How heterogeneous are diagnoses among patients referred for suspected treatment-resistant schizophrenia?

Among 392 suspected treatment-resistant schizophrenia referrals, 61.5% were reassessed outside schizophrenia, reshaping clozapine interpretation.

Direct answer

In a Portuguese specialist clinic, most patients referred for suspected treatment-resistant schizophrenia did not end up with a schizophrenia diagnosis after reassessment. Of 392 referrals, 61.5% were classified as organic psychotic disorder (F06) or unspecified psychosis (F29), and clozapine exposure tracked diagnosis rather than referral status.

1sources cited

This article was generated with WisPaper-powered search and paper analysis.

Referral for suspected TRS is not a diagnosis

The paper asks a narrow question with a practical edge: when clinicians refer someone for suspected treatment-resistant schizophrenia (TRS), what diagnoses do those patients actually carry after specialist reassessment? TRS here means insufficient response to at least two adequate antipsychotic trials, excluding clozapine, the definition the paper cites as most widely used. The answer is that the referred group is diagnostically mixed. Among 392 patients referred to a specialized outpatient clinic in Lisboa, Portugal, the largest group was organic psychotic disorder (ICD-10 F06) at 135 patients (34.4%), followed by unspecified psychosis (F29) at 106 (27.0%), schizophrenia (F20) at 76 (19.4%), and schizoaffective disorder (F25) at 75 (19.1%).

That distribution is the study's central fact. F06 and F29 together made up 61.5% of the sample, so a referral for suspected TRS was not equivalent to an established diagnosis of schizophrenia. The authors read this as evidence that referral pathways can mix persistent psychotic symptoms with secondary causes, unresolved diagnostic status, comorbidity, or incompletely characterized prior treatment response.

The classification method matters for interpreting every later comparison. Final diagnoses came from specialist longitudinal reassessment of electronic clinical records, supported by a proposed work-up of blood and urine tests, electroencephalography, brain computed tomography, and neuropsychological assessment. No standardized research diagnostic interview was used, so these are real-world clinical diagnoses rather than research-grade ones.

F29 mostly marks an unfinished diagnostic work-up

The F29 label is easy to misread as a mild or vague category. In this cohort it carried a specific operational meaning: F29 was assigned when the diagnostic work-up was not completed, most commonly because patients did not attend the proposed examinations. In those cases an organic contribution could be neither confirmed nor confidently excluded.

F06 worked the other way. It was assigned when the work-up was completed and documented clinically relevant abnormalities supporting an organic contribution to the psychosis. F20 and F25 were assigned when the work-up was completed without such abnormalities and the longitudinal presentation supported a primary psychotic disorder.

This asymmetry shapes the paper's interpretation. The authors describe F29 as reflecting unresolved diagnostic status rather than a uniformly characterized clinical group, and they note that routine lumbar puncture was unavailable, limiting systematic assessment of encephalitic causes. Any comparison involving F29 therefore compares a diagnostically uncertain group with groups whose work-up reached a conclusion.

Clozapine exposure tracks diagnosis, not referral

Across the whole sample, 140 patients (35.7%) were receiving clozapine and 252 (64.3%) were not; 212 (54.1%) were on antipsychotic monotherapy and 180 (45.9%) on combination therapy. Those totals hide a sharp split by diagnosis. Treatment distribution differed significantly across the four diagnostic categories (chi-square with 9 degrees of freedom = 45.24, p < 0.001).

Patients with F06 or F29 were overrepresented in the no-clozapine monotherapy group, while F20 and F25 patients were more often in the clozapine combination group, based on standardized Pearson residuals. Concretely, 72 of 135 F06 patients (53.3%) and 62 of 106 F29 patients (58.5%) were on no-clozapine monotherapy, whereas 26 of 76 F20 patients (34.2%) and 27 of 75 F25 patients (36.0%) were on clozapine combination therapy.

Antipsychotic burden followed the same pattern. Chlorpromazine equivalent doses differed across groups (p = 0.013) with medians of 500 mg/day for F06, 600 for F20, 600 for F25, and 502 for F29. Defined daily doses also differed (p = 0.006), with medians of 2.00 for F06, 2.00 for F20, 2.33 for F25, and 1.76 for F29. The authors describe higher exposure in F20 and F25 and lower treatment intensity in F06 and F29.

Only F29 keeps lower clozapine odds after adjustment

The descriptive split could reflect differences in age, sex, homelessness, or substance use. The multivariable model tested that. After adjustment for demographic, clinical, and substance-use variables, F29 remained independently associated with lower odds of clozapine exposure relative to F20, with an adjusted odds ratio of 0.39 (95% CI 0.20–0.75, p = 0.005). No other diagnostic category showed a statistically significant independent association.

The paper is careful about what this means. The authors write that lower clozapine odds in F29 may reflect differences in diagnostic formulation or clinical decision-making, while barriers to clozapine initiation may also have contributed in some cases. The retrospective design does not let them separate these explanations.

A separate comparison by clozapine exposure adds context rather than causation. Clozapine-treated patients had more antipsychotics (median 2 vs 1, p < 0.001), higher chlorpromazine equivalent doses (median 644 vs 504 mg/day, p < 0.001), more psychiatric admissions (p = 0.002), and longer cumulative hospitalization (p < 0.001). The authors present these as markers of a more intensive treatment trajectory, not as proof that clozapine exposure reflects greater underlying severity or formally established TRS.

Service contact differs while admission history does not

Not every clinical feature separated the diagnostic groups. Sex distribution did not differ significantly (p = 0.239), and neither did alcohol, cannabis, cocaine, heroin, or any drug use (all p > 0.05). Personality disorder, intellectual disability, and catatonia syndrome also showed no significant differences.

Two features did differ. Age varied across groups (p = 0.001), with F06 patients older than the others (mean 53.6 years versus roughly 48 in F20, F25, and F29). Homelessness was more frequent in F06 (57.0%) and F29 (55.7%) than in F20 (38.2%) and F25 (28.0%) (p < 0.001).

Service use split in a specific way. Consultations in the specialist clinic differed across groups (p = 0.002), with F20 highest and F29 lowest, and follow-up time in the clinic differed strongly (p < 0.001), again longest for F20 and F25 and markedly shorter for F29. By contrast, time since first contact with the hospital, number of admissions, and duration of hospitalizations did not differ significantly. The authors conclude that differences concentrated in follow-up within the TRS clinic rather than in earlier hospital history.

Clozapine exposure is not a TRS marker here

The most transferable caution in the paper concerns measurement. The retrospective records did not allow formal TRS status to be reconstructed for every patient, particularly because adherence and the adequacy of previous antipsychotic trials could not be fully characterized. Clozapine exposure is therefore interpreted as part of the observed treatment trajectory, not as evidence that TRS criteria were necessarily met.

The same logic applies to combination therapy. It was more prominent among F20 and F25 patients, but the authors frame it as a marker of therapeutic complexity rather than evidence of biological treatment resistance.

This matters for anyone using treatment patterns as a proxy. If cohorts defined by suspected TRS contain different diagnostic groups and different levels of treatment exposure, biological or clinical associations may partly reflect that heterogeneity rather than resistance itself. The authors make this point when discussing future biomarker work, arguing that such studies should combine biological measures with standardized assessment of diagnosis, treatment adequacy, adherence, and treatment response.

A single clinic limits how far the numbers travel

The study is a retrospective observational analysis of electronic records from one specialized outpatient clinic, covering the period up to the end of 2022. It included all 392 patients referred during that period, with no upper age limit and no patients under 18 because the service does not treat pediatric populations. The authors acknowledge that a single-service retrospective design may limit generalizability and depends on the completeness and accuracy of recorded information.

The diagnostic method sets another boundary. Reassessment relied on specialist longitudinal clinical review plus completion of the proposed work-up, not standardized research interviews, and not all patients completed the protocol. The paper states plainly that it does not estimate the prevalence of formally established TRS within the cohort and does not determine the mechanisms behind individual treatment non-response.

These limits qualify the percentages rather than erase them. The 61.5% figure describes the diagnostic composition of referrals to this clinic under this reassessment process; it is not a population prevalence estimate for TRS.

Reassess the diagnosis before escalating pharmacotherapy

The paper's practical priority follows from the comparison between F29 and F20. The strongest adjusted signal is not that clozapine use was low in general, but that F29 patients had roughly 0.39 times the odds of clozapine exposure compared with F20 patients after adjustment, while F06 and F25 differences did not survive as independent associations. Because F29 largely represented incomplete work-ups, that gap sits at the intersection of diagnostic uncertainty and treatment decisions.

The authors therefore argue that referral for suspected TRS should trigger two parallel tasks: assessing whether treatment resistance is adequately established, and reassessing the underlying diagnosis. That means reviewing longitudinal course, comorbidities, prior treatment adequacy, and potential secondary causes of psychosis, especially when investigations are incomplete.

Clozapine remains central for established TRS in the paper's framing, but its presence or absence should be read within the full diagnostic and treatment context. The decision-relevant split is diagnostic: a referral labeled TRS is more informative about what still needs to be checked than about which treatment the patient has already received.

Sources used in this answer

1

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

Among 392 patients referred for suspected treatment-resistant schizophrenia, 61.5% were reassessed as organic psychotic disorder (F06) or unspecified psychosis (F29); F29 remained independently associated with lower clozapine exposure odds than F20 (adjusted OR 0.39, 95% CI 0.20–0.75).