A 72-year-old woman suffers a life changing stroke and begins to agonize over the possible causes. Was it her slightly elevated blood pressure or cholesterol level? Perhaps it was the hormone replacement therapy she was on for several years following her hysterectomy. Or maybe it was years of inhaling second-hand smoke from her husband before his death from a heart attack a few years ago. Her various physicians are unable to give her a definitive answer as to what caused her stroke. It may be a combination of these factors or something else entirely. There is no cause that can be identified with any high degree of certainty.
In health and medicine, as in so much else in life, we seek certainty. We want to know that getting the right screen, lowering blood pressure and cholesterol, and taking the right medication will guarantee longer, healthier lives. If we are unfortunate enough to suffer a disease or condition, we want to know the exact cause and expect a treatment that will be successful most if not all of the time. We tend to forget that medicine is as much an art as it is a science and may be unaware that the correlation between such actions as lowering cholesterol, for example, and preventing a primary heart attack are not as high as we would like or as we have been lead to believe.
The ethical and health policy implications of such desires can be enormous. If you believe that mammograms for women lead to a significant reduction in deaths due to cancer, for example, and that this reduction outweighs the potential for risk or harm from such screenings, then you are more likely to argue that a just society would ensure that women have access to them. If not, then you are less likely to make this argument. The estimate of risk and benefit will inform your ethical and political stance.
At issue is our understanding that risk and benefit is a complex business, certainly more complex than is often presented by the media. That is why books such as The Illusion of Certainty are greatly needed. Written by two environmental engineers, the book discusses our desire for certainty with respect to healthcare and to various environmental health hazards. Satisfying this desire is, to say the least, elusive. An important distinction in the book is the difference between relative and absolute risk reduction. Relative risk reduction “reflects the number of people who benefit from an intervention, compared to the number of people who get the disease if they aren't treated (but not compared to the total number of people being considered).” When results are expressed as a relative risk reduction, “the absolute risk levels for the experimental and control groups are not known.” To use an example from the book, “if a new drug comes along that reduces the number of liver disease deaths from 6 out of 100 (6%) to 4 out of 100 (4%), then the relative risk reduction is 33 percent, so you could say that the drug cuts your chances of dying of this liver disease by 33 percent.” Of course, because the chance of dying from liver disease is fairly small among this group, reducing this chance by 33% is not, one could argue, especially beneficial.
Absolute risk reduction presents an arguably more accurate picture. This is the difference between the absolute risks in two groups. Using the same example, if 6 out of 100 (6%) people in a particular group will get a certain kind of liver disease and die from it, and this drug reduces the number to 4 out of 100 (4%), then absolute risk reduction is 6% − 4% or 2%. Absolute risk reduction “compares the number of people who will benefit from intervention (in this case, 2) to the total number of people being considered (in this case, 100).” Another way of looking at this is that in administering this drug to 100 people, 2 people will obtain an admittedly significant benefit, whereas 98 will not receive that benefit.
Newspaper headlines and television news reports, of course, often rely upon the more sensationalistic relative risk reduction figures. Reducing the risk of liver disease by a third is much more impressive and interesting than reducing it by 2%.
Armed with a clear understanding of the difference between relative and absolute risk reduction, the authors go about evaluating the evidence supporting various screenings, treatments, and environmental hazards. They discuss in short, concise, and highly readable chapters such topics as mammograms, prostate cancer screening, cholesterol reduction, colorectal cancer screening, smoking, and the health risks of radon. Certainly one of the highlights of their approach is a visualization tool they devised called the Risk Characterization Theater (RCT). This graphic shows a 1000 seat theater in which each seat can be colored in to show the absolute benefit or risk of a treatment, screening, or environmental hazard. Thus we see that according to their analysis of the studies that one colon cancer death will be avoided for every 1000 people screened. The RCT shows one filled-in seat in the 1000-seat theater.
Rifkin and Bouwer's analysis of studies of cholesterol and heart disease may be especially surprising even to those who follow the literature on healthcare. Their reading of the literature suggests that cholesterol levels cannot accurately predict whether someone will develop Coronary Heart Disease (CHD) because individuals with normal and elevated blood serum cholesterol have “essentially the same incidence of CHD.” According to their estimations, lowering cholesterol in a group of 1000 with cholesterol between 210 and 250 mg will result in avoiding at most 1 additional death among this group. It is important to note that this is an annual figure. Over a period of 5 years the number would be five fatal heart attacks avoided. Lowering cholesterol by taking statins presents a somewhat better picture, because statins seem to have effects beyond lowering cholesterol. On their estimation statins would result in approximately 30 out of 1000 experiencing a benefit over a 5-year timeframe. For those who already have cardiovascular disease the benefits are even greater, roughly 67 out of 1000 benefiting.
Because the correlations between lowering cholesterol and avoiding a heart attack hardly approach high levels, why, then, is there such an emphasis upon this by various reports from physicians, drug companies, and even government agencies? Self-interest is a factor, along with an inadequate understanding of absolute and relative risk. Important, too, is the effect of applying absolute risk reduction to large numbers of people. Avoiding 30 heart attacks over a 5-year period in a population of 1000 may not seem impressive, but avoiding 30,000 in a population of 1,000,000 gets attention. As an individual patient, the question is whether you want to take a medication that does have the possibility of adverse reactions on the assumption that you are not one of the 970,000 out of 1,000,000 that will not get the desired benefit from it.
Except for their discussion on the health effects of smoking, the other discussions by the authors tend to be somewhat counter to what has become conventional wisdom. Their risk characterization theater on smoking graphically illustrates the results of a British Doctors Study published in 2006. The graphic shows dramatic increases in deaths over time between smokers and nonsmokers. If you want to demonstrate to someone the dangers of cigarette smoking, I can think of no quicker or better depiction.
One of the drawbacks of this book might be that it is written by environmental engineers rather than physicians. Their discussions do not particularly suffer from a lack of clinical knowledge, though it is likely that the sections on environmental hazards will be less interesting to those focused on healthcare delivery, just as the chapters on screenings, tests, and treatments will be less interesting to those in fields related to the environment. Rifkin and Bouwer attempt to make up for their lack of clinical expertise by including a commentary from a physician, Dr. Bob Sheff. Although Sheff's main recommendation that patients partner with their physicians in order to better understand treatment options and their true risks and benefits has merit, I found this solution somewhat inadequate. In many respects physicians may be just as misled as patients, perhaps even more so because they are likely to be influenced by pharmaceutical companies and vendors in ways that patients are not. Additionally, it is difficult for a patient to form a relationship with a physician in the brief amount of time that is usually associated with an office visit. Addressing the uncertainty issues raised by Rifkin and Bouwer may entail looking hard at current forms of healthcare delivery and formulating new ones, such as approaches that include group interactions.
In any event, Rifkin and Bouwer have done a fine job of describing the roots of uncertainty as it relates to health. They have also adequately explored the state of our current knowledge with respect to various treatments, screenings, and environmental hazards. Although we can point to certain risk factors associated with a 72-year-old female suffering a stroke, we cannot say with complete certainty why the stroke happened or how it could have been avoided. Complete certainty will likely never be attainable. The question is whether the public's quest for and belief in certainty will prevent practical and ethical discussions of healthcare based upon probabilities rather than absolute knowledge.
