Tobacco Documents Online

ETS AND L UNG CAIVCER= SCOfiH RL VIEW 1994 (Volume 1) ~ ~ O ~ IV N C.j In, formartion provided by the W TOBLUXA) Manufacturers Association

ETS AIYD L LTNG CANCER - SCnTH REVTEW SECTION 1: Introduction. SECTIONI2: ETS and Lung,Cancer Epidenniological Database SECTIONI3: Substantive Published Literature related to Section 2. SECTIONI4: Studies/Analyses not considered in relation to The Summary Review of the Epidemiolbg,y (Section 2) . SECTION 5: Meta-Analysis and its Applications SECTION 6: Meta-Analysis of Lung Cancer Database (Section.2) SECTION 7: Confounding Variables possibly relevant to ETS and Lung Cancer: Examples: . Nutrition . Air Pollution . Life Style & Family Hist+vry SECTION 8: Mfsclassification. SECTION 9: Recent papers oni the Diagnosis of Lung Cancer SECTION 10: Epidemiolb.gy of Weak Associations SECTION 11: EPA Report - Critiques SECTION 12: An Overview of the Epidemiology

S~CTIO'1T7 Introduction

INTRODUCTION The material being submitted by the T'MA for this review follows the : request of the SCOTH Secretariat - namely concentrating on peer-reviewed'science, published since the Fourth~Iteport of 'the ISCSH (1988) or not taken into account in their report. The first two volumes of this submission focus on epidemiological studies and issues relatRng,directly, to the interpretation of such studies. Section 2 is a listing of 37 epidemiologiical studies relevant to ETS and lung eancer. The listing identifies each study by author, date, location and type. The number of lung cancer cases is shown for each study, together with relative risk and 95% confidence interval. Confounding factors claimed to have been~ considered are also shown. Section 3 comprises copies of each of the 37' stu.dies in chronological order andd referenced according to Section 2. Section 4 comprises a list of a further 15 epidemiological studies on ETS and lung, cancer considered by the statistician Mr P N Lee to be inadequate for the technical reasons statedi therein. These have also been excluded from other reviews, such as the EPA report. Section 5 provides some general (peer -reviewed) papers on meta-analysis and its application to ETS and lung eancer. Alithough~ there are considerable reservations about! the use of meta-analysis for combining, widely differing epidemiological studies, the technique has beenn frequently applied to the ETS and lung cancer epid'erniology. Section 6therefore : provides some meta-analysis calculations of various groupings: of the 37 studies, carried out by Mr Lee outlining the assumptions made. The analysis of spousall studies, published since the 1'988 ISCSH Fourthi Report, of the workplace studies, off exposure in childhood, and a footnote on social exposure are included.. A re-calculation of the EPA's meta-analysis was conducted by LeVois and Layard. This includes two studies which the EPA chose to exclude. It follows the procedure used by the EPA, but! unlike the EPA uses conventional l confidence : intervals. It i& included~ in Section 11, but might be consideredl alongside Section 6. Many of the published epidemiologieal studies, especially the earlier ones, made littlee or no al'lowance for confounding factors. Other researchers have tried to account for such~ factors: Section 7 includes & selection of some of the papers available on somee of the possible co ounding, factors, such: as diet, air pollution, family history and~ lifee style. It should be noted'. that many other suchi papers cani be provided, and that' bias due to confounding is especially important where weak associations are involved.

Among other issues which need to be considered in determining the value of these epidemiologiical studies, or any analysis of such studies, are misclassification andd misdiagnosis of lung cancer. Some papers covering these topics are included as Sections 8 and 9 respectively. There have been severaU papers produced on the epidemiology of weak associations, three of which are included as ~ Sectiion~ 10. Section 11 contains a series of articles critical of the EFA report. Most of these have not beenipeer-reviewed, but in effect they form an informal peer review of the EPA's process and' conclusions. We believe these deserve serious consideration because the EPA report admittedly used unconventional' statistical techniqves to, arrive at its conclusions. This section contains the re-calculation of the EPA's meta}analysis referred to earlier. Our own overview of the material submitted on the epidemiology is that although some studies have reporked' a statistical association, this can readily be explained by several factors crucial to low risk epidemiology, but not adequately considered. An article currently "in press", though not peer-reviewed, summarises most of the issues and is included as Section 12. This section also includes a note of our views on the problems of interpreting epidemiological studies. The second phase of our submissioN including such subjects as the chemistry and physics of ETS compared withi mainstream smoke, assessments of'exposure to ETS and problems with qpestionnaire-based studies will, we believe, support our views on the epidemiology,

SECT`ION 2 AND L UN~ EpidefnioZogical Database

TIZE'DATA1 Spousal smoking has been~ the prime index used' for the study of the possible effects of ETS. The : apparent reason is that this was alf that, was investigated in the early days. Three early studies, Hfrayama, Trichopoulo& and Garfinkel I all considered only this index. Two of these studies were prospective studies which interviewed all members of the household (above a given age). They did, not ask women if their husbands smoked - they merely related risk ini women to smoking, habits reported by the husband. Since that time most ETS' studies: have actually enquired about spousal smoking. Others have asked about exposure fromi other household members. Some studies have also reported results on workplace and chil hood exposure. It' is notable that the studies on childhood and workplace exposure, which seem to have attracted less attention from reviewers than the data on spousall smoking, show no real indication of any increase in risk. The tables and figures that follow summarise lhe key evidence in relation to smoking by the husband, smoking by the wife, ETS exposure in the ! workplace, and ETS exposure in childhood. The tables show, for each successive study, the individual relative risk estimates and 95% confidence limits, which are also plotted graphically in the figures. In the figures, the circles indicate the relative risk estimates and the confidence limits are indicated by the error bars. The width of the error bars is proportional to the inverse of the variance of the relative : risk estimate, which is closely related l to the :number of lung cancers studied. By this means of presentation, large studies, which contribute more to the overall evidence, are seen more clearly than smalill studies. The tables and figures are based on results from a total of 37 studies (1-37). Section 4 explains why results from certain other studies, which might have been thought report relevant data, are not included in the tables andl fagures!

G H- n I C o' M I z A W N.W+ O'O 00 J p. U A W N.N. O~ 00 J 6~ M A W N~-.• ~Q 10 Oe ~l O~ lA A W N.+ ~§+. ~~ ~'F c N9a32f ~ ~ •'~6~+ eR --------~~--~--------------~~----- N , Vi (Il ~ ~,'1 Vl N~?S '~y 8 Vl ~ s 8 Vl -, i/N 'n' V1 N N Vl ~ N sN > Y> ij{f »j~{ a9w~ '>CCC6660sOS > > ~ »>'~ > >ay > § 88 V.+~ O U tvn owe O N -A N.+ ~n V ~ y ~+ W w ~: ~O O ~p ~D . ~p N A A1 ir O O F, O 9 2 N AtO A A O•A•• :j N A W - .. ~ .. .. - .. .. ;ooq~o3~og~a~oo~ 0'3 .- o {,~ xd~r. -0~r~~3sW~' T ~ N 7 ..$I w Pfl vw Maa19 sr 5 arR rw g w sc I a- "SE "g I sI § R I I K I I I 5 I I

Figure 1- LUNG CANCER AND HUSBAND'S SMOKING 1 r- 0 RR 0.50 0.20 0. 101 1 1 u 1 1 1 1 1 111 1 111 1 1 1 1 1 11 1 1i 1 1 - i 2 4 6 7 8 9 12 15A 18 21 22 25 27 28 29 30 32 33 36 37 3 5 10 13 16 19 23 26 31 34 11 14 17 20 24 35

TABLE 2 Relative risk of ltme cancer amOna'lifelbne nonsmokinR men in relmion to smoking by the wife Study Number of bmg Reibtive risk (95% eonfidence: Signi- Ref Author Year L,oation Type oaooers limits fimnce Nates 3 Coaea 1983 UUSA E 1.97(0.3E-10.32) 0c 5 Buf(kr 1984 UUSA , cc 11 O:S1(0.14-1.79) orc 6 Hirayama 1984 Japan P' 64 213(1.051.76) + al 7' Kabot' 1 1994 UUSA cc 12 1.00(0.245:07) orc 10' Aluba 1936 Japan cc 19 1.g0(0.40-7.00) bi 11 L.ee 1986 UK cc 15 1.30(0:384.39) a 26 Kabat',2' 1990 UUSA cc 36 1.20(0.54-267) urc 31 Jneckel' 19911 Germany 268(0.59-12.36) urc. LMiex of expoam based on ever smoking by the spouse or if, data for that , index atr not milable, the neatesr equivalent, ag. livingg with'a'smokec Study aathor is name of first author in publinti bn from whidh data extracted, see tefetenees. The study'year is the year of that' publianioa Stuey type CC oase co.nhol P prospentiive. Si2nificanoe + statistically signiticant'inoreace at 95% confidence level Ndtt a aajusted'.far age; b adjutted'for age and other risk factors; u unadjisted;, r relative risk calwlated'from tabular data pteseated; c confidence limits nlauloted from tabular data prraenterJ: 1 95% limits estimated from 90'96' limits presented; Two stud'ies (23,23) provided suitable data only far sexes combined. As'ttie great majo[ity, of'the cases waultl'have been ~ women. results have been induded in Table I and Figure 1 and'are not'ocasideted bae. One study (14) providad suitable data only for females and for sexes combined. Again the sexes'.eombined result (telative risk 2:6„95% limits 1.0y6:5, naes til) is nd', aarsiduad I in the main body of this table or in Figote 2