Which Antidepressant Is Right for Me?

If you are reading this, you are probably in one of two situations. Either you are about to start an antidepressant for the first time and want to know how the choice is made, or you have already tried one that did not work and you are wondering what comes next.

Either way, the honest answer is that no antidepressant is best for everyone. Large comparative trials have never produced a clear winner, which is why guidelines generally treat the first-line options as roughly equivalent on average and expect the choice to be individualised. What follows is the reasoning a prescriber actually works through, in roughly the order they work through it, and where your genetics fit in.

This is the first filter and it does most of the work. Depression, generalised anxiety, panic disorder, OCD, PTSD and premenstrual dysphoric disorder do not all respond to the same drugs, and some antidepressants carry approvals or evidence for one and not another. OCD, for example, tends to need higher doses and a longer trial than depression does. If there is a bipolar history, an antidepressant on its own can be the wrong choice entirely. Nothing about your genetics changes any of this.

This is usually the second filter, and it is where most real-world decisions are made. The common first-line options differ meaningfully in what they tend to cause: sexual side effects, weight change, sedation versus activation, and how rough the first two weeks feel. A drug that is well tolerated by most people but hits the one side effect you cannot accept is the wrong drug for you, regardless of what anything else says.

Some of this is negotiable. Sedation can be an advantage if you are not sleeping. Activation can be an advantage if you are flattened.

Antidepressants interact with a lot of things, and some of them interact with each other. A few are strong inhibitors of the same liver enzymes that clear other drugs, which can push an unrelated medication to unexpectedly high levels. Combining a serotonergic antidepressant with certain migraine drugs, tramadol, linezolid or another serotonergic agent raises the risk of serotonin syndrome.

This is also where genetics and drug interactions meet: a medication that blocks one of your metabolising enzymes can make you behave, temporarily, like someone born with a slow version of it. That effect is called phenoconversion, and it can override your genotype entirely.

A drug that worked for you before is usually tried again first, and a drug that clearly failed at an adequate dose for an adequate duration usually goes to the back of the queue. Family history carries some weight too, though less than people expect. Prior response is one of the strongest predictors available and it costs nothing to ask about.

This is where pharmacogenetics enters, and it enters here rather than first for a reason: it tells you about exposure, not about whether a drug will suit you.^[1] Two enzymes do most of the work across this class, CYP2C19 and CYP2D6, and the antidepressants split fairly cleanly between them. If you clear one of those enzymes unusually fast or unusually slowly, the drugs that depend on it are the ones most likely to give you either no benefit or an outsized dose of side effects.

The practical value is that the class is large enough to route around a problem. If CYP2D6 is the issue, several options do not lean on it.

Most first-line antidepressants are available as inexpensive generics, but not all of them are, and formularies differ. Dosing frequency, whether a drug can be taken with food, and how badly it is affected by a missed dose all matter more to whether a treatment works than they get credit for, because the best-matched drug in the world does nothing if it is not taken.

Two liver enzymes account for most of the pharmacogenetic guidance in this class, and which one matters depends entirely on which antidepressant is being considered. This is why genetics is more useful for choosing among antidepressants than for judging any one of them in isolation.

The pattern worth taking away is that the antidepressants are not affected uniformly. A CYP2D6 poor metabolizer has a genuine issue with the tricyclics and with paroxetine, and no particular issue with the drugs cleared by CYP2C19. Someone who is an ultrarapid CYP2C19 metabolizer has the mirror-image problem. Because the class is large, a result that rules out several options usually still leaves several others.

What genetics cannot tell you is whether a drug will lift your mood. It describes what your body does to the drug, not what the drug does to your brain.

Testing is most useful at two moments. The first is after one antidepressant has failed at an adequate dose for six to eight weeks, particularly if the failure looked like no effect at all rather than a partial response, or if side effects appeared at a dose most people tolerate. Both patterns are consistent with an exposure problem, which is exactly what a pharmacogenetic result speaks to.

The second is before starting, if you already know you are going to be choosing among several options and would rather not discover an enzyme problem by trial and error. It is worth being realistic about the value here: a result narrows the field and explains failures, but it does not hand you an answer, and a normal result across both enzymes is a common and unhelpful outcome.

• Antidepressant Not Working? What to Know Before You Give Up
• Anxiety Medication Not Working? Genetics May Be Part of the Problem
• Looking for a GeneSight Alternative? Here's the Short Answer
• Looking for a Genomind Alternative? Here's What to Know
• Escitalopram Not Working? Why Lexapro Might Not Be Helping
• 23andMe Pharmacogenetics: How to Get a Drug Response Report From Your Existing Data

No. Comparative trials have not produced a clear overall winner among the common first-line options, which is why guidelines expect the choice to be individualised rather than standardised. The differences that matter in practice are side-effect profiles, interactions and how your body handles a particular drug.

Not directly. Pharmacogenetic testing tells you how quickly you process specific drugs, which predicts exposure and therefore the risk of no effect or excess side effects. It does not measure whether a drug will improve your mood. It is best understood as a way to narrow the field and explain past failures rather than as a way to pick a winner.

No. Response to one antidepressant is a weak predictor of response to another, particularly when the two are cleared by different enzymes. A meaningful share of people who do not respond to a first drug respond to a second. How the first one failed is the useful clue, so it is worth describing precisely.

Most prescribers look for six to eight weeks at an adequate dose. The first two weeks can feel worse rather than better as your body adjusts. If you are only a few weeks in, it is usually too early to judge, though intolerable side effects are a reason to check in sooner.

They matter more. Tricyclics have a narrow margin between an effective dose and a toxic one, so the same metabolizer status that would prompt a modest dose adjustment for an SSRI may prompt guidelines to suggest considering a different drug altogether.^[2] That is one reason tricyclics are not usually the first thing tried.