Overdiagnosis: It's Not Just For Men

I could just as easily titled this post "Mammography: Too Much of a Good Thing".  Why?  As with most tests, there is some amount of imprecision.  This is where statistics come into play in determining accuracy and predictive value and all that kind of stuff.  But remember, as I teach the residents, order a test because it will change (or make an impact upon) your decision (-making process).  However, as this latest study demonstrates, we should order a test because it will be of benefit to the patient.  In the broadest terms, we should do things that will improve all-cause mortality.  This is where screening for prostate cancer w/PSA (prostate specific antigen) has been getting us into trouble of late.  Whereas we're now able to detect prostate cancer at earlier stages, we haven't demonstrated any reductions in all-cause mortality, much less cancer related mortality.  Why?  Because we can't predict which cancers will kill us as opposed to remaining quiescent.

It turns out that we're running into that same situation w/mammography.  When we find an abnormality on mammography, we can't determine whether it's the kind that will kill or remain localized, since not all breast cancers grow (quickly).  In a retrospective study published yesterday in Annals of Internal Medicine, the authors evaluated Norwegian women over a 10 year period of time when screening mammography was gradually implemented biennially for those 50-69yo.  Using 2 different models, they estimated that overdiagnosis of breast cancer occurred in 15-25% of women screened.  As a consequence, many women received unnecessary treatment for a condition that would not have otherwise affected them.  While the editorialists noted that due to the retrospective nature of the study, not all confounders could be accounted for, they also suggested perhaps an even higher rate of overdiagnosis in the States given our tendency to start annual screening mammography a decade earlier.  And in case you're like me and think that this conclusion is fluke, consider this: a meta-analysis last April of 7 studies involving 600K women came to a startling similar estimate of 30% overdiagnosis.

So what are we to do?  Just like w/PSA screening, the milk has been spilled.  I doubt that we have the fortitude to stop screening for either prostate or breast cancer with our current methodologies even when faced with such a high rate of overdiagnosis of lesions that won't kill us.  Instead, we should use this study to galvanize our efforts to better separate out the cancers that will kill us from those that will die with us.  In the meantime, we'll need to set aside some time to discuss the implications of screening and the potential for overdiagnosis.

Follow @alvinblin
Tweet