitting
nonsignificant results.[53] [54] There is, however, no evidence
that these issues have prevented publication of negative studies on ETS.
The tobacco industry has used the possibility of publication bias
to argue that reviews of the literature on the health effects of ETS
are invalid[12-15,20,24,25,33] [55] [56] In the case of
Flue-Cured Tobacco Cooperative Stabilization Corp v EPA, the
tobacco industry claims that publication bias invalidates the EPA risk
assessment because negative studies have been excluded.[20]
Our findings suggest that negative studies, such as those published in
symposia, should be carefully scrutinized before they are included in
reviews or considered by the courts because the negative conclusions
may not be supported by any statistical analyses. The recent US Supreme
Court decision in Daubert v Merrell Dow Pharmaceuticals Inc
regarding the standards that govern the admissibility of scientific
evidence at a trial suggests guidelines regarding the consideration of
data for public health policies (New York Times. June 30,
1993:A8).[57] The court ruled that federal judges must ensure
that scientific evidence and testimony admitted in trials are "not
only relevant, but reliable." Judges were urged to consider
"whether the theory in question can be (and has been) tested, whether
it has been subjected to peer review and publication, its known or
potential error rate, and the existence and standards controlling its
operation, and whether it has attracted widespread acceptance within a
relevant scientific community."[57] For negative
unpublished data, we recommend that the courts, as well as all
scientists, assess whether the data have been submitted for publication
in a peer-reviewed journal, whether the article presents original
research, whether the article actually reports statistical tests, the
quality of the methods, and the source of funding for the research.
From the Institute for Health Policy Studies (Drs Bero, Glantz, and
Rennie) and Division of Cardiology (Dr Glantz), School of Medicine, and
Division of Clinical Pharmacy, School of Pharmacy (Dr Bero), University
of California-San Francisco.
Presented in part at the Second International Congress on Peer Review
in Biomedical Publication, Chicago, Ill, September 11, 1993.
This study was supported by funds provided by the Cigarette and
Tobacco Surtax Fund of the State of California through the
Tobacco-Related Disease Research Program of the University of
California under awards 2KT0072 (Dr Bero) and 1RT520 (Dr Glantz).
We thank Mildred Cho, PhD, for comments and Phillip Lollar for
administrative help.
Reprint requests to Institute for Health Policy Studies, University of
California-San Francisco, 1388 Sutter St, 11th Floor, San Francisco,
CA 94109 (Dr Bero).
References
1. Sterling TD. Publication decisions and their possible
effects on inferences drawn from tests of significance--or vice versa.
J Am Stat Assoc. 1959;54:30-34.
2. Dickersin K. The existence of publication bias and risk
factors for its occurrence. JAMA. 1990;263:1385-1389.
3. Easterbrook PJ, Berlin J, Copalan R, Matthews DR.
Publication bias in clinical research. Lancet.
1991;337:867-872.
4. Rennie D, Flanagin A. Publication bias: the triumph of
hope over experience. JAMA. 1992;267:411-412.
5. L'Abbe KA, Detsky AS, O'Rourke K. Meta-analysis in
clinical research. Ann Intern Med. 1987;107:224-233.
6. Cook DJ, Guyatt GH, Ryan G, et al. Should unpublished
data be included in meta-analyses? current convictions and
controversies. JAMA. 1993;269:2749-2753.
7. The Health Consequences of Involuntary Smoking: A
Report of the Surgeon General. Washington, DC: US Dept of Health
and Human Services; 1986:332. DHHS publication CDC 87-8398.
8. National Research Council. Environmental Tobacco
Smoke: Measuring Exposures and Assessing Health Effects.
Washington, DC: National Academy Press; 1986.
9. Respiratory Health Effects of Passive Smoking: Lung
Cancer and Other Disorders. Washington, DC: Indoor Air Division,
Office of Atmospheric and Indoor Air Programs, Office of Air and
Radiation, US Environmental Protection Agency; 1992.
10. Glantz SA, Parmley WW. Passive smoking and heart
disease: epidemiology, physiology and biochemistry.
Circulation. 1991;83:1-12.
11. Steenland K. Passive smoking and risk of heart disease.
JAMA. 1992;267:94-99.
12. Vandenbroucke JP. Passive smoking and lung cancer: a
publication bias? BMJ. 1988;296:391-392.
13. Mantel N. The passive smoking myth. In: Bieva CJ,
Courtois Y, Govaerts M, eds. The Present and Future of Indoor Air
Quality. Amsterdam, the Netherlands: Excerpta Medica;
1989:155-159.
14. Kilpatrick SJ. The epidemiology of environmental
tobacco smoke (ETS) and the weight of evidence argument. Int
Surg. 1992;77:131-133.
15. Sugita M, Kanamori M, Izuno T, Miyakawa M. Estimating a
summarized odds ratio whilst eliminating publication bias in
meta-analysis. Jpn J Clin Oncol. 1992;22:354-358.
16. Armitage AK. Environmental tobacco smoke and coronary
heart disease. J Smoking Rel Disord. 1993;4:27-36.
17. Caldwell ADS. Give a dog-end a bad name. J
Smoking Rel Disord. 1992;4:1-2.
18. Science Does Not Support Draft Documents
Undergoing Expert Review. Washington, DC: Tobacco Institute;
1990.
19. Huber GL, Brockie RE, Mahajan VK. Passive smoking: how great a hazard? Consumers Res. July 1991:10-15, 33-34.
20. Flue-Cured Tobacco Cooperative
Stabilization Corporation and the Council for Burley Tobacco, Inc, and Universial
Leaf Tobacco Company, Inc, and Philip Morris, Inc, and R.J. Reynolds
Tobacco Company and Gallins Vending Company v United States
Environmental Protection Agency and Carol Browner. 1993.
21. Bero LA, Glantz SG. Tobacco industry response to a risk assessment of environmental tobacco smoke. Tobacco Control.
1993;2:103-113.
22. Bero LA, Galbraith A, Rennie D. Sponsored symposia on environmental tobacco smoke. JAMA. 1994;271:612-617.
23. Chalmers TC, Frank CS, Reitman D. Minimizing the three stages of publication bias. JAMA. 1990;263:1392-1395.
24. Cigarette Smoke and the Nonsmoker.
Washington, DC: Tobacco Institute; 1984.
25. Tobacco Smoke and the Non-smoker: Scientific
Integrity at the Crossroads. Washington, DC: Tobacco Institute;
1986.
26. Smoking in the Workplace: Some
Considerations. Washington, DC: Tobacco Institute; undated.
27. Smokers' Rights in the Workplace.
Washington, DC: Tobacco Institute; undated.
28. Smoking Restrictions: The Hidden Threat to Public Health. Washington, DC: Tobacco Institute; undated.
29. McNamee R. Outcome of retrospective cohort studies and study size: a publication bias? Br J Ind Med. 1989;46:143.
30. Weetman DF. Is successful prevention newsworthy?
Indoor Environ. 1992;1:197-198.
31. Wells AJ. Passive smoking and lung cancer: a
publication bias? BMJ. 1988;296:1128.
32. Woodward A, McMichael AJ. Passive smoking and cancer risks: the nature and uses of epidemiological evidence. Eur J
Cancer. 1991;27:1472-1479.
33. Lee PN. Environmental Tobacco Smoke and
Mortality. Basel, Switzerland: S Karger AG; 1992.
34. Health Effects of Passive Smoking: Assessment of Lung Cancer in Adults and Respiratory Disorders in Children.
Washington, DC: Indoor Air Division, Office of Atmospheric and Indoor
Air Programs, Office of Air and Radiation, US Environmental Protection
Agency; 1990.
35. Humble CG, Samet JM, Pathak DR. Marriage to a smoker and lung cancer risk. Am J Public Health. 1987;77:598-602.
36. Brownson RC, Reif JS, Keefe TJ, Ferguson SW, Pritzl JA. Risk factors for adenocarcinoma of the lung. Am J Epidemiol.
1987;125:25-34.
37. Varela L. Assessment of the Association Between
Passive Smoking and Lung Cancer. New Haven, Conn: Yale University;
1987. Dissertation.
38. Svensson C, Pershagen G, Klominek J. Smoking and Passive Smoking in Relation to Lung Cancer in Women. Stockholm,
Sweden: Dept of Epidemiology, National Institute of Environmental
Medicine; 1988.
39. Lam WK. A Clinical and Epidemiological Study
of Carcinoma of Lung in Hong Kong. Hong Kong: University of Hong
Kong; 1985. Dissertation.
40. Sterling T, Sterling D, Weinkam J. Comparison of risk of chronic conditions and cancer between homemakers and otherwise
unemployed women. Submitted for publication.
41. Sterling TD, Rosenbaum WL, Weinkam JJ. Bias in the diagnosis of lung cancer and its possible consequences in establishing
smoking related risks. Submitted for publication.
42. Feuer G, Ecobichon DJ. Assessment of health effects of environmental tobacco smoke. Submitted for publication.
43. Oldaker GB. ETS in passenger cabins of commercial
aircraft, offices and restaurants: a review of the literature.
Submitted for publication.
44. Wu AHT. A Case-Control Study of Lung Cancer
in Women. Institution unknown; 1983. Dissertation.
45. Butler TL. The relationship of passive smoking to
various health outcomes among Seventh Day Adventists in California. Los
Angeles: University of California; 1988. Dissertation.
46. Sterling TD, Rosenbaum WL, Weinkam JJ. Bias in the attribution of lung cancer as a cause of death and its possible
consequences for calculating smoking-related risks.
Epidemiology. 1992;3:11-16.
47. Oldaker GB. ETS in passenger cabins of commercial
aircraft. J Natl Cancer Inst. 1989;81:1424-1426.
48. Chalmers TC. Informed consent, clinical research and the practice of medicine. Trans Am Clin Climatol Assoc.
1982;8:388-391.
49. Davidson RA. Source of funding and outcome of clinical trials. J Gen Intern Med. 1986;1:155-158.
50. Szklo M. Issues in publication and interpretation of
research findings. J Clin Epidemiol. 1991;44(suppl 1):109S-113S.
51. Catanoso J, Wireback T. What did Reynolds know?
News Rec. September 26, 1992:A1-A2.
52. Dickersin K, Min Y-I, Meinert CL. Factors influencing
publication of research results: follow-up of applications submitted to
two institutional review boards. JAMA. 1992;267:374-378.
53. Rothman KJ. Significance questing. Ann Intern
Med. 1986;105:445-447.
54. Pocock SJ, Hughes MD. Estimation issues in
clinical trials and overviews. Stat Med. 1990;9:657-671.
55. Mailing to the Tobacco Action Network.
Washington, DC: Tobacco Institute; 1990.
56. Environmental Tobacco Smoke and Health: The
Consensus. Washington, DC: Tobacco Institute; undated.
57. Daubert v Merrell Dow Pharmaceuticals Inc. US Supreme Court. 1993:No. 92-102.
Table of Contents
