Book review: Overdiagnosed

Title: Overdiagnosed. Making people sick in the pursuit of health (2011)

Author: H. Gilbert Welch (with Lisa Schwartz and Steven Woloshin)

Publisher: Beacon Press, Boston

Overdiagnosed is an attempt to change the conventional wisdom - the prevailing paradigm - that early diagnosis is always a good thing and therefore that the best test is the one that finds the most disease. He shows us the downside of our attempt to make sure that nobody misses out on any potential benefit by failing to be diagnosed. He shows us that the dogma of early diagnosis is maintained by overestimated benefits and a general disregard of the harms.

Dr Gilbert Welch is a physician and epidemiologist based at The Dartmouth Institute where he is Professor of Medicine. His area of expertise is overdiagnosis, the topic of this book and of his previous book, Should I Be Tested For Cancer? No, And Here’s Why (UC Press, 2004).

Overdiagnosis occurs when people without symptoms, either by falling on the wrong side of a test threshold (e.g. blood sugar, cholesterol, bone density score) or undergoing screening, are diagnosed with a condition that either would not have caused a problem, would have caused a problem later but did not benefit from early diagnosis, or receive a ‘false-positive’ diagnosis (leading to anxiety, further tests and treatment with no potential benefit but with potential harm).

This area has been covered before, even in my blog (here) but Dr Welch spells it out very clearly, and in a way that is easily accessible by the (non-medical) public.

There is an acceptance that screening can benefit some patients, and that in severe cases (like severe hypertension or diabetes) early diagnosis can lead to significant health benefits. However, it is explained that there is often a spectrum of patients; at one end the benefits are great, at the other end the benefits are marginal. Throughout the spectrum, though, the harms from overdiagnosis are the same so that at some point the benefits and harms are balanced leaving a group that has a net benefit and a group that has a net harm. Finding that crossover point might be difficult, but at least we should acknowledge that it exists.

There are many reasons given for the current thinking that more diagnosis is better. The influence of the pharmaceutical and device industries and patient demand are two such reasons. One reason, however, deserves more attention. It is the fact that doctors want to find out everything they can about the patient and that the most important thing is not to miss any opportunity to diagnose. The drivers behind this are firstly the fact that doctors often truly believe that every diagnosis made and treatment provided will lead to benefit (the prevailing wisdom), and secondly, the perverse nature of the medicolegal system is such that doctors are heavily penalised if they ‘miss’ something, but never penalised (for malpractice, anyway) for ordering too many tests (or overtreating, for that matter). This is one of my personal gripes: that the system rewards intervention (“at least he tried”) and frowns upon conservatism (“he didn’t even order an MRI”).

Dr Welch shows us what happens when that MRI (X-ray, blood test, or whatever) is ordered; we are more likely to receive a diagnosis and subsequent treatment. Whether or not that is in the best interests of the patient is not as clear as it is assumed it to be.

The concept that the book tries to get across is counter-intuitive. It even goes against one of my mottos, that there is no such thing as ‘bad’ knowledge; that the more we know about something the better. The book provides many examples of where it would be better not to get tested in the first place. I have personally seen many cases of anxiety, unnecessary treatment and harm come from tests that have been shown to provide no benefit to overall or disease-specific survival. The problems start from getting the test performed in the first place.

Details such as the illusion of better survival statistics in cancer screening due to lead-time bias, overdiagnosis and extrapolation are all explained but my only criticism of the book is that the author does not go far enough. For example, he shows how improvements in the disease-specific survival are overestimated, but frequently fails to mention that overall survival (the main aim of early diagnosis) is often completely unchanged (see my previous post). He is also very generous in his calculations of net benefit and harm; assuming best-case scenarios for early diagnosis. I accept that this makes his point stronger when the calculations still show a questionable (or no) benefit, but overall he is too lenient. I would have used both barrels.

Other areas covered include: the false economics of medicine in general, and of early diagnosis in particular (which does not save money, it leads to increased direct and indirect costs, often without any health benefits); the various positive feedback mechanisms that drive overdiagnosis; genetic testing and the confusion of DNA with disease; and the failure to distinguish between an epidemic of disease and an epidemic of diagnosis.

The one quote I would like to mention is:

… ironically, pursuing health [as opposed to healthcare] requires not paying too much attention to it.

The conclusion is that the benefits of our vigorous attempts to find disease (or pre-disease) have been systematically exaggerated and the harms have been minimised or not disclosed. We (particularly the public, not just the doctors) need to weigh the benefits against the harms before embarking on expensive, time-consuming and stressful diagnostic procedures instead of blindly assuming that labelling more people with disease will improve health.