Radiology's Overdiagnosis Problem

Saurabh Jha, MBBS, MRCS; Tricia Ward


August 21, 2015

Editor's Note: Saurabh Jha, MBBS, MRCS, guest-edited the August 2015 print issue of Academic Radiology devoted to overdiagnosis in radiology. Below is our interview with Dr Jha about his goals in compiling this special edition of the journal.

Medscape: Why did you want to devote an entire issue of a radiology journal to the topic of overdiagnosis?

Dr Jha: My intention was to get the issue to the forefront in radiology. One of the pioneers who drew attention to this problem was the radiologist William Black, in a famous essay with Welch published in the New England Journal of Medicine.[1] But since then, much of the discussion on this topic has come from outside radiology.

I spoke to the editor of Academic Radiology, Sam Baum, who has had personal experience with the healthcare system and feels that this is a topic that needs to be highlighted by radiologists primarily. The important thing for radiologists to do is to acknowledge that we are in the thick of overdiagnosis, whether we like it or not. If we are involved in diagnosing, we are involved in overdiagnosing as well.

To get the broader radiology community to acknowledge this, it was important for a peer-reviewed publication to come out on the topic, and Academic Radiology delivered.

Medscape: You contributed some perspective articles[2,3] to the journal in which you discuss over diagnosis vs false-positives; can you expand on this for our audience?

Dr Jha: With overdiagnosis, we are lowering the bar for what is considered disease. A false-positive is when you call something a disease that is not really a disease. A false-positive is a limitation of the test, whereas overdiagnosis is "exuberance" of the test, so to speak.

The best example is with pulmonary embolism (PE). As we go from the main pulmonary artery to the subsegmental pulmonary arteries, which are very small, you could diagnose a false-positive subsegmental PE—you called it, but it wasn't really there. Or you could have a true subsegmental PE, but that is an overdiagnosis because what you are calling disease is of such little significance that it is equivalent to cracking a small nut with a sledgehammer. The sledgehammer can do more harm to the nut itself. It can be an overreaction to a problem that is quiescent. In this case, the true-positive is in fact an overdiagnosis.

False-positives are a different entity, but the two are linked. They both lead to overmedicalization. If you have a very big PE in the main pulmonary artery (a submassive PE) that is catastrophic, it can lead to death, whereas with a smaller subsegmental PE, you could cause more problems by treating it with anticoagulation vs leaving it alone. The specificity of the test is a lot lower for these smaller arteries compared with the bigger ones, so as the tendency toward overdiagnosing increases, so too does the number of false-positives.

The distinction between overdiagnosis and false-positives is important, but it is a distinction without a difference in a way.

Medscape: The point has been made that the real problem is not overdiagnosis, but rather overtreatment. Would you agree?

Dr Jha: The rationale for that argument is that we radiologists are not responsible for the problems; it's all you overtreaters. There are problems with that logic. If something isn't important to treat, why is it important to diagnose? We make that distinction all of the time, when we read films. We see a cyst in the kidney, and we don't make a song and dance about it because no good comes from the treatment.

We do make a fuss about other things we see, such as cancer or dissection of the aorta—because there will be problems if these are untreated, and treating them can reduce these problems.

The distinction between overdiagnosis and overtreatment is a false one. One leads to the other. If radiologists say, "It's not our problem; it's yours," then what we are really saying is, "You're the clinicians; we're just the quantifiers." Obviously, we're more than that—we make value judgments on what we see, and we amplify when we think a finding is important on the basis of the consequences of that entity not being treated.

Medscape: Do you think the problems is that radiologists are looking for more disease (the more you look, the more you find), or is it that better technologies allow you to see things than you could not see before (such as a very small subsegmental PE)?

Dr Jha: It's a combination of both. We certainly are looking for more. With PE or example, many people who come into the emergency department or other places have a blood test for D-dimer, and when the blood test is positive, they are categorized as having suspected PE. Patients who have cancer undergo more follow-up imaging, and technologies have improved.

Problems occur when we find something that we don't know anything about. Why don't we just ignore it? Why do we feel compelled to overtreat? That bias is the biggest driver of overdiagnosis and overtreatment. Just because a new technology enables me to see something new, it doesn't mean that I should start jumping off the building every time I encounter something I'm uncertain about. Why don't I just ignore it?

I think the best analogy comes from the legal system. They asked, "Is it better that 10 guilty men go free or that one innocent man is sent to the gallows?" In effect, they have said, "We know that information isn't perfect; we know there is a chance we are going to err. So we are going to err toward letting the guilty men go free rather than letting the innocent get strung up." The burden of proof is to prove that someone is guilty, not to prove that one is innocent.

In medicine, we have to prove that somebody is healthy rather than prove that they are diseased. In effect, we have said, "It is better that 10 people who are not going to be harmed by subsegmental PE get overtreated with anticoagulation (and possibly bleed) than it is that one person gets undertreated and harmed by the PE." That is a medical/ethical issue that isn't really explained by the incentive system. I think it goes beyond that.

Medscape: Do you think part of it is a fear of litigation? That is, missing a diagnosis is seen as a mistake, whereas overtreating wouldn't be seen as an error if it's well intentioned?

Dr Jha: The key phrase is "well intentioned." We think "well-intentioned" means erring toward, "This subsegmental PE could harm you," as opposed to erring toward the fact that I am going to put you on anticoagulation, which could harm you.

I certainly think the litigation culture is a large part of it. I sometimes wonder whether our medical culture drives the litigation culture. I'm sure they are both recursive, but if I were to identify what came first, the chicken or the egg, I would say it's the medical culture.

There is so much power in hindsight. We feel there is good intention in overtreating. It means, "I care for you." Whereas if I ignore the subsegmental PE, I sound flippant. The care and the intervention are so interlinked psychologically that it is hard to disassociate one from the other. Subtractive medicine is very hard to propagate culturally; it's very much a can-do culture.

I trained in the National Health Service in the United Kingdom. I heard surgeons say to the family that we are not going to intervene, we think it would be useless, a waste of your time and our time. That is not the culture in the United States.

Mind you, overdiagnosis is not entirely an American phenomenon. It is predominant in the United Kingdom. It is predominant in South Korea, where they started screening ultrasonography for thyroid cancer and found all of these small papillary carcinomas of the thyroid, which aren't going to do anything.[4] All of these patients had hemi-thyroidectomies, and so on and so forth.

It's an incredible waste of money, but also an incredible imposition of morbidity on people for no good whatsoever. It's like scoring an "own goal" in soccer, and we keep scoring it.

Medscape: How do you propose reducing the problem of overdiagnosis and overtreatment? Is this something the professional societies and academic groups should address?

Dr Jha: We need to have honesty. Using the subsegmental PE example, we need to stand up and say that we are overtreating subsegmental PE, because we are overdiagnosing it. To reduce the overtreatment, we are going to have to stop calling subsegmental PE and accept that there will be somebody who will be harmed as a result of that. That is the trade-off.

First, you have to acknowledge that there is overtreatment from overdiagnosis. (Whether you think that is a problem or not is a different matter). Second, you have to acknowledge the trade-off.

Third, the specialty societies have to agree to not hedge our bets on subsegmental PE. Getting people to agree to that would be a huge victory. The solution lies in raising the threshold for defining the risk, and accepting that an increase in the threshold will lead to some people (who would have benefited from the lower threshold) being harmed. You can't have it both ways. You have to be honest with the trade-off.

Medscape: In not hedging bets, you are referring to the typical statement saying "isolated embolus to subsegmental pulmonary arteries cannot be excluded" on a multidetector CT report.[5]

Dr Jha: The specialty societies should come out and say that that statement shouldn't be made. They should say it in no uncertain terms. They shouldn't couch it in legalese language that allows people to be liable in the event that they don't put the statement on a report. We need to have unity among the professional societies.

Some people will say, we need precision medicine. Yes, we do—but that might come in 10 years, or that might come in 200 years. Meanwhile, you still have to practice medicine.

Medscape: Radiologists have patients referred to them for screening; how can they control uniform thresholds for screening tests, such as mammography or lung cancer screening?

Dr Jha: Increasing the threshold for diagnosis is where radiology can make a difference. The best example is lung cancer screening.

The American College of Radiology increased the precise threshold for follow-up of nodules[6]; they did that with the explicit recognition that this would lower the sensitivity of the test, and the number of false-positives fell dramatically. They also excluded findings that were characteristic of an entity previously known as "bronchoalveolar cell carcinoma," which we now read as "adenocarcinoma in situ"—again, with the acknowledgement that there is the occasional case that could become invasive.

A lot of diagnosis of or calling disease falls upon imaging, because you can't biopsy everything. Imaging ends up being the definer of disease. If we are going to stop overtreatment, we have to stop calling certain findings "cancer." We will have to think of new names for them, and both radiologists and pathologists will have to work at that.

Thresholds are always going to be arbitrary. You can never have a line that perfectly divides health from disease. Another example where radiology did that is with thyroid nodules. If I recall correctly, nodules < 1.5 cm were not to be followed up, and you can bury it in the body of the report rather than make a big deal of it. [Editor's Note: In patients < 35 years of age, the size cutoff for workup is ≥ 1 cm. In patients aged ≥ 35 years, the size cutoff for workup is ≥ 1.5 cm].[7]

The thyroid size criteria are a gateway to overdiagnosis of early carcinoma of the thyroid. If you can control that gateway, you can control the number of overdiagnosed patients.

The one area where it is probably not going to change anytime soon is with breast cancer screening. If you mention overdiagnosis, it's a bit like saying "Voldemort"—he who must not be named. That is the big difference between lung cancer screening and breast cancer screening.

With lung cancer screening, the American College of Radiology decided to tackle bronchoalveolar cell carcinoma as the lowest-hanging fruit, the one that gets the most overdiagnosed and leads to controversy and dilutes the survival effect. I think that was a wise thing to do. If they can stick to that and not have indication drift, then I think lung cancer screening could very well be a success story for a long time to come.

Medscape: Is there a role for radiology in determining whether patients get certain tests?

Dr Jha: Absolutely. We have to be realistic, and it is going to be impossible for radiologists to curb the tendency toward suspecting PE in very low-risk people. Often this is triggered by the D-dimer test. When the D-dimer test is negative, you can say the patient doesn't have thromboembolic disease. But to achieve that degree of negative predictive value, we had to pay a price in terms of specificity.

It's easy to do this test—so many more people are having it, and what ended up being a gatekeeper for diagnosis of PE became an invitation card for more follow-up. Radiology cannot control that, but what we can control, given the body of evidence, is some common shared risk when someone asks us to exclude PE. You can't get paid for doing the study and not share the risk. We can go up to the segmental artery and leave it at that, which would be a tremendous service to both the physicians and the patients. But in terms of dealing with the actual referral base, that really is not in our jurisdiction.

Medscape: Do you see a role for patient discussions—for example, educating patients so that they can see the downsides of overscreening or overdiagnosis? In breast cancer screening in particular, there is a huge patient advocacy component.

Dr Jha: I'm going to try not to be pessimistic. You probably read the Wall Street Journal article showing a rising rate of bilateral mastectomies. This is very much patient-driven. They are going against their physician's suggestion. They are going against the oncologist's discussion. I'm not sure that patient discussion will lead to a reduction in overdiagnosis, but that does not mean that the discussion should not happen. Being informed is very important for its own sake.

I'm not sure it would be feasible to have it on a case-by-case basis, because that would mean that someone would have to spend time explaining the overdiagnosis and explaining the overtreatment. They would also have to explain that we don't know if you are going to have, for example, a subsegmental PE or not.

Letting the patient make their own decision is appealing, but we don't have all of the right facts. A problem could occur if the patient says, "If you had told me the risk was 6%, I might have had the test, but you told me it was 2% when in reality it was 6%." That could open an entirely new can of worms.

The important message that patients should get is that there is a chance that this test will lead to you being treated without any benefit. In the case of lung cancer, that means a lumpectomy or a biopsy; in the case of breast cancer, lumpectomy, radiation, or even a mastectomy. We don't know 100% that it's going to be you, and we don't know that it won't be you. That is the level of informed consent that is both desirable and possible. I'm not sure any more than that is feasible.

With the breast density movement, it is unfair for the primary care physicians to pick up the slack, to be honest. It should be radiologists who are having these discussions. They better know the literature, and they are most aware of the options. Certainly, radiologists can make a huge difference and should participate in that.

In other scenarios, such as subsegmental PEs, it would be difficult to quantify the risk before the actual test. After the results of the test, it would be reasonable to talk about the risks of anticoagulation vs the risks of being untreated for segmental PE.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.