The research drew on two of the largest biomarker studies in depression research: EMBARC, the Establishing Moderators and Biosignatures of Antidepressant Response in Clinical Care trial conducted in the United States, and CAN-BIND-1, the Canadian Biomarker Integration Network in Depression. Together, the cohorts included 386 patients aged 18 to 65, with 257 women and 129 men. Participants were randomly assigned to receive the selective serotonin reuptake inhibitors sertraline or escitalopram, or a placebo, with 123 patients on sertraline, 138 on escitalopram and 125 on placebo. Crucially, each patient underwent resting-state functional MRI scans both before treatment and again after one to two weeks of medication, allowing the researchers to compute how thousands of pairwise connectivity relationships between brain regions changed as treatment began.

Functional connectivity, measured as the temporal correlation of spontaneous blood-oxygen-level-dependent signal between brain areas, provides a window into the brain’s intrinsic organization without requiring patients to perform any task. The team faced a familiar obstacle in this kind of data: the signal-to-noise ratio of individual connectivity edges is low, and naive machine learning models tend to latch onto features that do not replicate. To address this, the researchers developed an innovative analytical strategy combining predictive and contrastive machine learning frameworks, constrained to connectivity changes with sufficient reliability. The contrastive component allowed them to statistically disentangle effects that are common to all treated patients from effects specific to drug or to placebo, a decomposition that has rarely been attempted at this scale.

The first major finding was a universal signature. Regardless of whether patients received sertraline, escitalopram or placebo, and regardless of whether their symptoms ultimately improved, nearly all medicated patients showed increased connectivity within a system linking visual cortex, the precuneus and the thalamus. The precuneus, a hub of the brain’s default mode network implicated in self-referential thought and memory, and the thalamus, the brain’s central relay station, are both known to interact with visual processing regions. Earlier work has shown that placebo treatment can alter primary visual cortex activity and connectivity, and this new result places those observations within a much larger and more systematic framework. The consistency of this visual-precuneus-thalamus change across two independent cohorts suggests it reflects a generalizable response to entering pharmacological treatment for depression, not a drug-specific mechanism.

By contrast, the neural correlates of placebo-related symptom improvement were centered elsewhere. The study found that striatal and attention networks mediated the placebo-driven reduction in depressive symptoms. The striatum, a subcortical structure central to reward processing and motivational learning, has repeatedly been implicated in placebo phenomena, including a well-known finding that reward-related ventral striatal activity distinguishes sertraline responders from placebo responders. The involvement of attention networks aligns with the psychological literature on expectation, which holds that placebo effects arise when anticipation and attention reorient the brain’s evaluative machinery. In practical terms, patients whose early connectivity shifts in these networks were pronounced were the ones whose symptoms improved most from expectation and context, independent of any pharmacological action.

The true drug effects were narrower and more selective than many researchers expected. Connectivity changes specific to sertraline and escitalopram converged on the amygdala, the midcingulate cortex, the orbitofrontal cortex and the cerebellum. These are regions with wellestablished roles in emotional regulation and depression: the amygdala generates threat and salience responses, the orbitofrontal cortex evaluates reward and punishment, the midcingulate cortex integrates motivation and control, and the cerebellum, long dismissed as purely motor, is increasingly recognized as a participant in cognitive and affective circuitry, with altered cerebellar-cerebral connectivity reliably distinguishing patients with depression. Notably, these drug-specific changes appeared in only a subset of the patients actually taking antidepressants. Pharmacological treatment, in other words, leaves a detectable early neural fingerprint in some brains but not in others, a neural reflection of the heterogeneity that has always frustrated clinicians.

The most clinically consequential finding emerged from that heterogeneity. When patients on sertraline did not show the drug-specific amygdala, midcingulate, orbitofrontal or cerebellar connectivity changes, their responses could be predicted using a model trained on placebo response signatures. This implies that a substantial share of what looks like antidepressant response in the clinic may actually be placebo response occurring in medicated patients. Given that systematic reviews of antidepressant trials have long documented substantial and growing placebo response rates, this study provides a mechanistic account of why: expectation engages striatal and attentional circuitry, and patients who are predisposed to engage that circuitry will improve whether or not the drug’s molecular mechanisms take hold in their emotional-regulation circuits.

Methodologically, the study’s strength lies in its cross-cohort validation. Patterns identified in one cohort were tested in the other, and the universal visual-precuneus-thalamus changes, the placebo mediators and the drug-specific effects all replicated across the EMBARC and CAN-BIND-1 datasets, which used different scanners, sites and clinical protocols. The analysis pipeline itself was rigorous, employing established preprocessing tools including fMRIPrep, boundary-based registration, ICA-based motion artifact removal and standard strategies to control the spurious correlations introduced by subject head motion. The predictive models were evaluated with cross-validated Pearson correlations between predicted and observed symptom change, and hyperparameters governing sparsity were tuned to avoid overfitting. The code was released publicly through Code Ocean, and the EMBARC data are available through the National Institute of Mental Health Data Archive, inviting independent scrutiny.

The implications for patient care are considerable. Today, deciding whether an antidepressant is working typically requires six to eight weeks of observation, and roughly half of patients discontinue treatment early, often because of side effects before any benefit arrives. If early connectivity changes measured after one week could be incorporated into an interactive treatment optimization framework, clinicians could potentially distinguish, within days, patients whose brains are responding to the drug’s pharmacology from patients whose trajectory depends on placebo-related circuitry, who might instead benefit from psychotherapy, neuromodulation or placebo-enhanced care strategies. This aligns with a broader movement in psychiatry toward biosignature-guided treatment, including prior work from overlapping research groups demonstrating that electroencephalographic signatures and structure-function covariation patterns can predict antidepressant response.

Important caveats remain. The study examined only two serotonergic medications over a brief window, and it remains unknown whether the same signatures generalize to other antidepressant classes, to longer treatment durations, or to adolescent and older populations. Resting-state fMRI measures indirect hemodynamic activity rather than neural firing, and even carefully denoised connectivity estimates carry residual uncertainty at the individual level. The authors also note that drug-specific connectivity changes were present in a subset rather than all medicated responders, so the absence of such changes does not guarantee nonresponse. Still, by systematically partitioning early brain changes into universal, placebo-mediated and drug-specific components, the study delivers what the field has lacked: a mechanistic map of how treatment for depression begins in the brain, weeks before the patient reports feeling better. It transforms the placebo from a statistical nuisance into a defined neural process, and it moves precision psychiatry a tangible step closer to the clinic.

Subject of Research: Early treatment-induced changes in brain functional connectivity in major depressive disorder following antidepressant or placebo administration

Article Title: Early brain functional connectivity changes induced by antidepressants and placebo

Article References: Tong, X., Fonzo, G. A., Carlisle, N. B., Xie, H., Berdichevsky, Y., Keller, C. J., Oathes, D. J., Nemeroff, C. B., Lin, F. V., & Zhang, Y. (2026). Early brain functional connectivity changes induced by antidepressants and placebo. Nature Mental Health. https://doi.org/10.1038/s44220-026-00729-y

Image Credits: AI Generated

DOI: 10.1038/s44220-026-00729-y

Keywords: major depressive disorder, functional connectivity, antidepressants, placebo effect, sertraline, escitalopram, machine learning, resting-state fMRI, EMBARC, CAN-BIND, biomarkers, precision psychiatry

The trial enrolled 262 patients with symptomatic paroxysmal or persistent atrial fibrillation and randomised them in a 2:1 ratio to pulmonary vein isolation or to a sham procedure. The primary endpoint was the between-group difference in the change from baseline to six months in the AFEQT summary score, a well-established questionnaire measuring atrial fibrillation-related quality of life. At six months, the mean AFEQT score improved from 61 to 81 in the ablation group and from 59 to 75 in the sham group. Both improvements were clinically meaningful, but the difference in change between the groups was not statistically significant. Meanwhile, ablation demonstrated clear objective efficacy: atrial fibrillation was detected in 21 percent of ablated patients versus 41 percent of sham patients, an absolute difference of 20 percent, with a favourable safety profile.

What distinguishes PVI-SHAM-AF from earlier research is the rigor of its control condition. The sham intervention was not an inert experience but an active clinical pathway encompassing hospital admission, deep sedation and analgesia, venous access, a stay in the catheterization laboratory, cardioversion for patients who arrived in atrial fibrillation, and structured follow-up. The trial therefore tested catheter ablation against this comprehensive procedural and clinical-care package, not against no treatment at all. The investigators emphasise that the study did not evaluate conventional non-invasive ambulatory care. Within the six-month observation window, however, the superior rhythm control achieved by ablation simply did not translate into a statistically significant improvement in patient-reported quality of life beyond what the active sham pathway delivered.

Compared with the two previous sham-controlled ablation trials, SHAM-PVI and PFA-SHAM, PVI-SHAM-AF was substantially larger, exceeding the combined sample size of both predecessors. Baseline quality of life was impaired to a similar degree as in the major open-label trials CABANA and EARLY-AF. Intriguingly, while the improvement within the ablation arms was broadly consistent across all these studies, the improvement observed in the sham arm of PVI-SHAM-AF was markedly lower than in SHAM-PVI and PFA-SHAM. The editorial authors suggest this discrepancy indicates that patient management in PVI-SHAM-AF may have been more comprehensive than in the other two sham trials, intensifying the care received by the control group and thereby narrowing the apparent treatment effect on symptoms.

The findings echo a broader pattern in cardiovascular medicine. Previous reports have documented clinically relevant placebo effects in randomised trials of invasive procedures across settings as varied as resistant hypertension, percutaneous coronary intervention and vertebroplasty for osteoporotic vertebral fractures. A systematic review and meta-analysis found that only about a third of placebo-controlled trials of cardiovascular interventions demonstrated significant superiority of the invasive procedure over placebo. Against this backdrop, PVI-SHAM-AF is less an outlier than a striking confirmation that subjective, patient-reported endpoints are acutely susceptible to contextual healing, expectation and the attention that accompanies any hospital-based intervention.

Why did both groups improve so substantially? The editorial points to a likely combination of placebo and Hawthorne effects, regression to the mean, and optimised concomitant care driven by protocol-adherent assessments and intensive follow-up. Symptoms in atrial fibrillation, the authors argue, may not be determined solely by arrhythmia physiology but also by symptom perception, atrial or ventricular ectopy, anxiety and depression. Quantifying the individual contributions of these mechanisms is difficult, and PVI-SHAM-AF was not designed to isolate them. Disentangling how much of the improvement stems from the procedure itself, from heightened clinical attention, or from psychological factors remains a critical objective for future research and could reshape how symptom-directed therapies are evaluated.

Do the results mean ablation fails to relieve symptoms, or that it should be withheld from patients with preserved left ventricular function? The editorial authors firmly reject both conclusions. The trial demonstrates that ablation reduces atrial fibrillation recurrence and burden safely, and it does not challenge indications in settings where reducing arrhythmia burden may improve clinically relevant outcomes, such as tachycardia-induced or atrial fibrillation-aggravated cardiomyopathy and heart failure with reduced ejection fraction, populations that were not the focus of this study. The results are confined to patient-reported outcomes rather than morbidity or mortality, reflecting a familiar divergence in cardiology where effects on hard endpoints and on how patients feel can part ways, as seen with beta-blockers, inotropes and implantable cardioverter-defibrillators in heart failure.

Questions about patient selection and follow-up duration also deserve scrutiny. Data on the 937 patients who declined randomisation were insufficient for comparison, and willingness to accept an invasive sham procedure may have selected individuals with particular symptom burdens, expectations or treatment preferences; patients with severe symptoms might have declined to guarantee receiving ablation. Yet the mean baseline AFEQT score of 60 mirrored those in CABANA, EARLY-AF and STOP AF First, beta-blocker use remained high and balanced between groups, and no treatment effect heterogeneity by baseline score was observed. The six-month primary endpoint was chosen deliberately to minimise the impact of repeat procedures, but the authors concede it may have been too short to capture the full treatment effect, and it does not exclude longer-term benefits in rhythm control, medication use, healthcare utilisation or clinical outcomes. Twelve-month follow-up is ongoing.

For clinical practice, the implications centre on shared decision-making. PVI-SHAM-AF shows that the objective rhythm-control benefits of catheter ablation should be clearly distinguished from its effects on patient-reported quality of life at six months, information that is essential for counselling patients whose primary goal is symptom relief. The results also underscore the therapeutic power of comprehensive, attentive clinical care. Crucially, the authors stress that invasive sham procedures are research tools for delineating true treatment effects and must never be translated into clinical practice. Instead, the challenge is to identify which components of the sham-associated care pathway improved symptoms and how those elements can be incorporated into routine, non-invasive care.

Perhaps the trial’s most enduring lesson is methodological. The editorial demonstrates that randomised, blinded evaluation of an established, guideline-recommended invasive procedure is accepted by patients, accepted by many expert physicians, and feasible across multiple centres and countries. Sham-controlled designs are particularly important when evaluating subjective or patient-reported outcomes, because such endpoints are vulnerable to placebo effects, expectations, regression to the mean and the effects of intensified clinical care. Just as placebo controls are indispensable for drug therapies, the incremental effect of an invasive or surgical procedure on how patients feel cannot be reliably determined without an appropriate sham control. PVI-SHAM-AF may thus be remembered less for what it says about ablation alone and more for how it resets the standard of evidence for procedural medicine.

Subject of Research: A sham-controlled randomised trial evaluating catheter ablation for symptom relief and rhythm control in patients with symptomatic atrial fibrillation.

Article Title: PVI-SHAM-AF: what does this trial tell us about catheter ablation and symptoms in patients with atrial fibrillation?

Article References: Laufs, U., Böhm, M., Mahfoud, F., & Wachter, R. (2026). PVI-SHAM-AF: what does this trial tell us about catheter ablation and symptoms in patients with atrial fibrillation?. Clinical Research in Cardiology. https://doi.org/10.1007/s00392-026-03016-z

Image Credits: AI Generated

DOI: 10.1007/s00392-026-03016-z

Keywords: atrial fibrillation, catheter ablation, pulmonary vein isolation, sham-controlled trial, quality of life, AFEQT score, placebo effect, rhythm control, patient-reported outcomes, PVI-SHAM-AF, cardiology, clinical trial