And yet another wrinkle arises from the ambiguity of some test samples: In more cases than you'd expect, pathologists disagree on whether a particular image shows cancer or not. In one especially telling study published in 1996 in Human Pathology, a panel of eight highly credentialed pathologists reviewed a series of 37 slides to determine which represented melanoma, or skin cancer, and which showed benign lesions. Surprisingly, for a significant minority of slides—38 percent of them—two or more pathologists disagreed with the majority opinion. This finding and others like it are important because they highlight the subjectivity associated with cancer diagnosis; subtle, cellular alterations may fall into a gray area between cancer and not-cancer, particularly for early stage cases—the very realm in which screening is routinely used.
A Risk of Overtreatment
Once an abnormality is labeled cancer, it is difficult for the patient and physician to "do nothing." In addition to subtle pressures on the doctor (the fear of malpractice, the prevailing standard of aggressive care), it is simply not possible to know whether a given, tiny lesion represents a case of overdiagnosis or a clinically significant finding. The inclination is usually to err on the side of caution and to proceed with treatment. But treatments can be harsh, ranging from surgery (in the case of women with DCIS) to radiation and chemotherapy, all of which have severe side effects and which no one would want to undertake unnecessarily.
The furor over mammography—triggered in part by a meta-analysis published in the Lancet in 2000—focused on the contention that screening did not reduce women's overall mortality; worse still, in one study, mortality for the screened women was actually slightly higher than for the control group, a finding attributed to overtreatment—in this case, increased heart attacks caused by radiation therapy. Radiation protocols have now been changed to minimize the heart region's exposure. But a fundamental question remains: Are there cases in which the cure is worse that the disease?
The answer is clearly yes—at least some of the time. In prostate cancer, abnormal cells are often slow-growing, and treatment can cause impotence and incontinence. Dr. Stephen Taplin, a senior scientist at the National Cancer Institute, helpfully compares some prostate cancers to gray hair, more a byproduct of aging than a life-threatening issue. "If I made men impotent and incontinent because they had gray hair, there wouldn't be any question I'd be hurting them," he said in a phone interview. But if these problems occurred as a result of aggressive cancer treatment, "most men would say, 'Doctor, you've saved me!'"
A final, terrible irony is that while screening leads to overdiagnosis and overtreatment, it also misses some cancers—and these tend to be the most aggressive, fastest-growing ones. This is because fast-growing abnormalities (for any type of cancer) are statistically more likely to develop during the window between tests, causing problems before, say, an annual screening can provide fair warning.
So how to explain the public's simple faith in testing—the significant gap between scientific evidence and popular perception? Many social and cultural pressures conspire here—and a complete catalog would no doubt include everything from physician report cards (which rate physicians according to the percentage of patients screened) to celebrity testimonials, from the financial interests of some doctors and health networks invested in imaging to the well-intentioned work of advocacy groups, particularly for women.
But one factor that has really muddied the debate deserves special mention: the "five-year survival rate," as it is called, often deployed to support screening. ("Five-year survival" represents the percentage of people diagnosed with a particular kind of cancer at a particular time who are still alive five years later.) As Dr. H. Gilbert Welch points out in his excellent book, Should I Be Tested for Cancer? this is a highly misleading statistic for the following reason: When cancers are detected early, the five-year survival rate (dated from the time of diagnosis) will necessarily improve—even if patients live no longer than they would have otherwise. Overdiagnosis further inflates the figure since more people are identified with nonprogressive cases who can be expected to live longer.
The public-health challenge, then, is to convey a more balanced, realistic message about cancer testing so that people will be receptive to negative as well as positive news about particular tests. There is good evidence to support regular Pap smears in women, no good evidence (at least not yet) to support routine lung-cancer screening for former smokers. Thus, it is particularly disheartening to see a large-scale screening program such as the New York Early Lung Cancer Action Program (full disclosure: NY-ELCAP is led by researchers at Cornell Medical School, where this writer studied medicine), which lacks a control group and so will not be able to clarify whether screening 10,000 former smokers actually saves lives. (A devastating critique of NY-ELCAP by Dr. Steven Woloshin, along with Schwartz and Welch, is available in the Lancet, but not for free.) That this massive project is heavily funded by New York's tobacco settlement fund provides a further, unfortunate twist.
Ultimately, the public needs to set aside automatic enthusiasm for screening and develop a new kind of savvy—one that balances hope with a certain dose of healthy skepticism and leads people to embark on testing only after considering a host of variables, both personal and scientific. As it turns out, in cancer screening, as in so much else, there really isn't a free lunch.