Wednesday, October 28, 2026
Two papers, a psychopharmacology pearl, a psychotherapy pearl and the news.
AI-generated content. This briefing was written by Claude Opus 5.5 and may contain errors. Please verify against the primary source before acting clinically.
| The Paper |
| New research from the psychiatric journals |
Multimodal Machine Learning–Based Prediction of Relapse in Major Depressive Disorder
Lea Teutenberg, Nils R. Winter, Paula Usemann, et al. (37 authors; senior author Tilo Kircher). JAMA Psychiatry. October 2026.
Researchers from the Marburg-Münster Affective Disorders Cohort followed 564 adults with current or lifetime major depressive disorder for about two years; 270 relapsed. At baseline each participant contributed up to nine data modalities: clinical and psychometric measures, four structural and functional MRI modalities, genetics, neurocognition, blood markers, and gut microbiome. Machine-learning models were trained to classify who would relapse, with balanced accuracy as the main metric.
Across models, mean balanced accuracy ranged from 47% to 64%. The best model used only a few established clinical relapse risk factors, and adding imaging, genetic, blood, or microbiome data on top of it did not improve prediction.
|
In practice
Even with nine data streams, the scan, the genotype and the microbiome added nothing to a careful clinical history. For now, relapse risk comes from what you already ask about, and no current biomarker panel justifies a different maintenance plan for an individual patient. |
Full text doi:10.1001/jamapsychiatry.2026.3232
| Outside Psychiatry |
| A general-medicine paper that ends up in our clinic |
Smoking Cessation Among Underserved Patients Referred for Lung Cancer Screening: A Randomized Clinical Trial
Joanna L. Hart, Anil Vachani, Christine M. Neslund-Dudas, et al. (22 authors). JAMA. October 2026.
This pragmatic trial at five centers in four US health systems randomized 3,259 adults who smoked and had been referred for lung cancer screening, all Black, Hispanic, rural, or of low socioeconomic status. Arms were usual ask-advise-refer care; usual care plus free pharmacotherapy (nicotine replacement, with reimbursement for varenicline or bupropion); free pharmacotherapy plus up to $600 in incentives paid only for biochemically confirmed abstinence; or all of that plus a mobile tool prompting patients to think about their future health.
Biochemically verified sustained abstinence at 6 months was 4.3% with usual care, 5.1% with free pharmacotherapy, 8.8% with incentives, and 7.2% with the added mobile tool. Incentives beat usual care by 4.6 percentage points (95% CI, 2.1–7.0), while free medication alone did not differ from usual care.
|
In practice
People with serious mental illness smoke at far higher rates and die younger, largely of tobacco-related disease, so this matters to us. The lesson travels: handing out cessation medication does little unless someone takes it, and verified-abstinence rewards (the same logic as contingency management for stimulant use) roughly doubled quit rates. The trial did not report a psychiatric subgroup, so the size of the effect in our patients remains untested. |
Full text doi:10.1001/jama.2026.17192
| Past the Boards |
| One advanced clinical point, usually pharmacology |
Fluvoxamine multiplies duloxetine exposure sixfold, so the combination is avoided rather than dose-adjusted
Duloxetine is cleared by CYP1A2 and CYP2D6, but 1A2 does most of the work. In the label's interaction study, fluvoxamine 100 mg (a potent 1A2 inhibitor) raised duloxetine AUC about 6-fold, Cmax about 2.5-fold, and half-life about 3-fold. Paroxetine, blocking only the 2D6 route, raised AUC by about 60%. The difference is large enough that the current US label says to avoid co-administration with potent 1A2 inhibitors such as fluvoxamine and ciprofloxacin, rather than lowering the dose.
The practical traps are a fluvoxamine augmentation in OCD layered onto duloxetine prescribed for pain, and a ciprofloxacin course for a UTI. Smoking works the other way. Tobacco smoke induces 1A2 and cuts duloxetine bioavailability by about a third. The label does not adjust dosing for smokers, but a patient who quits abruptly on an inpatient unit loses that induction. Duloxetine is itself a moderate 2D6 inhibitor and tripled desipramine exposure in the label's study, so watch 2D6 substrates added on top of it as well.
| In the Room |
| A psychotherapy technique, shown in a vignette |
Concordant or complementary: name which identification you are in before you interpret
Racker split countertransference into two kinds. In concordant identification the therapist feels what the patient's self feels. In complementary identification the therapist is pulled into the role of the patient's internal object: the critical parent, the neglectful caretaker, the one who gives up. Both are information, but they point to different interventions. Concordant feeling supports empathic clarification of the patient's experience. Complementary feeling tells you which object relation is being lived out in the room, and that is where transference interpretation becomes possible. The common mistake is acting on a complementary pull while believing it is clinical judgment. Examples include the sudden urge to set a firm limit, to reassure, or to end the session a minute early. The discipline is to ask, before you act: whose feeling is this, and which role am I being handed?
|
Vignette
A 34-year-old man describes, in a flat voice, his third missed deadline at work, then says, "I know, I know, you're going to tell me I'm sabotaging myself." The resident feels a sharp, unfamiliar impatience and is about to point out the pattern. She notices that the impatience is not hers. It has the tone of the father the patient described weeks earlier, who "never had to raise his voice." Rather than deliver the confrontation the patient had already scripted for her, she says: "You've already written my line, and it's a disappointed one. I notice I was halfway to saying it. What happens in you while you wait for me to say it?" He pauses, then says it is almost a relief to be scolded, because then at least someone is paying attention. The resident had been cast in a complementary role, and she used the pull to interpret it rather than play it out. |
| *** |
| That's the five. Go see your first patient. |