Clear, engaging, and practical updates on clinical psychiatry.

Earn CME for listening to the podcast with a multimedia subscription. Listen for free here or using Apple Podcasts, Android, or Stitcher.

Previous Post
Next Post

A Genetic Test that Didn’t Fail

Podcast, Volume , Number ,

Print Friendly, PDF & Email

Genetic testing is all the rage in psychiatry, but the test that might work best is the one that no one is talking about.

Published On: 10/04/2021

Duration: 15 minutes, 29 seconds

Related Article: Genetic Testing in Depression,” The Carlat Psychiatry Report, September 2021

Chris Aiken, MD, and Kellie Newsome, PMHNP,  have disclosed no relevant financial or other interests in any commercial companies pertaining to this educational activity.


Dr. Aiken: Last month, we covered a new randomized controlled trial of pharmacogenetic testing in depression. The study failed to find any meaningful difference with the genetic test, even in patients who had not responded to one or more antidepressant trials. We’ve covered a lot of these studies in the journal, and all of them have met a similar fate. The better the design of the study, the more likely it is to fail. Among the casualties of GeneSight’s AssureRx, NeuroIDgenetix, Neuropharmagen, and – in last month’s trial – Genomind’s Genecept.

Kellie Newsome: So why do we keep covering these disappointments? Well, for one thing the genetic testing industry keeps promoting these trials as successes, with glossy brochures that misrepresent the outcomes. In each of these trials, the test failed on the primary outcome, but it did make a small difference on some of the secondary outcomes. This is not unusual in statistical research like the type we do in psychiatry. Everytime you measure something in a study it’s like rolling the dice, and there’s a 1 in 20 chance that you might get a positive result by luck alone…. A false positive. The best way around this is to name your primary outcome in advance, much like a pool player has to name which pocket she plans to land the 8 ball in at the end of the game. You know how it goes in pool – if you miss that pocket but by some fair fortune the ball goes into another hole – you still lose the game. If we hold pool players to that standard, we ought to do the same for clinical research as well.

Dr. Aiken: That’s why secondary outcomes are not meant to inform practice. They are meant to inform future research, to test whether they are solid, repeatable results or just random statistical noise. That has not been done, but let’s suppose it has and these secondary outcomes really are valid. Still, they are not that impressive. In the case of GeneSight, you’d need to test 19 patients to bring 1 to remission based on the secondary outcomes. 

Kellie Newsome: But in fairness, sometimes the tests don’t even change the treatment, either because the results are normal or the patient isn’t taking any medications that are affected by the test. So one thing you’ll hear emphasized at industry sponsored talks on genetic testing is that the tests make a big difference when the the test results suggest that the patient’s medications are out of sync with their genes. Sounds intuitive – who could argue with that.

Dr. Aiken: Still, it’s not the slam dunk it sounds like. In the large genesight study, you’d need to test 13 patients to bring one to remission after excluding all the patients whose testing didn’t change their results. So it only brings the number needed to test down from 19 to 13. And, again, these are secondary outcomes so it’s not even clear these small numbers are based on a real difference.

Kellie Newsome: But the news is not all bad. In recent issues we’ve focused on specific genetic tests that can make a difference when your patient is taking certain medications – even if the larger genetic panels come up short. And there is one genetic panel that has a positive trial – and it involves an entirely different kind of gene.



Dr. Aiken: Most genetic tests look at a combination of pharmacodynamic genes – the ones that shape the brain’s response to medications – and pharmacokinetic genes, specifically the genes for the CYP enzymes that metabolize many psychiatric medications, influencing whether they’ll reach unusually high or low levels in the body. To put it another way, pharmacodynamic genes purport to influence medication selection – whether the patient will respond to a tricyclic or and SSRI – while pharmacokinetic genes influence medication dosing – whether they’ll need unusually high or low doses to achieve an adequate serum level.

But the genetic test with the positive study didn’t look at any of those. It looked at genes that code for the p-glycoprotein transporter. This transporter acts like a bouncer at the blood brain barrier, kicking unwanted chemicals out of the CNS. The problem is, it also kicks some psychiatric medications out of the CNS, and when it does the patient will need a high dose to have a chance at responding to the drug.

The genetic test in this study looked at two genes that code for the p-glycoprotein transporter: ABCB1 and ABCC1. We’ll call them the ABC genes for short. In some patients, these genes upregulate the p-glycoprotein transporter, causing it to kick drugs out of the CNS more aggressively so they require higher doses. In others it is downregulated, so that lower doses are required to achieve adequate levels in the CNS. But most people have average genes and need average doses. 

Now that you have an idea of what p-glycoprotein does, let’s look at the actual study.


The CNS Dose Study  

Kellie Newsome: The study was conducted by Ajeet B. Singh, an Australian psychiatrist who helped start the company that sells this genetic test. That company is CNS Dose, and the specific test they use is the called the amplichip test. The full CNSDose panel includes a few other genetic tests, like the popular CYP genes (2D6 and 2C19), as well as UGT1A1, which metabolizes desvenlafaxine. But Dr. Singh’s study only looked at the p-glycoprotein genes ABCB1 and ABCC1

Like most genetic studies, it was industry sponsored, but otherwise it checks all the right boxes on study design. It was a randomized trial where half the patients had gene-guided dosing and the other half had treatment as usual. It was double blind – meaning that the patients and raters didn’t know who got the test. That’s important because we’ve noticed a trend where unblinded studies come up positive, but when they are repeated with a blinded design the results don’t look so good – suggesting that simply ordering the test has a powerful placebo effect for the patient.

Dr. Aiken: The study involved 148 patients with major depression who were treated for 3 months, and unlike a lot of genetic trials they didn’t limit the sample to patients who had recently failed an antidepressant. That’s a strength here, because patients who failed their first trial are more likely to have genetic abnormalities, making it easier to come up with positive results with that enriched sample.

Another difference with this study is that the test was only used to inform dosing. The doctors were free to use any antidepressant they wanted, and the test told them whether they were likely to need a low, medium, or high dose. That’s also interesting, because we’d expect the inclusion of antidepressants that aren’t affected by p-glycoprotein to water down the results, and yet the results were still positive.

Kellie Newsome: Just how positive? The numbers here are pretty big. Those who got the gene-guided test were nearly 3 times more likely to remit from depression – 72% vs. 28%. Put another way, for every 3 patients you tested, 1 would achieve remission by virtue of having the genetic test – the number needed to test was 3.

Dr. Aiken: So what’s the catch? Well, this is only one study, and we’d like to see it replicated, preferably by a group with no stake in the product. There is a second study of CNSDose, but it had a very different design. Still, it’s worth reviewing because it tells us a little more about how the test influences dosing.

This 2017 study followed 119 patients with depression as they started treatment with desvenlafaxine, brand name Pristiq. This antidepressant is one whose dosing is likely to be affected by two genes in the CNSPanel – the UGT1A1 metabolizes desvenlafaxine in the liver, and the p-glycoprotein gene kicks desvenlafaxine out of the brain. In the study, doctors were encouraged to raise desvenlafaxine’s dose up 200 mg/day based on clinical response, but to raise it no faster than by 50 mg every 2 weeks. At the end of the study, 80% of the patients remitted, which is a pretty high number. They then ran the CNSDose test on those who remitted to see which dose panel would have recommended for each patient, and 90% of the time it did. 

These results are very informative, even if you don’t use the CNSDose test, because the PDR states that there is no evidence that desvenlafaxine works any better once you get beyond the 50 mg/day dose. People rarely dose this antidepressant beyond 100 mg/day, but in the clinical trials they did go up to 400mg/day and in Australia it’s approved up to a dose of 200mg/day.  In this study, the CNSDose test suggested high doses, at or above 150 mg/day, in 15% of patients. It suggested low doses at or below 50mg per day, in 20% of patients, and for the rest the recommended dose was between 50mg and 100mg a day.

Kellie Newsome: And here’s a paradox. Clinicians who use popular genetic tests like GeneSight often find that desvenlafaxine is the only antidepressant that almost never runs into conflict with the test results. And yet, it’s one of the antidepressants whose dosing is most impacted by the p-glycoprotein gene, which isn’t included in most genetic panels.

Dr. Aiken: The bottom line is that if you’re going to order a genetic panel, CNSDose is the only one we know of that can claim positive results, but those results only inform antidepressant dosing, not selection. And, if you don’t order the test, you might still get there by dosing a little higher than usual in patients who don’t respond, but you won’t know who is going to respond to a higher dose. That’s what inspired Dr. Singh to embark on this journey. Early in his career, he saw a patient with severe depression who had not responded to any treatments except venlafaxine. The problem was, he only partially responded at a dose of 225 mg/day. Dr. Singh decided to go beyond the max dose – gradually raising venlafaxine to 450 mg/day, at which point the patient fully remitted. The treatment was a success, but he knew that he could have saved this patient from a lot of pain – including the loss of his job, his marriage, and a suicide attempt – if he had found this strategy earlier. This got him interested in identifying patients with metabolic differences ahead of time through genetic testing.

Kellie Newsome: Dr. Singh has developed a new version of the AmpliChip panel which includes more genes – 16 in total – called AmplisEvo. The test is available for $195 dollars through BasePair Genomics. 

Dr. Aiken: P-glycoprotein is not as well studied as the CYP enzymes, so we’re not always sure which antidepressants are affected by it and how much. But many antidepressants seem to pass through its grip in one way or another, as do some antipsychotics, mainly risperidone and paliperidone. Besides genetic variations, drug interactions can also speed up or slow down the transporter, so understanding how to adjust the dosing can get complicated pretty quickly.

But let’s end on a simple note. We counted 4 antidepressants that p-glycoprotein does not touch. They are bupropion, duloxetine, fluoxetine, and mirtazapine. Testing for p-glycoprotein variations is not likely to be very informative for these meds, and they may be worth trying in treatment resistant cases as an alternative to raising the dose of a p-glycoprotein metabolized med.

And now for the word of the day…. Schizo-obsessive disorder

Obsessive compulsive symptoms are very common in schizophrenia – we see them on some level in 25% of patients with schizophrenia. Usually they show up when the patient is not in a full psychotic episode, though they can overlap with the psychosis. The word for this overlap is Schizo-obsessive disorder, or schizo-OCD, and next week we’ll interview Michael Poyurovsky on this complex comorbidity.

Got feedback? Take the podcast survey.


Leave A Comment