Preventive Medicine Column

Dr. David L. Katz

A study just published in the New England Journal of Medicine suggests that mammography routinely finds cancers that would be better left unfound- cancers that would not progress, and do not need treatment.

The authors make this case by examining time trends in the frequency with which both early stage and late stage breast cancer is diagnosed.  They argue, reasonably, that the benefit of screening is finding early stage cancers that would progress to late stage cancers if not detected.  And so, each early stage cancer found through screening should- they tell us- be one less late stage cancer.  Over the past several decades, that has not been the case; the rise in incidence of early stage cancers is much larger than the fall in incidence of late stage cancers.

On this basis, the authors conclude: the discrepancy between the two represents over-diagnosis.  They go on to suggest that the decisive decline in breast cancer mortality over recent years is due entirely to better treatments, not to the early detection offered by screening.

The researchers in this case may be right, which is why their paper was published in the New England Journal.  But there are important ways in which they may be wrong.

Some early stage cancers may progress despite treatment.  Many late stage cancers may be treated effectively- but the treatments required may be far more onerous than treatment at earlier stages.  Mortality is an important measure- but so is survivable misery.

And then there is the fact that some studies have told us mammography does, indeed, reduce mortality.  Other studies, all but indistinguishable, have refuted it.

There are two basic problems with cancer screening in general- problems that pertain as much or more to prostate cancer as to breast cancer.  The first is the challenge of accurate detection, and the second is the challenge of accurate prediction.  Let’s deal with them in turn.

The very point of screening is to find cancers when they are tiny and subtle, not when they are large, obvious masses eroding through other body parts.  The earlier a cancer is found, the harder it is to see both because it’s tiny, and because it may look a whole lot like the healthy tissue around it.

The challenge of finding something tiny and subtle is met, in statistical parlance, with sensitivity.  Sensitivity is the ability, in this context, for a medical test to find what’s there.

But there is a problem with extreme sensitivity.  In cancer screening, false positives- finding what isn’t really there- are the price paid for finding the cancers that are truly there, and need to be found.

The other goal of cancer screening is to rule out disease when it’s absent.  In statistical terms, specificity is the tendency for a test to give a negative result when disease is truly absent. In cancer screening, false negatives are when a test that says there is no cancer- when, in fact, there is.  And the greater the specificity, the more often this happens.

The solution to these problems is improved technology.  In the case of mammograms, this would mean enhanced imaging, or computer-aided interpretation of the images.  It might mean alternatives to mammography- such as ultrasound, or thermography.  Or combinations.

The second challenge is prediction.  The new study doesn’t really highlight the flaws in mammography.  Rather, it suggests we don’t know what to do with the information the test gives us.  Some early stage breast cancers, in particular ductal carcinomas in situ, are destined never to progress.  This is true of many prostate cancers as well, and the reason for formal recommendations against routine prostate cancer screening.  We don’t want just to find cancer early; we want to change health outcomes for the better by finding cancer early.  That doesn’t happen when cancer is found that would never have progressed if left alone.

The solution here is deeper knowledge at the cellular level.  Gene variants can help anticipate cancer behavior.  A combination of reliable detection through better technology, and then better information from biopsy specimens, should lead us in the direction of treatment when it’s needed- away from it when it isn’t.

Mammography specifically, and cancer screening in general, is often something of a muddle.  It doesn’t help for us to refute this, or simply rant in favor of our preconceived notions.  We should acknowledge the trade-offs, work toward better screening methods, and in the interim- muddle through with what we’ve got.




Dr. David L. Katz;