In a previous installment of the Dex Diaries, readers learned that our concerns about prenatal dexamethasone for CAH centered on an apparent bait-and-switch whereby prospective pregnant patients and their doctors were told by a clinician-researcher that the intervention was “safe for mother and child” while the same clinician-researcher obtained NIH funding to see if it really was safe. In today’s Dex Diary, Dr. Aron Sousa — the person who suggested the Dex Diaries series — articulates many physicians’ frustrations with this scene.
Alice asked me to contribute a little something to the Dex Diaries, and since I have lived a good deal of this project, I am taking her up on it.
First off, some disclaimers: I have a lot riding on this issue. I am a physician, and my job as a senior associate dean for medical education is to train future physicians. So I have high expectations of my profession and its future. Doctoring is a noble profession, but it only remains so if its members rout out mediocrity and plain old bad behavior. And I am Alice's mate — I want her to do the right thing, to be careful, to be forthright, and use her considerable energy and talent for the public good. That’s why I’ve helped and supported her throughout this project.
Among all the complexities of off-label use of prenatal dexamethasone for CAH lays the issue of mediocre science. In this case, the right methods have not been used to study the right population to answer the right questions.
The right question for this kind of medicine is actually pretty simple and is taught, in essence, to all medical students and residents in this country. Good evidence-based questions can be formed using the mnemonic PICO: Patient, Intervention, Comparison, Outcome (see Evidence-based Medicine: How to Practice and Teach EBM, by Sackett et al., 5th ed., 1997, p. 27).
The patients in this case are women whose offspring bear significant risk for CAH and those women’s fetuses. The intervention is prenatal dexamethasone starting before the formation of genitalia, about week seven of the pregnancy. The comparison to the dexamethasone should be a placebo, because there is no competing intervention. That is to say, there is no other intervention available for what this off-label use is intending to do — prevent genital virilization — so there’s nothing but placebo to compare the dexamethasone to if you want an actual controlled study. And a study that isn’t well controlled is pretty useless.
I’ll allow that the desired outcomes are tricky to specify here, because we should care about how this treatment turns out for at least the mothers, the babies, and the people the babies become. In evidence-based medicine, we try to track the outcomes patients really care about. The mothers and the people the babies become probably care about outcomes like death, physical disability, mental disability, future medical treatment, and quality of life. (That’s a lot of possible outcomes to track, which speaks to how many risks this intervention entails.)
So, on a good day, any second-year medical student should be able to formulate the following medical research question about this intervention: Does the use of prenatal dexamethasone, in pregnant women at risk for having a child with CAH, starting in the seventh week of pregnancy, lead to less mental and physical disability, less need for medical treatment, and better quality of life when compared to placebo for the mothers, the babies, and the people the babies become?
Remember, this off-label maternal-fetal intervention has been in use for over twenty years. But today, if you run a PubMed search of “dexamethasone AND placebo AND congenital adrenal hyperplasia” you get a big fat zero. “No items found.” Same thing when you run the search on clinical trials. “No items found.” (That’s true with or without MESH headings, capitalization, etc.)
How are we even having an argument about this intervention? The standard of evidence for therapy is a randomized controlled trial (RCT). For prenatal dexamethasone, there should be a study in which mothers are randomized to get either dexamethasone or placebo (the control) and then the mothers and their offspring are prospectively followed for years to see which group does better for the outcomes we care about. Without this kind of evidence, doctors have no business giving pregnant women drugs that are chosen to actually change the fetus.The mind reels.
In this case, a well done RCT would also be blinded — the researchers, the patients, and the people assessing the outcomes would not know whether the mothers took the placebo or dexamethasone until the study was complete. The NIH grant materials obtained under the Freedom of Information Act show that there are plenty of patients who could have been enrolled in this kind of study to demonstrate the benefit claimed by practitioners. (Download for free the article that shows this.) Instead, the 2010 systematic review and meta-analysis on this intervention confirms how problematic this has all been. With generosity, I say, mediocrity rules.
Mediocre science is poor form, and prenatal dex for CAH is rife with mediocre science. But the really troubling piece that Alice, Ellen Feder, and Anne Tamar-Mattis found in this case was not mediocre science. The really troubling piece was apparent misdirection.
Here’s how the circle of misdirection can work in cases like this: A team of researchers gets some NIH money to study a population using a mediocre study design — say, a short-term observational drug trial. They choose a sample of convenient people who agree to use a medication already available for other uses, so they don't need to bother with any of the messiness of phase 1, phase 2, and phase 3 trials that new drugs normally require.
They get some promising results with the off-label use. The researchers use the promising results to tell patients of their wonderful new therapy — looks like it works! looks like it’s safe! — and this brings in more patients. The researchers then tell the NIH that they have all these “treated” patients to study and that with some more money, they can really answer whether this intervention works and is safe.
And with more mediocre study design — now retrospective sampling — the patients who happen to end up in the retrospective sample yield promising results. The researchers now claim to patients: New results! Looks good! More new patients (and their unsuspecting doctors) are told of the great new therapy, creating more potential subjects for retrospective studies. And here’s where prescription strength misdirection comes in: The researchers tell the NIH that the data is not perfectly clear and more study needs to be done to really get an answer. The patients are drawn in by promising results; the NIH funding is drawn in by perpetual uncertainty.
The cycle continues and we sitting on the sidelines are none the wiser. The mediocre study design yields more promising but inconclusive results, which leads to more patients, which leads to more NIH funding, and so on for years and perhaps millions of dollars and a PubMed search that still will not answer the second-year medical student’s most basic question. It’s like a perpetual motion machine.
Of course, perpetual motion machines are not actually possible — unless you're perpetually inadequately studying your own subjects.
My thanks to Aron for his contributions. In the next Dex Diary, drawing on her own experiences, Ellen Feder explores how prospective parents approach prenatal medications.