I've always liked technical papers that have a graphic in them that summarizes the whole paper. Dr. Welch and his coauthors' sobering look at over-diagnosis can be summarized with one graph that they present in the opening chapters:



We've been screening for more and more in asymptomatic patients in the last 50 years, based on the premise that it would be better to catch things early and treat them. So, have patient outcomes improved? After several million people have been affected by this onslaught, the answer is no. Welch explains why and why it's often a bad idea to screen well people.

I spent 35 years working in a hospital laboratory. Our relationship with the clinicians was usually cordial and sometimes actually cooperative, but now and then it was adversarial. Problems included how to deal with the request for an inappropriate test (tests that didn't really exist; tests that couldn't be controlled; tests that were incredibly expensive - but the clinician's attitude was that neither they nor the patient was paying for it, so...; tests whose results could not be interpreted; tests whose results could not help the patient no matter what the result was), how to explain that if every test were done "stat", then none of them would be, and how to explain that if more and more biopsies were done for an "abnormal" test (e.g. a radiograph) with creeping criteria for positivity, then one might expect that most of those patients would not have abnormal pathologic findings. Many of these problems came down to a lack of knowledge of the statistical properties of reality. Dr. Welch covers several of these problems directly and some of them by implication. His book is a breath of fresh air.