Research models, and the research that is done to support those models, often have very little predictive value to the real world.
The following was originally posted to the HCPLive network blog Thought Broadcast.
After my recent post on Silenor, the new sleeping drug, I received many nice—and not-so-nice—comments and emails from people who challenged my conclusions. The “not-so-nice” responses—which I did not publish–were, for the most part, from readers who were clearly invested in Somaxon Pharmaceuticals and must have feared that my comments would further weaken the value of their stock (as if my input was that powerful – lol). [BTW I have no financial interest, long or short, in Somaxon.]
But others were professional and well-reasoned, and emphasized the true need for a medication like Silenor: an effective sleeping medication with little daytime “hangover” effect and few other adverse effects. Some comments (which you can read on my page) were from users who seem to have had good experiences with Silenor.
Assuming these comments are truly from satisfied users of Silenor (a big assumption, as this is still the internet!!), I started to think about something that every psychopharmacologist (including myself) struggles with on a regular basis: namely, the fact that some people respond exactly the way we expect them to respond, based on our biochemical hypotheses, whereas others have completely different responses. Or to put it another way, our models—and the research that is done to support those models—often have very little predictive value when you translate this information to the real world. And nowhere is this more apparent than in psychiatry.
In medical school and residency, I was a bit of a pharmacology nut. (I’d use the word “wonk,” but I always thought that word was reserved for “insiders,” of which I am decidedly not one.) Pharmacology, indeed, is a fascinating subject. Bioavailability, pharmacokinetics, receptor-binding affinities, metabolism, drug-drug interactions—each of these is a discipline unto itself, with its own language, experimental methods, and predictive power.
In psychiatry, the intricacies of pharmacology become even more elaborate. There’s a lot of research (not to mention pages and pages of textbook chapters and lots of drug company promotional materials) describing the details drug dosing regimens, drug-drug interactions, receptor-binding affinities and their putative relationships with symptom profiles, etc. Just to pick out some familiar examples, the dosing strategies for Seroquel XR and Abilify seem quite detailed indeed: the predicted effects of very low doses are thought to be different from “mid-range” doses, which are different still from higher doses. And so on.
The truth of the matter is, these subtleties often make very little difference in the real world. That’s not to say that the theories of action are wrong. (Maybe Seroquel XR really is an effective norepinephrine reuptake inhibitor at moderate concentrations.) Instead, it speaks to the enormous variability among human beings, and the fact that our real-world patients sometimes don’t fit the models we construct for them.
Back to Silenor. I received a detailed email from Martin Scharf, a professor of psychiatry at Wright State University who has spent decades researching sleep medications. In fact, he was involved in several of the studies on low-dose doxepin which set the stage for Silenor. (See some of his work here and here.) In his email he claimed that I “misse[d] the point of the uniqueness of the dosage selection” for Silenor, citing how the low dose (i.e., <10 mg/d) makes the drug more specific for the histamine receptor and avoids side effects that can be attributed to its interaction with other receptor subtypes. Thus Silenor causes “virtually no next day effects,” whereas “at the 10 mg dose there is little doubt as to next day effects.” [Indeed, I should point out here that the same arguments were made, in great detail, in the Stephen Stahl article that I cited in my post.]
My response to Dr Scharf is: OK. But what do patients actually experience? In fact, I would take issue with the comment that “there is little doubt as to next day effects” because I have patients taking 25 or 50 mg doxepin who do not report next day effects (and some MDs who responded to my first post said the same thing). Or if they do experience next-day effects, perhaps they’re not bothersome and they’re only apparent upon questioning. In other words, the numbers predict one thing, but what do patients report?
The human body is a laboratory, and each drug we prescribe is an “experiment” in that laboratory. For many people, however, the laboratory doesn’t resemble the pristine one in the clinical trials. People take extra doses of drugs, or skip doses altogether. They’ll take drugs at different times than prescribed. Once inside the body (and even this process is highly variable), the drugs are metabolized differently, or they’ll react with existing chemicals (foods, other drugs, neurotransmitter receptors, different polymorphisms of receptors, etc) in ways that we can only begin to understand, due to this substantial combinatorial complexity.
Unfortunately (or fortunately, depending on how you look at it), bringing a drug like Silenor to market requires that the molecule be tested in very rigorous (and therefore “pristine”) conditions in both animals and humans. Dr Scharf and his colleagues have done painstaking work in showing precisely how drugs like Silenor work—ideally. But in the real world, “all bets are off,” as they say. Patients taking 25 or 50 mg doxepin often feel “just fine” and have no QT prolongation or weight gain, regardless of what the clinical trials would predict. Similarly, patients taking 6 mg might report absolutely no benefit at all.
I should emphasize, I do NOT advocate reckless prescribing—e.g., giving “PRN” prescriptions to everyone and ask them to adjust the dose according to what makes them feel better, or prescribing without consideration of all of the potential biological effects of our drugs. We clinicians do have a responsibility to look out for drug-drug interactions (which can be lethal at times), to check drug levels (for certain medications, even regardless of clinical response) and to prescribe potentially abusable medications cautiously. Knowledge of all of the principles of pharmacology is necessary to ensure the safe use of medications, and taking advantage of this knowledge can improve patient outcomes.
But in the real world it is all about the patient, and if they “do well” on doses of medications that seem to make no sense at all (and all safety measures are in place), isn’t that what counts most? And, moreover, shouldn’t that lead us to question the hypotheses that led us to those doses in the first place?