“
Two decades ago the federal government invited 150,000 men and women to participate in an experiment of screening for cancer in four organs: prostate, lung, colon, and ovary. The volunteers were less likely to smoke, more likely to exercise, had higher socioeconomic status, and fewer medical problems than members of the general population. Those are the kinds of people who seek preventive intervention. Of course, they are going to do better. Had the study not been randomized, the investigators might have concluded that screening was the best thing since sliced bread. Regardless of which group they were randomly assigned to, the participants had substantially lower death rates than the general population—for all cancers (even those other than prostate, lung, colon, and ovary), for heart disease, and for injury. In other words, the volunteers were healthier than average. With randomization, the study showed that only one of the four screenings (for colon cancer) was beneficial. Without it, the study might have concluded that prostate cancer screening not only lowered the risk of death from prostate cancer but also deaths from leukemia, heart attack, and car accidents (although you would hope someone would raise the biological plausibility criterion here).
”
”
H. Gilbert Welch (Less Medicine, More Health: 7 Assumptions That Drive Too Much Medical Care)