Neurology Center

One of the difficulties doctors face when prescribing antidepressants is that they're unpredictable.

One person might do well on a certain drug, but the next person might get no benefit from the exact same pills. Finding the right drug for each patient is often a matter of trying different ones until one works.

So a genetic test to work out whether a certain drug will help a particular person would be really useful. Not to mention really profitable for whoever patented it. Three recent papers, published in three major journals, all claim to have found genes that predict antidepressant response. Great! The problem is, they were different genes.

First up, American team Binder et al looked at about 200 variants in 10 genes involved in the corticosteroid stress response pathway. They found one, in a gene called CRHBP, that was significantly associated with poor response to the popular SSRI antidepressant citalopram (Celexa), using the large STAR*D project data set. But this was only true of African-Americans and Latinos, not whites.

Garriock et al used the exact same dataset, but they did a genome-wide association study (GWAS), which looks at variants across the whole genome, unlike Binder et al who focussed on a small number of specific candidate genes. Sadly no variants were statistically significantly correlated with response to citalopram, although in a GWAS, the threshold for genome-wide significance is very high due to multiple comparisons correction. Some were close to being significant, but they weren't obviously related to CRHBP, and most weren't anything to do with the brain.

Uher et al did another GWAS of response to escitalopram and nortriptyline in a different sample, the European GENDEP study. Escitalopram is extremely similar to citalopram, the drug in the STAR*D studies; nortriptyline however is very different. They found one genome-wide significant hit. A variant in a gene called UST was associated with response to nortriptyline, but not escitalopram. No variants were associated with response to escitalopram, although one in the gene IL11 was close. There were some other nearly-significant results, but they didn't overlap with either of the STAR*D studies.

Finally, one of the STAR*D studies found a variant significantly linked to tolerability (side effects) of citalopram. GENDEP didn't look at this.

The UST link to nortriptyline finding is the strongest thing here, but for citalopram / escitalopram, no consistent pharmacogenetic results emerged at all. What does this mean? Well, it's possible that there just aren't any genes for citalopram response, but that seems unlikely. Even if you believe that antidepressants only work as placebos, you'd expect there would be genes that alter placebo responses, or at the very least, that affect side-effects and hence the active placebo improvement.

The thing is that the "antidepressant response" in these studies isn't really that: it's a mix of many factors. We know that a lot of the improvement would have happened even with placebo pills, so much of it isn't a pharmacological effect. There are probably genes associated with placebo improvement, but they might not be the same ones that are associated with drug improvement and a gene might even have opposite effects that cancel out (better drug effect, worse placebo effect). Some of the recorded improvement won't even be real improvement at all, just people saying they feel better because they know they're expected to.

If I were looking for the genes for SSRI response, not that I plan to, here's what I'd do. To stack the odds in my favour, I'd forget people with an moderate or partial response, and focus on those who either do really well, or those who get no benefit at all, with a certain drug. I'd also want to exclude people who respond really well, but not due to the specific effects of the drug.

That would be hard but one angle would be to only include people whose improvement is specifically reversed by acute tryptophan depletion, which reduces serotonin levels thus counteracting SSRIs. This would be a hard study to do, though not impossible. (In fact there are dozens of patients on record who meet my criteria, and their blood samples are probably still sitting in freezers in labs around the world... maybe someone should dig them out).

Still, even if you did find some genes that way, would they be useful? We'd have had to go to such lengths to find them, that they're not going to help doctors decide what to do with the average patient who comes through the door with depression. That's true, but they might just help us to work out who will respond to SSRIs, as opposed to other drugs.

Binder EB, Owens MJ, Liu W, Deveau TC, Rush AJ, Trivedi MH, Fava M, Bradley B, Ressler KJ, & Nemeroff CB (2010). Association of polymorphisms in genes regulating the corticotropin-releasing factor system with antidepressant treatment response. Archives of general psychiatry, 67 (4), 369-79 PMID: 20368512

Uher, R., Perroud, N., Ng, M., Hauser, J., Henigsberg, N., Maier, W., Mors, O., Placentino, A., Rietschel, M., Souery, D., Zagar, T., Czerski, P., Jerman, B., Larsen, E., Schulze, T., Zobel, A., Cohen-Woods, S., Pirlo, K., Butler, A., Muglia, P., Barnes, M., Lathrop, M., Farmer, A., Breen, G., Aitchison, K., Craig, I., Lewis, C., & McGuffin, P. (2010). Genome-Wide Pharmacogenetics of Antidepressant Response in the GENDEP Project American Journal of Psychiatry DOI: 10.1176/appi.ajp.2009.09070932

Garriock, H., Kraft, J., Shyn, S., Peters, E., Yokoyama, J., Jenkins, G., Reinalda, M., Slager, S., McGrath, P., & Hamilton, S. (2010). A Genomewide Association Study of Citalopram Response in Major Depressive Disorder Biological Psychiatry, 67 (2), 133-138 DOI: 10.1016/j.biopsych.2009.08.029