Final summary Overdiagnosis.
OVER-DIAGNOSIS EPIDEMIC – In the final instalment of our series, Paul Glasziou talks about the way forward. Over-diagnosis is a significant problem that’s already common in some areas of medicine, such as screening and some mental health conditions. It is a problem now but a bigger threat for the future…
OVER-DIAGNOSIS EPIDEMIC – In the final instalment of our series, Paul Glasziou talks about the way forward.
Over-diagnosis is a significant problem that’s already common in some areas of medicine, such as screening and some mental health conditions. It is a problem now but a bigger threat for the future: better imaging technologies, more biomarkers, more genetic tests and so on, will gradually expand the amount of over-diagnosis that’s possible.
There are several drivers for over-diagnosis. Of course there’s screening and new screening methods, but there are also areas that are not technology-driven. For instance, the diagnostic criteria in psychiatry where the changing definitions of mental illnesses have expanded the number of children and adults classed as having a mental illness.
A slippery slope
An interesting example of how over-diagnosis thrives was presented in a recent issue of the British Medical Journal (BMJ). Researchers did a study of magnetic resonance imaging (MRI) scans on people’s knees. Participants were middle-aged or elderly people. The study authors discovered that most people with knee symptoms had some sort of abnormality in their knee.
But then they also did a series of MRI scans on people with no symptoms at all, and discovered that about 85% of them had some sort of abnormality in their knee as well. So we can imagine a person with some minor symptoms getting an MRI scan done and discovering something that actually had nothing to do with their exact symptoms.
These new diagnostic technologies are one cause of the increasing rates of over-diagnosis. Our ability to detect earlier means that we’re going to detect disease in more people, some of whom would never have been disturbed by their apparent “disease”. A prime example of this dilemma is prostate cancer where if you looked carefully enough, you’d find around 70% of 70-year-olds have something a pathologist looking at under a microscope might call a prostate cancer. But most of those so-called cancers would never disturb the man in his lifetime.
Part of the problem is our ignorance about natural variation that were difficult to examine in live people. But technology such as MRI scans now makes that easier. But we discover things that surprise us, such as a large proportion of people have abnormalities. A lot of abnormalities are more normal than we expected. No one body is ideal – so someone many years ago said that “a normal person is someone who just hasn’t had sufficient tests”.
The first step to reducing the epidemic of over-diagnosis is raising awareness of the problem and its drivers (that’s partly what our conference next year is about). The next step will be to understand how the health care system develops these over-diagnoses and where they arise from, so we can try to improve the process by which that happens.
There isn’t a single simple solution to this complex problem. One piece of our work is to break down and sub-divide the problem of over-diagnosis – to understand the different causes and categories and how one might approach each.
The first article in this series discussed the three different categories of over-diagnosis. For screening, the best probable solution is that you develop cures for the diseases. We don’t screen for testicular cancer, because the cure rate is now so high that screening is not necessary. When we get to close to 100% cure rates for cancers, we won’t need to screen at all.
The second problem is changing disease definitions. Where we drop the diagnostic threshold, we get milder forms of disease – where people benefit less but still get an equal amount of harm from treatment and labelling. Eventually we get to such mild conditions that the likely adverse affects outweigh any benefits. For diseases defined by moveable thresholds we need to have an internationally agreed set of rules about when you change or lower thresholds.
At the moment, there isn’t a clearly laid out international agreement. Usually a steering committee gets together and decides that change a disease definition is appropriate but it does so without clear guiding principles. The groups that define what a disease is needs to have people with appropriate expertise involved, and we need to set some clear guidelines about the constituency of such committees. You need medical expertise but you also need other types of expertise that generally aren’t in these committees, for example, epidemiologists, consumer representatives, and perhaps sociologists. And you also need a minimal, or no, conflicts of interest in those groups.
The third category was the creation of new diseases. About four new diseases are being defined every week. In fact, most diseases were probably defined or described in the last century; we’re rapidly increasing the number of conditions we classify as disease. Most of these are legitimate but there are some that are probably less legitimately defined. We need to come to some agreement about what really constitutes a disease, and that’s particularly important for mental health conditions.
Over-diagnosis is consuming huge resources. It is shifting our attention from very ill people who really need the miracles of modern medical technology, and distracting us with conditions and interventions that provide marginal benefit, no benefit, or even net harm, as we move diagnosis towards milder forms of illness.
This series has been a great start for creating some awareness of the problem of over-diagnosis, but clearly it’s just a start. There needs to be a much wider awareness and attention given to this as a growing problem within medicine, because it has implications for the whole future and affordability of medicine.
This is final part of our series on over-diagnosis, click on the links below to read other articles: