Oral Comments to CAL-EPA on Section 7.2 of External Review Draft - Excerpt: ETS and Lung CancerPresented March 25, 1996, Berkeley, California My name is William Butler. I hold a Ph.D. in Biostatistics and have been teaching and doing research in biostatistics and epidemiology for more than 20 years. I am here today as a scientist resident in California who had conducted original epidemiologic data analysis on ETS and health at the request of the RJ Reynolds Tobacco Company. The views I express are my own and not necessarily those of RJ Reynolds Tobacco Company. I'm here today to show you that CAL-EPA has been seriously mislead by the analyses and conclusions presented in the two largest U.S. epidemiologic studies of ETS and lung cancer. The Brownson and Fontham studies show no association between ETS and lung cancer, when analyzed properly according to standard epidemiologic practice. The results from these two studies are plainly inconsistent with the conclusions of the U.S. EPA's 1992 risk assessment and CAL-EPA's preliminary report. The inconsistency between the stated conclusions in Brownson and Fontham and the patterns of association present in their data demonstrates that CAL-EPA cannot rely on the published reports of individual studies or on summaries and reviews by others such as the U.S. EPA. First, I'll show you why the Brownson study is a negative study. 1 Re-Analysis of the Raw Data of the Brownson Study
|
| All Respondents | |||||||||
| (Brownson et al., 1992, Table 2) | Self-Respondents | ||||||||
| Spousal Smoking | No. of Controls | No. of Cases | OR* | 95% CI | No. of Cases | OR* | 95% CI | ||
| Never | 568 | 213 | 1.0 | -- | 80 | 1.0 | -- | ||
| Ever | 598 | 218 | 1.0 | 0.8,1.2 | 74 | 0.9 | 0.6,1.2 | ||
| Cigarette pack-years | |||||||||
| 0 | 568 | 213 | 1.0 | -- | 80 | 1.0 | -- | ||
| 0<<15 | 128 | 32 | 0.7 | 0.5,1.1 | 18 | 1.0 | 0.6,1.7 | ||
| 15<<40 | 200 | 54 | 0.7 | 0.5,1.0 | 20 | 0.7 | 0.4,1.2 | ||
| 40 < | 216 | 110 | 1.3 | 1.0,1.7 | 32 | 1.0 | 0.6,1.6 | ||
|
* Adjusted for age and history of previous lung disease The two odds ratio presented in the article by Brownson et al. that mislead CAL-EPA is presented in the set of columns for All Respondents; that is, 1.3 for "Highest Exposure Category vs Never". This odds ratio is the only possible result in Brownson et al. to support their stated conclusion that "Our and other recent studies suggest a small but consistent increased risk of lung cancer from passive smoking." (abstract) However, this assertion is not supported by or consistent with the results of the more valid analyses that include only self-respondents. I obtained the raw data for the Brownson study on a computer diskette prepared by NCI, and I analyzed these data on the association between spousal smoking and lung cancer for subjects who were self-respondents and lifetime never smokers. The epidemiologic analysis of self-respondents, when possible, is preferred because it avoids the potential bias introduced by the less reliable and less valid information from surrogate respondents who are less knowledgeable about the subject's actual exposures. The preference for results of analyses only for self-respondents is standard epidemiologic practice and recognized by CAL-EPA in its draft report. Brownson et al. state that they performed statistical analyses only for self-respondents, but they did not provide the quantitative results from those analyses to be used by the CAL-EPA. I provide the results of those analyses here today for CAL-EPA's use. As can be seen in the set of columns for self-respondents, there is no association between lung cancer and "Ever vs Never" exposure; that is, the odds ratio equals 0.9. Also, there is no pattern of association between lung cancer and the categories of ETS exposure; that is, all odds ratios for exposure categories are equal to or less than 1.0 for self-respondents. Coupled with the findings reported in Brownson et al., the results from the analysis of the Brownson data demonstrate that there is no association between lung cancer and spousal smoking status among never smoking women. 2 |
Re-Analysis of the Raw Data of the Brownson Study,
|
| Self-Respondents | Surrogate-Respondents | ||||||||
| Occupational ETS Exposure | No. of Controls | No. of Cases | OR** | 95% CI | No. of Cases | OR** | 95% CI | ||
| Never | 560 | 66 | 1.0 | -- | 85 | 1.0 | -- | ||
| Ever | 434 | 59 | 1.1 | 0.8,1.7 | 56 | 0.9 | 0.6,1.3 | ||
|
* Among women who worked outside the home Further, when you analyze the Brownson data in the same manner for occupational ETS exposure, the results are the same. There is no association between lung cancer and "Ever vs Never" occupational exposure among the self-respondents (that is, Odds Ratio equals 1.1) or among the surrogate respondents (that is, Odds Ratio equals 0.9). These findings are consistent with Brownson's conclusion that "there is no elevated lung cancer risk associated with passive smoke exposure in the workplace." (p. 1527) The Brownson data, when analyzed as described above, show that there is no association between lung cancer and ETS, regardless of whether the source is spousal smoking or workplace ETS. CAL-EPA should acknowledge that the Brownson study is a negative study whose results contradict the U.S. EPA's conclusion that ETS is a carcinogen. Next, I'll show you why the Fontham study is a negative study. Data from Fontham et al., (1994, Table 8),
|
| (a) Case/Controls | ||||||||
| Adult ETS Exposure | Childhood ETS Exposure | |||||||
| No | Yes | |||||||
| No | 23/71 | 5/44 | ||||||
| Yes | 118/364 | 235/724 | ||||||
| (b)Odds Ratios & 95% Confidence Intervals | ||||||||
| Adult ETS Exposure | Childhood ETS Exposure | |||||||
| No | Yes | |||||||
| No | 1.00 Baseline |
0.35 (0.12, 0.99) |
||||||
| Yes | 1.00 (0.60, 1.67) |
1.00 (0.61, 1.64) |
||||||
|
The Fontham study examined the joint exposure to childhood and adult ETS exposure. The CAL-EPA report cites the results of Fontham's analysis of that joint exposure as presented in Table 8 of the article of Fontham et al. CAL-EPA was seriously misled by the manner in which Fontham et al conducted and presented those analyses because Fontham et al. did not acknowledge the presence of a statistical interaction that supports the absence of an association between ETS and lung cancer. I re-analyzed the data from Table 8 of Fontham et al. (1994) on the joint exposure to ETS during childhood and adulthood. I used a single baseline group (that is, those with neither childhood nor adult ETS exposure) for all combinations of exposure. The use of a single baseline group is recommended in standard epidemiologic textbooks such as Kleinbaum, Kupper and Morgenstern; Breslow and Day; Schlesselman; and others. Entries in the first table indicate the number of cases and controls for each combination of childhood and adult ETS exposure as presented in Table 8 of Fontham et al. The baseline group of women with neither childhood nor adult ETS exposure is in the upper left hand corner. 3 Based on this single baseline group, the odds ratios for the two categories of women with adult exposure equal 1.0; that is, no association between adult ETS exposure and lung cancer - regardless of whether or not the women had childhood ETS exposure. Fontham incorrectly interprets these data to indicate that adult ETS exposure is associated with higher lung cancer risk and that the elevations in risk for women exposed during childhood were twice as high as those of women not exposed during childhood. Specifically, using Fontham's approach of stratum-specific baseline groups, one obtains an odds ratio = 2.86 = 1/0.35 for adult ETS exposure among those with childhood ETS exposure and an odds ration of 1.00 among those with no childhood ETS exposure. Fontham incorrectly interprets this first odds ratio to indicate that those with both childhood and adult ETS exposure are at about twice the risk of lung cancer relative to those with adult but no childhood ETS exposure. However, it is clear from the analyses presented here (and not presented in Fontham et al.) that this interpretation is wrong. The increased odds ratio for those with both childhood and adult ETS exposure results not from increased risk associated with both exposures but from a significantly lower risk among those with childhood but no adult ETS exposure. This is an unexpected finding that, most likely, reflects a shortcoming of the Fontham data set due to bias in sample selection or data collection. The data presented here clearly indicate that there is no increased risk of lung cancer associated with adult or childhood ETS exposure, and any analyses that implies such an association distorts the patterns present in this data set. |
| Brownson Study: | No Association of Lung Cancer Risk with Spousal of Workplace ETS. Exposure Among Self-Respondents Who Are Lifetime Never Smokers | ||||
| Fontham Study: | No Association of Lung Cancer Risk with Adult ETS Exposure During Childhood. | ||||
| U.S. EPA (1992): | The two largest U.S. studies were published post-1991 and generate results consistent with each other and contrary to U.S. EPA's conclusion that ETS is a carcinogen. | ||||
| Review of Studies: | CAL-EPA must re-analyze the data of the epidemiologic studies it uses to assess the association between ETS and lung cancer. | ||||
|
In summary, the two largest U.S. epidemiologic studies of ETS and lung cancer were published after the U.S. EPA's risk assessment. These two studies benefit from methodologies designed to address, at least partially, some of the shortcomings of all the previous research. My analysis demonstrates that the data from these two studies are in agreement: There is no association between lung cancer and adult ETS exposure, a finding that contradicts the stated conclusions of each study. These discrepancies demonstrate the errors that can occur if one relies solely on published results and conclusions. 4 My analysis also shows that the data from the two largest U.S. epidemiologic studies are not consistent with U.S. EPA's conclusion that ETS is a lung carcinogen or with CAL-EPA's conclusion that post-1991 epidemiologic studies support a casual relationship between ETS and lung cancer. In light of these analyses, it is clear that CAL-EPA must re-analyze the data of any of the epidemiologic studies that it uses to assess the association between ETS and lung cancer. Chapter 7: Carcinogenic Effects of Exposure to
Environmental Tobacco Smoke Oral Comments by William J. Butler, Ph.D. |
|
|