Thirty-two percent. In the LQD trial—the pragmatic Lithium versus Quetiapine study—that's the fraction of patients who met criteria for treatment-resistant depression while never having received psychological therapy during their current depressive episode.
They qualified as resistant to treatment they never got.
Hold that number against a single clause in the Marques consensus vignette. To run their exercise on next-step treatment selection, the authors stipulated that the patient had already completed "an adequate course of psychotherapy," which conveniently removed therapy from the menu of options under discussion.
It reads as housekeeping. It is actually a confession.
That one phrase encodes the entire field's working habit: psychotherapy as background scenery, a prerequisite you assume happened, not a trial you count. The vignette didn't debate whether therapy was adequate—it declared it, so the panel could get on with ranking drugs and devices.
De Oliveira Reis and colleagues pulled the pin on that clause. In a letter to the Journal of Clinical Psychiatry, they reportedly noted that Marques' own team admitted it is "uncommon to encounter patients" in real practice who arrive with structured psychotherapy exposure alongside full medication adherence.
Read that twice. The consensus assumed a condition the authors concede they rarely see.
Which means the vignette's tidy patient—six months of weekly therapy, fully medication-adherent—is a laboratory animal, not a clinic patient. And the whole architecture of "what to do next" was built on top of an assumption the field knows to be usually false.
The vignette's tidy patient—six months of weekly therapy, fully medication-adherent—is a laboratory animal, not a clinic patient, built on an assumption the field knows to be usually false.
Here is the twist that makes this more than a squabble over a hypothetical: both sides agree.
Marques, Frye, and Singh, replying in the same issue, don't defend a drugs-only construct. They write that TRD "should not be limited to failure of pharmacotherapy alone," cite the evidence that combined treatment is more effective than either modality alone in achieving response, and endorse operationalizing psychotherapy adequacy in future staging.
So there is no live disagreement about the principle. Nobody in the exchange is arguing psychotherapy shouldn't count.
The gap is between what everyone concedes and what the definitions actually do.
And it's not that the tool doesn't exist. Marques' reply points directly at it: the Antidepressant Treatment History Form-Short Form-2, the ATHF-SF2, already defines a failed evidence-based psychotherapy trial with the same rigor applied to a drug.
The criteria are specific. Participation in a recognized therapy—acceptance and commitment therapy, behavioral therapy, CBT, problem-solving therapy, interpersonal therapy, short-term psychodynamic therapy, among others—for at least 12 sessions over 15 weeks, targeting depressive symptoms, without marked clinical improvement, and with substantial adherence.
Twelve sessions. Fifteen weeks. Documented adherence. No marked improvement. That is a failed trial, defined as tightly as any adequate-dose, adequate-duration antidepressant course.
So the measuring instrument is built, validated, and sitting on the shelf. The gatekeeping definitions simply haven't picked it up.
Look at the definitions that actually control access. The EMA, reportedly drawing on Souery's model, requires lack of meaningful improvement after two antidepressants from different classes, each at adequate dose and duration with documented adherence.
Two drugs. Count them. Nowhere in the regulatory definitions described here does a psychotherapy trial appear.
This is not a minor omission. These are the definitions that gate access to esketamine, to TMS, to the resource-intensive interventions—and they award eligibility on a partial audit of the patient's treatment history.
Brown and colleagues catalogued 155 distinct definitions of TRD in the literature, diverging on nearly every parameter: number of required failures, minimum trial duration, dose thresholds, and—crucially—whether psychotherapy and neuromodulation count at all. The same person qualifies as resistant in one setting and not in another.
But the regulatory definitions, the ones with teeth, have converged on the narrowest version: pharmacology only. Escalation eligibility gets stamped once two pills fail, regardless of whether a single adequate course of therapy was ever logged.
Now put a real patient inside that machinery.
Mr A is 54, a veteran, on his fourth-plus antidepressant—SSRIs, SNRIs, tricyclics, antipsychotic augmentation, none ever producing remission. He is currently on duloxetine at 90 mg. By any drug-counting definition, he is textbook TRD.
He also, per the case, "recently started cognitive-behavioral therapy for his depression, but he is doubtful about whether this will be helpful, and he has missed several sessions."
And so he wonders: should he skip the CBT and undergo TMS instead?
Read that sentence against the ATHF-SF2 criteria. Recently started. Several sessions missed. Doubtful it will help. That is not 12 sessions over 15 weeks with substantial adherence—it is the textbook picture of an inadequate psychotherapy trial, a therapy course that hasn't happened yet in any countable sense.
Under a drugs-only staging rule, none of that matters. Mr A already clears the bar. The half-attended CBT is invisible to the definition; his four failed medications are the whole story the system reads.
So the machine is available to him now, before an adequate psychosocial trial has even been logged—and the definition offers no reason to wait, because psychotherapy was never on the ledger to begin with.
This is where the design flaw stops being academic. A patient who dutifully completed 12 adherent sessions and a patient who missed most of his are, to the FDA and EMA definitions, indistinguishable. Both are simply "two-drug failures." One got an adequate psychosocial trial; the other didn't; the staging rule cannot tell them apart.
The construct, as De Oliveira Reis put it, risks remaining "defined more by pharmacologic sequencing than by comprehensive treatment exposure."
There is a name for part of what this produces: pseudoresistance. Apparent failure driven not by true refractoriness but by inadequate treatment—wrong dose, poor adherence, or, here, a psychotherapy trial that never reached threshold. Among the 30–60% of patients who show incomplete response, many reportedly reflect pseudoresistance rather than genuine TRD.
A drugs-only definition can't screen for the psychotherapy version of that error, because it doesn't look at psychotherapy at all.
None of this argues that TMS doesn't work, or that Mr A should be denied a device. It's an argument about sequencing—about what a "failed trial" is allowed to mean before someone qualifies to be escalated.
The stakes scale with the hardware. When escalation meant another pill or an augmentation agent, an over-inclusive definition wasted a prescription. When escalation means a machine that delivers stimulation to the brain—TMS, and beyond it the implantable end of the spectrum—the same partial audit routes patients toward invasive, high-cost interventions on incomplete evidence of what they've actually tried.
The tools to close the gap already exist. The ATHF-SF2 defines a failed psychotherapy trial as precisely as a failed drug trial. The consensus authors and their critics agree it should count.
What's missing is the one step that matters: writing psychotherapy adequacy into the definitions that actually decide who gets the machine. Until that happens, "treatment-resistant" will keep certifying resistance to a course of treatment a third of these patients never received—and the hardware entering the clinic will keep making that omission more consequential, one patient at a time.