Bad Screens

 There was an article in the New York Times this weekend about some genetic screening tests that have a high rate of false positives. I worked through the math in my head to some extent, and I'm not surprised at the results. I'm more surprised that anyone is using these screening tests because the results don't seem very helpful.

The issue is that we now know that some fetal cells end up in the maternal system during pregnancy. It's not a lot of cells, but technology to find and analyze nucleic acids is very sensitive, so someone came up the with idea of taking a blood sample from the mother and using things like PCR and antisense tagging to check for presence or absence of chromosomes or bits of chromosomes in fetal cells or fragments of nucleic acids found in the maternal blood sample.

I guess this works great for identifying Down's syndrome in a pregnancy. There are some false positives, but few false negatives, and the true positive rate is high enough that most people think the method is useful and successful for early detection.

Other conditions caused by chromosomal abnormalities, though, don't have the same success rate, partly just because the conditions are much less common. If the number of false positives for there conditions is the same as for Down's syndrome, but the true positives are much less frequent, then the probability of a positive test result being accurate drops dramatically. So some of the tests the article discussed had false positives in 80-90% of positive test results. It's just math.

So I guess one could ask, how useful is the test if a positive result in wrong 90% of the time? How valuable is a diagnosis of such conditions compared to the distress caused by the false positive results? I would think the doctors ordering such tests and the providers, should be telling the patients about these statistics before carrying out the test, and maybe should have a standard follow-up protocol for every positive test result. But if further testing is the consequence of any positive test, is there really a point to the test? Or is there something that can narrow the field, a way to identify increased risk for a rare condition? My guess is that many patients getting such results have a limited understanding of the statistics and what they mean, and may not have much understanding of the conditions, either.

It seems that some care should be taken in making recommendations about such screening tests, and some deep discussion about clinical value is needed. Does widespread screening for very rare conditions help anyone? I mean besides the companies doing the screening tests. From the point of view of value to society, how do we balance the increased knowledge gained from the screening with the dramatically increased stress for those who get a false positive result, and have to worry over whether it is real or not until they get a more reliable result, along with the stress of worrying over complications from the more reliable by higher risk testing to determine if the first result is true or false? We should, I guess, add in the value to all the people who get true negative results that their babies don't have these rather serious conditions, but before the screening, hardly anyone worried over those conditions anyway because they are rare.

It doesn't seem like any easy question to me. Maybe we need to introduce Artificial Intelligence or an Expert System, or something, to helps with such decisions. I hear those are pretty good.