Tobacco Documents Online

LUNG CANCER AND EXPOSURE TO ENVIRONMENTAL TOBACCO SMOKE APPENDIX 1 REVIEW OF INDIVIDUAL STUDIES R.L.T4'JEEDIE, PHD DSC K.L. MENGERSEN, PHD BOND UNIVERSITY

1 SUMMARY OF ACTIVITTES AND INTRODUCTION...... 1 2 COHORT STUDIES ........................................................... 4 2.1 HIRAYAMA (1981-1984) ...........................................:...................................4 1 Hypothesis Testing .................................................................................4 2 Representativeness of the Data ...........................................................4 3 Bias and Misclassification ....................................................................5 4 Indicator of ETS ...................................................................................... 6 5 Confounding Factors ............................................................................. 6 6 Data Dredging .......................................................................................... 7 7 Errors .................................................................................................... .....7 8. Statistical Methodology .....:...............................................:.................. 8 2.2 ...................................................................... GILLIS et al (1984) ............:...... 11 .3 1 Hypothesis Testing ...............................................................................11 2 Representativeness of the Data .......................................................••11 3 Bias and Misciassif'icatior ....................................................................11 4 Indicator of ETS ....................................................................................11 5 Confounding Factors ...........................................................................11 GARFIIVKEL (1980,1981) ............................................................................12 3 1 Hypothesis Testing ...............................................................................12 5 Confounding Factors .............:.............................................................12 CASE CONTROL STUDIES ............................................ 13 3.1 P.IQBA et al (1986) ........................................................................................13 1 Hypothesis Testing ...............................................................................13 2 Representativeness of the Data .........................................................13 3 Bias and Misdassification ...................................................................13 4 Indicator of ETS ....................................................................................14 5 Confounding Factors......... ._....__ ......................................................14 6 Data Dredging ............................ .......................................................... 14 8 Statistical Methodology .......................................................................14 3.2 BROWNSON et al (1987) ............................................................................15 1 Hypothesis Testing ...............................................................................15 2 Representativeness of the Data ..................................................:......15 3 Bias and Misclassification ...................................................................15 4 Indicator of ETS ....................................................................................15 5 Confounding Factors ...........................................................................15 6 Data Dredging ........................................................................................15 8 Statis tical Methodology .......................................................................16 ii

3.3 CHAN and FUNG (1982) .................................................:..........................17 1 Hypothesis Testing ....................................:.......................................:..17 2 Representativeness of the Data .........................................................17 3 Bias and Misclassif3cation ...................................................................17 4 Indicator of E'I5 .................. ................................................................... ..17 5 Confounding Factors ...........................................................................17 3.4 CORREA et al (1983,1984) ...........................................................................18 1 Hypothesis Testing ...............................................................................18 2 Representativeness of the Data .........................................................18 3 Bias and Misclassification ................... ........ ........ .-........................... 18 4 Indicator of ETS ....................................................................................18 5 Confounding Factors ...........................................................................18 8 Statistical Methodology .......................................................................19 3.5 DALAGER et al (1986) ..................................................................................20 1 Hypothesis Testing ............................................................... ................ 20 2 Representativeness of the Data ..............................................••••••.••••20 3 Bias and Misclassification ...................................................................20 4 Indicator of ETS ......................:........................:.................................... 21 5 Confounding Factars..... ...... _ ..............................................................21 6 Data Dredging ........................................................................................ 21 7 Errors ......................................................... .....:..... ......... ... : ...................... 21 8 Statistical Methodology ....................................................................... 21 3.6 GAO et al (1987) .........................................................e...................................22 1 Hypothesis Testing............................................................................... 22 2 Representativeness of the Data .........................................................22 3 Bias and Misclassification ...................................................................22 4 Indicator of Exposure to ETS ..............................................................23 5 Confounding Factors ........................................................................... 23 6 Data Dredging ........................................................................................ 23 8 Statistical Methodology and Results ................................................23 3.7 GARFINKELet al (1985) ..............................................................................25 1 Hypothesis Testing ............................................................................... 25 3 Bias and Misclassification ...................................................................25 4 Indicator of ETS ................................:................................................... 25 6 Data Dredging ........................................................................................ 26 8 Statistical Methodology ....................................................................... 26 iii

3.8 GENG et al (1988) ..........................................................................................27 1 Hypothesis Testing ...............................................................................27 2 Representativeness of the Data ..................................................••.••••27 3 Bias and Misdassification ...............::..................................................28 4 Indicator of Exposure to ETS ..............................................................28 5 Confounding Factors ........................:.................................................. 28 7 Errors ................................................................................ ..................................... 28 8 Statistical Methodology and Results .................................:..............28 3.9 HUMBLE et ai (1987) ...............................................:....................................29 1 Hypothesis Testing ............................................................................... 29 2 Representativeness of the Data .........................................................29 3 Bias and Misclassification ...................................................................29 4 Indicator of ETS .................................................... ................:............... 29 5 " Confounding Factors ........................................................................... 29 6 Data Dredging ....:.....................................................:............................. 30 8 Statistical Methodology ....................................................................... 30 3.10 INOUE and HIIZAYAMA (1988) ........................:............:..........................31 1 Hypothesis Testing ............................................................................... 31 2 Representativeness of the Data ................................................ ......... 31 3 Bias and Misclassification........... _._ ...................................................32 4 Indicator of Exposure to ETS .............................................................. 32 5 Confounding Factors ........................................................................... 32 8 Statistical Methodology and ?.esuits ................................................32 3.11 KABAT and WYNDER (1984) ...................................................................34 3 Bias and Iviisdassification ...................................................................34 4 Indicator of ETS .................................................................................... 34 5 Confoun ding Factors ........................................................................... 34 6 Data Dredging ........................................................................................34 3.12 KATADA et al (1988) ...................................................................................35 1 Hypothesis Testing ............................................................................... 35 2 Representativeness of the Data .........................................................35 3 Bias and Misdassification ...................................................................35 4 Indicator of Exposure to ETS ..............................................................36 5 Confounding Factors .............................................................:............. 36 8 Statistical Methodology and Results ........................................ iv

3.13 KOO et al (1983,1984,1985,1987) ...............................................................37 ................................. 37 - 3 Bias and Misdassification .................................. 4 Indicator of ET5 ...........................:...........................................:............ 37 5 Confounding Factors ...........................................................................37 6 Data Dredging ....:........:........................................................................... 37 8 Statistical Methodology ...................................................::................... 37 3.14 LAM et al (9187) ............................................................................................ 39 2 Representativeness of the Data .........................................................39 3 Bias and Misclassification ..............................:....................................39 4 Indicator of Exposure to ETS ..............................................................39 5 Confounding Faztors ............................ ...... ....................................... 39 8 Statistical Methodoiogy ....................................................................... 40 3.15 LEE et al (1986) ...............................................................................................41 _4 Indicator of ETS .................................................................................... 41 8 Statistical Methodology ........................................................................41 3.16 MIL.LER (1983) ............... ................................................................................. ................42 3 Bias and Misclassification ...................................................................42 4 Indicator of ETS .................................................................................... 42 5 Confounding Factors .........................:.................................................42 7 F.rrors.. _..... _ ......................................................... _................................. 42 3.17 PERSHAGE.N et al (1987) ............................................................................43 2 Representativeness of the Data .....................................................:..43 3 Bias and Misdassification ...................................................................43 4. Indicator of ETS ...................................................................................43 5 Confounding Factors ....................................................................... .... 44 6 Statistical Methodology ..................... .................................................. 44 3.18 SANDLER et al (1985) ........................................................................:.........45 3 Bias and Misclassif•ication ...................................................................45 4 Indicator of ETS ...:..............................:.................................................45 ............................ 45 5 Confounding Factors ............................................... 6 Data Dredging ............................................... _....................................... 45 8 Statistical Methodology .......................................................................46 v

3.19 SHIIvIIZt.J et al (1988) ....................................................................................47 1 Hypothesis Testing ............................. .................................................. 47 2 Representativeness of the Data ........................................... .............. 47 3 Bias and Misclassification ......................................:............................47 4 Indicator of Exposure to ETS ...............................................................48 5 Confounding Factors ........................................................................... 48 6 Data Dredging ........................................................................................48 8 Statistical Methodology and Results .......:........................................48 3.20 SVENSSON et al (1989) ..........................................................................:....49 1 Hypothesis Testing ...............................................................................49 2 Representativeness of the Data .................:..................••.--•.••••••-•••••.•49 3 Bias and Misclassificat:on ...................................................................49 4 Indicator of Exposure to ETS ..............................................................50 5 Confounding Factors ........................................................................... 50 6 Data Dredging ........................................................................................ 50 8 Statistical Methodology and Results ................................................51 3.21 TRICHOPOUTAS et al (1981,1983) ............................................................52 2 Representativeness of the Data ........:................................................52 3 Bias and Misclassification ...................................................................52 4- Indicator of ETS ....................................................................................53 5 Confounding Factors ........................................................................... 53 6 Data Dredging ........................................................................................ 53 7 Errors ....................................................:.... .............................. .. .... .... .. .... 53 8 Statis tical Methodology ....................................................................... 53 3.22 VARELA (1987) ............................................................................................. 55 2 Representativeness of the Data ...................................................•.....55 3 Bias and Misclassification ...................................................................56 4 Indicator of ETS .................................................................................... 56 5 Confounding Factors .......................:.........................:......................... 56 6 Data Dredging ........................................................................................ 57 7 Errors and_ Omissions ..........................................................................57 8 Statistical Methodology ....................................................................... 57 3.23 WU et al (1985) ..............................................................................................58 2 Representativeness of the Data .........................................................58 3 Bias and Misclassification ...................................................................58 4 Indicator of ETS .................................................................................... 58 5 Confounding Factors ...........................................................................58 3 BIBLIOGRAPHY ............................................................... 59 vi

t 1 SUIViMARY_OF ACTIVITIES AND INTRODUCTION We have undertaken a detailed study of the statistical analysis carried out on the papers on Exposure to ETS and Lung Cancer -in the attached bibliography. The papers under consideration include three cohort studies and 22 case-control studies. Our study-involved the following i Checking (recalculating) the odds ratios and associated confidence intervals, by the methods used by the respective authors. This was done for all papers, wherever the published data were sufficient for it to be possible. ii Calculating the confidence intervals for odds ratios using the Cornfield- Method, where possible. This method yields confidence intervals which are-considered to be more accurate in general than other approximate methods (see Breslow and Day [1980] pages 133-4). Many existing studies use simpler but less accurate methods. iii Examination and cross-checking of data tables for consistency. This revealed for example, discrepancies in Hirayama's reported data from 1981 to 1984, and incorrect confidence limits in Trichopoulos's (1983) reported study. i v A consideration of the appropriateness of a the'statistical methods employed with regard to the data collected, their representativeness, subdivision into subgroups, the parameters measured and other areas needed for accurate study Ail these activities are discussed in the following sections. Before presenting the details of our study of the individual papers some general comments are in order. The most widely discussed work is that by T. Hirayama, which includes a number of papers and short articles or letters on data collected in a cohort study over the 16 year period 1965 to 1981. Of all the case-control studies carried out, that by Trichopoulos et al (1981, 1983) has received widest comment and very much more than other case-control studies. - It is important to note that the works of Hirayama and Trichopoulos sparked interest and debate in the area of exposure to CTS and lung cancer but studies by others since then have largely failed to produce significant results, especially outside East Asia. As noted in Section 1 above, there are some general problems such as misdassification,- confounding, interviewee honesty, interviewer bias, definition and measurement of exposure to ETS and the representativeness of the data which are common to many of the papers under consideration. 1

Hence it is worth defining and discussing these concepts before the detailed analysis of individual papers, in order to avoid repetitious comments. - a Misclassification: this occurs when subjects, people_in a study, are incorrectly classified. In particular, for this study, misdassifying smokers as nonsmokers and cancers as non- cancers, is capable of causing error. - -, It is generally thought that misclassification in the present context is most important through smokers claiming to be nonsmokers. This could be substantial in societies where smoking by women is not an accepted custom, hence leading to women who smoke claiming to be nonsmokers. - The effect of misclassification on the results of a study have been researched by a number- of authors, most notably Lee (1986), Ahborn and Uberla- (1988) and Wald et al (1986). Section 4 of this- report covers this issue in some detail. -- b Confounding: this occurs -when sources or factors other than exposure to ETS and random variation cannot be eliminated as affecting the result or outcome. Some uncontrolled factors common in many of these studies are pollution in the workplace (occupation), family history, diet, type of medical care available, and lifestyle. - - c Interviewee Honesty: as mentioned under misclassification, subjects may intentionally incorrectly state their smoking habits due to family, peer or society pressure, or a relative who is supplying the information may be unsure or ignorant of the facts. - - - - d Interviewer - Bias: an interviewer can intentionally or unintentionally influence a subject's response to a question by the manner in which the question is posed, or can even lead a subject to a particular response. In many of the studies considered here the data are collected through interview. Also a questionnaire can be constructed so as to lead the respondent in a particular way. e Definition and Measurement of Exposure to ETS: the definition of a person as being exposed to E-TS (a "passive smoker"), in all studies is someone who is a nonsmoker married to a smoker. Also the smoking habit of the spouse of a nonsmoker was used as a measure of the exposure and degree of exposure to ETS by the nonsmoker. - The average number of cigarettes smoked per day by the spouse of a nonsmoker was the most commonly-used measure of the degree of exposure to ETS, although other measures such as pack years, or hours per day exposed were occasionally used. Clearly this is neither a concise nor reliable method of measuring exposure to ETS. It is open to distortion from interviewer and interviewee bias. 2

Friedman,- Petitti and Bawol (1983) reported that 40 to 50% of - persons with nonsmoking spouses reported some exposure to ETS while 30 to 35% of those married to smokers reported no exposure. f Representative Data: if the conclusions of a study are to be extrapolated to the general population then a cohort or prospective study should be as representative as possible and constructed for the purpose of testing some particular hypothesis. Case-control studies must adhere to the -same constraints mentioned above, but particular care must be taken to ensure that the controls are representative. Great care needs to be taken with regard to misclassification, confounding effects and so on as mentioned earlier which may give rise to unrepresentative data and hence results-which do not apply to - the general population. Many of the areas of potential error are present in all studies we have investigated. - - We group, systematically, our comments into 8 categories: 1 Hypothesis Testing - 2 Representativeness of the Data 3 Bias and Misclassification 4 Indicator of Environmental Tobacco Smoke (ETS) 5 -Confounding Factors - 6 Data Dredging 7 Errors - 8 Statistical Methodology. In some cases, where studies are small or have little background published, we only comment on the major area of concern. Omission of some areas does not necessarily reflect endorsement of the study in such an area: for example, we comment only briefly on Lee et al (1986), not because it is completely acceptable, but because it gives negative results and so provides little support for an association between exposure to ETS and lung cancer. This reflects our goal, which is to provide an overview of the strength and validity of results which do tend to provide such support. 3

COHORT STUDIES 2.1 HIRAYAMA (1981-1984) The results of this study have been published in 4 reports (Hirayama, 1981, 1983, 1984a, 1984b).- The 1981 publication contains data from a prospective study from 1965 to 1979 while the 1983 and 1984 publications have data from 1965 to 1981 for the same study. Altogether, married, nonsmoking women aged 40 and above (n=91540) from a cohort of 265118 adults_ in 29 Health Centre Districts in Japan were followed for-16 years. These studies -reported a statistically significant increase in risk associated with exposure to ETS. Overall, however, we found the study to have major flaws, principally with respect to inconsistencies in the data, errors in the calculations, and defects in the design and conduct of the study. The extent of these flaws is sufficient to cast serious doubt upon the validity of the conclusions drawn by the author. These studies have been debated by many authors and much of what we present below is noted elsewhere also. In this critique of Hirayama's work we have (because of its importance and size) given comments on methodological problems at a level of detail not given elsewhere; this avoids repetition, but we do note that we could have spent an equal amount of space on such criticisms of other studies. 1 Hypothesis Testing The data were obtained from a study not specifically designed to investigate whether there are links between exposure to ETS and lung cancer. The results can, therefore, be used only to generate such a hypothesis; they cannot be used to prove or disprove it. There are a wide number of hypotheses put forward in this study which make its validity for scientific proof suspect. - 2 itepresentativeness of the Data - 71 here are two questions to be considered here: • Is the sample of cases representative of the population to whom the result is intended to be extrapolated? • Is the sample of controls representative of the population from whom the cases are drawn in order for the analysis of relative risks to be valid? - The first concern is real when considering Hirayama's study. The 29 health centres used in the study were not randomly chosen. MacDonald (British Medical Journal, 3rd October 1981, pp.915,916; 28th November 1981, p.1465) raises some concerns over this, pointing out that no reason or methodology was given by Hirayama for the selection of these particular centres. Hirayama merely states that "the satisfactory representativeness of the sample was confirmed after the survey" (British Medical Journal, 3 October 4