Loading evaluation data…
In antipsychotic trials for acute schizophrenia, dropout reasons are typically classified into inefficacy, adverse events (AEs), and other reasons. AEs can include not only somatic, but also psychiatric AEs, which may reflect inefficacy. However, studies seldom report them separately. We examined detailed reasons for dropouts attributed to AEs and assessed the impact of reclassification. We searched Vivli for randomized controlled trials with individual-participant-data on adults with acute schizophrenia (up to April 9, 2025). When dropout was attributed to an AE, we cross-checked reports and recategorized the corresponding dropouts into somatic, psychiatric, and other AEs using MedDRA(v28.1). We performed random-effects meta-analyses to compare antipsychotics and placebo for i) the study-reported dropouts due to AEs, and for ii) dropouts reclassified as due to somatic AEs. Seventeen trials (6823 participants; 70% male; mean age 40.9 years; mean Positive and Negative Syndrome Score 93.1) were included. Studies reported that 1369 dropouts due to inefficacy, 445 dropouts due to AEs, and 1173 dropouts due to other reasons. We reclassified the study-reported dropouts due to AEs as due to psychiatric (n = 223), somatic (n = 207), and other AEs (n = 15). We did not find evidence of difference between antipsychotics and placebo with respect to study-defined dropouts due to AEs (OR, 0.93 [95% CI; 0.72, 1.12]). However, we found evidence that dropouts reclassified as somatic AEs were more frequent with antipsychotics (OR, 1.54 [95% CI; 1.04, 2.27]). We conclude that combining somatic with psychiatric AEs may mask the true impact of antipsychotics on tolerability. Future trials should report these categories separately.
However, we found evidence that dropouts reclassified as somatic AEs were more frequent with antipsychotics (OR, 1.54 [95% CI; 1.04, 2.27]).
post hoc reclassification without risk-of-bias integration and with a confidence interval close to the null
We did not find evidence of difference between antipsychotics and placebo with respect to study-defined dropouts due to AEs (OR, 0.93 [95% CI; 0.72, 1.12]).
the confidence interval spans one under both conflicting point estimates, supporting absence of evidence rather than equivalence
We conclude that combining somatic with psychiatric AEs may mask the true impact of antipsychotics on tolerability.
post hoc evidence from selected trials supports a classification concern but not a general tolerability conclusion
We reclassified the study-reported dropouts due to AEs as due to psychiatric (n = 223), somatic (n = 207), and other AEs (n = 15).
reported category counts sum to 445, with classification judgments limiting their interpretation
Derived from the full evaluation — not a separate score.
Strengths
Reclassification distinguishes somatic events from psychiatric events within the same conventional adverse-event dropout category, addressing a problem that prior aggregate-data reviews could not directly examine.
↳ Introduction; Results
Methods specifies MedDRA version 28.1, category rules, and second-reviewer finalization of ambiguous mappings. Eligibility screening was independently performed with arbitration when needed.
↳ Experimental procedures
Results places conventional and somatic-only dropout comparisons alongside absolute-risk translations, making the consequence of changing the outcome definition readily interpretable.
↳ Results, Fig. 3
Limitations
No risk-of-bias assessment of the included randomized trials is reported or incorporated into the conclusions. This matters because investigator-assigned dropout reasons underpin the reclassified outcome.
↳ Experimental procedures; Discussion
The Discussion does not adequately qualify its broader tolerability recommendations against the post hoc design, selected IPD sources, and consequential classification judgments.
↳ Experimental procedures; Results; Discussion
The abstract reports conventional dropout OR 0.93 versus 0.90 in Results, and one allocation split totals 6855 despite a stated sample of 6823. These reporting errors do not establish reversal of the comparative findings but impede verification.
↳ Abstract; Results
Participant-level reclassification provides a substantive explanation for why conventional adverse-event discontinuation can obscure a somatic dropout signal. Independent screening and explicit MedDRA rules support the workflow, but the absence of trial-level risk-of-bias assessment is the dominant methodological constraint. The somatic estimate is close to the null boundary, making the Discussion's clear-evidence language stronger than the synthesis warrants. The reporting recommendation has a visible application pathway, while selective IPD coverage and limited discussion of scope restrain broader conclusions.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Sound3.1
Confidence mediumParticipant-level reclassification addresses a previously documented ambiguity in adverse-event discontinuation outcomes and reveals a different somatic-only comparison. The advance is focused rather than a general validation across antipsychotics.
Independent screening, explicit MedDRA assignments, and random-effects synthesis provide a traceable workflow. Missing risk-of-bias assessment binds the score at 2.8, with certainty language, classification choices, and restricted IPD coverage also limiting interpretation.
The conventional-versus-reclassified comparison is compact and supported by absolute-risk translations and a glossary. Conflicting numerical statements and stronger Discussion language reduce precision without obscuring the main argument.
The paper links its question to the Cochrane risperidone review and earlier antipsychotic syntheses. It does not adequately discuss how post hoc analysis, selected IPD availability, and classification judgments constrain its broader recommendations.
Caveats4 of 4 checks
The current investigation is explicitly identified as post hoc within a registered parent project. Conflicts of interest and conditional data availability are disclosed, with no substantiated conduct-related reliability finding.
Conflict of interest: declared · Preregistration: not stated · Data availability: declared
17 references in manuscript 16 of 16 checkable references found in an index 1 are books, websites or datasets — counted, but not index-checkable
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The recommendation concerns detailed reasons the paper says are already collected, reducing the data-collection burden. It remains a reporting recommendation rather than an implementation package or established standard.
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
When considered separately, alcohol, cannabis, amphetamines, and other stimulants showed similar no-effect results, while nicotine use indicated a small effect.
Tentative2026 · Clinical medicine
Individuals with PTE in a civilian population were seizure-free as often as those with other types of epilepsy, but less commonly employed and less likely to hold a university degree.
TentativeAntipsychotic–benzodiazepine combinations might be among the most effective options for rapid tranquilisation in patients with psychomotor agitation but carry a risk of hypotension
Tentative