The Last Squirmish Between Irving Selikoff and Sir Richard Doll

In one of his last publications before he died, Dr. Selikoff reflected on the ethical dimensions of epidemiology. He recounted the development of our understanding of the lung cancer hazards of asbestos and smoking, and noted that there had been “random instances” of lung cancer cases reported among asbestos workers in the 1930s and 1949s, but “[w]ith the continued growth of the asbestos industry, it was deemed wise to epidemiologically examine the proposed association. This was done in an elegant, innovative, well-considered study by Richard Doll [7], a study which anyone of us would have been proud to report in 1955.” Irving J. Selikoff, “Statistical Compassion,” 55 J. Clin. Epidemol. 141S, 142S (1991).

Despite his praise for Doll’s work, Selikoff goes on to downplay Doll’s achievement by explaining how Doll supposedly missed a synergistic multiplicative interaction between asbestos exposure and smoking, which Selikoff claimed to have found a decade later:

“Not only was the association [with smoking] not yet established, indicating the need for its investigation in cohort studies, but smoking histories were not available (and indeed, many of the workers involved may not have smoked cigarettes, having begun their asbestos exposure at a time when cigarette smoking was considerably less common, even among blue collar workers). We would want such information now, but these studies were accomplished at an earlier, less informed, time.”

Id. at 143S

This short passage is revealing. In 1955, epidemiology was still a relatively young science, and it was Doll who energetically was developing and implementing its methods. Doll’s use of his cohort study was not undertaken just because it was deemed “wise,” but because the method had evolved to the point that Doll could cast offer the asbestos company in question a reasonably rigorous method of answering their “wise” concern.

Contrary to Selikoff’s suggestions, the smoking association was better established in 1955, when Doll published, than was the asbestosis association. By the time Doll published his famous paper on the association between asbestosis and lung cancer, he had published three studies on the association between smoking and lung cancer. Interestingly, Doll later acknowledged that his failure to obtain smoking histories was purely an oversight. By the time Selikoff undertook his studies of asbestos insulators in the late 1950s, a wise investigator would have known that he needed to be very careful smoking histories to study the role of asbestos in an exposed cohort.

Perhaps more revealing yet, however, was Selikoff’s counterfactual assertion that Doll’s 1955 study was conducted too early to assess the role of tobacco in lung cancers observed in the early 1950s. By the early 1950s, cigarette smoking was well established in both in the U.K., and in the U.S., and had been so for several decades. Here are the data for the United States:

 

Correlation between smoking and lung cancer in US males, showing a 20-year time lag between increased smoking rates and increased incidence of lung cancer.

Correlation between smoking and lung cancer in US males, showing a 20-year time lag between increased smoking rates and increased incidence of lung cancer.

National Cancer Institute Figure 2003

And here are the data from the United Kingdom:

 

Figure 1

Figure 1

Figure 1, from Robert Platt, et al., Smoking and Health: A Report of The Royal College of Physicians of London on Smoking in relation to Cancer of the Lung and Other Diseases 3 (1962).