Genetic counselors have a love/hate relationship with numbers. We sometimes sniff with contempt at numbers , and nobly proclaim that psychosocial issues are the real business of genetic counseling. On the other hand, patients and referring physicians demand concrete answers, so we spend an inordinate amount of time discussing statistics and risk figures. And, truth be told, hard data is less complicated to address than complex psychological issues. With numbers, as Casey Stengel used to say, “You can look it up” – but counseling requires us to use our skills to fly by the seat of our pants.
This focus on numbers has led to the particularly absurd practice of evaluating the effectiveness of genetic counseling by measuring accuracy of patient recall. Not surprisingly, patients often demonstrate poor recall of any but the simplest facts. Surely by now we have learned that patients are not calculating machines! Whatever information we provide passes through patients’ complicated emotional, neurobiological, and cultural filters and out the other end comes a jumbled understanding of technical information that plays out in a complex psychosocial milieu. Yet despite their fairly consistent inability to remember much of what we tell them, patients usually make good decisions. Many researchers are at last questioning the validity of using recall to measure effectiveness – it’s a counseling session, not a final exam, for chrissakes’ almighty.
I am relieved that researchers are developing alternative ways of evaluating genetic counseling. What intrigues me about numbers, though, is a question that almost no researcher has asked – “Why do genetic counselors (or any health professional) choose to use a certain set of numbers when counseling patients?” It’s a given that all research studies are flawed, and therefore the numbers generated by studies are flawed, some tragically so. I would like to believe that we carefully evaluate and compare studies, and choose those with the soundest methodologies and largest sample sizes to generate risk figures for use in genetic counseling. But I am not convinced this is always the case. Let me illustrate with 3 common examples from genetic counseling practice.
1) The single most widely cited genetic counseling statistic – for at least 30 years – is the mythical 0.5% miscarriage rate of amniocentesis. This number is well enshrined in text books, journal articles, and our collective memory. It’s usually presented to patients, with some variation, as “Amnocentesis has a half-percent miscarriage rate.” This may then be followed by a statement like “But at our center, the miscarriage rate is X” (inevitably some number lower than 0.5%), or perhaps “More recent studies have shown a lower rate.” These “statements of fact” are so incorrect that they border on falsehoods. First off, as I have pointed out (ad nauseum, to some) no study has ever shown a 0.5% miscarriage, period, end of story. Second, no individual center or physician knows their own miscarriage rate. The only way to determine a center’s or individual physician’s pregnancy loss rate is through a randomized controlled study within a center. This requires sample sizes – and raises ethical issues – well beyond the means of even the largest and best financed clinics. In the absence of such a study, providing a center- or physician-specific risk is biased guesswork at best, and fraudulent deception at worst. Third, while some recent studies have shown low rates of fetal loss, the very first multi-institutional US study of amniocentesis published in 1976 found no statistically significant difference in the loss rate between the amnio/no amnio groups. The most honest and accurate thing you could say to patients is”Studies have shown that amniocentesis has a loss rate anywhere from no increased risk on up to about 1%. There is no reason to believe that this center’s loss rate is different than published data, although we lack the statistical evidence from our own center to prove that assertion”. How can you justify saying anything other than that?
2) Maternal serum screening for aneuploidy is a mainstay of patient referrals to genetic counselors. Parents are put through the emotional ringer, and make life-changing decisions based on the results of these tests. The aneuploidy risk is usually stated as a precise-sounding fraction, e.g. 1/127. To soften the blow, we may re-state it as a “barely 1%”, or re-frame it as more than a 99% chance that the fetus does not have an aneuploidy. But no matter how you say it, the statistic itself still carries an air of truth and authority, etched in stone and handed down to us from the Lab Gods. Yet a single blood sample from one patient analyzed in different labs can result in very, very different aneuploidy risks from each lab. So which number is right? Which lab do we swear by and why?
3) In cancer counseling, it is common practice to assess the likelihood of carrying a BRCA or other gene mutation by using one of the many risk assessment models – BRCAPro, FHAT, PAT, BOADICEA, etc. While each model has its strengths and weaknesses, it is pretty clear that the same family, when assessed by multiple models, can wind up with very different BRCA carrier probabilities. And although the adherents, er, uh, I mean, supporters of a particular model will tout its strengths, when it comes down to it, all models perform about the same, and the models usually perform extra poorly at the upper and lower risk thresholds. Why use one number over the other? In fact, why use any number at all (other than for research purposes)?
I suspect that which numbers we use tells us more about ourselves than about any absolute or approximate reality. I think it also makes us very uncomfortable when the pillars of truth are found to be structurally unsound; we are not comfortable “going there.” Do we use a particular set of numbers because that’s what our bosses told us to do? Because that’s what we learned in school? Because we just read an article by some respected and clever researcher, so we assume the numbers he or she uses must be pretty good? Are there deeper and subtler reasons? For example, I have argued that 0ur imperfect collective memory of the risks of amniocentesis has allowed us to construct a more palatable reason for offering amniocentesis at age 35, a justification based on a semi-fictional medical risk/benefit assessment rather than based on the more realistic reason of economic gains of preventing births of babies with Down syndrome.
Why do you think you use the numbers you do? Share your thoughts on this, and let’s get a lively debate going.