There's been a recent explosion in products that use genetic testing to determine which antidepressant is right for you. Since finding the right antidepressant can be a long process of trial and error, these tests seem very exciting.

Dozens of studies - usually conducted by the manufacturers - purport that they work. And countless excellent psychiatrists - again usually working for the manufacturers - have given their endorsements. But much of the field remains skeptical.  Dr. Allen Francis, chair of the DSM-IV task force, writes that "So far, genetic studies have no proven role in guiding medication choice in psychiatry - except for profits gained by labs doing premature testing." And Daniel Carlat, editor-in-chief of the Carlat Report, reports that "if we were to hold the GeneSight test to the usual standards we require for making medication decisions, we’d conclude that there’s very little reliable evidence that it works."

We agree with these voices of skepticism.

Prediction of drug response falls into two categories - pharmacokinetics and pharmacodynamics. "Pharmacokinetics" means the way your body metabolizes the drug.  For example, whether you're a "fast metabolizer" whose liver destroys the drug before it has time to take action. "Pharmacodynamics" means the way the drug works on your body.  For example, whether an SSRI antidepressant effectively increases your serotonin levels. Most genetic testing products claim to be able to predict both of these domains.

But experts in pharmacokinetics say that the few genes that these products test (usually genes coding for liver enzymes) can reliably predict response to antidepressants. From the Handbook of Pharmacogenomics and Stratified Medicine:

"Available data do not support a correlation between [antidepressant] plasma levels and the response of most [antidepressants] (with the exception of [tricyclics]) and this is probably linked to the lack of association between response and CYP450 genetic polymorphisms found for the most part by previous studies. Therefore, evidence does not as yet exist to support the recommendation of CYP genotyping in clinical practice, since the effect of CYP variants on clinical outcomes is still not completely clear for the most studied isoenzymes."

In other words, the plasma level of most antidepressants (the parameter that liver enzymes affect) doesn't really seem linked to how well they work. This is probably because antidepressants don't work through the plasma, rather they work in the brain, which is separated from the plasma by the blood-brain barrier. Although being in the plasma is a necessary first step to getting in the brain, plasma levels themselves, and the liver enzymes that help determine them, are not robust predictors of response. Studies investigating the link between genes coding for liver enzymes and antidepressant response have uniformly come up negative.

The pharmacodynamic picture is even more complicated. Most consumer genetic tests only look at one pharmacodynamic gene related to efficacy - SLC6A4. While two meta-analyses (linked here and here) have found it matters for antidepressant response, two other meta-analyses (linked here and here) have found that it doesn't, leaving its status questionable. Even the positive studies find it only explains about 3.6% of antidepressant response - not really enough to significantly matter. But this is the only pharmacodynamic gene linked to efficacy that most of these products test. So these products probably have little ability to predict pharmacodynamic response either.

Another issue with these tests, is that so many genes are involved in the antidepressant response, that likely, no gene in isolation has a huge impact on the efficacy of these drugs. Tansey et al found that no single chromosome explained more than 5% or so of antidepressant response.  But chromosomes have thousands of genes on them, and each gene can have hundreds of relevant sites. They warned that antidepressant response was likely to be "polygenic", ie involve very many genes of small effect. A well-validated genetic test, the massively polygenic score for heart disease, looks at 6.6 million variants. Most antidepressant tests look at between ten and twenty. If the trait really is polygenic, the tests may be useless.

Representatives of genetic testing companies have noted that all public research on gene-response correlation has been disappointing, but claim to have done their own proprietary research. Since this research is not released, we cannot be sure one way or the other. Given these companies are unregulated, there is literally nothing stopping them from potentially reporting biased information. The current claim is that their in they have solved this previously unsolvable problem of linking genes to antidepressant response.  However, they are keeping their solution secret, even as public study after public study comes up negative. The only evidence they provide is from in-house studies by scientists they hired purporting to show their product works. These studies are usually open-label, not placebo-controlled, and otherwise violate the rules of good research.

As a result, we may not recommend these tests to our patients. Instead, we recommend careful titration of and experimentation with antidepressants.

Right now, the best test for antidepressant response is through trial and error.  The best test for your personal metabolism of antidepressants is to take antidepressants at different doses and see which one works for you. We realize this is not a lot of comfort for patients who have had to try many different antidepressants are are tired of constant experimentation. A good psychiatrist can sometimes use features of the patient's condition or family history, as well as response from prior psychotropics to make predictions about which medication to try next.