How Much Are We Over-Diagnosing Cancer?

The word about 'overdiagnosis' is a regular feature in medical journals, stories are found at least weekly in major newspapers, and patients are starting to question whether cancer screening tests are really right for them.


Ok, no no, we are a long way from finding the right balance of too much and too little medicine. But now that we accept that 'too much medicine' is a real thing, we need to figure out just how big the problem is.

Peter Ubel (@peterubel) is a physician and behavioural scientist, and author of Critical Decisions (see this and related books on our list)

He has attempted to lay out the way in which we can quantify (and clarify) the times where we inappropriately give a person the label of 'cancer.'

He states clearly that misdiagnosis, while unfortunate, isn't overdiagnosis. He also says that false-positives, while they can lead to harmful results, are not overdiagnosis.

What is is then? Whole conferences (eg. Preventing Overdiagnosis) have been devoted to defining it. 

Overdiagnosis, according to Ubel, occurs when we detect things that would never have caused a problem for the patient. He gives the example of a tiny breast cancer that would never have been noticeable in an elderly woman (who would undoubtedly die of something else first). When trying to change the culture to encourage people to stay away from screening tests that will lead to overdiagnosis, we are up against several challenges. One of those is the fact that early diagnosis can sometimes make it seem like we live longer if we detect the cancer earlier, though finding it early doesn't improve or save our life (lead-time bias, which is explained in the article).

Ultimately, in order to quantify the prevalence of overdiagnosis, we will need population-level data after a screening program has been introduced, and the data will need to be measured for long enough that any of the lead time bias effect will have passed.

Read more of Dr Ubel's explanation, How Much Are We Over-Diagnosis Cancer? in Forbes.