Some Studies that Went Wrong!
© 1999, Lauren A. Colby. Version 2.3
Some Studies that Went Wrong!
In February, 1991, a paper was published in the Journal, Cancer, entitled "Comparative Epidemiology of Cancer Between the United States and Japan". The authors, Ernst L. Wynder, M.D., et al, started out with the assumption that smoking causes lung cancer. In fact, Dr. Wynder has been crusading against smoking since the 1950's and the authors' report was paid for by the anti-smoking National Cancer Institute. As we will see, the authors took some liberties with the figures presented in their report, so as to try to make the data fit their preconception that smoking causes lung cancer, but eventually had to admit that the data did not support that assumption.
The highlight of the Wynder Report is a graph, which purports to show per capita cigarette consumption in the United States and Japan for the time period 1920 to 1985. The graph relies upon data, plotted at five year intervals. It purports to show a sharp dip in consumption during World War II, to less than 1/3 the pre-war rate. Also, amazingly, it purports to show that for the entire time period between 1920 and 1985, per capita cigarette consumption in Japan was always less than in the U.S.
As we will shortly see, the authors of the Wynder Report presented data which they themselves acknowledged to be in contradiction with their own graph. Before discussing that matter, however, it will be helpful to consider a basic problem in epidemiology, i.e., the difficulty of comparing data for two differing populations.
Suppose we postulate that people who eat jellybeans are prone to develop more carbuncles than people who don't. To test the theory, we decide to study jellybean consumption in two different countries: country A and country B. Both countries have populations of 1,000,000 divided equally between men and women. Jellybean consumption in both countries is 1,000,000 beans per day, yielding a per capita consumption figure of one jellybean per person per day.
There is, however, a difference. In Country A, only men eat jellybeans, while in Country B, both men and women eat jellybeans. Obviously, in Country A, the jellybean consumption for men is 2 per day, while in Country B it is one. In Country A, the daily jellybean consumption for women is zero, while in country B it is one. Any comparison of the two countries must take this into account.
Dr. Wynder and his colleagues presented data on relative smoking rates for men and women in Japan and the United States. The rates, expressed in terms of the percentage of each sex who smoke, are as follows:
Likewise, the authors presented statistics for 1970, 1980, and 1986, showing that Japanese males smoke more cigarettes per day than U.S. males, while Japanese females smoke fewer cigarettes per day than their counterparts in the U.S. Clearly, the total consumption figures given in the graph, accompanying the report, need to be adjusted to take into account the differing rates of smoking among males and females in Japan and the US. Otherwise, the authors are comparing apples and oranges. No adjustment was made but, if one had been made, so as to compare only Japanese males with US males, the graph would most assuredly have shown much higher per capita consumption in Japan than in the U.S. This is so because, in Japan, where few women smoke, the large number of non-smoking women "waters down" or dilutes the per capita consumption figures for the population, taken as a whole.
Other data presented in the report compared the lung cancer death rates in Japan and the U.S. For some reason, the authors elected to give figures only for white U.S. males and females, excluding African Americans and American Indians (probably, the inclusion of that data would have interfered with some pre-conceived notions). Whatever the case, the male lung cancer death rates, age adjusted, for 100,000 of population were presented as follows:
At the time of the release of the report, there were interviews on TV with Japanese doctors, who sought to explain the high rate of smoking in Japan and the low rate of lung cancer by declaring that in Japan, cigarettes were hard to get, during World War II. The graph, prepared by Dr. Wynder and his colleagues, seems to support that claim, showing as it does a big dip in cigarette consumption during the War.
The graph is, however, tricky. The data is plotted at five year intervals, and 1945, the last year of the war, is one of the years used. Simple interpolation was used to indicate the data between 1940 and 1945, and between 1945 and 1950; in other words, the authors drew two straight lines, one between 1940 and 1945, and another between 1945 and 1950.
Actually, data is available for annual cigarette consumption in Japan for every year from 1920 to 1990, based upon sales. Those figures come from a book, International Smoking Statistics, published by the Oxford University Press in 1993. The figures show that in Japan, as in the U.S., there was a switch from machine made cigarettes to hand rolled cigarettes during World War II. Taking that into account (which the Wynder authors apparently did not), the Japanese consumed 71,158 million cigarettes (of both kinds) in the last year before the War, 1941. Consumption continued unabated until 1944, when 64,280 million cigarettes were consumed. In 1945, consumption dropped to 31,021 million cigarettes. It then rose steadily until 1950, when 75,138 million cigarettes were consumed. So there was a dip, but it lasted only 5 years, and was not nearly as pronounced or as lengthy as the Wynder chart would make it seem.
The bottom line, however, was the concession of Dr. Wynder that the data did not support smoking as a cause of lung cancer in Japan. That concession did not come without a few confusing gyrations. In discussing cancer of the larynx, the authors say that "The age adjusted mortality rates for laryngeal cancer during 1955 are higher in U.S. Whites than in the Japanese. These differences can be partially explained by the higher levels of cigarette consumption and alcohol consumption in the U.S.". The authors discuss cancer of the esophagus, saying that "In spite of the higher tobacco and alcohol consumption in the U.S., Japanese males have higher esophageal cancer mortality rates, which suggests that other risk factors are of importance". Thus, in their discussions of these two types of cancer, the authors assert that smoking and alcohol use are greater in the U.S. than in Japan, using that "fact" in one instance to justify their preconceived belief as to the cause of laryngeal cancer, and dismissing the "fact" as irrelevant when it comes to the other cancer (esophagus), where the figures just don't bear out the preconception.
When it comes to lung cancer, however, the authors state that during 1955 to 1985, lung cancer death rates are "higher in US White men than in Japanese men which is discrepant with the higher prevalence of cigarette smoking among Japanese males for the same period of time". Exactly! According to the authors' own figures, the lung cancer rate among Japanese males is less than one third the rate among US White males, and as early as 1955, 81.4% of Japanese men were smokers (compared to 52.6% in the U.S.). That is, indeed, a big discrepancy. The Wynder authors must have had to write that word "discrepant" through gritted teeth, but at least they had the honesty to do it.
On January 13, 1995, the Wall Street Journal reported another study, this one involving animals and funded, in part, by the U.S. National Institutes of Health. According to the report, the study was inspired when a researcher in Buffalo, John Pauly, was studying some tissue from a smoker and lung cancer victim and found a tiny particle of cellulose acetate, the material used to make cigarette filters. He apparently decided that pieces of cigarette filters, imbedded in the lungs, are the cause of lung cancer and decided to do an experiment with mice.
He implanted pieces of filters, coated with cigarette tar, in the lungs of six mice and found that they remained intact in the lungs for six months. This finding was haled as a great break-through, demonstrating that pieces of cigarette filters may lodge in the lungs and cause cancer. What this ignores, however, is a simple fact: no cancers were found in the mice! What the study really proves, therefore, is merely that implanting pieces of cigarette filters, drenched with tar, in the lungs of mice does the mice no apparent harm!
Before leaving this subject, i.e., studies which don't bear out the smoking/lung cancer connection, it's worth mentioning a couple of studies that involve Native Americans. Some of the heaviest smokers (and drinkers) in America are to be found among the Native Americans. In fact, a 1992 study by the CDC showed that 39.5% of American Indians smoked, as opposed to 25.6% of the general population. Knowing this, I have been looking for some statistics on lung cancer among Native Americans.
Turns out there have been at least two such studies. The first was conducted by J.M. Samet, et al, of the University of New Mexico School of Medicine and published at Am J Public Health Sept., 1988, 79(9) 1182-86. The study dealt with both Hispanics and Native Americans. The authors concluded that in the study \ period (1958-82), "[in whites] age adjusted mortality rates from lung cancer and from chronic obstructive pulmonary disease increased progressively in males and females. Mortality rates for both diseases increased in Hispanics during the study period, but the most recent rates for Hispanics were well below those for Other Whites....in Native Americans, rates for both diseases were low throughout the study period, and did not show consistent temporal trends."
The second study was conducted by M.C. Mahoney, et al., of the New York State Department of Health, using data from Native Americans in upstate New York, during the time period 1980-86. It is published in the Int J Epidemiology, June, 1989, 18 (2) 403-412. The authors came to the same conclusion as Samet, et al. They stated that the principal causes of death among the Native Americans were TB, diabetes, pneumonia and cirrhosis. However, "fewer than expected malignant deaths occurred among both Native males and females [and]... A deficit of deaths was observed for colon and lung cancer deaths among Native males and for colon and breast cancer deaths among Native Females...".
In short, Native Americans smoke more than the general population but suffer from less cancer and, in particular, less lung cancer.