BOOK REVIEW: Changing the Odds: Cancer Prevention Through Personal Choice and Public Policy

October 1, 1996

The premise of Steen's book is that nearly everyone is confused about cancer prevention. The public, for whom the book is written, is the most confused, but cannot be faulted. According to Steen, the responsibility for their confusion lies with scientists and the press. Scientists make mistakes by reporting preliminary data from inadequate and mostly "workmanlike, undistinguished" (page 3) studies. The press, in turn, repeats these mistakes, adds some of its own, and so oversimplifies a complex topic that the public, in the end, receives "unconnected facts, partial truths, and outright lies" (page 3).

The premise of Steen's book is that nearly everyone is confusedabout cancer prevention. The public, for whom the book is written,is the most confused, but cannot be faulted. According to Steen,the responsibility for their confusion lies with scientists andthe press. Scientists make mistakes by reporting preliminary datafrom inadequate and mostly "workmanlike, undistinguished"(page 3) studies. The press, in turn, repeats these mistakes,adds some of its own, and so oversimplifies a complex topic thatthe public, in the end, receives "unconnected facts, partialtruths, and outright lies" (page 3).

To this morass of disinformation, Steen brings his solution. Hewill sort out the risks by reviewing what the jacket cover describesas "over 2,000 primary sources and references." Afterpresenting evidence and ranking over 200 different risks, Steenrecommends to the confused public what course they can take toprevent cancer.

It is a noble task, and Steen should not be faulted for tryingto accomplish it. For some long-accepted prevention messages,Steen is right on the mark. Smoking is responsible for many cancersand deserves emphasis. Unfortunately, Steen's thesis is not wellsupported by this example: The public is not likely confused aboutthe dangers of smoking; it is a public health message now 30 yearsold. The point is that we should not criticize Steen for recitingthe obvious, nor can we evaluate his book on this basis. Rather,we must evaluate it on the greater task--the selection and reviewof evidence, the ranking of factors, and the recommendation ofstrategies--keeping in mind that Steen has criticized scientistsand the press for failing to carry out their responsibilities.

My first concern relates to the strategy by which Steen has selectedevidence. He has chosen a tiny fraction of the studies publishedon topics without informing the reader of the process that heused to make those selections. Consider for example, the associationbetween smoking and cervical cancer. From 1966 through 1995, 39case-control studies, 5 cohort studies, 10 comprehensive reviews,2 meta-analyses, and at least a dozen studies of biologic parametershave been published. From these 70 publications, Steen selects2: a case-control study showing a positive effect and a clinicalstudy showing that, in smokers, nicotine--a marker of exposure--reachesthe cervical epithelium. Although the results of these two studiescoincide with the accepted public health notion that smoking isunhealthy, there is no consensus in the scientific community onwhether smoking has an independent causal role in the etiologyof cervical cancer.

This example brings up an important concern: to what extent hasSteen accurately represented the current status of scientificthinking on the other 200 topics included in his book? The publichas no way of knowing.

The ranking process used in the book is also troublesome. Steenuses the size of the P value as his measure of validity, thusignoring the more likely challenge; namely, confounding. Steenalso groups cancers by incidence, and uses the group number (1through 5) as a multiplier applied to the relative risk from thestudy with the smallest P value to provide the ranking number.But prostate and bladder cancers are grouped together (group #4)despite the fact that the incidence of prostate cancer (132,000cases) is much closer to that of colon cancer in group #5 (156,000cases) than it is to bladder cancer (51,600 cases) in group #4,which, itself, is much closer to uterine cancer (45,500 cases)in group #3. The result is a ranking of risks that is nearly asarbitrary as the group cut-off points themselves.

A Frustrating List

Perhaps most frustrating, however, is the fact that Steen's finallist of risk factors for preventable cancers (page 269) and hislist of "what to do" to prevent those cancers do notmatch up. He describes several strategies (pages 273 and 274),from stopping smoking to learning the 10 warning signs of cancer,but ignores at least 9 items listed earlier as preventable risks.These include: exposure to DDT, stress, use of oral contraceptives,eating more carbohydrates, eating fewer calories, having children,and exposure to black hair dyes. All were identified by Steen'smethod of selecting evidence and his ranking scheme.

It is unclear why these items were not transferred from the listof risks to the list of preventive strategies. Perhaps some donot offer preventive promise. If so, then it is reasonable toquestion the way in which they were identified. On the other hand,perhaps Steen simply overlooked them or made some other unspecifiedjudgment regarding their validity or reasonableness. In eithercase, this book provides the public--for whom it was written--withone more reason to be confused about what should be done to preventcancer.