Diagnostic Drift in the Reporting of Cancer Incidence

Giving a patient a label that says “ cancer” is a fateful event, and there are strong reasons for minimizing the number of errors, both false-negative and false-positive. The first step toward reducing both kinds of errors is, surely, to improve agreement among pathologists about how they read specific specimens. A second, and even more important, area needs further development. Over a period of many decades, we have become accustomed to thinking of the interpretation of the pathologist as the “ gold standard” against which other diagnostic measures are evaluated. However, we must question this practice in light of developments in cancer screening and improved understanding of the biology of proliferative lesions. The work of Wells et al. (1) and the commentary of Page et al. (2) do not deal with changes over time in the boundaries that define what we call “ cancer.” At any one fixed time, agreement will be good if most pathologists have changed their standards in much the same way, but apparent trends in cancer incidence rates may still be created or masked. For example, the reported incidence rate for cancer of the prostate approximately doubled over a 10-year period, without notable impact then or later on the mortality rate. This was a reflection of the efficacy of the prostate-specific antigen test in finding lesions that were then biopsied and reported as cancer but have not (at least, not yet) displayed the morbidity and mortality characteristic of prostate cancer. A 1974 spike in the reported incidence of cancer of the breast came at a time when the wives of the U.S. president and vice-president had breast cancer and coincided with the first broad public attention to mammography. That spike was not followed by any change in mortality detectable at the national level. Similarly, there is evidence that lung cancer screening programs lead to the detection and treatment as cancer of many lesions with prognoses far better than the usual forms of lung cancer. There are parallel chains of evidence for cancer of the thyroid and ovary. Whether these newly detected lesions should even be called “ cancer” is not always clear. Microscopic appearance may be the best surrogate yet found for the future biologic behavior of a lesion, but “ best” in this context does not necessarily mean good enough. There is little point in argument about whether the new approach to the diagnosis of cancer leads to over-diagnosis, or the old one to under-diagnosis; the important thing is that they lead to incommensurable results, and the broadening definition of cancer can create an impression of an alarming rise in incidence with little change in the biologic situation. What is needed is a new way to define a “ gold standard,” suitable for routine use in the practice of oncology, that is a better predictor of future biologic behavior than is microscopic appearance. This is likely to be a difficult task, but the time to start is now.