Choosing between SSRIs, SNRIs, and what comes next

The decision tree for picking an antidepressant in 2026 isn't really a tree, it's more like a workflow, and the workflow is a lot simpler than the size of…

Sections

1. First trial, never been on anything
2. First SSRI didn’t work, now what
3. Two failed trials, now we’re somewhere different
4. Reconsider the diagnosis (the move that mostly gets skipped)
5. Augmentation options, with what they’re actually trading
6. Ketamine, esketamine, and TMS
7. What’s nice to hear about all this
8. The pattern that ends up being not-TRD
9. Bottom line
10. Sources

The decision tree for picking an antidepressant in 2026 isn’t really a tree, it’s more like a workflow, and the workflow is a lot simpler than the size of the drug menu would suggest. Here’s how it actually plays out when a guy sits down who’s never been on a psych med before.

First trial, never been on anything

SSRI. Lexapro or sertraline, usually. Both are well-tolerated, broadly effective, generic and cheap, and have decades of data behind them. Lexapro tends to be slightly cleaner on side effects, sertraline has a wider dose range to play with, both are reasonable defaults. The choice between them usually comes down to small factors… if the depression has prominent OCD features sertraline can be pushed higher and tends to do better, if the depression is anxiety-dominant Lexapro is slightly preferred. Either is defensible, and a prescriber who has a strong opinion about which one of these is “really” the right first SSRI is overthinking it.

I don’t start a first-ever antidepressant patient on Effexor, Cymbalta, or Wellbutrin in most cases. The SNRIs have more pronounced side effects on the way up and a noticeably harder discontinuation profile (Effexor is the worst offender, and we’ll get to that), and there’s no good reason to start with a harder drug when the easier one works for most people. Wellbutrin is a reasonable first-line for the subset of patients with depression-plus-low-motivation-plus-low-energy and no significant anxiety, but it’s a smaller subset than the general SSRI-appropriate population.

Trial length: eight to twelve weeks at a real dose. Not at the starting dose, the real dose. Most failures of first SSRIs are dose failures masquerading as drug failures, where somebody sat on Lexapro 10 mg for six weeks, decided it wasn’t working, switched, and never found out whether Lexapro 20 mg for eight weeks would’ve done the job. The starting dose is where you ramp from. The therapeutic dose is usually double or triple that.

First SSRI didn’t work, now what

Two real options. Switch to a different SSRI in the same class, or switch class to an SNRI or Wellbutrin.

Switching within the SSRI class (Lexapro to sertraline, sertraline to Prozac, etc) used to be considered a weak move because the drugs are mechanistically pretty similar, but the actual data shows within-class switching is meaningfully effective, probably because individual response to specific SSRIs is more variable than the mechanism would predict. Different SSRIs hit slightly different secondary targets in different proportions (the term sigma receptors will appear in textbooks here, all you need to know is that the SSRIs aren’t strictly interchangeable even though they look like they should be), and patients sometimes respond well to a second SSRI after the first one did nothing for them. Which is one of those clinical facts that the residency lectures don’t quite manage to convey in a useful way.

Switching class is the other option. SSRI to SNRI (Lexapro to Effexor or Cymbalta) makes sense particularly if the depression has a physical component (chronic pain, the bone-deep-tired version of fatigue) that the SSRI didn’t touch. SSRI to Wellbutrin makes sense if the depression has prominent low motivation, anhedonia, or fatigue and not much anxiety.

The choice between within-class and across-class isn’t a rule, it’s a judgment call. I tend to within-class first if the first trial had a partial response, on the theory that we got something useful out of the serotonin mechanism and might do better with a slightly different SSRI flavor. I tend to switch class if the first trial gave zero response at a real dose, on the theory that the patient’s depression isn’t serotonin-driven and needs a different lever.

Two failed trials, now we’re somewhere different

Technically you’re now in treatment-resistant depression (TRD) territory, where TRD just means two adequate antidepressant trials in a row that didn’t get you better. Two trials isn’t actually that many, and a lot of patients have been called treatment-resistant when really they had two short or under-dosed trials that weren’t actually adequate. But if both trials were real (eight-plus weeks at a real dose), then yeah, this is a depression that’s not responding to standard first-line treatment.

At this point the decision menu opens up. Options: augment with a second drug, try a third class entirely (a tricyclic like nortriptyline, or in highly selected cases an MAOI), try ketamine or esketamine, consider TMS (transcranial magnetic stimulation, the magnet-pulse-over-the-prefrontal-cortex procedure), consider ECT in severe cases, or reconsider the diagnosis. The last one is the move that gets skipped, and it’s usually the one that matters most.

Augmentation is usually the next move. The most-evidenced augmentation options in 2026 are lithium at low dose (300 to 600 mg, way lower than bipolar dosing), Abilify (aripiprazole, an atypical antipsychotic which at low doses like 2 to 5 mg is FDA-approved as an antidepressant add-on, not used here at the high doses you’d see in actual psychosis treatment), Wellbutrin added on top of an SSRI, and T3 thyroid hormone (Cytomel, low dose) which is an older strategy that still works for specific patients.

Reconsider the diagnosis (the move that mostly gets skipped)

At two failed trials this is the move that gets skipped the most often, and it’s usually the one that ends up actually solving the problem. A surprising number of “treatment-resistant depression” cases turn out to be something else that got missed. Bipolar II with hypomanic episodes the patient thought were just “really productive weeks.” ADHD where the depression is secondary to years of cognitive struggle that the patient never named as ADHD. Alcohol use disorder where the patient is drinking four to six beers a night and hasn’t connected the dots. Untreated sleep apnea where the daytime fog and the mood drop both come from getting zero deep sleep for years. Hypothyroidism. Untreated grief. Each of these looks like depression that isn’t responding to antidepressants, because the antidepressant is treating a problem the patient doesn’t actually have.

This comes up a lot. The kind of guy who shows up on his third SSRI frustrated that nothing’s worked usually turns out to have an actual story that the prior prescribers didn’t dig for. His sleep is wrecked because of severe undiagnosed apnea. He’s drinking five drinks a night and didn’t think to mention it because the drinking pre-dated the depression. His mom had bipolar II and he had a couple of weeks in his twenties where he didn’t sleep much and felt incredible and started three projects, which he remembers as good weeks. The third antidepressant isn’t going to help that guy. A CPAP machine is going to help him, or a Lamictal trial is, or putting down the bottle is.

If you’ve failed two antidepressant trials, the next appointment should be re-taking the history from scratch with fresh eyes. Sleep, drinking, family history of bipolar, energy patterns over years not weeks, any episodes of unusually elevated mood that didn’t get labeled bipolar at the time. The diagnosis often shifts at this point, and the shift is most of why the treatment finally starts working.

Augmentation options, with what they’re actually trading

Lithium at 300 to 600 mg added to an SSRI is the oldest and most-evidenced augmentation strategy. Modest dose, requires labs (a lithium level, thyroid panel, and kidney function check every six months or so), but the response rate in partial responders is genuinely good and the drug has independent suicide-reduction data that no other augmenting agent has. Underused now, mostly because the labs are a hassle and the prescribers who came up in the SSRI era don’t reach for it the way the prescribers who came up before SSRIs did.

Abilify at 2 to 5 mg is the most-used augmentation in current American practice. Works, often quick, side effects at this dose are usually manageable. The trade-offs are metabolic effects (weight, blood sugar, cholesterol) with longer use, and the small risk of akathisia (an inner restlessness, like your skin won’t stop wanting to crawl out from under itself) which is genuinely miserable when it happens. Worth a try in the right patient, worth being honest about the trade-offs.

Wellbutrin added to an SSRI is the most common combination I personally end up with. Cleans up sexual side effects (see the Wellbutrin post for the whole conversation), adds energy and motivation, augments the antidepressant effect, generally well-tolerated. Not as well-studied as lithium or Abilify for augmentation specifically, but the practical clinical experience is good.

Thyroid augmentation with T3 (Cytomel) at 25 to 50 micrograms is an old strategy that’s been mostly forgotten. Some patients respond to it specifically. Worth considering particularly in patients with even subclinical thyroid issues on labs, the kind where the TSH is technically in range but on the higher end and the patient’s labs would have been called hypothyroid forty years ago.

Ketamine, esketamine, and TMS

Ketamine (IV or oral) and esketamine (Spravato, the FDA-approved nasal spray version) are real options for TRD. Response rates in treatment-resistant patients are forty to sixty percent, which is meaningfully better than another SSRI trial in this population. The catch is the effect is short-lived without maintenance dosing, the cost is significant, insurance coverage is uneven, and the practical access is limited to clinics that have the setup. Worth pursuing if you’re stuck. Not a first-line.

TMS is FDA-approved for TRD and works for a meaningful subset of patients. Thirty to thirty-six sessions over six to nine weeks, in-person, daily, half an hour at a time. The time commitment is substantial and that’s the main practical barrier, not the procedure itself, which is uneventful. Response rates around thirty to fifty percent in TRD populations. Insurance generally covers it now after two failed antidepressant trials, which wasn’t the case ten years ago.

ECT (electroconvulsive therapy, the controlled-seizure-under-anesthesia treatment that everybody thinks of as the One Flew Over the Cuckoo’s Nest thing and which doesn’t resemble that movie in any way) is the most effective treatment we have for severe depression, and it’s dramatically underused because of cultural baggage from a procedure that was retired forty years ago. The modern version is done under general anesthesia, the cognitive side effects are real but usually transient (some memory loss around the treatment window that mostly clears in a few weeks), and the response rate in severe TRD is around seventy percent. Higher than anything else we have. The barrier is patient hesitation, not effectiveness. Which is unfortunate, because there’s a guy reading this who needs it and isn’t going to ask.

What’s nice to hear about all this

The workflow above looks intimidating because it has a lot of options, but the people-getting-better end of it is real. Most patients who finish a thoughtful version of this process end up significantly better, sometimes in remission, sometimes at a partial response that’s still a huge improvement over baseline. The drug menu is bigger now than it has ever been, the augmentation options are well-studied, and the procedures (TMS, ECT, ketamine) have all gotten substantially less scary in their modern forms than the reputation that’s been chasing them for forty years. The grim version of the depression-doesn’t-respond conversation has a lot more rungs in the ladder than most patients are aware of, and most of them work for somebody.

The pattern that ends up being not-TRD

The most common surprise after re-taking the history is that the guy who’s been on three different antidepressants over four years with partial response on each wasn’t actually treatment-resistant in any meaningful sense, he had a different problem the antidepressants weren’t equipped to treat. The kind of guy who comes in looking like a TRD case has often been drinking four to six beers most nights for years and not connecting it to the depression. Sometimes he turns out to have sleep apnea on a sleep study that nobody had ordered. Sometimes there’s a family history of bipolar on one parent’s side that he didn’t bring up because he didn’t think it was relevant, and a couple of weeks in his late twenties that look a lot like hypomania in retrospect.

The treatment plan that ends up working for that guy isn’t a fourth antidepressant. It’s Lamictal, a CPAP, and cutting the drinking, in some order. Two years later he’s in remission, the antidepressants were never the problem, the diagnosis was. Which is the lesson that ends up being the actual content of this post, dressed up as a decision tree.

A surprising fraction of “treatment-resistant” depression turns out to be something else that got missed.

Bottom line

First trial, SSRI at a real dose for a real length of time. First failure, switch class or stay in class based on the response pattern. Second failure, re-take the history before declaring TRD, then augment, switch class again, or consider ketamine, TMS, or ECT depending on severity. Most failed antidepressant trials are failed by inadequate dose, inadequate duration, or missed diagnosis, in roughly that order. The drug menu in 2026 is bigger than ever, the workflow is essentially unchanged, and the move that most patients need isn’t a different drug, it’s a more honest version of the appointment that gets the diagnosis right.

Sources

1. Cipriani A, Furukawa TA, Salanti G, et al. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet. 2018;391(10128):1357-1366. PMID 29477251.
2. Rush AJ, Trivedi MH, Wisniewski SR, et al. Acute and longer-term outcomes in depressed outpatients requiring one or several treatment steps: a STAR*D report. Am J Psychiatry. 2006;163(11):1905-1917. PMID 17074942.
3. Papakostas GI. Managing partial response or nonresponse: switching, augmentation, and combination strategies for major depressive disorder. J Clin Psychiatry. 2009;70(Suppl 6):16-25. PMID 19922740.
4. Furukawa TA, Cipriani A, Cowen PJ, Leucht S, Egger M, Salanti G. Optimal dose of selective serotonin reuptake inhibitors, venlafaxine, and mirtazapine in major depression: a systematic review and dose-response meta-analysis. Lancet Psychiatry. 2019;6(7):601-609. PMID 31178367.