Tobacco Documents Online

4 Gone are the days where one needed a statistically significant relationship to reach a positive conclusion! P N Lee 25.11.99.

2 Relative risks (95% CI) Any exposure Earlier study Lung cancer Earlier study Adenocarcinoma This study Adenocarcinoma Childhood exposure (from parents) 0.78(0.74-0.96) 0.80(0.61-1.05) 0.6(0.3-1.2) Spousal exposure 1.16(0.93-1.44) 1.08(0.82-1.42) 1.0(0.5-1.8) Workplace exposure 1.17(0.94-1.45) 1.06(0.81-1.40) 1.5(0.8-3.0) Spousal or workplace exposure 1.14(0.88-1.47) 1.01(0.73-1.40) 1.2(0.6-2.5) The studies also showed similar tendencies in having more extreme results for the highest duration tertile. Thus comparing the highest 10% with the lowest 75% (including non-exposed) we have Relative risks (95% CI) Hin.hest duration of exposure Earlier study Lung cancer Earlier study Adenocarcinoma This study Adenocarcinoma Childhood exposure (from parents) 0.66(0.41-L06) 0.61(0.32-1.16) 0.3(0.1-0.9) Spousal exposure 1.89(1.19-3.00) 1.70(0.95-3.04) 1.8(0.3-12) Workplace exposure 2.07(1.33-3.21) 1.70(0.97-3.00) 1.7(0.4-6.6) Spousal or workplace exposure 1.56(1.07-2.28) 1.58(0.98-2.54) 1.8(0.5-6.2) Certainly the results are much what one would expect in a replicate smaller study essentially identical in design to the earlier study. The study is too small to affect materially meta-analyses of the overall evidence with spousal and workplace exposure and I do not propose to repeat my views here on the difficulties in interpreting the positive association seen in meta-analyses as a cause and effect relationship. However, two points are worth making. N 01 O M CA CO CD Q First, it is interesting that we have yet another study showing a negative association of pp t0 lung cancer risk with paternal ETS exposure in childhood (RR = 0.7, 95% CI = 0.3-1.3). Apart

3 from the rather unusual study of Rapiti et al in India (see Review 1064), which found an implausibly high relative risk of 17 (95% CI = 5.7-47) in women and no association in men (RR =1.07, 95% CI = 0.36-3.21), all the other studies with relevant data (Fontham, Sobue, Svensson, Zaridze and the first Boffetta study in both sexes) all reported relative risks in the range 0.7 to 0.9. While there are some studies of total ETS exposure from household members in childhood that report relative risks greater than 1, the overall pattern of results for childhood exposure appears to indicate a negative rather than a positive association. Boffetta et al state in the summary that "bias and confounding cannot be excluded as explanations for the apparent decrease in risk from childhood exposure," but their discussion does not identify the actual source of bias and confounding that is responsible. They tend to reject recall (reporting) bias and selection bias as implausible explanations on the grounds that if they did apply they might affect results for adult ETS exposure too. Negative confounding by an unknown risk factor is what they are left with as a possible explanation, and they note also that "if the protective effect is real, its underlying mechanism remains to be elucidated." In my view there is a need for a more detailed review of the evidence on childhood ETS exposure. Second, I did not find the discussion of the available literature regarding the association of ETS with adenocarcinoma gave a totally clear picture. The general impression given is that, with one or two exceptions, the relative risks for adenocarcinoma associated with ETS exposure (mainly from the spouse) are of the order of 1.3-2.1, and that this increase is less than that for squamous or small-cell carcinoma. This does not really fit in with meta-analyses showing that the overall association of lung cancer with spousal exposure is about 1.2 (without corrections for misclassification bias and uncontrolled confounding). Again formal review and meta-analyses are needed to gain a clearer impression, but IARC do not seem to want to get involved in meta- analyses. Overall, two prevailing impressions come from this paper. First that the study was essentially null. Second that the authors' write-up tended to play down the negative associations with childhood ETS exposure (although almost significant and consistent with other literature) and be happy to claim that nonsignificant positive associations with adult ETS exposure (actually not an association at all with spousal exposure) "confirm previous reports of a weak effect."

IIt 3~~ REVIEW 1078 CONFIDENTIAL Subject ref 8b "Exposure to environmental tobacco smoke and risk of adenocarcinoma of the lung" P Boffetta et al International Journal of Cancer (1999), 83, 635-639 In 1998 (see Review 1008) Boffetta and his colleagues published the results of the IARC West European Multicentre case-control study of ETS and lung cancer, involving 650 cases (and 1542 controls) in nonsmokers in Germany, Italy, Portugal, UK, Spain, Sweden and France, the study period ending in 1994. The current paper describes results of another multicentre case- control study, involving 70 cases of adenocarcinoma (and 178 controls) in Germany, Italy, Sweden, France, Poland, Romania and Russia. As before, the controls were a mixture of hospital patients and healthy individuals, and the same questions on ETS exposure in childhood (from the parents), and during adulthood from the spouse and workplace were asked. As shown in Table 1, the distribution of cases and controls varied markedly by age, sex and country. However, all analyses were adjusted by these three factors. Results are shown for childhood exposure in Table II, for spousal exposure in Table III, for workplace exposure in Table IV and for spousal or workplace exposure in Table V. The first point that stands out from the results is that none of the pairwise comparisons or trends are statistically significant, with the exception of the reduced risk of lung cancer associated with long-term childhood exposure. The point estimates for ever versus never exposed for each of the four analyses show a nonsignificantly reduced risk for childhood exposure, no association for spousal exposure, and nonsignificantly increased risks for workplace exposure and for spousal/workplace exposure. The pattern is not dissimilar from that seen in the earlier larger study (see table below), which also reported reduced risks for childhood exposure. However, the estimate for workplace exposure is rather higher than reported in the earlier study.