Focusing on overdiagnosis as a driver of too much medicine

Real but elusive trigger of too much medicine

Overdiagnosis is a driver of “too much medicine.” But the uncertainties inherent in overdiagnosis set it apart from other drivers—such as overtesting and overtreatment—making it harder to explain and to tackle.

Overtesting and overtreatment can be identified in a given patient. There is a consensus based on solid evidence that a patient with low back pain but without specific neurological signs or deficits who undergoes magnetic resonance imaging of the spine has been overtested. Similarly, a child who is prescribed antibiotics for earache or upper respiratory symptoms in the absence of fever has been overtreated.

Identifying overdiagnosis in the context of screening is more challenging, because a lesion thus identified fulfils accepted criteria for diagnosis.4 But there is no way to know whether it has been overdiagnosed. An asymptomatic prostate nodule on biopsy, for example, may meet histological criteria for cancer. High quality evidence shows, however, that only some—perhaps very few—of such lesions will progress, and we cannot know for certain what will happen in each specific case. Overdiagnosis is easier to identify in well documented autopsy series, observational comparisons across geography or time, and long term follow-ups of screening trials, for example.

Our ability to diagnose often outpaces our understanding of prognosis, making some degree of overdiagnosis inevitable. We are increasingly able to find tumours, aneurysms, and a host of risk factors for disease. But not every tumour progresses, not every aneurysm grows, and not every person with high cholesterol has a heart attack. Technological advances, such as advanced imaging and genetic testing, aggravate the problem, helping us find less severe abnormalities earlier, long before we know what they mean or whether they need to be treated.

Harm is seen as benefit

Ironically, even though it causes harm, the effects of overdiagnosis look like benefits. People with disease that is overdiagnosed do well because, by definition, their disease was non-progressive. They are “cured” when cure was not necessary in the first place. This creates a cycle that reinforces efforts leading to more overdiagnosis. A screening test that results in substantial overdiagnosis improves survival statistics by diluting the diagnosed pool with many non-progressive cases, which makes screening seem more effective than it is. The spurious rise in incidence makes the case for screening more compelling, thus heightening people’s sense of risk—a phenomenon known as the popularity paradox.6

Seeing through the cycle is challenging. Recognising the distorting effect of overdiagnosis on the perception of success for individuals is the first step. This is not easy, as individual clinical observations of good individual outcomes are compelling. Strategies to improve understanding of these counterintuitive concepts by the public, health professionals, politicians, and the media are important areas of study and effort.

The US Preventive Services Taskforce, which develops and reports screening recommendations, emphasises (albeit briefly) that overdiagnosis is an important harm of breast and prostate cancer screening in materials for the public and professionals.7 Similarly, the UK’s NHS breast screening leaflet describes (but does not name) overdiagnosis and quantifies the chance that a screen detected breast cancer “would never have caused harm.”8 Efforts to engage and educate journalists about overdiagnosis could also help dampen unrealistic expectations about screening.

What is needed

Improving prognostic science to better distinguish indolent from progressive disease could reduce overdiagnosis. Longitudinal cohort studies to understand the natural history of a “new” diagnosis are needed, especially when new technology with greater resolution is introduced. In some cases, these studies are considered unethical if an effective accepted treatment exists. Randomised controlled trials of cancer screening are often rejected (by health authorities, patient organisations, and specialists, for example) because the benefits of screening seem obvious to the general population, health professionals, and policy makers. But new screening tests or sets of diagnostic criteria must be compared with standard management in randomised trials before they are put into practice, using health outcomes as the endpoint.9

We need tools to help recognise overdiagnosis in individuals.1011 This is an active area of research in cancer diagnosis and early detection, where new molecular tools and informatics may help to distinguish progressive from non-progressive lesions detected on screening.12 This science is still in its infancy, so it can be misunderstood or misapplied. This could make things worse by promoting undue optimism about the usefulness of new biomarkers and by introducing confusion if tests with uncertain clinical utility are widely used. Cardiovascular risk estimation provides a cautionary example, where risk factors meant to predict the chance of future events (such as high cholesterol) are sometimes elevated in public messaging as representing actual “disease” states themselves.

None of this will be easy, but it starts with recognising the unique aspects of overdiagnosis as a major contributor to too much medicine and working to identify it, communicate its effects, and explore new research directions.