How Understanding Thyroid Cancer in Belgium Can Help Us Mitigate the Problem of Increasing Incidence

There is an epidemic of thyroid cancer in the south of Belgium. Or is there? Van den Bruel et al (1) report that the incidence of thyroid cancer in the north is 4.1 per 100 000 people, and in the south, the rate is twice as high, 8.3 per 100 000 people. Although this could represent an epidemic of disease in one part of the country, it is also possible that the difference is caused by geographic variation leading to overdiagnosis. Van den Bruel et al (1) examined this hypothesis by looking closely at geographic variation in rates of various activities that contribute to the detection of thyroid cancer. But what exactly are the concepts of overdiagnosis and geographic variation, why do they matter, and how do they relate in this case? This report is complex, so to examine the concepts, we will focus on just two findings: rates of imaging studies and rates of needle biopsy before surgery. These two rates are of particular interest because they provide clues about what we can do to alleviate the problem of thyroid cancer overdiagnosis. Overdiagnosis is the detection of a disease that was never destined to affect the person in their lifetime (2). It is a problem because patients diagnosed with the disease can only be harmed by treatment, not helped, because treatment was not necessary. Overdiagnosis of cancer is possible when two conditions are present: a subclinical reservoir of disease and activities that uncover it. In thyroid cancer, the presence of a subclinical reservoir has long been acknowledged. In 1947, thyroid cancer was described as an incidental autopsy finding in people who had died of other causes, never knowing they had the disease (3). The subclinical reservoir is large; in 1985, a very careful autopsy study done by Harach et al (4), in Finland, showed that 36% of people not known to have thyroid cancer during their lifetime nevertheless had one or more foci of thyroid cancer. The hypothesis of cancer overdiagnosis can be strongly supported when 3 things are observed in a population: 1) incidence rates are rising, 2) mortality from the disease stays stable, and 3) the proportion of cancers diagnosed in the early stages increases. The presence of thyroid cancer overdiagnosis in Belgium is supported by the stable to falling mortality rates in combination with a rising incidence rate, particularly among small cancers (5). The purpose of Van den Bruel's work was to examine clinical patterns of care, so they could understand the reasons for the overdiagnosis they presume is likely present. Some will argue, perhaps with some validity, that the true incidence of thyroid cancer is also increasing, because the entire increase in incidence is not due to small cancers. However, a recent paper by Malone et al (6) from the United States suggests that even some of the larger tumors being found in recent years may be part of the overdiagnosis phenomenon. In that paper, almost half of stage III and IV well-differentiated cancers were initially discovered on imaging studies. Even if a concurrent true increase in thyroid cancer incidence is occurring, the lion's share of the increased incidence we have been observing is due to small tumors. Van den Bruel et all (1) were doing the right thing in examining the causes of overdiagnosis, because patients stand to derive real benefit from our lessening of inappropriate identification and aggressive treatment of subclinical disease. The strategy they used, examining geographic variation, helps us to both understand the ways that subclinical thyroid cancer is being uncovered and figure out how we can lessen the detection of the subclinical reservoir. Geographic variation in care is the idea that rates of some types of healthcare vary for reasons that cannot be explained by disease incidence or severity. The problem was first described by Wennberg and Gittelsohn (7) in 1973 when they reported on the variation in rates of tonsillectomy across different hospital service areas in the state of Vermont, without any reason to believe that tonsil-related illness or morbidity differed. Some reasons put forward for geographic variation are service availability, physician practice patterns, and economic incentives. Variation in care is a problem because it implies that care is being administered differently in different areas, which means some people are probably getting too much healthcare, whereas others might not be getting enough. In the 40 years since Wennberg published his first study, researchers at Dartmouth have cataloged wide variations in healthcare spending (8), procedure rates (9), drug prescription rates (10), and even average numbers of diagnoses per person across the United States (11), to name just a few examples. At its core, geographic variations in care observed in developed countries generally means that people are getting care for reasons unrelated to clinical need; some other force is at work that does not reflect the pure interests of the patient. This should gravely concern any physician. Belgium has considerable geographic variation, as Van den Bruel et al (1) lay out so clearly. Imaging rates for computed tomography/magnetic resonance imaging and positron emission tomography are 62 per 1000 person years in the region with low thyroid cancer incidence, but 86 per 1000 person years in the region with high incidence. It is highly unlikely that the large rate differences in imaging between the two regions of Belgium are related to disease incidence or severity. More likely, they are due to variations in how physicians practice, their tendency to order tests for a given clinical scenario. These types of differences in individual physician practice style have been established by previous research (12). Imaging studies that include the thyroid gland but are not thyroid-directed, such as chest computed tomography scans for the workup of pulmonary embolism or blunt trauma, are an important source of the incidental detection of thyroid cancer, particularly cancers that are too small for a patient or clinician to notice (13). Thus, it is likely that some of the thyroid cancers identified in the high-incidence region were first found on imaging done for other reasons. In another, more complex example, needle biopsy rates before surgery are much higher (36%) in the low-incidence region than in the high-incidence region, where rates of needle biopsy before surgery are only about 17%. This is surprising. One would have guessed that an area with high cancer incidence rates would also have high rates of needle biopsy before surgery, just as had been observed in the imaging example outlined above. Instead, it appears that many more people in the high-incidence region are simply sent to surgery directly. It is unlikely that this is due to differing levels of access to healthcare. Belgium has universal health insurance and presumably roughly equitable access to healthcare services across the country. Again, the most plausible reason for the variation is that at the level of the individual physician, there are big differences in personal practice. One possibility that might be put forward to explain the variable rates of needle biopsy before surgery in this study is that some physicians in the high-incidence region do not know how to do needle biopsy or do not believe in it. Another possibility is that referral patterns bypass the endocrinologist in the high-incidence region, sending patients directly to surgery. This latter scenario presumes endocrinologists are the ones doing needle biopsy, which may or may not be the case. Understanding the reasons would require more in-depth analysis of physician practices at a micro level. How do low rates of needle biopsy before surgery translate to higher incidence rates of thyroid cancer? The data in Van den Bruel's work show that in the high-incidence region, there is much more thyroid surgery overall, perhaps because it is being substituted for needle biopsy. The proportion of people identified on pathology as having cancer are similar in both the low- and high-cancer-incidence regions, but because so much more surgery is done in the high-incidence region, a higher number of cancers are uncovered from the subclinical reservoir, which creates the higher observed incidence rates. The time for arguing that the entire increase in cancer incidence is due to a true rise in disease is over. This paper, along with all the other accumulated data from studies of risk factor exposures and other healthcare systems make this clear. Because we are detecting a subclinical reservoir of disease, many patients are being diagnosed with thyroid cancer and treated unnecessarily. Although this by itself should be cause for alarm, there is another reason to be concerned. Treating people who don't need it also adversely affects those whose tumors are on a trajectory to cause harm. For instance, if several people with small incidentally detected cancers are placed into a clinical trial with a patient whose small tumor had come to attention because it was aggressive and had caused a vocal cord paralysis, the trial results will be falsely improved over what they should have been. This kind of unintended result threatens all patients, present and future. The question we need to address now is what are we going to do to solve the problem of overdiagnosis of thyroid cancer? Happily, Van den Bruel et al (1) have provided us with new tools by showing the power of outlining geographic variation in the clinical pathway that leads to the diagnosis of thyroid cancer. The differences in thyroid cancer rates between the north and south of Belgium can largely be explained by practice variation at the level of the individual physician. To understand these variations and lessen them will require diving down into the smallest unit of analysis: examining physicians' beliefs, practices, local networks, and referral patterns. The research methods to delineate these types of networks exist and are just beginning to be exploited (14). Our next steps are to get to this lowest level and effect change by having one person talk to the next. The next steps in the research agenda are to get to this lowest level, to clearly demonstrate these patterns. But we can effect change right now through our own activities. If leaders in the field of thyroid cancer talk openly in their own practice circles about the problem of overdiagnosis, proposing their own ideas for how to address it, they will influence those around them. If that happens, we will have taken the first steps toward mitigating this difficult problem. This work was supported by The Department of Veterans Affairs. The views expressed do not necessarily represent the views of the Department of Veterans Affairs or the United States Government. Disclosure Summary: The author has nothing to disclose.