Monday, March 22, 2010

Cervical cancer prediction even when a true cause is known

The problem with identifying cause, in this case of a disease, often seems to be that the cause identified is only a partial predictor of the effect. That is, it is only one of an assembly of other contributors, mainly not measured or not known.

We've mentioned a couple of publicized examples of testing for prostate and breast cancers. The positive PSA or mammogram identifies a possible risk factor, but with low probability that serious outcomes will eventually arise. A new example is testing for HPV (human papilloma virus) as a causal risk factor for cervical cancer in women, as described in a recent news story. HPV seems to be a genuine cause, but the probability that someone infected with HPV (which is common) will eventually get cervical cancer, is small. So is the test worth the cost and the uncertainty of what to do with a positive test?

There is no easy answer to such questions. But one thing that is clear: the question itself is worth taking seriously, as a research problem that is challenging but very important. Such research involves both the actual causal story--the primary science itself, as well as clear policy implications: some way to decide what to do with incomplete understanding of the risk factors at play.

No comments: