Advertisement

Responding to the Challenge of Overdiagnosis

      The possibility of overdiagnosis first became apparent to me early in my career when I ordered a sinus film on myself. I was seeing patients in the Indian Health Service and was frustrated by how many sinus films were read as abnormal. Otherwise I felt fine; the film was just my idea of an experiment. I was rewarded with a finding: a polyp secondary to chronic maxillary sinusitis.
      A few years later, soon after I joined the staff of the White River Junction VA, the possibility became more real. One of my patients called me because of persistent hoarseness. Our Ear, Nose & Throat specialist found and removed a small vocal cord tumor but also ordered a chest x-ray. The chest x-ray was read, showing a possible widening of the mediastinum, and our radiologist recommended a chest computed tomography (CT). The CT showed a normal mediastinum but also a golf ball–sized mass in the right kidney—with all the radiologic features of renal cell carcinoma. That hoarseness could produce kidney cancer was never covered in my training.
      Then, virtually my entire patient panel of elderly men seemed to develop prostate cancer (and typically I was not even ordering the test responsible for the epidemic: the prostate-specific antigen).
      Despite my now 25-year-old diagnosis, I have never had a problem with my sinuses. My patient's hoarseness quickly resolved after the tumor removal, but he chose to keep his kidney and lived another decade to worry about it. (He died of pneumonia, had a 5-cm renal cell carcinoma on autopsy, but no metastatic disease.) And, a lot of prostates came out (or were radiated) for a disease that was not going to cause problems, although the treatment certainly did.
      Go figure. May be this kind of stuff only happens in government-run health care, but I do not think so.
      These formative experiences suggest some of the causes of overdiagnosis: indiscriminant test ordering, clinical cascades ending in incidental detection, and population-based screening. It is useful to separate the role of the radiologist from that of the ordering clinician.
      The responsibility for mitigating overdiagnosis secondary to indiscriminant test ordering clearly lies with the ordering clinician. I am tempted to say any clown who orders a sinus film on himself when he feels well—yet writes “sinus pain” under the reason for the examination—deserves to get chronic sinusitis. (Let the record show that I have never ordered a subsequent image on myself.) The American Board of Internal Medicine Foundation has now engaged some 60 specialty societies in Choosing Wisely—a campaign largely directed toward promoting prudent test ordering (

      American Board of Internal Medicine Foundation. Choosing Wisely. Available at: http://www.choosingwisely.org/. Accessed April 13, 2015.

      ).
      Ordering clinicians and radiologists share responsibility for mitigating overdiagnosis secondary to clinical cascades ending in incidental detection. Clinicians should prepare patients for the possibility of incidental detection before the imaging examination and that many unsuspected abnormalities may be better left alone. Radiologists should support clinicians in this effort by being clear when the pursuit of diagnostic certainty is likely not in the patient's best interest.
      The responsibility for mitigating overdiagnosis secondary to population-based screening clearly lies with the radiologist—with the proviso being that the screening test involves an image. Therefore, while you are off the hook for overdiagnosis after prostate cancer screening, you are not for breast and lung cancer screening.
      Almost 25 years ago, it was a radiologist who informed me about two diagnostic principles that were also not covered in my training (
      • Black W.C.
      • Welch H.G.
      Advances in diagnostic imaging and overestimations of disease prevalence and the benefits of therapy.
      ). First, the vast reservoir of abnormalities—including abnormalities labeled “cancer”—implied that whenever we doctors look harder, we will find more. Second, finding more implies that we are detecting milder and milder forms of disease. Thus, looking harder produces two misleading pieces of feedback that push us to look harder still: there is more disease than previously recognized (so we better find it), and the typical patient now appears to do better (so we must be helping people).
      It was powerful feedback that helped usher in the new era of anticipatory medicine. Instead of simply determining what was currently wrong with our patients, we became increasingly focused on what might go wrong in the future. In the era of anticipatory medicine, the task for radiologists was straightforward: look harder, find more, and patients will appear to do better—having fewer complications from their “disease” and surviving longer after their diagnosis.
      But the increasing recognition of the harms of overdiagnosis and overtreatment has made it clear that radiologists need to take on a more nuanced task and a more challenging one. The goal cannot simply be to detect more abnormalities but to identify the abnormalities that matter. It is too easy to find more cancer; the question is “who can find the cancers that matter?”
      It is all about the diagnostic threshold, the threshold to say something is wrong and the threshold to recommend something be done about it. Do not pretend the two can be separated: like it or not—what you radiologists say influences what we clinicians do. In the past, physicians worried only about errors in one direction: missing something that might be important. The response was to lower diagnostic thresholds, by using higher resolution technologies and lower thresholds for interpretation. Now, we realize that relentless focus on the error of missing things has created a new one: finding things that are not important.
      The value of raising diagnostic thresholds in radiology is perhaps best understood in the context of lung cancer screening. Investigators in both the Danish and Dutch–Belgian trials are careful to simply observe small abnormalities in an effort to minimize the number of biopsies and the number of patients treated for lung cancer (
      • Pedersen J.H.
      • Ashraf H.
      • Dirksen A.
      • et al.
      The Danish randomized lung cancer CT screening trial—overall design and results of the prevalence round.
      ,
      • van Klaveren R.J.
      • Oudkerk M.
      • Prokop M.
      • et al.
      Management of lung nodules detected by volume CT scanning.
      ). These higher thresholds were recently incorporated in the American College of Radiology's assessment categories for Lung CT Screening Reporting and Data System (

      American College of Radiology. Lung CT Screening Reporting and Data System (Lung-RADS(tm)). Available at: http://www.acr.org/Quality-Safety/Resources/LungRADS. Accessed April 13, 2015.

      ). It is an example that should be emulated more broadly.
      The next generation of radiologists will face increasing volumes of data: there will always be more views, more slices, and more pixels to look at. Their task will increasingly become sorting the wheat from the chaff, minimizing the cascades of diagnostic testing and the side effects of excessive intervention. It is a task not unlike that of an editor: pruning the data back to reveal the useful information.
      Do not make the mistake of thinking, this is just about saving money (not that there is anything wrong with the effort to make medicine more sustainable). It is about better balancing the benefits and harms for patients. And, do not say we are powerless because of the lawyers. Ask yourself what is the right thing to do for patients and help your profession set a balanced standard of practice.
      Finally, do not make the mistake of thinking the problem of anticipatory medicine is confined to radiology. All physicians need to shift their thresholds for diagnosis and intervention knowing that patients with little chance to benefit from a diagnosis and treatment are also the ones at the highest risk for net harm.

      References

      1. American Board of Internal Medicine Foundation. Choosing Wisely. Available at: http://www.choosingwisely.org/. Accessed April 13, 2015.

        • Black W.C.
        • Welch H.G.
        Advances in diagnostic imaging and overestimations of disease prevalence and the benefits of therapy.
        N Engl J Med. 1993; 328: 1237-1243
        • Pedersen J.H.
        • Ashraf H.
        • Dirksen A.
        • et al.
        The Danish randomized lung cancer CT screening trial—overall design and results of the prevalence round.
        J Thorac Oncol. 2009; 4: 608-614
        • van Klaveren R.J.
        • Oudkerk M.
        • Prokop M.
        • et al.
        Management of lung nodules detected by volume CT scanning.
        N Engl J Med. 2009; 361: 2221-2229
      2. American College of Radiology. Lung CT Screening Reporting and Data System (Lung-RADS(tm)). Available at: http://www.acr.org/Quality-Safety/Resources/LungRADS. Accessed April 13, 2015.