Scrutiny-Dependent Cancer and Self-fulfilling Risk Factors

Physicists have long understood that the act of observation can affect the phenomenon being observed. Measuring pressure reduces the air pressure in an automobile tire; measuring voltage reduces the amount of voltage in a circuit. Physicians would do well to understand how the act of observation can affect the apparent incidence of canceras well as apparent risk factors for this disease. That the number of cases of cancer diagnosed is sensitive to the degree of scrutiny is a familiar fact in prostate cancer. Consider the remarkable volatility in the reported incidence of prostate cancer in the United States over the past 40 years (Figure). No known tumor biology or carcinogenic process can explain its rapid rise and quick decline. Instead, the changing incidence is the result of medical practice: the growth of transurethral resection of the prostate as therapy for benign prostatic hyperplasia; the promotion of prostate-specific antigen testinga simple blood test that was frequently offered for free; a period of retrenchment when primary care providers and urologists agreed that screening made no sense in men with limited life expectancy; and, finally, the reaction to recommendations against screening. Figure. Prostate cancer incidence during 19752014, age-adjusted to the 2000 U.S. standard population. Data from the Surveillance, Epidemiology, and End Results database. PSA= prostate-specific antigen; TURP= transurethral resection of the prostate. Prostate cancer is thus the poster child for scrutiny-dependent cancer: When physicians look harder for it, more cases become apparent; when we look less hard, fewer cases become apparent. Indolent prostate cancer is common. Increased scrutiny reveals more cases of cancer, and those revealed are more likely to represent overdiagnosis. Yet, any type of cancer with a substantial disease reservoir of indolent, subclinical forms can be subject to the degree of observer scrutinythe combined effect of the frequency of diagnostic and screening examinations (including the physical examination, imaging studies, and laboratory testing), their ability to detect small irregularities, and the threshold to label these irregularities as cancer. The degree of scrutiny can also affect the assessment of risk factors for prostate cancer. A recent analysis of 2 trials showed that family history is a relatively weak risk factor when all men undergo biopsy (relative risk,< 1.5), but the apparent relative risk is greater than 2.0 in men exposed to current screening and biopsy practices (1). In the accompanying editorial, Albertsen (2) delineated the likely mechanism operating in current practice that produces this result: Men with a family history of prostate cancer are frequently encouraged to undergo [prostate-specific antigen] testing. This, in turn, leads to a higher biopsy rate, which, because of the large pool of subclinical disease, leads to higher rates of prostate cancer diagnosis in this group of men, thereby confirming' the original hypothesis. In short, the risk factor of family history is spuriously strengthened because men with a family history are exposed to greater scrutiny. Indolent thyroid cancer is common, making it another scrutiny-dependent cancer. One of the most prominent risk factors for this disease is sex. According to the SEER (Surveillance, Epidemiology, and End Results) database, women are approximately 3 times more likely than men to be diagnosed with thyroid cancer. However, this risk factor may be more apparent than real, because thyroid cancer mortality in men and women is approximately identical in the United States. Women may simply be more likely to connect with the health care system; are thus more likely to have thyroid nodules detected and evaluated; andgiven the substantial disease reservoirare more likely to have cases of small papillary thyroid cancer diagnosed. Melanoma is also sensitive to the degree of scrutiny. Many consider sun exposure an important risk factor for this disease. However, among patients with melanoma, high sun exposure is associated with halving the risk for death from this disease (that is, the case-fatality rate in patients with melanoma with high sun exposure is one half that of those with low sun exposure) (3). What might explain this paradox? Sun exposure could be a self-fulfilling risk factor. Persons with sun-damaged skin (for example, those with multiple actinic keratoses) may be more likely to see dermatologists, undergo biopsy, and be diagnosed with a biologically favorable melanoma. These factors would explain the lower case-fatality rates associated with sun exposure. For breast cancer, the degree of scrutiny has been shown to affect not only the apparent incidence of disease but also the underlying tumor biology (4). The more scrutiny, the more likely that biologically indolent cases of breast cancer (that is, those that are low grade and luminal) are found. Relative to cancer found by women themselves, mammographically detected cancer is more likely to be indolent. Cancer detected on ultrasonography or magnetic resonance imaging is even more likely to be indolent than that found on mammography. So, as physicians look harder, we not only find more cases of cancer but those that we do find become less importantthat is, more likely to grow slowly and to represent overdiagnosis. These relationships may help explain recent data showing that women living in the highest quintile of neighborhood socioeconomic status (SES) had twice the rate of breast cancer diagnosis as women in the lowest quintile after all the classic breast cancer risk factors were controlled for (5). That high SES is an independent risk factor for breast cancer seems unlikely; that women in high SES neighborhoods are more connected to health care and more likely to undergo not only mammography but also ultrasonography and magnetic resonance imaging is more probable. This phenomenon may also erroneously strengthen risk factors related to parity and be responsible for the paradoxical finding that women diagnosed with in situ breast cancer have 5- and 10-year relative survival rates greater than 100%. Emphasizing that observer scrutiny can have various effects on the assessment of risk factors is important. It may apparently strengthen a genuine risk factor (for example, family history in prostate cancer), and it may spuriously produce a risk factor that does not, in fact, exist (for example, neighborhood SES and breast cancer). However, scrutiny can also apparently weaken a genuine risk factor. Despite a wealth of epidemiologic data showing that the risk for smokers dying of lung cancer is at least 15 times higher than that of never-smokers, population-based computed tomography screening in Japan reported that the risk for lung cancer detection in smokers and never-smokers was virtually identical (6). The degree of scrutiny is thus an important confounder of the relationship between a purported risk factor and the diagnosis of cancer. This phenomenon may prove equally relevant in the assessment of genetic risk factors for this disease. If testing for genetic variants that are apparently associated with cancer becomes commonplace and more aggressive screening is recommend to those with positive results on genetic testing, then the risk associated with these variants will be spuriously strengthened. To avoid such misleading feedback loops, risk factor epidemiology must shift from diagnosis to harder outcomesthose more directly related to the disease process. Although late-stage and metastatic disease are potential candidates, death from cancer is the most meaningful and least ambiguous outcome. As cancer diagnosis becomes increasingly sensitive to observer scrutiny, we hope that those investigating the risk for this disease focus on risk factors for death from cancer, not cancer diagnosis.