Urban Residency Confounds Most Secondhand Smoke Studies
Antismoking activists tout secondhand smoke studies as scientific evidence that smoking bans are necessary to protect the public from a serious health hazard. The consensus claim being that secondhand smoke exposure increases risk of dying from lung cancer 20% and that it causes 35,000 deaths from heart disease each year.
Researchers that publish these studies admit they are subject to potential confounding from a host of other risks that might be more particular to nonsmokers exposed to more smoke than the general public, or those with less exposure. They usually make the tenuous conclusion that the possible confounding is not enough to dispute their findings.
However, a fairly thorough summary of commonly recognized confounders is presented in the EPA report: "Respiratory Health Effects of Passive Smoking", Dec 1992. Among these are history of lung disease, family history of lung disease, heat sources for cooking, cooking with oil, occupation, and diet. Others have claimed recall and misclassification bias are potential confounders.* Even without this possibility, Robert Nilsson, of Stockholm University did conclude these confounders are sufficient to explain most of the increased risk found in second hand smoke studies. ( Ecotoxicology and Environmental Safety, Vol 34, #1, June 1996) None, however, have considered urban residency.
*(This misclassification might be underestimated because some current smokers may have claimed to be nonsmokers, or smokers who had recently quit, for the purpose of saving money on health insurance premiums. Later, when diagnosed with lung cancer or heart disease, they would have to maintain the fraud, or conceal the fact they had resumed smoking, to avoid probable loss of health coverage just when most urgently needed. Having heard the publicity about secondhand smoke, they would then claim massive secondhand smoke exposure to explain contacting these diseases. In any regard, we cannot assume recall bias or misclassification are similar to those encountered in other kinds of epidemiological studies.)
Lastly, the most common criticism of secondhand smoke studies is that the statistical results are too weak to draw any conclusion. Consider that many secondhand smoke studies report negative results, some in data subsets, which range from RR= 0.9 to 0.6. This would indicate a "protective effect" from secondhand smoke, although no sensible plausibility has ever been proposed. For example, the WHO 1998 study found never-smokers who grew up in smoking homes had 78% as much lung cancer as never-smokers who grew up in nonsmoking homes; Kabat 2003 found nonsmoking males married to smokers had 94% as much heart disease and 75% as much lung cancer as nonsmoking males married to nonsmokers. Brownson, 1992 reported many subsets with 70% risk. EPA 1992 reported: 81% risk Buffler 1984; 75% risk Chan1982; 86% risk Janerich 1990; 79% risk Kabat 1984; 74% risk Liu 1991; and 79% risk Wu-Williams 1990. Considering these results are much more likely to be statistical "bounce", than proof secondhand smoke is healthful , it appears the standard relative error of secondhand smoke studies is about 20%. Further, there may be many more of these studies which have never been published. Dr, Geof Givens, applied a statistical routine which found it is likely there are many more unpublished secondhand smoke studies with insignificant results and that the estimated excess risk may thus be overstated by around 30%. . Indeed, Enstrom and Kabat encountered peer personal attacks and suspension of research funding from the University of California's Tobacco-Related Disease Research Program, when they announced the results of their statistically insignificant secondhand smoke study. Levy and Marimont report the National Cancer Institute's own guidelines consider any relative risk less than two insignificant. Most secondhand smoke studies report results much lower than two. (Lies, Damned Lies & 400,000 Smoking-Related Deaths)
Non-profit epidemiologists such as Garfinkel, J. Nat. Cancer Inst. 1981, followed the NCI guidelines. Overall, he found nonsmoking wives of smokers of less than 20 cigarettes/day had 27% more lung cancer; those of smokers over 20 cigarettes/day had 10% more lung cancer. Because these results were less than two, and because they followed an inverse trend with smoking habits of husbands, he concluded the results were insignificant. He also performed a matched analysis of a subset his data. When the data were corrected for urban/rural residency, race, education, and occupation, he found an elevated risk of 37% when the husband smoked less than 20 cigarettes/day and an elevated risk of 4% when the husband smoked more than 20 cigarettes/day. Note the impact of correcting the data for some other risk factors, and why any risk less than two stands a good chance of being compromised by secondary factors.
Individual people are too diverse to "average together" in such studies. Association does not prove cause. After all the years of being warned of the hazards of red meat diets, suddenly Dr. Atkins announces such diets do produce weight loss, and in a quick study, the American Heart Association finds less heart disease among those consuming diets rich in red meat and saturated fats. Consider, another example.. the relative risk of magnetic field exposure causes cancer was three. Yet no plausibility was ever found. According to Dr. Steven Milloy, in "Science Without Sense" , 1995, The Cato Institute, other relative risks never found to be plausible are wearing a brassiere/ breast cancer, RR=12,500 ; eating12+ hotdogs-a-month/ leukemia, RR=9.5 ; douching/cervical cancer, RR=4.0. Nilsson reported the editor of the journal "Epidemiology", Ken Rothman admitted of secondhand smoke studies, "We're pushing the edge of what can be done with epidemiology".
It has always been known smoking patterns vary with degree of urbanization. The International Encyclopedia of the Social Sciences, vol 14, 1968, p 337 reported urban and rural non-farm dwellers smoked at the same rate, but farm workers smoked less. By 1990, this pattern had shifted so that urban dwellers smoked about 33% more than suburban dwellers, while rural dwellers had intermediate smoking rates.
By, 1998, another shift had occurred. Today, rural dwellers have the highest smoking rates (27% female, 31% male), suburban dwellers the lowest (20% female, 24-25% male) , and urban dwellers are intermediate. ( Health, United States, 2001) However, most secondhand smoke studies were completed prior to this final shift.
Within the last five years, an explosion of concern about urban pollution being linked to increased lung cancer and heart disease has occurred. This concern should have been addressed years ago. "Geographic Patterns in the Risk of Dying and Associated Factors", U.S. Dept of Health and Human Services, reported that between 1960 and 1970 that the highest rates for most cancers were in cities. "Persons at High Risk of Cancer" reported that during the 1970's, the urban/rural risk of lung cancer from all causes was 1.89 for males, 1.64 for females. Adjusting for 33% excess smoking in cities, this elevated risk then becomes a 43% excess for both sexes.
Recently, these claims have been taken more seriously. Arden Pope, of Brigham Young University found after controlling for smoking, urban dwellers are 16% more at risk of dying of lung cancer from urban pollution. Overall, minus the greater smoking rates in cities, he summarized the scientific literature as indicating a greater risk of dying from lung cancer of 30 to 50% higher in urban areas.* Others have summarized this effect as manifesting itself internationally too, finding an excess overall elevated risk of up to 50%. Later, Pope also found the same is true of heart disease, with, after controlling for smoking, and other assorted risk factors, an elevated risk of 8% to 18% for every ten micrograms/ cubic meter of air of fine particulates. Since a typical city averages about twenty ug/cM the elevated risk then becomes 16% to 36% higher. (Cardiovascular Mortality and Long-Term Exposure to Particulate Air Pollution)
*This agrees well with my adjusted elevated risk of 43% for lung cancer, calculated above.
How would these facts affect secondhand smoke studies?
Most studies of secondhand smoke have been performed in teaching hospitals which are almost all located in large metropolitan areas. In obtaining study participants, then, one would expect nonsmoking spouses of smokers to follow the same urban/suburban residency patterns as smokers do. This would hold especially true for cohort studies where all participants are healthy at the beginning of the study. Rural resident nonsmokers would be less likely to participate because they tend to seek their medical services in small community hospitals, but could be included in case-controlled studies from cancer registries, or from being transferred to a large urban hospital when their health deteriorates substantially. But since these rural residents are less likely to suffer diseases blamed on secondhand smoke due to their living more free of urban pollutants, they are in any case underrepresented even in case-controlled studies. Studies of workplace secondhand smoke would not necessarily follow this same pattern, but they generally find weaker conclusions than the spousal studies do. Anecdotal reports indicate that workplace smoking policies for urban health care facilities ( which are big employers) were the first to ban smoking, suburban employers were second to restrict or ban smoking, while general urban employers ( such as manufacturing) came third. Commuting occurs to such a large degree that the mixing of suburban dwellers working in urban settings and urban dwellers in the suburbs is common. This mixing may indeed be the reason workplace secondhand smoke studies find weaker conclusions than spousal studies.
Taking 33% excess smoking in urban areas as a typical value, representing the period from 1980 to 1995 when most secondhand smoke studies were conducted, then we would expect nonsmoking spouses of smokers to be 33% more often urban residents, as well. During 1980, less differential between urban/suburban residency existed, but by 1995 more differential existed. This, of course assumes the medley of secondhand smoke studies is truly a representative population sampling, otherwise, one must question the validity of any of the data. For the reasons stated previously, rural residency would be underrepresented. But since for most of the period 1980 to 1995, researchers didn't consider urban/suburban residency an important confounder when sampling populations for participants, they wouldn't be concerned, or even record residency status or history of residency status. Pope stated urban lung cancer risk was studied during the 1970s, but was all but forgotten during the 1980s because everyone was so focused on smoking. ( first reference cited above) Indeed sample questioners such as those found in the EPA report don't inquire about urban/suburban residency.
If one were to conduct a lung cancer study of nonsmokers who lived in nonsmoking households, and it happened the population samples consisted of 33% more urban residents in one group, as opposed to the other, then one would find an excess lung cancer rate of 14% in that group. ( 43% X 33% = 14% ) In fact this is what has inadvertently happened during the years most secondhand smoke studies were conducted. This 14% is 70% of the consensus excess lung cancer risk claimed from living with a smoker. ( which is 20%.) Considering the standard errors that are recognized by the authors when conducting such studies, ( usually 30%) an additional positive determinate error of 70% means there is a 40% chance the measured lung cancer risk of a nonsmoker living with a smoker is less than zero.
Similarly, for non-rheumatic heart disease deaths. The secondhand smoke claims concerning heart disease deaths are actually more vague. Unlike lung cancer which is widely believed to be caused mostly by tobacco, many other well established causes of heart disease exist. Obesity and stress may be evenly distributed between urban and suburban dwellers. Stress and smoking may act synergistically to produce more heart disease in cities, but this could be partially offset by smokers generally being less obese. Considering Pope's average increased 26% risk of dying from heart disease due to urban pollution, a hypothetical study of nonsmokers living in smoke-free households which inadvertently contained 33% more urban residents in one group would find (26% X 33% =) 8.7% more such deaths in that group. According to the U.S. death tables, during 1995, 489,200 deaths due to ischemic heart disease occurred. 8.7% of that number is 42,580 deaths which is in excess of the 35,000 deaths each year usually claimed caused by secondhand smoke. It is easy to see how 8.7% more deaths among nonsmoking spouses of smokers could lead to the claim of 35,000 more heart disease deaths overall. However, part of Pope's average increased risk may be due to excess smoking-stress synergy in urban areas (discussed above), which could reduce my calculation of 8.7% more deaths in a hypothetical study to a number that when multiplied by 489,200 yields a product closer to the 35,000 deaths usually claimed caused by secondhand smoke.
Data gathered back in the 1970s, when those who smoke were evenly distributed between urban and suburban residency, produced secondhand smoke studies which generally found lower risk. As the distribution gravitated towards urban residency, the overall risks found in secondhand smoke studies which began gathering data after this gravitation, increased accordingly, but unfortunately, efforts were not made to overcome the play of urban-suburban residency as a confounding factor. Saracci wrote in "Tobacco: A Major International Health Hazard" , IARC Scientific Publications No. 74 , 1986 that "(a) the published studies, when considered by themselves, are compatible with absence of excess risk of lung cancer due to passive exposure to ETS or with the presence of a small excess risk; (b) in the light of the other available evidence, external to the studies, the interpretation favourable to the presence of a risk becomes definitely more plausible than the alternative; ( c) under these circumstances further epidemiological studies aiming at a direct estimate of the risk may be justified provided steps are taken to overcome, in the study design and conduct, the play of biases, some of which have been alluded to. Unless this is done, the studies stand a good chance of contributing results of a confusing rather than of a clarifying nature."
Case-in-point is the recent study published that found no risk from secondhand smoke, based on Californians who participated in the American Cancer Society CPS-1 study begun in 1959. Kabat, et. al. In 1959, no such urban/suburban smoking differential existed. Later, when the ACS initiated a second study in 1982, CPS-2 did find positive risk from exposure to secondhand smoke.
Urban residency is an important factor that must be accounted for in order to determine if secondhand smoke actually causes disease. I cannot claim every study has been affected the same way by urban residency, but overall, the medley of studies that have been taken together to claim secondhand smoke causes a 20% elevated lung cancer risk and 35,000 deaths from heart disease in nonsmokers probably, on average are affected the way I describe. The authors would have to examine the raw data, and either assign a corrective factor based on residency, or, in the manner of Garfinkel, compare nonsmoking spouses of nonsmokers and smokers within the framework of residency. This assessment of urban/suburban living appears to explain virtually all the excess risk of lung cancer and heart disease associated with secondhand smoke exposure. It is possible future research of urban pollution will uncover other health risks that are currently being blamed on secondhand smoke.