The study, conducted by João Gama-Marques and colleagues at Lisbon’s specialised resistant schizophrenia outpatient service and published in the journal Schizophrenia, offers an unusually clear window into a problem that clinicians have long suspected but rarely quantified so directly. When patients who have failed multiple antipsychotic trials arrive at a tertiary referral centre, the assumption is often that they represent a relatively uniform population: people whose schizophrenia has proven stubbornly resistant to first- and second-line medication. The new findings challenge that assumption at its foundation, showing that the very label of suspected treatment resistance may bundle together diagnostically distinct groups of patients whose illnesses arise from fundamentally different mechanisms.

Understanding why this matters requires appreciating what treatment-resistant schizophrenia actually is. Roughly one in three people with schizophrenia does not respond adequately to standard antipsychotic medications, even when these drugs are taken at appropriate doses for sufficient durations. For this population, clozapine stands alone as the only medication with robust evidence of efficacy where other antipsychotics have failed. Yet clozapine is dramatically underused worldwide, constrained by its requirement for regular blood monitoring due to the risk of agranulocytosis, a potentially dangerous depletion of white blood cells, along with a burdensome side-effect profile that includes metabolic changes, sedation, constipation, and seizures. Guidelines from multiple countries therefore reserve clozapine for patients who meet strict criteria for genuine treatment resistance, making accurate diagnosis a gatekeeping step in whether the drug is ever offered.

This is precisely where diagnostic heterogeneity becomes clinically consequential. In the Portuguese cohort, the researchers reviewed electronic clinical records and reclassified every referred patient according to ICD-10 diagnostic codes after specialist reassessment. Patients with confirmed schizophrenia (F20) and schizoaffective disorder (F25) emerged as the groups with the highest clozapine exposure and the greatest overall antipsychotic burden, a pattern consistent with the expectation that clinicians intensify pharmacological treatment when a primary psychotic disorder genuinely resists standard regimens. By contrast, patients classified in the organic psychotic disorder (F06) and unspecified psychosis (F29) categories showed markedly lower pharmacological intensification, suggesting that once reassessment revealed a different diagnostic picture, treating clinicians were more cautious about escalating medication.

The statistical analysis sharpened this picture further. When the investigators examined factors associated with clozapine exposure across the diagnostic groups, an unspecified psychosis diagnosis (F29) was independently associated with lower odds of clozapine exposure compared with schizophrenia (F20). In other words, even after accounting for other variables, the diagnostic label itself predicted whether patients received the drug of choice for genuine treatment resistance. This relationship carries a double significance. On one hand, it may reflect appropriate clinical restraint: clozapine’s evidence base rests on trials conducted in schizophrenia, and extending it to psychoses of uncertain or organic origin carries uncertain benefit and real risk. On the other hand, it raises the possibility that some patients within these heterogeneous groups might be missing out on effective treatment because diagnostic uncertainty, rather than treatment resistance itself, is shaping therapeutic decisions.

Organic psychotic disorders deserve particular scrutiny in this context. The F06 category encompasses psychotic symptoms arising directly from brain injury, neurological disease, or other medical conditions affecting the brain. Psychosis in these settings can mimic schizophrenia closely, featuring hallucinations and delusions that look superficially identical, yet the underlying pathology, and the appropriate treatment target, may be entirely different. Antiepileptic management, treatment of an underlying infection or autoimmune process, or adjustments to a causative medication may resolve symptoms that no antipsychotic could touch. A patient with such a condition who has failed two or more antipsychotic trials is not treatment-resistant in the schizophrenia sense; they have been treated for the wrong illness. The Lisbon findings suggest that this scenario occurs often enough among TRS referrals to reshape how such referrals should be evaluated.

The unspecified psychosis category (F29) presents a related but distinct challenge. It functions as a residual diagnosis, applied when psychotic symptoms are evident but the information available is insufficient to determine whether the illness is schizophrenia, an affective psychosis, an organic condition, or something else. That such a category featured prominently among referrals for suspected treatment resistance tells its own story: patients who have bounced through multiple treatment failures often accumulate fragmented clinical histories, and the diagnostic signal may become progressively noisier as episodes, medications, and institutional transitions pile up. Reassessment by specialists with the time and expertise to reconstruct that history appears to be a powerful tool for restoring diagnostic clarity, and the study’s authors argue that this reassessment step is genuinely important rather than merely administrative.

The study’s most consequential conclusion may be its caution about how clozapine use itself is interpreted. In health-services research, clozapine prescription is frequently used as a proxy marker: if a patient is on clozapine, researchers often infer that treatment resistance was recognised and appropriately treated. The new findings show why this inference is unsafe. Because clozapine exposure tracks the diagnostic category so closely, and because diagnostic categories themselves vary so widely among TRS referrals, clozapine use cannot be read as a stand-alone indicator of treatment-resistant schizophrenia status. Two patients on clozapine may sit in entirely different diagnostic and therapeutic contexts, and a registry that counts them identically will misdescribe the very phenomenon it aims to measure. The authors explicitly recommend that clozapine exposure be interpreted within the diagnostic and treatment context rather than in isolation.

These results also speak to a broader and increasingly urgent conversation about diagnostic precision in psychiatry. Schizophrenia, as defined by current classification systems, is almost certainly a heterogeneous syndrome rather than a single disease, and the boundaries separating it from schizoaffective disorder, bipolar psychosis, organic psychoses, and undefined psychotic states remain contested. Studies of referred populations, by concentrating the most complex and treatment-refractory cases, act as stress tests for these boundaries. When nearly a diagnostic category’s worth of referred patients shift classification under specialist scrutiny, the finding suggests that referral labels travel poorly: what a community clinic records as resistant schizophrenia may be something quite different by the time it reaches a specialised service with resources for thorough reassessment. For epidemiologists, this means that prevalence estimates for treatment-resistant schizophrenia derived from administrative data may be systematically distorted. For trial designers, it means that enrolling patients based on referral status alone risks diluting study samples with diagnostically mixed populations, potentially obscuring treatment effects.

For clinicians, the practical message is a call for humility and thoroughness at the point of referral evaluation. Before concluding that a patient has treatment-resistant schizophrenia, the assessment should verify the diagnosis itself, not merely the adequacy of prior medication trials. Checking adherence, ruling out organic causes, reviewing substance use, and reconstructing the full treatment history are all standard recommendations, but the Lisbon data demonstrate empirically that diagnostic reclassification changes outcomes in a substantial share of real referrals. The work, supported by Portuguese national funding through the Foundation for Science and Technology, ultimately reframes suspected treatment resistance not as a single clinical problem but as a gateway question that demands diagnostic precision first. Only when the underlying condition is correctly identified can decisions about clozapine, the most powerful and most demanding tool in the antipsychotic arsenal, be made on solid ground, ensuring that the patients who genuinely need this last-line medication receive it, and that those whose psychoses stem from other causes are redirected toward treatments that might actually help them.

Subject of Research: Diagnostic reclassification and clozapine use among patients referred for suspected treatment-resistant schizophrenia

Article Title: Diagnostic heterogeneity and clozapine use among patients referred for suspected treatment-resistant schizophrenia

Article References: Gama-Marques, J., Schumacher, M. M., & Henriques-Calado, J. (2026). Diagnostic heterogeneity and clozapine use among patients referred for suspected treatment-resistant schizophrenia. Schizophrenia. https://doi.org/10.1038/s41537-026-00812-4

Image Credits: AI Generated

DOI: 10.1038/s41537-026-00812-4

Keywords: treatment-resistant schizophrenia, clozapine, diagnostic heterogeneity, ICD-10, organic psychotic disorders, unspecified psychosis, schizoaffective disorder, antipsychotic treatment, psychiatric diagnosis, schizophrenia research, referral clinics, psychosis

The protein at the center of the study, Golga3, is better known to cell biologists as a structural component of the Golgi apparatus, the membrane-bound organelle that packages and ships proteins inside cells. But Golga3 has a second, more surprising job. It helps regulate the stability of an enzyme called serine racemase, which converts the amino acid L-serine into its mirror-image form, D-serine. D-serine is no molecular footnote: it binds to the glycine site of the N-methyl-D-aspartate receptor, a glutamate-sensitive receptor that is essential for learning, memory, and synaptic plasticity. By inhibiting the ubiquitin-tagged, proteasome-driven degradation of serine racemase, Golga3 effectively keeps D-serine production running. Less Golga3, the reasoning goes, could mean less serine racemase, and ultimately less D-serine available to support NMDA receptor function.

That chain of logic matters because the NMDA hypofunction hypothesis has become one of the leading neurobiological accounts of schizophrenia. Decades of evidence, from the psychotomimetic effects of ketamine and phencyclidine, which block the NMDA receptor, to postmortem and genetic studies, suggest that weakened signaling through this receptor contributes to the positive symptoms of psychosis, the negative symptoms such as social withdrawal and blunted affect, and the cognitive difficulties that so often define the long-term course of the illness. D-serine itself has been tested as an add-on treatment in clinical trials, with mixed but sometimes encouraging results. If Golga3 sits upstream of D-serine availability, its reduction could represent a measurable molecular signature of the same pathway.

To test that idea, a team led by Eda Uzun Uysal and Bahri İnce of the Bakirkoy Prof. Dr. Mazhar Osman Training and Research Hospital for Psychiatry, Neurology and Neurosurgery in Istanbul, together with Ömer Alper Uysal of Eyüpsultan State Hospital, recruited three carefully matched groups. The first comprised 24 patients with schizophrenia who responded to standard antipsychotic monotherapy, a category researchers call non-treatment-resistant schizophrenia. The second comprised 24 patients whose illness had proved resistant to conventional treatment and who were being managed with clozapine, the gold-standard medication for treatment-resistant schizophrenia. The third group consisted of 26 healthy controls. Critically, the groups were matched for age, sex, education, and smoking status, factors that can otherwise confound comparisons of blood-based biomarkers.

The measurement protocol was deliberately standardized. Venous blood samples were drawn in the morning after an overnight fast, reducing the chance that circadian rhythms or recent meals would skew protein levels. Plasma concentrations of Golga3 were then quantified using an enzyme-linked immunosorbent assay, or ELISA, a technique that uses antibodies to capture and quantify a specific protein in a biological sample. Alongside the blood work, the clinicians assessed symptom severity using the Positive and Negative Syndrome Scale, the standard instrument for rating psychotic symptoms, and the Global Assessment Scale, which captures overall functional impairment. This dual approach allowed the researchers to ask not only whether Golga3 differed between groups, but whether its levels tracked with the clinical picture.

The answer to the first question was a clear yes. Plasma Golga3 levels varied significantly across the three groups, with an overall statistical significance of p less than 0.001. Post-hoc comparisons showed that both patient groups differed from the healthy controls: patients on standard antipsychotic monotherapy had significantly lower Golga3 than controls, with a p value of 0.001, and so did the clozapine-treated, treatment-resistant patients, with a p value of 0.002. Strikingly, however, the two patient groups were statistically indistinguishable from each other, with a p value of 0.722. Whatever is driving the reduction in Golga3, it appears to be shared across the illness rather than tied specifically to treatment resistance or to clozapine exposure.

The answer to the second question was more sobering. The researchers found no significant correlations between plasma Golga3 levels and any of the clinical variables they measured, including symptom scores on the PANSS and functional ratings on the GAS. In other words, patients with lower Golga3 were not reliably more symptomatic or more impaired than patients whose levels were closer to the control range. That absence of a dose-response relationship is a common feature of early biomarker studies and tempers any temptation to treat Golga3 as a diagnostic or prognostic test. It suggests instead that reduced Golga3 may mark a pathway-level vulnerability that is present in schizophrenia broadly, rather than a dial that tracks day-to-day symptom fluctuation.

The authors themselves are careful about scope. They describe the findings as preliminary and emphasize that the study was designed to detect whether changes in the Golga3-serine racemase pathway are present in schizophrenia at all, not to establish the pathway’s functional consequences. Because the study measured Golga3 but not D-serine or serine racemase directly, the intermediate steps of the hypothesized cascade remain unverified in this cohort. It is possible that reduced plasma Golga3 reflects altered Golgi function, altered protein degradation, or even a peripheral readout of a central nervous system process, but the study cannot yet distinguish among these interpretations. The authors call explicitly for future work examining D-serine levels and related metabolic markers to clarify what the reduction means biologically.

Several limitations frame how the results should be read. The sample sizes, while adequate for detecting group differences of the magnitude observed, are modest, and case-control designs of this kind capture a snapshot rather than a trajectory. All patients were taking antipsychotic medication, so the possibility that treatment influences Golga3 levels cannot be excluded, although the similarity between patients on monotherapy and those on clozapine offers some reassurance. Plasma is also a peripheral compartment; whether blood levels mirror what happens in the brain, where serine racemase and D-serine do their synaptic work, is an open question that will require imaging, postmortem, or cerebrospinal fluid studies. Replication in independent, ideally medication-naive or first-episode cohorts would strengthen confidence that the finding is intrinsic to the illness rather than an artifact of sampling or treatment.

Even with those caveats, the study is notable for pointing a spotlight at the Golgi apparatus, an organelle rarely invoked in mainstream schizophrenia research, and for connecting it to the glutamatergic machinery that many researchers consider central to the disorder. Biomarkers that can be measured in a fasting blood draw are attractive targets for psychiatry, a field still lacking objective laboratory tests for its major diagnoses. If follow-up studies confirm that Golga3, serine racemase, and D-serine move together in patients, and if the reduction proves specific to schizophrenia rather than psychosis in general, the pathway could eventually inform both patient stratification and drug development aimed at enhancing NMDA receptor function. For now, the message is one of cautious excitement: a single protein, measured in blood, has opened a new window onto the biology of schizophrenia, and the view through it is only beginning to come into focus.

Subject of Research: Reduced plasma Golga3 and the Golga3-serine racemase-D-serine pathway in schizophrenia

Article Title: Reduced plasma Golga3 levels in patients with schizophrenia: a case-control study

Article References: Uzun Uysal, E., Uysal, Ö. A., & İnce, B. (2026). Reduced plasma Golga3 levels in patients with schizophrenia: a case-control study. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08685-8

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08685-8

Keywords: schizophrenia, Golga3, D-serine, serine racemase, NMDA hypofunction, glutamate, plasma biomarkers, psychosis, treatment-resistant schizophrenia, clozapine, psychopharmacology, case-control study