ENVIRONMENTAL TOBACCO SMOKE (ETS) AND RISK OF LUNG CANCER- HOW CONVINCING IS THE EVIDENCE? Alan W Katzenstein 40 (_n CD CD cz %10 %10 BATCo document for Legal Services : Health Canada 19 October 1999 HOW CONVINCING IS THE EVIDENCE? Does exposure to environmental tobacco smoke (ETS) in- crease the non-smoker's risk of developing lung cancer? The scientific evidence does not support such a claim, even though press releases and statements tit the popular inedia have tended to create that impression. THE NATURE OF THE EVIDENCE Despite statements that nonsmokers exposed to EIS in- cur increased risk of lung cancer, there appears to be not a single scientific study that has assessed the risk to non- smokers by actual measurement of exposure to ETS in the home, in the workplace or in public. The 1986 report of the National Academy of Sciences on ETS com- mented: "The studies do not directly address chronic health effects in individuals who are exposed at work or have occasional exposures in the home or elsewhere." Two studies have attempted to estimate total exposure through questionnaires, but neither study found a statis- tically significant increased risk of lung cancer on the basis of total ETS exposure. Except for the two studies noted above, all the studies listed by Surgeon General C. E. Koop in Congressional testimony in 1985 and 1986 and the studies cited in the 1986 reports of the Office of Technology Assessment (OTA), the National Academv of Sciences (NAS), and the Surgeon General pertain only to the risk of lung cancer for nonsmokers married to smokers in relation to the risk for nonsmokers with nonsmoking spouses. In other words, the risks have been estimated primarily in rela- tion to the spouses' smoking habits. THE QUALITY OF THE EVIDENCE All of the reports in the literature have been reviewed by independent investigators, as is customary and appropri- ate in the scientific process. Most of the studies have been criticized for one or more significant flaws in meth- odology, such as too few subjects to permit statistically reliable findings, failure to provide appropriate controls Ln r-D CD 0__ \.O G_*1 r\J BATCo document for Legal Services: Health Canada 19 October 1999 against which comparisons can be made, failure to verify the origin of the primary cancer in all subjects, and failure to account for life style factors that might influence the results. Numerous reviewers have commented on the possibility of inaccurate classification of smokers and of spouses in terms of their smoking habits, which could result in sig- nificant errors in the risk statistics. The NAS report comments at length on this weakness in the evidence, going so far as to revise its calculations downward to ac- count for possible bias or misclassification. ,U of the reports said to find increased risk for the ex- posed nonsmoker have been found to be significantly flawed and have been widely criticized. In 1985, the International Agency for Research on Can- cer reviewed the quality of the evidence in the eight re- ports then available and concluded that "each is compati- ble either with an increase or with an absence of risk." In other words, none established increased risk with anv certainty. The 1986 report of the Office of Technology Assessment noted that "al] the lung cancer studies have some meth- odologic weaknesses." Hghlights of the critical comments by various reviewers of the studies listed in Surgeon General Koop's congres- sional testimony of June, 1986, are summarized in Ap- pendix 2. THE QUANTITY OF THE EVIDENCE More than 30 reports covering 19 separate studies have been listed in the various reviews and testimony on the risk of lung cancer in relation to ETS. 9 2 studies have been so severely criticized for inap- propriateness that they were excluded from the data base in the OTA study and were not even fisted in the NAS report. 0 3 studies were mentioned and discounted by NAS because of technical weaknesses. 0 13 studies survived critical assessment by NAS, 12 of which are included also in the Surgeon General's Report. (Jn C) (Z CC) "0 t- l'O 011% BATCo document for Legal Services: Health Canada 19 October 1999 The Surgeon General's Report inexplicably fails. to cite one of the studies in the NAS report but includes another that was excluded by NAS because raw data were not provided to permit critical assessment. Considered by the populations studied: 0 6 studies were based on US. subjects. 0 5 studies were based on European subjects. 0 3 studies were based on Asian subjects, including one study based on atomic bomb survivors now aver- aging more than 70 years of age. The 13 studies fisted by NAS cover a total of 677 cancer subjects, with studies ranging from as few as 10 subjects to as many as 146. OVERALL RISK ESTIMATES NAS concluded, "the risk of lung cancer is roughly 3017o higher for nonsmoking spouses of smokers than it is for nonsmoking spouses of nonsmokers." Then, recognizing the likelihood of bias or misclassification of subjects, NAS lowered the overall estimate to about 25% higher risk for exposed nonsmokers. However, the lack of statistically significant risk findings for all but two of the studies, along with the flaws identi- fied in each of the individual studies, as summarized in Appendix 2. makes it highly questionable whether an in- crease in overall risk has really been established. NAS further noted that "the estimate of the increased risk from the American studies is lower than the average for all the studies. . RISKS FOR AMERICANS When considering only the studies of American subjects, NAS concluded that the risk for nonsmokers with a spouse who smokes is only 14% higher than the risk for a nonsmoker with a nonsmoking spouse. (There is no indi- cation whether NAS made adjustmentsforpossible bias or misclass~fication in theAmerican studies.) However, indi- vidually and as a group, the risk levels from these studies ~_n CD C) OD .10 Z~_. BATCo document for Legal Services : Health Canada 19 October 1999 are not statistically significant; that is, statisticians con- sider the results might well arise by chance alone. Of the 5 studies of American subjects included in the NAS report, 3 provide data separately for males and fe- males, yielding 8 data sets. (See Table 1.) 0 In 4 of those data sets, the risk for nonsmokers is increased. 0 In 1 set, there is neither increased nor decreased risk. 0 In 3 sets, the risk is decreased. Tl~iis 4-1-3 pattern is remarkably consistent with what statisticians would expect from a series of studies where no real difference in risk actually exists. The study excluded from the NAS Est but included in the Surgeon General's report similarly shows an increased risk that is not statistically significant. CONCLUSION A total of 14 studies on lung cancer risk for the non- smoker in relation to spousal smoking survive critical as- sessment despite their technical flaws. Of the 6 studies based on American subjects, none shows a statistically significant increase in risk for the nonsmoker with a spouse who smokes. Based on the evidence to date, the concern about the risk of lung cancer for nonsmoking Americans appears to be overstated and unsupported. Katzenstein Associates Larchmont, NY 10538 March 1987 Ln CD CD Co -10 ON ~_n BATCo document for Legal Services : Health Canada 19 October 1999 Table I LUNG CANCER RISK ESTIMATES AND CONFIDENCE LIMITS FOR STUDIES WITH AMERICAN SUBJECTS AS CITED BY NATIONAL ACADEMY OF SCIENCES AND SURGEON GENERAUS REPORT Females Males Confidence Confidence Study N Risk Limits N Risk Limits F!rom NAS Report' Kabat & %nder 13 0.79 0.26 2.43 5 1.00 0.20 4.40 Buffler 33 0.80 0.32 1.99 5 0.50 0.14 1.79 Garfinkel (19851 92 1.12 0.74 1.69 Garfinkel (1981) 89 1.18 0-90 1.54 Correa 14 2.03 0.83 5.03 22.29 0.31 16.50 Overall risk ratio ._for American studies2 1.14 0.921.4 0 From Surgeon General's Report' WU 29 Smoked 1-20 yrs. L4 0.4 4.9 Smoked 21 + yrs. 1.2 0.4 3.7 Prfer-ces: I. Enifironme, tal 7bbacco&noA-c-A1easurzng and Assesstng Health Effects. 1986. No t io nal Academy of &?en ces. Tabte & 4 2. Ibid., p. 231 3. The Health Consequences oflnvoluniary Smok-zng---A report of the Surgeon General. 1986. Tabk-5 8 and 9 I Notes: -Ri4k " value is the ratto oflung cancer riskfDr nonsmokers whose spouses 47noke in relation to the riskfor nonsmokers married to nons?rwkers. "Coztfidence limils" are the values between which the risk value can be eapecled tofiall 95% of the lime based on the variability of the underlying data When the 95% cort(Wence limits are both greater and Iess than 1.00, the risk value fs considered not statistically significant, ie., the rewa are laely to be due to chance and do not support a causal relationship. Ln CD C:D -z~:b G-11 BATCo document for Legal Services : Health Canada 19 October 1999 Figure 1 LUNG CANCER RISK AND CONFIDENCE LIMITS FOR STUDIES WITH AMERICAN S UBJECTS AS CITED BY NATIONAL ACADEMY OF SCIENCES AND SURGEON GENERAIISREPORT Prom NAS Report Kabat and %~nder Females 1 0 Males Buffler Females Males Garfinkel (198.5) Females Garfinkel (1981) Females Correa Females Males Overall from NAS Table 12.4 From Surgeon General's Report wu Smoked 1-20 vrs. Smoked 21 + yrs. 0.5 1.0 1.5 2.0 1.5 3.0 4.0 Risk Ratio Relative risk 95% Confidence limits Data and sourcesfrom Table I 0 CID -_O 0~1 --j BATCo document for Legal Services: Health Canada 19 October 1999 Appendix 1 In Congressional testimony in June. 1986, Surgeon General C. E. RR values are'risk rados." reflecting the risk of lung cancer for non- Koopcited 14 studieson the link between risk of lungcancerand non- smokers whose spouses smoke compared to the risk for nonsmokers smokers' exposure to spouses who smoke. (See belota) Three of the marned to nonsmokers. Forexample. an RR - L25 would represent studies were characerized as not showing positive correlation be- a 25% increase in risk for the exposed nonsmokers. while an RR = tween lung cancer risk and exposure to spousal smoking. Of the 11 0. 80 would represent a risk 20% less than that of a nonsmoker w ith studies said to show "positive correlation," all have been reviewed by nonsmoking spouse. independent investigators and found flawed as noted in Appendix 2. SUMMARY OF STUDIES ON PASSIVE SMOKING AND CANCER LISTED BY U.S. SURGEON GENERAL C. E. KOOP (JUNE 17,1986) Statisticall V A`orker Country Risk Ratin Significant Size of SZudV Chan (19791 Hong Kong 0.85 =31 F Case contrf,l Phillips ( 19801 USA 2.4F, 1.5M F 5_1,LAiO F. I7.isio M ?.rQ,,VeCL1Ve Garfinkel (195111 U, SA 1.2 N 0 150.1)(10 F ~,nspcctive HiukVama (19511, JVLn ISF, 2.9M F. M .90.000 K 21.-.060 \1 ?rospective Richopoulos (19~1) Greece 2.4 F :302 F Case control Correa (1914) USA 2.0m, 21F F I.S9 M. 1-55 Ca5e Control Gillis (1983) Scotland 12M. LOF NO 6 M. S F Case contrul llir:l.vlm;l (1933) Japan 2.3-3.4 - 2S F ?.,,,;I I PC dVP Hirayama (190 Japan 1.4-1.9 - 2OU F Fruspective Knoth (1983.1 Germany 3.0 - 39 F Case control Koo (1983) Hong Kong - - 120 F Case control Kabat (1984) USA IAM. 0.917 M . 37 M. 117 F Caw control Miller (1984) USA 1.9 F 438 F Ca~e Control Repace ~ 19841 USA 1.7 180.000 F Reanalysis Sandler (1985) USA 1.7-4.6 M, F 420 M. F Case control Sandler (IfIS.5) USA 2.0 . M. F 466 M, F Case control -s! control Garfinkel (195.;- USA 2.1 F 134 F Czi, Source., Slavwwnt cz:C. Dereti Aloop, Af.D., Sur~qcon General, bejbre the Subcommittee on Ecalth and the EnrironmenC Committee on Energy and Commerce. 1:S. ;Yowv q'Representativer (June 12~ 1986) UT CD CD CD ~1z a,- Co BATCo document for Legal Services : Health Canada 19 October 1999 Appendix 2 COMMENTS FROM THE LITERATURE ON THE STUDIES CITED BY SURGEON GENERALKOOP *Abbyeria,wns: SG = Surgeon Cepteral OTA = Office of Tecknolq& Assessment report NAS = Nationat Acade?P~y qf Sciences rcport tSources arr identified in the reference list. 0 0 Lil CD CD CXD CN 11Z BATCo document for Legal Services : Health Canada 19 October 1999 Cited in Surgeon General Testimonv* Comments From the Literaturet Koo (1983)-SG OTA noted RR = 1.28 for cases exposed less than cites no RR, no 35.000 hours, and RR = 0.96 for cases exposed significance more than 36,000 hours. Baiter noted no signifi- indicated. cant difference in lung cancer incidence for casese~xposed at honic, in the workplace, or both, nor h * v degree of exposure. NAS mentions and ex. cludvs this study. SIG Report shows results nut statistically significant. Correa (1983)-SG OTA and Weiss note small numberof cases. Baiter lists RR = 2.0 and OTA note statistically significant relationship for males, 2.1 was only for nonsmokers whose husbands for females: smobed more than 41 pack-years. Lebowitz notes statistically possible error or bias in estimating degree of ex- significant for posure; also notes anomalous finding of "in- females creased risk in smokers whose Mothers smoked ... but no elevated risk for nonsmokers:' Sandler (19,95)-SG W'iiss notes "almost. certain misclassification of lists expo.;ure.*' OTA and Baiter note too few cases to RR ~ 1.7 - 4.6, evaluate Lebowitz notes that there was "no sig. significant for nificant increase in lung cancer risk but in- males and crea.-S in tion-tobacco m-lated cancers." NAS females. mennons and (!xcludes. Excluded in SG Report. Knoth (1983)--SG OTA and Baiter noted lack of control groups. lists RR = 3.0; Weiss noced -possibie misciassification of expo- no indication sure: no control 8roup." Lebowitz noted 'a high of significance likelihood of miscl assificati on /bias in this study.' NAS mentions and excludes. Excluded in SG Re- port- Miller (198~)-SG Results cited pertain only ro women not em- lists RR = 1.9, ployed outside the home. For women employed statistical[ * V outside. RR = 0.80: for all women in the study, significant. RR = 1.4, not statistically significant. OTA noted raiiure to control for age differences between cases and controls. NAS mentions and excludes. Excluded in SG Report. Url C:> C) CC) _~D BATCo document for Legal Services : Health Canada 19 October 1999 Cited in Surgeon General Testimony Comments From the Literature Phillips (1980)-SG Study pertains entirely to Seventh Day Adventists lists RR = 2.4 (SDA). Environmental Protection Agency's Car- for females, L5 cinogen Assessment Group noted~ 'SDAs are a for males; very unique group and may difrer from the gen- statistically eral populationwith regard to lungcancer risk by significant for more than just their lack of exposure to passive females. smoking. In addition to not smoking, SDAs do not drink. and they maintain rather strict diets." OTA also considered SDAs inappropriate for assessing risks of ETS exposure. NAS does not mention Phillips study. Excluded in SG Report. Repace (1985)-SG Calculations are based on data from Phillips lists RR = 1.7, (1980) study. Repace is an analyst at EPA. EPA not statisticalLy Carcinogen Assessment Group criticized study significant. based on SDAs as noted above and faulted the Re- pace study for unsupported age-adjustments to lung cancer rates. OTA noted that some of the as- sumptions are "inappropriate--- because of SDA lifestyle differences-, excluded Repace report from further consideration. NAS does not men- tion Ri-pacL stud~y at all. Excluded from SG Re- port. Kabat (1984)-SG DTA found RR = 1.0 for males and 0.8 for females lists RR = 1.4 based on spousal smoking. For exposure at home. for males, 0.9 RR = 1.3 ) for males and 0.9 for mmales. For expo- for females; sure at work, RR = 3.3 for males and 01 for fe- statistically males. OTA noted "authors consider data on pas- si:znificant for SiVe SMDking'preliminary*." NAS found RR = 0.79 males. for females. L.00 for males. SG Report lists RR 0.9 overall. 0 (_n c~ "0 BATCo document for Legal Services : Health Canada 19 October 1999 Cited in Surgeon General Testimony Comments From the Literature 7)-khopoulw (1981) Numerous reviewers have noted that many of the -SG lists lung cancer cases were not histologically con- RR = 2.4 for firmed; SG Report states 65% histologically con- females, firmed. Lebowitz states'only about one-fourth of statistically the lung cancer cases had histological confirma- significant. tion:Axbowitz also notes that cases and controls were matched on some factors but not on others. Balter noted a dose-response trend is significant when data are anal),zed based on husbands smok- ing 1-20 and 21 + cigarettes/day, but trend is no[ sipnificariL when based on 1-10, 11-20, 21-30 and 31 - cigarettes/day. RR = 2.9 for 31 + cigaret. tes,"dky but RR = 4.3 for 21-30 cigarettes/day, creating doubt about the significance of pooling data. SG Report cites RR only for pooled data. ffirayama (1981, Study has been wideLy criticized. EPA meino on 1983)-SG Repace notes that Hirayama report had been "se- lists RR = 1.8 vereiv criticized" in letters in British Medical for females Journal where it was published. Rylander noted and 2.9 for wide criticism *frorn the point of view of ques- males, both tionnaire reliability, absence of histological diag. statisticalLv nosis. statistical treatment, grouping of smoking significant. habits among husbands, and confounding factors such as air pollution from heating and/or cook- ing" az wril as anomalous results among sub- groups and possible changes in classification of subjects and spoirses over time. Lehnwitz points up numerous inconsistencies in the findings as well as other weaknesses in the treatment of data. Garfinkel (1985)- Garfinkel ct al. Stated "elevated risk ranging SG lists from 13 to 31% not statistically significant RR = 2.1 for No consistently higher risk for certain age groups females, or by histological types or by exposure at home or statistically at work. Exposure in other areas carried a higher significant. OR I odds ratios], but this finding is difficult to interpret.' Data show smaller RR for 7. + hours/ day expostire to spouse's smoking than for 3-5 hours/day, as well as smaller RR for 25 years of exposure than for 5 years of exposure, NAS listed RR = 1.12 for females, not significant on basis of confidence limits. (JŪl (D cz:i Co ,,,0 BATCo document for Legal Services : Health Canada 19 October 1999 References: Arundel, A., T. Irwin, and T Sterling. Nonsmoker lung cancer risk from tobacco smoke exposure: An evaluation of Repace and Lowrey's phenomenological model. jour- nal of Environmental Science and Health C4(1)-. 93-118 (1986) Baiter, N. J., S. J. Kilpatrick '. P Witorsch, and S. S. Schwartz, Causal relationship between environmental tobacco smoke and lung cancer In non-smokers: A criti- cal review of the literature. To be published, Proceed- ings of the Air Pol.lution Control Association. (1986) Burch, R R. J. Health risks of passive smoking: Prob- lems of interpretation. Environment International 12:23-28 (1986) Environmental Protection Agency, Carcinogen Assess- ment Group. H. J. Gibb. Repace and Lowrey's estimate of the nonsmoker's lung cancer risk from passive smok- ing. (Undated memorandum) Johnson, C. Letter. Environment International 12:21 (1986) Kilpatrick. S. J. Letter. Environment International 23:29-31 (1986) Lebowitz, M. D. The potential association of lung cancer with passive smoking. Environment International 21:3-9 (1986) Office of Technology Assessment, U~ S. Congress. Pas- sive Smoking in the Workplace: Selected Issues. Staff paper. (May, 1986) Rylander, R. Environmental tobacco smoke and lung cancer. European journal of Respiratory Diseases. Supp. No. 133, 65:127-133 (1984) Weiss, S. T. Passive smoking and lung cancer (Editorial). American Review of Respiratory Diseases 133:1-3 (1986) (-n CD CD OQ "0 -t'. .10 --j L^1 BATCo document for Legal Services: Health Canada 19 October 1999 Katzenstein Associates 51 Rockwood Drive Larchmont, MY 10-538 DK OE.-- .:~Ow U-1 CD CD Co ~Iz Il- BATCo document for Legal Services: Health Canada 19 October 1999