One of my favorite topics to blog about over the last six or seven years has been the topic of overdiagnosis and overtreatment. These are two interrelated phenomena that most people are blissfully unaware of. Unfortunately, I’d also say that the majority of physicians are only marginally more aware than the public about these confounders of screening programs, if even that.
Overdiagnosis has long been appreciated to be a major impediment to translating programs to screen for disease into better outcomes in a number of diseases but has only recently really seeped into the public consciousness, beginning in particular in 2009 when the United States Preventative Services Task Force (USPSTF) issued mammography recommendations that pushed back the recommended age to start screening to 50. Certainly, the concept of overdiagnosis is counterintuitive. After all, why do we screen for disease in asymptomatic people? The reason is simple—and maddeningly intuitive. We screen for disease based on the belief that catching potentially deadly diseases like cancer early, before they produce clinical symptoms, will allow earlier intervention and save lives. It seems blindingly obvious that this should be the case, doesn’t it? Unfortunately, real life biology and pathophysiology aren’t quite so neat and tidy, and the relationship between early detection and improved survival is muddied by phenomena such as lead time bias and the Will Rogers effect, in addition to overdiagnosis.
What is overdiagnosis? In brief, it is the detection of pathology or disease that, if left untreated, would never endanger the life of a patient or even harm him. Note that overdiagnosis is not the same thing as a false positive. A false positive occurs when a test detects disease that isn’t really there; in contrast with overdiagnosis there is definite pathology. The disease being screened for is there, at least in an early form. It’s just that, at the very early stage detected, it’s either not progressive or so indolent that the patient will grow old and die of something else before it would ever cause a problem. Indeed, it’s been estimated that as many as one in three breast cancers detected by mammography in asymptomatic women might be overdiagnosed and that one in five might spontaneously regress. However, because we don’t know which ones are unlikely to cause harm and haven’t worked out a safe method of observing them and intervening if they look as though they are progressing, we are obligated to treat them all when discovered. The problem of overdiagnosis has led to multiple alterations in what once were considered definitive recommendations for screening mammography, first by the USPSTF and most recently by the American Cancer Society.
You can see how this might be the case if you consider two examples I discussed in an old post. First, in autopsy series, foci of prostate cancer can be found in at least 75% of men over the age of 80; yet the vast majority of men who die after age 80 die of something other than prostate cancer. Second, in autopsy series, thyroid cancer can easily be found in 36% of adults, and investigators have estimated that if the slices were made thinly enough for microscopic examination they could have “found” thyroid cancer in close to 100% of adults between 50-70, even though clinically apparent thyroid cancer receiving treatment only had a prevalence in the population studied of around 0.1%.
Speaking of thyroid cancer…
Fukushima and the “epidemic” of thyroid cancer in children
If there were ever a better example of the adage, oft-stated by yours truly, that the more intensively you look for a disease the more of it you will find, it’s hard to find a better example of this than the aftermath of the March 2011 meltdowns at the Fukushima Daiichi Nuclear Power Plant in Japan as a result of damage sustained a tsunami that hit Japan. It was the largest nuclear disaster since Chernobyl in 1986, and its effects are still being felt.
Now here’s where overdiagnosis is illustrated. An article in Science from earlier this month reported:
Months after the disaster, Fukushima Prefecture set about examining the thyroids of hundreds of thousands of children and teens for signs of radiation-related cancers. The screening effort was unprecedented, and no one knew what to expect. So when the first round of exams started turning up thyroid abnormalities in nearly half of the kids, of whom more than 100 were later diagnosed with thyroid cancer, a firestorm erupted.
Does this sound familiar? First, consider that the amount of radiation emitted by the Fukushima reactors was one-tenth that emitted by Chernobyl and that, as reported in this article, the World Health Organization (WHO) estimated that the 12 to 25 mSv of exposure in the first year after the accident in the hardest hit areas might result in tiny increases in cancer incidence. By way of comparison, people receive around 2.4 mSv of background radiation per year and a chest X-ray delivers around 0.1 mSv (here is a a handy visual comparison from xkcd). So why did the Fukushima Prefecture undertake this screening program? Here’s why:
Memories of Chernobyl got Japanese authorities worrying about thyroid cancer. The fallout from that April 1986 accident included radioactive iodine, which settled across swathes of Belarus, Russia, and Ukraine, contaminating pastures grazed by dairy cows. Children who drank the tainted milk accumulated the radioactive iodine in their thyroids. (Adult thyroids absorb less iodine.) A 2006 World Health Organization (WHO) study found that in the most contaminated areas, there had been about 5000 thyroid cancer cases among those who were under 18 at the time of the accident, though the report noted that more cases could emerge over time. The United Nations in 2006 attributed 15 childhood thyroid cancer deaths to Chernobyl. Caught early, the cancer is almost always cured by removal of the thyroid gland.
The WHO had also estimated that the thyroid-equivalent doses in 2011 were between 100 and 200 mSv in the worst-hit areas and 10-100 mSv elsewhere in the Fukushima Prefecture as delivered by inhalation, external exposure from groundshine (radiation emitted from isotopes that land on the ground), and ingestion. So it was not unreasonable that there might be an increase in thyroid cancer among people who were under 18 at the time of the disaster. However, the WHO also noted in 2013 that intensive screening for thyroid cancer was likely to increase the prevalence of thyroid cancer solely through a screening effect, which is basically another term for overdiagnosis.
The results of the screening program thus far were reported in an October 2015 article published in Epidemiology by an environmental epidemiologist at Okayama University named Toshihide Tsuda, entitled “Thyroid Cancer Detection by Ultrasound Among Residents Ages 18 Years and Younger in Fukushima, Japan: 2011 to 2014.” In it, Tsuda et al examined data from the screening program. Basically, all residents 18 years old and younger in March 2011 were screened by ultrasound during the 2011-2013 fiscal years according to the nearest area first, in 2011; the “middle area” second in 2012; and the “least contaminated area” in 2013. “Nearest area” denotes the area closest to the plant which was the most contaminated, while “middle area” and “least contaminated” area denote the areas further away with decreasing levels of contamination. A second round of screening began in April 2014 and is scheduled to be completed this month.
Positive findings were handled thusly:
Subjects with positive findings received a secondary examination, and if necessary, underwent fine needle aspiration. When cancer cells were detected, the patient was followed and operated on at an appropriate time. The excised thyroid tissue was examined histologically. Explanations about medical decisions, such as timing of fine needle aspiration and surgery, were not made publicly available by the prefecture. In addition to the progressive course of the disease, a patient’s school schedule was also considered in the timing of procedures because of the need for hospitalization. Based on information from Fukushima Prefecture, most fine needle aspirations and surgeries were performed by doctors from Fukushima Medical University.
Overall, of 367,687 residents were screened in 2011, 298,577 (81%) underwent the first round screening by the end of December 2014, with the proportion of residents of the respective areas who had undergone screening being 88% in 2011; 87% in 2012; and 74% in 2013. Basically, Tsuda et al found between 0 and 605 cases per million in the various areas
Among 2,251 ultrasound screen-positive cases by the end of December 2014, 2,067 cases were examined in secondary examinations, which detected 110 thyroid cancer cases, as indicated by the presence of cancer cells by cytology after fine needle aspiration. Among the 110 cases, 87 cases were operated by the end of December 2014: 86 cases were histologically confirmed (83 papillary carcinomas and three poorly differentiated carcinomas), and one case was diagnosed as a benign tumor.
Overall, according to Tsuda et al, this finding represents an approximately 30-fold increase in the number of thyroid cancer cases among children and adolescents in the Fukushima Prefecture. Not surprisingly, there was considerable alarm after the publication of these results that still persists until today. Just last month, for instance, The Telegraph published the alarming headline “Fukushima disaster: Children cancer rates rise with 16 new cases.” Back around the time this study was first published online, it wasn’t uncommon to see articles and posts with titles like ‘Fukushima: “Alarming” Rise in Child Thyroid Cancer Rates‘ and “Researcher: Children’s cancer linked to Fukushima radiation.” Even the occasional medical site fell for the hype.
There is almost certainly no “thyroid cancer epidemic” in Fukushima
This study is, not surprisingly, being held up as evidence that there is a radiation-induced “epidemic” of thyroid cancer in Fukushima Prefecture, all due to the meltdown of the nuclear reactor. But is there? The authors of the study itself acknowledge that at least some of the increase could be due to the “screening effect” (or, as I call it, overdiagnosis) detecting silent thyroid cancers in children but blithely dismiss this possibility as unlikely “magnitude of the IRRs was too large to be explained only by this bias.” This, of course, leads to the question, “How could Tsuda know this?” The answer to that question is obviously, “He can’t.” First, he doesn’t compare his results to the known prevalence of small, subclinical thyroid cancer in children, which, admittedly, is difficult to know because autopsy series of children are hard to carry out. Second, he doesn’t know that the apparent prevalence of a disease can’t be increased by 30-fold by vigorous screening. We know, for instance, that the prevalence of ductal carcinoma in situ has increased at least 16-fold since mammographic screening began in the 1980s, all due to mammographic screening, as I have discussed on several occasions before.
Perhaps his strongest argument is that 40 of 54 cases operated on in the Fukushima Prefecture had positive lymph nodes. However, looking at the report cited by Tsuda and using a bit of the old Google Translate (the document is in Japanese), I see that 52 of the 54 cases showed papillary carcinoma, which is the most common (and least aggressive) form, and it’s not clear to me, due to the limitations of Google Translate, exactly what was found in all these cases. On the other hand, it strains credulity to believe that the relatively small dose of radiation to which the children of these areas were exposed could result in thyroid cancers, including node-positive thyroid cancers, in such a short period of time. Remember, these are primarily the results of the first screening, some of which occurred less than a year after the meltdowns. In cancer biology, that’s a short period of time.
It turns out that the reaction to Tsuda’s paper among epidemiologists wasn’t so favorable, though. For example:
Scientists emphatically disagree. “The evidence suggests that the great majority and perhaps all of the cases so far discovered are not due to radiation,” says Dillwyn Williams, a thyroid cancer specialist at University of Cambridge in the United Kingdom. In journal papers and in a series of letters published last month in Epidemiology, scientists have attacked the alarmist interpretations. Many acknowledge that baseline data from noncontaminated areas were needed from the outset and that the public should have been better educated to understand results and, perhaps, to accept watchful waiting as an alternative to immediate surgery. But most also say the findings hint at a medical puzzle: Why are thyroid abnormalities so common in children? The “surprising” results of the screening, Williams says, show that “many more thyroid carcinomas than were previously realized must originate in early life.”
Indeed, at the time Tsuda’s report was published, there was even an accompanying editorial by Scott Davis, who noted:
As stated at the beginning, this commentary is intended to broaden the context or perspective from which the report by Tsuda and colleagues can be evaluated. I have highlighted some of the major challenges faced in a large-scale disaster of the kind experienced in Fukushima, particularly regarding the collection of data from individuals. In this respect, it was not possible to collect the detailed data needed to estimate an individual radiation dose. Therefore, the findings cannot contribute to the two most urgent scientific questions: the characteristics of the dose response curve at low doses, and the details of the role of other factors that might modify the risk of thyroid cancer associated with radiation exposure. Similarly, these findings do not add anything new regarding radiation-induced (or related) thyroid cancer.
Davis further noted that, “given the preliminary geographic dose estimates, any excess of cases due to radiation from the Fukushima plant would be too small to detect using epidemiologic methods,” and that the primary usefulness of Tsuda’s work is to help the Japanese government decide how to allocate health care resources for the victims of Fukushima.
The letters received by Epidemiology from radiation biologists, physicists, and epidemiologists were scathing. For instance, it was noted that the authors used an unvalidated method to estimate IRRs, that the study design was not clearly labeled as an ecological study (for more on ecological studies and the ecological fallacy that vastly overestimates correlations, go here), and, of course, that there could indeed be a massive increase in prevalence due solely to a screening effect, as Richard Wakeford and colleagues point out:
Thyroid disease screening with ultrasound can have a dramatic effect on the detection of thyroid nodules. A 15-fold increase in the incidence of thyroid cancer occurred in South Korea after the introduction of a national cancer screening program in 1999, with the incidence rate in regions increasing in direct proportion to the proportion of screened people. Consequently, it is inappropriate to compare the data from the Fukushima screening program with cancer registry data from the rest of Japan where there is, in general, no such large-scale screening. The proper comparison is between different screened areas within Fukushima Prefecture, since significant radioactive contamination from the accident was confined to a relatively small part of the prefecture.
There is no statistically discernible difference in thyroid cancer prevalence between the low, intermediate and high contamination areas of Fukushima Prefecture. The prevalence ratio (PR) for the highest to lowest contamination areas was 1.08 (95% CI: 0.60, 1.96), and the highest prevalence was seen in the area with an intermediate level of contamination (PR= 1.21 (95% CI: 0.80, 1.82)). Further, the measured levels of radioactivity in thyroids in Fukushima Prefecture were far lower4 than would be needed to elevate cancer rates as much as Tsuda et al.1 claim.
But here’s the letter that really demonstrates that Tsuda’s study does not identify that there is any sort of thyroid cancer “epidemic.” It’s by Noboru Takamura at the Department of Global Health, Medicine and Welfare, Atomic Bomb Disease Institute, Nagasaki University:
We recently conducted thyroid ultrasound screening, using the same procedures as the Fukushima Health Management Survey, in 4,365 children aged 3–18 years from three Japanese prefectures, and confirmed one patient with papillary thyroid cancer (prevalence, 230 per million). Furthermore, we recently reviewed findings of thyroid ultrasound screening conducted in Japan.3 In one survey, 9,988 students underwent thyroid screening and four students (including one foreign student) were subsequently diagnosed with thyroid cancer (prevalence, 300 per million). In another study at Okayama University that examined 2,307 students, three patients with thyroid cancer were found (prevalence, 1,300 per million), while at Keio High School, of 2,868 female students examined, one was found to have thyroid cancer (prevalence, 350 per million). These results show that the prevalence of thyroid cancer detected by advanced ultrasound techniques in other areas of Japan does not differ meaningfully from that in Fukushima Prefecture.
In other words, putting it all together, when advanced ultrasound techniques are used to screen for thyroid cancer in children in Japan, there is no detectable difference between the prevalence of thyroid cancer in children in the three areas Tsuda et al surveyed, and there is no detectable difference between the prevalence of thyroid cancer detected this way in the Fukushima Prefecture and that observed in other studies examining students in other parts of Japan. Tsuda’s study is not good evidence that radiation from Fukushima has created an “epidemic” of thyroid cancer.
The health authorities in the Fukushima Prefecture were well-intentioned, and, in reality, I can’t really fault them for wanting to determine whether the nuclear accident at Fukushima would result in a detectable increase in thyroid cancer in children, given that there is good evidence that radiation to the neck during childhood can increase the risk of this particular cancer. However, it’s clear that many of the doctors involved, although they recognized the issue of overdiagnosis as a potential confounder, didn’t understand just how much overdiagnosis (or, if you prefer, the screening effect) can increase the apparent prevalence of the disease being screened for. If you don’t go into such a massive population screening effort with your eyes wide open, you can easily be fooled into thinking a disease has become a lot more common when in fact all that’s happened is that you’re finding more of it because you’re looking for it a lot harder.
Unfortunately, such findings can cause harm. Besides the anxiety finding such lesions causes patients and their families, virtually all of the children in whom small cancers were found underwent total thyroidectomy, some with removal of the lymph nodes in their central neck. Thyroidectomy is a safe operation, but nonetheless carries the risk of damage to the recurrent laryngeal nerve, which can cause hoarseness due to paralysis of one of the vocal cords, or to the superior thyroid nerve, which can cause permanent changes to the voice. There is also a risk of damage to the parathyroids and permanent hypoparathyroidism. These risks are small but real. Finally, a guaranteed outcome of a total thyroidectomy is the life-long need to take thyroid hormone supplements. Given how uncommon thyroid cancer is throughout life, it’s highly likely that the vast majority of these children did not need to have their thyroids removed.
Dillwyn Williams is quoted near the end of the article as saying that the “surprising” result of Tsuda’s study is that “many more thyroid carcinomas than were previously realized must originate in early life.” I’m not sure why Williams considers that finding so surprising. I suppose it could be considered a surprise that so many more children have small cancers in their thyroid glands than expected, but it shouldn’t be as big a surprise as Tsuda, Williams, and many others view it. When you look for disease carefully and aggressively, you will always find a lot more of it. Then you’re confronted with the dilemma of what to do with it, given that it is subclinical and you don’t know if it will ever progress to cause harm to the patient.
None of this is to say that screening programs don’t save lives. Clearly they do. It’s just that it’s rarely as many lives saved as people assume, because the relationship between early detection and saving lives is nowhere near as clear as people view it. The thyroid screening program launched in the wake of the Fukushima disaster illustrates that very starkly.