Overdiagnosis - Part 1
I'm a pathologist, which means my job is all about diagnosing disease. Pathologists see parts of your body and categorize what's wrong with you. We write up a report that tells your doctor the name of your problem. Our interpretation of your medical tests is considered the final decision, after the informed guesswork of the internists and radiologists. That's what I like about Pathology - I get to see, from the inside out - what is actually happening. The truth.
But lately I've been reading a lot about Overdiagnosis and it's usual progenitor, Overtesting. As a pathologist, I've got a perspective I could add to this topic. To get it right, I'm still reading more about the subject. I'm going to get down some of my initial ideas and questions here in my blog in several posts. Most of what I'm writing will be a public health, population-based information - not my specialty. So if you are reading this as an expert, I'd appreciate your insights.
More and more people have a diagnosis of cancer. We think of this as Incidence - when the disease occurs. They may have been exposed to a risk factor (i.e. smoking, ionizing radiation), or they could just be older. Even our general success treating cardiovascular disease contributes to an overall increased burden of cancer - people live longer and are therefore more likely to get cancer.
But people might have a diagnosis of cancer because they've been overdiagnosed. This perspective makes it clear that incidence is not just about cancer occurrence, but more about the rate of cancer diagnosis.
What is overdiagnosis?
Overdiagnosis is the diagnosis of cancer in people who would never have experienced symptoms or harm had the cancer remained undetected and untreated.
How does it happen?
It can happen through screening programs. The best example of this is the spike in prostate cancer diagnoses following widespread screening for prostate cancer with PSA testing. It can also happen through incidental detection during investigations of unrelated problems - someone undergoes an abdominal CT scan following a car accident and receives a diagnosis of kidney cancer.
Why does it matter?
We don't communicate this very well in medicine or in the media, but here is a unpopular fact: every medical test and procedure has a risk. These risks are generally considered acceptable when they are outweighed by the benefit. Ionizing radiation from mammography, for example, is considered okay because the procedure identifies tumors (more about that in another post). Interventions can be fatal. The history of modern medicine is littered with examples of unintentional harm done to patients in the name of testing and treatment. There is a name for harm done to patients when in medical care: Iatrogenic (from the Greek physician-caused).
There are many unappreciated risks from overdiagnosis. For example, there is a higher rate of suicide amongst men diagnosed with prostate cancer. People are anxious, disfigured and ashamed. Many women diagnosed with breast cancer reshape their identity around this diagnosis.
Ultimately, when someone is unnecessarily diagnosed with a cancer, they can only be harmed by treatment, not helped.
How do you measure overdiagnosis?
Let's be clear - I'm interested in this topic as a vulnerable human but also as a pathologist. I make the diagnoses. This question for me becomes:
When I diagnose a melanoma, can I know that it is an 'overdiagnosis'?
Melanoma has a lot of peculiarities about diagnosis (most based on the fact that it is so small), that make it extra difficult to determine this answer, but for the most part, this is not a phenomenon that can be measured behind the microscope*. No, overdiagnosis is determined by the differences in the lifetime risks of being diagnosed with specific cancers over time. By this reckoning, overdiagnosis is the total excess lifetime risk of cancer.
"Although it is difficult to know which persons are subject to overdiagnosis, overdiagnosis is easily appreciated at a population scale."
A recent article in the New England Journal of Medicine (NEJM) by Welch, Kramer and Barnett, is a helpful starting point for better understanding overdiagnosis. The article, entitled Epidemiological Signatures in Cancer, demonstrates with data from the national cancer registry called the SEER program (Surveillance, Epidemiology and End Results) that what we think of as true cancer occurrence is influenced by diagnostic practices. (A summary available without a subscription can be found here.)
Many cancers detailed in the report have responded well to treatments and interventions - Hodgkins Lymphoma and Chronic Myelogenous Leukemia are two examples. Our efforts towards prevention work well too - the incidence and mortality of cervical and oral cancers has declined. When the graphs show similar curves we know that we are documenting a true occurrence pattern, even if the direction is not the way we want to see it. For example, both the incidence and mortality of liver cancer has increased together.
The cancers that demonstrate a clear pattern characteristic of overdiagnosis are thyroid, kidney, melanoma, breast and prostate. In the NEJM special report, the authors classify cancers that are both too small to produce symptoms or a mass, and also grow slowly (indolent). This creates a reservoir of tumors that are particularly sensitive to diagnostic scrutiny. We pride ourselves on improving diagnostic scrutiny; we test for these tumors more frequently, with more sensitive instruments, and we redefine the thresholds to label findings as abnormal. But the numbers show us another story - these are cancers not destined to cause death.
I'd urge you to look through the article and the links I've found below for contextual understanding. But one graph from the article I thought helpful to share to help make the point. We are looking at the rate (top) and relative rate (bottom) of disease incidence over time, from 1975-2015.
Looking at the graph on the right side, we can easily see that Melanoma is overdiagnosed. The incidence is soaring (rate per 100,000) and the mortality rate is flat. If lifetime mortality has changed little in absolute terms, then many of these melanomas did not need to be diagnosed as such.
I'm going to keep examining the literature on overdiagnosis. My next post will be also be about overdiagnosis.
*Some people disagree with that:
"Overuse is harmful because all medical procedures carry some risk; underuse is harmful because it leaves illness and suffering untreated." https://lowninstitute.org/projects/right-care-series-in-the-lancet/
The US Preventative Services Task Force does NOT recommend routine skin care screening.
"Evidence to assess the net benefit of screening for skin cancer with a clinical visual skin examination is limited. Direct evidence on the effectiveness of screening in reducing melanoma morbidity and mortality is limited to a single fair-quality ecologic study with important methodological limitations. Information on harms is similarly sparse. The potential for harm clearly exists, including a high rate of unnecessary biopsies, possibly resulting in cosmetic or, more rarely, functional adverse effects, and the risk of overdiagnosis and overtreatment.
Conclusions and Recommendation The USPSTF concludes that the current evidence is insufficient to assess the balance of benefits and harms of visual skin examination by a clinician to screen for skin cancer in adults (I statement)."