Relapse in psychosis is unfortunately a common and distressing experience for people with serious mental health conditions, most notably schizophrenia. Researchers commonly rely on relapse in psychosis as an outcome when determining whether an intervention has been effective, with one estimate suggesting over 50,000 studies exploring it, in some way, between 1975 and 2020 (Kiraz and Demir, 2021). Yet, perhaps unexpectedly, there has been no clear or consistent definition of relapse across studies until now (Moncrieff et al., 2020).
Accepting maintenance treatment with antipsychotic medication can be a difficult decision, as many drugs have significant side effects and long-term use may impact physical health. Therefore, clear information about the risk of relapse is crucial, and this depends on having a correct and consistent definition of relapse in research. However, relapse definitions vary widely across trials and over time, complicating comparisons and potentially introducing heterogeneity and bias in systematic reviews and meta-analyses. The ultimate effect is less valid evidence to inform treatment and prevent relapse.
Percentage change in symptom severity has been widely used as a marker but has been shown to be unreliable (Siafis et al., 2024). Previous consensus efforts have relied on small expert groups within single countries, raising questions about generalisability.
A systematic review and Delphi study has been conducted to address these limitations (Howes et al, 2025).
Relapse in schizophrenia is frequently used as an outcome in research, but there has been no consensus on what it means or how to measure it.
Methods
Systematic review
The authors searched PubMed, PsycINFO, and EMBASE from 2012 to 2024. The systematic review was reported according to the (PRISMA) statement, but no protocol was publicly available.
Studies were included if they met the following eligibility criteria:
- Published in English in a peer-reviewed journal between 2012 and 2024
- Reported results of randomised controlled trials on antipsychotic drugs
- Study population with schizophrenia and/or schizoaffective disorder
- Study population aged 18 or above
- Reported “relapse” or “deterioration” as an outcome.
Data from included publications were extracted by two independent researchers. Relapse outcomes were classified as:
- Criteria using validated assessment tools
- Criteria that used a clinician’s judgment (including Clinical Global Impressions Scale (CGI).
Consensus Criteria Development
The Treatment Response and Resistance in Psychosis (TRRIP) network comprises psychiatrists, researchers, industry representatives, as well as other specialists. A working group of volunteers provided feedback.
The Delphi method was applied in the following steps:
- Phase 1: Initial scoping literature review of criteria used in randomised controlled trials, presented to the group
- Phase 2: Based on phase 1, a questionnaire to define relapse criteria was developed
- Phase 3: First anonymous survey of TRRIP members using the questionnaire to identify significant areas of consensus
- Phase 4: The results from phase 3 were presented and discussed at a meeting
- Phase 5: Second anonymous survey was conducted that presented the results of the initial survey to determine whether respondents agreed with the majority opinion and identify remaining areas of disagreement
- Phase 6: Based on phase 5, consensus criteria were developed by the core group
- Phase 7: Patients and carer representatives reviewed the criteria
- Phase 8: All TRRIP members reviewed and endorsed the criteria
- Phase 9: Review of changes made because of peer review.
Delphi methods are a structured technique for building expert consensus on a topic where evidence is uncertain or incomplete.
Results
Meta analysis
The search yielded 1,572 publications, of which 26 met the inclusion criteria. Across these studies, operationalised relapse criteria were commonly applied; however, 85% of trials also permitted relapse to be determined based on clinical judgment. Among studies that defined relapse in terms of symptom change, 68% used relative change, such as a ≥30% increase in symptom severity ratings.
Delphi study
Relapse in clinical practice is typically understood as a return or worsening of schizophrenia symptoms following a period of stable improvement. In this study, relapse was defined using three components:
- An initial symptomatic phase, with symptoms above a critical threshold;
- A stable baseline phase, in which symptoms remain below that threshold; and
- A subsequent worsening, where symptoms again rise above a specified threshold.
The authors proposed two versions of the definition: a minimum criterion and an optimal criterion. These are presented in the table below.
| Category and characteristic | Minimum requirement | Optimum requirement |
| General | ||
| Diagnosis | Meets validated diagnostic criteria (e.g., ICD-11 or DSM-5) for schizophrenia | Diagnosis determined by validated tool e.g., DSM or ICD structured clinical interviews |
| Functioning | Not required | Measurement of function using validated scale (e.g., SOFAS) during all periods |
| Prior illness episode (pre-baseline) | ||
| Ascertainment | Retrospectively determined | Prospectively determined |
| Measurement | Not required | Validated symptom rating scale (e.g., PANSS) |
| Data source | Contemporaneous clinical notes and/or ≥2 of patient, carer, clinician report | Structured clinical interviews designed for the rating scale |
| Severity | Clear history of ≥1: relevant symptom(s) of ≥ moderate severity, admission to psychiatric hospital; daily community care by mental health professionals; or significantly disturbed, risky, or dangerous behavior | ≥1 symptom item on a validated rating scale of at least moderate severity (i.e., ≥4 on a PANSS item) |
| Minimum duration | ≥1 week for each severity indicator (unless hospitalised) | ≥1 week for each severity indicator (unless hospitalised) |
| Baseline period criteria | ||
| Ascertainment | Retrospectively determined; assessed at ≥2 time points | Prospectively determined; assessed at ≥3 time points |
| Measurement | Not required | Validated symptom rating scale (e.g., the PANSS) |
| Data source | Contemporaneous clinical notes and/or ≥2 of patient, carer, clinician report | Structured clinical interviews designed for the rating scale |
| Severity | Symptoms within domain ≤ moderate (e.g., ≤4 on a PANSS item) | Symptoms within domain ≤ moderate (e.g., ≤4 on a PANSS item) |
| Minimum duration | ≥12 weeks for each severity indicator | ≥12 weeks for each severity indicator |
| Relapse criteria | ||
| Ascertainment | Prospectively determined | Prospectively determined |
| Measurement | Overall illness severity using the CGI-S | Validated symptom rating scale (e.g., the PANSS) |
| Data source | Contemporaneous clinical notes and/or ≥2 of patient, carer, clinician report | Structured clinical interviews designed for the rating scale |
| Severity | Documented deterioration in mental state leading to an increase in CGI-S score ≥1, to a level of ≥4 | Absolute symptom increase of ≥1 item by ≥2 points in relevant domain (e.g., P1, P2, P3, G5, G9 for positive domain) to a rating equivalent to ≥ moderate severity (i.e., 4 on the PANSS) according to scale(s) used
For domain-specific relapse, total score not to be employed as a criterion For “general” relapse, a total increase equivalent to 12 points on the PANSS can be used |
| Minimum duration | ≥1 week, unless acute deterioration requires immediate intervention due to “severe” symptoms | ≥1 week, unless acute deterioration requires immediate intervention due to “severe” symptoms |
Conclusions
The authors conclude that they:
…identified considerable variability and conceptual issues with the criteria used in studies of relapse in patients with schizophrenia in the past decade. These issues make comparisons of study results difficult and, in some instances, raise questions about the validity of the definitions and reported results.
Based on the newly developed relapse criteria, they propose that:
If adopted, these recommendations should improve the validity and reporting of relapse studies in schizophrenia and facilitate comparisons between them.
If applied, the recommendations from this paper could improve the validity and reporting of schizophrenia relapse findings.
Strengths and limitations
A major strength of this review and Delphi study is the inclusion of contributors from a wide range of countries. The work is also grounded in a thorough literature review as well as earlier research in which relapse definitions were informed by data linkage. However, the composition of the author group does not reflect meaningful diversity: the vast majority appear to be middle-aged, white men with substantial industry funding. Transparency would have been improved if demographic characteristics of all Delphi panel members had been explicitly reported.
The authors note that a key shortcoming of earlier approaches has been the limited involvement of people with lived experience. In this Delphi process, however, “patients and carers” were asked only to review criteria and seem to have participated as unpaid volunteers, in contrast to the heavily industry-funded stakeholders who shaped the actual decision-making. It is also necessary to acknowledge the dilemma about whether representatives from the pharmaceutical industry should participate in academic research processes, including consensus-building exercises such as Delphi studies. Some argue that industry representatives offer valuable practical insight into treatment development and regulatory landscapes; others caution that their conflicts of interest risk shaping conceptual or methodological decisions in ways misaligned with public or patient interests. This underscores the need for careful governance and transparency when deciding on the composition of participants.
This imbalance raises significant ethical concerns, particularly given that the inclusion of lived-experience voices is central to the legitimacy of the work. Furthermore, the overall lack of diversity regarding gender, lived experience, and representation from the Global South, which substantially limits the broader relevance and value of the study.
A lack of diversity in those included limits the broader relevance and value of the study.
Implications for practice
Despite these significant limitations, the paper has the potential to improve clinical research practice if trial reports adopt the proposed reporting guidelines. Standardised reporting may reduce heterogeneity and restrict the degree of statistical flexibility that can inadvertently facilitate questionable research practices, including p-hacking (manipulating data analysis until a statistically significant result is achieved). If followed, more reliable results may be available for people who are making choices about their long-term treatment.
However, the gap in knowledge the authors describe in the paper concerning lack of inclusion of lived experience seems to be unaddressed or addressed in a tokenistic way. As such, it remains a gap in our common knowledge that should be addressed in a new Delphi study.
The attempt to include lived experience in this paper may be described as tokenistic, which remains a significant barrier to genuinely collaborative mental health research. Lived-experience participation is too often limited to superficial consultation, with little opportunity to influence core conceptual, methodological, or analytical decisions. Such practices can create an impression of inclusivity without altering the underlying power dynamics that shape how relapse is defined and studied. Tokenism not only undermines the quality and relevance of research but can also be ethically problematic: it instrumentalises lived experience while denying it epistemic authority, which is worse than pure exclusion.
Excluding lived experience voices can be compared to asking only men to define concepts related to the female body, or inviting only white participants to determine how best to address challenges faced by people of colour. In each case, those with the most direct and embodied knowledge are sidelined, while others are positioned as authorities on experiences they do not themselves inhabit.
There are two primary reasons why the meaningful inclusion of lived-experience perspectives is essential in research on relapse criteria: ethical responsibility and scientific value (Speyer et al., 2025). Ethically, those who are most affected by relapse and its consequences have a legitimate claim to participate in shaping how it is defined and studied. Excluding their voices reinforces existing power imbalances, risks misrepresenting their experiences, and may result in criteria that inadvertently perpetuate stigma or overlook what matters most to service users.
Scientifically, lived-experience perspectives provide forms of knowledge that enhance the relevance of research. They offer insight into the phenomenology of relapse, contextual influences that traditional measures may obscure, and pragmatic understandings of how changes in symptoms translate into daily life. Incorporating these perspectives strengthens conceptual clarity, improves the ecological validity of operational definitions, and supports the development of more robust and person-centred outcome measures. Together, these ethical and scientific imperatives make lived-experience involvement indispensable rather than optional (Speyer & Ustrup, 2025).
To build on the current study, one could consider a new Delphi process that is led by people with lived experience, with professional stakeholders such as clinicians, researchers and service providers participating as members of the panel rather than shaping the overall direction. A Delphi process guided by users would help shift the balance of influence, ensuring that the concerns, priorities and interpretive frameworks that steer the consensus emerge from those who are most directly affected by relapse and by the research that defines it.
It is also essential that such a study makes a deliberate effort to involve women, people of colour and other groups who have been marginalised in mental health research. Their experiences are shaped by social, cultural and structural conditions that often go unrecognised in standard research methodologies. Including a broad and diverse range of participants would therefore enhance both the equity and the scientific value of the Delphi process, supporting the development of relapse criteria that are more inclusive, more contextually grounded and more representative of the full range of lived experience.
While the new criteria have the potential to improve research their validity is limited by a lack of inclusion in their development
Statement of interests
Helene Speyer – I declare that I have no conflicts of interest related to this work. I have no personal or professional involvement in the study, no financial interests, and no roles or affiliations that could influence its content. AI tools were used to proofread.
Edited by
Simon Bradstreet. No AI tools were used in editing.
Links
Primary paper
Oliver Howes, Bernard Bukala, Eric Chen, Christoph Correll, Alkomiet Hasan, William Honer, John Kane, Stefan Leucht, Spyridon Siafis, Ofer Agid, Dickens Akena, Celso Arango, Lukoye Atwoli, Thomas Barnes, Michael Birnbaum, Istvan Bitter, Alan Breier, Robert Buchanan, Leslie Citrome, David Cotter, Nicolas Crossley, Michael Davidson, Andrea de Bartolomeis, Lynn DeLisi, Sonia Dollfus, Serdar Dursun, Bjørn Ebdrup, Helio Elkis, Robin Emsley, Peter Falkai, Emilio Fernández-Egea, Wolfgang Fleischhacker, Oliver Freudenreich, Ary Gadelha, Wolfgang Gaebel, Ariel Graff-Guerrero, Ad Gridley, Jaime Hallak, Philipp Homan, René Kahn, Stefan Kaiser, Maria Kapi, James Kennedy, Euitae Kim, Bruce Kinon, Jun Kwon, Stephen Lawrie, Jimmy Lee, F Leweke, Tao Li, Jan Libiger, Stephen Marder, Ingrid Melle, Herbert Meltzer, Armida Mucci, Dieter Naber, Shinchiro Nakajima, Jimmi Nielsen, Oisín O’Brien, Akin Ojagbemi, Wolfgang Omlor, Christos Pantelis, Jozef Peuskens, Thomas Raedler, Mao-Sheng Ran, Tiago Marques, Gary Remington, Susan Rossell, Jose Rubio, Gabriele Sachs, James Scott, Tianmei Si, Dan Siskind, Cynthia Siu, Iris Sommer, Takefumi Suzuki, Hiroyoshi Takeuchi, Rajiv Tandon, David Taylor, Solomon Teferra, Neil Thomas, Jari Tiihonen, Hiroyoki Uchida, Alp Ucok, Daniel Umbricht, Ganesan Venkatasubramanian, Elias Wagner, James Walters, Chuanyue Wang, Mark Weiser, Charlie Wright, Xin Yu, Robert McCutcheon. (2025) Relapse in Schizophrenia: A Systematic Review of Criteria for Clinical Studies and International Consensus Guidelines to Improve Them. Am J Psychiatry. 2025 Nov 1;182(11):969-983.
Other references
Kiraz, S., Demir, E. Global Scientific Outputs of Schizophrenia Publications From 1975 to 2020: a Bibliometric Analysis. Psychiatr Q 92, 1725–1744 (2021).
Moncrieff, J., Crellin, N. E., Long, M. A., Cooper, R. E., & Stockmann, T. (2020). Definitions of relapse in trials comparing antipsychotic maintenance with discontinuation or reduction for schizophrenia spectrum disorders: A systematic review. Schizophrenia Research, 225, 47–54.
Siafis, S., Brandt, L., McCutcheon, R. A., Gutwinski, S., Schneider-Thoma, J., Bighelli, I., Kane, J. M., Arango, C… Leucht, S. (2024). Relapse in clinically stable adult patients with schizophrenia or schizoaffective disorder: evidence-based criteria derived by equipercentile linking and diagnostic test accuracy meta-analysis. The Lancet Psychiatry, 11(1), 36–46.
Speyer, H., Roe, D., & Slade, M. (2025). Recovery-oriented psychiatry: oxymoron or catalyst for change? The Lancet Psychiatry. 12(10), 795-802
Speyer, H., & Ustrup, M. (2025). Embracing dissensus in lived experience research: the power of conflicting experiential knowledge. The Lancet Psychiatry, 12(4), 310–316.