Psychopharmacology
Precision Medicine’s Cultural Limits
Even if we had an Oracle Chip, we'd still need doctors.
Posted March 26, 2015
Something a patient said this week made a strong impression on me. My patients experience the world differently than I do. When they refuse a treatment, medication or psychotherapy—a common occurrence—I find myself having to convince them to do the “right” thing.
“But doctor, everyone is different.”
The irony is that I am entirely on their side even though it seems as if I am fighting against them. It is this plea that, in many ways, drives my scientific research agenda. Current practice guidelines are designed using large, randomized, controlled clinical trials. The analysis of such trials, in most cases, draws a single conclusion based on a single primary outcome that was pre-specified. For example, you have high blood pressure. Let’s try giving lots and lots of people either a beta-blocker or a placebo. The results: The group that takes the medication has a lower rate of death. Therefore, we should give everyone with high blood pressure a medication: the standard of care.
For an individual patient, however, the medication may or may not work, because “everyone is different” in terms of the underlying cause of high blood pressure, their genetic makeup in response to the medication, and other “biological” factors. The premise of Precision Medicine is that if we measure the right things (i.e. your genetics, your blood, your brain, your kidney, your liver, etc.) and use advanced computer algorithms to analyze them, maybe we can make a prediction at an individual level, and assign the right treatment to the right patient.
The problem is, applying this in the clinic is complicated.
To illustrate this, let’s take a sci-fi approach to a problem in the treatment of depression.
It is year 2080, after years of neuroscience and genetics research and billions of dollars of public and private research funding, a product called Oracle Chip finally arrives. It takes a drop of patient’s blood and a scan of the patients brain, and gives me a probability measure of whether the patient would achieve remission with two new antidepressants with their main side effects, Xeonazril (weight gain) and Flozarac (sexual side effects). The process involves induced pleuripotent stems cells and optogenetics and deep learning of connectomics and numerous other fancy techniques and discoveries that resulted in multiple prestigious prizes with new buzzwords and patents in numerous fields.
A bumbling psychiatrist takes Oracle Chip to the clinic.
Case 1: 45-year-old man. Oracle Chip says, Xeonazril (30%), Flozarac (90%). The patient is a recent divorcee who absolutely refuses any medication that would interfere with his sex drive even though I tell him Flozarac has a phenomenally better efficacy for him. “Doc, I need my sex drive,” he tells me, “now that I finally left my unhappy marriage.”
Case 2: 22-year-old woman. Oracle Chip says, Xeonazril (90%), Flozarac (30%). The patient comes from a culture that has a significant stigma against mental health services, especially medications. I tell her that Xeonazril, in her case, has a phenomenal efficacy because of who she is. She tells me, “but I am also a daughter, and my parents would be disappointed in me if I took a medication for my problems.”
What is the final treatment plan? Should be pretty obvious even if you don’t have an MD. Case 1: Xeonazril; Case 2: no meds. 6 months later: case 1: gained 20 lbs, very little solution of symptoms. Case 2: very little resolution of symptoms.
What in the world? Precision Medicine didn’t do anything! We have all the information but it still didn’t resolve this issue that “everyone is different” and I can’t coax patients to do the “right” thing. I suspect this is why the recent pragmatic Precision Medicine Coumadin trial failed. Lots of more informed people have written on that trial, so I will say that this is just a hunch.
There are two things that “everyone is different” encapsulates:
(1) Everyone has a different biology—Precision Medicine will address this.
(2) Everyone has a different set of preferences, and these preferences change in time—Precision Medicine, as it stands, will not address this. Physicians and perhaps others (i.e. lawyers) will still need to be there for this.
There might be complicated computational ways around these issues by adding additional predictors, such as patient preference, into the model. However, this is incompatible with current principle of random assignment for establishing causality in clinical trials. We also don’t know when the underlying reasons for patient preference (or, as a behavioral measure, adherence) are part of that causality in outcome.
We will need some innovative methods in clinical research.