Sample characteristics

A total of 7978 records were identified through the literature search. After screening, 1242 full-text articles were assessed for eligibility. After excluding those not meeting inclusion criteria, 83 studies were included, reporting on 83 independent EIS cohorts (see Fig. 1 for PRISMA flowchart). The key characteristics of the included studies are outlined in Supplementary Table 4, and Supplementary Table 5 provides details on the primary reasons for article exclusion.

Fig. 1

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flowchart. ^{Reference Haddaway, Page, Pritchard and McGuinness67}

The overall database comprised 30 946 individuals with a FEP. The mean age of the participants at baseline was 24.95 years (s.d. = 3.89; range 18.05–38.3 years) and 34.78% were female. Most studies were conducted in Europe (53.01%), followed by North America (27.71%) and Oceania (7.23%).

Although most included diagnoses were reported at baseline (n = 78, 93.98%), there was considerable variation in follow-up duration of the included cohorts, with mean follow-up of 41.17 months (s.d. = 53.98 months; range 1–252 months; median = 18 months).

There were no significant differences in sociodemographic characteristics between EIS that included affective psychotic disorders in their initial admission criteria (n = 71) and those that did not (n = 12). The percentage of females was similar in both groups (no: 36.06%, yes: 33.11%), as was mean age (no: 24.09 years, s.d. = 4.05; yes: 25.50 years, s.d. = 2.82). Likewise, GAF scores showed no significant differences (no: 42.82, s.d. = 9.83; yes: 38.73, s.d. = 7.35). Among the 12 EIS that explicitly excluded affective psychotic disorders in their admission criteria, none contributed cases of bipolar disorder or MDD-PF to the pooled analyses. However, these cohorts did include individuals diagnosed with schizoaffective disorder at first presentation, consistent with its a priori classification as a non-affective psychotic disorder in this review. Detailed cohort-level diagnostic contributions are reported in Supplementary Table 4.

Proportion of affective psychotic disorder

Most diagnoses in included studies were reported at baseline (93.98%). The estimated pooled proportion of affective psychotic disorder, including bipolar disorder and MDD-PF, was 18% (k = 70; 95% CI 15.44–20.60%; I ² = 95.6%; median 19.5%; interquartile range (IQR) 11.2–25.3%), with a wide 95% prediction interval (3.6–39.4%), reflecting substantial between-service heterogeneity. Individual study estimates varied considerably, ranging from 1 to 57%, reflecting high heterogeneity (Fig. 2; Supplementary Fig. 1). For each pooled estimate, the between-study variance (τ ²) and study-level weights are reported in the Supplementary Material, alongside forest and funnel plots (Supplementary Figs 1–12), which illustrate that heterogeneity arises from a wide dispersion of cohort-level proportions rather than from a small number of highly influential studies. Leave-one-out analyses showed that the pooled proportion was highly stable, with exclusion of individual cohorts resulting in only minimal changes in the pooled estimate (range 17.4–18.3%), indicating that no single study disproportionately influenced the overall result.

Fig. 2

Proportion of psychotic disorder. Affective psychotic disorders include bipolar disorder with psychotic symptoms and major depressive disorder with psychotic features. Non-affective psychotic disorders include schizophrenia, schizophreniform disorder, schizoaffective disorder, delusional disorder, brief psychotic disorder and other primary psychotic disorders. Pooled proportions are category-specific and derived from partially overlapping sets of studies; they are not mutually exclusive and should not be summed or interpreted as components of a single total.

A sensitivity analysis restricted to the 71 studies that did not explicitly exclude affective psychotic disorders yielded nearly identical estimates (e.g. bipolar disorder with psychotic features: 12.4%; MDD-PF: 7.7%), indicating that the inclusion of the remaining 12 studies did not materially affect the pooled proportions.

The proportion of bipolar disorder type 1 within FEP populations was 12.00% (k = 44; 95% CI 9.39–14.80%; I ² = 93.8%; median 12.2%; IQR 7.5–17.7%), with a wide 95% prediction interval (1.6–28.9%) and substantial between-study variability (Fig. 2; Supplementary Fig. 2). Only two studies estimated proportion of bipolar disorder type 2, at 14.58% (k = 2; 95% CI 11.97–17.39%) (Fig. 2; Supplementary Fig. 3). The proportion of MDD-PF, based on 28 studies, was estimated at 7.19% (95% CI 5.39–9.21%), with considerable between-study variability (I ² = 91.1%) (Fig. 2; Supplementary Fig. 4).

Proportion of non-affective psychotic disorder

The proportion of individuals with schizoaffective disorder was estimated at 7.40% (k = 49; 95% CI 5.78–9.19%; I ² = 92.5%) (Fig. 2; Supplementary Fig. 5). A meta-analysis of 79 studies estimated the proportion of schizophrenia at 45.48% (95% CI 40.33–50.68%; I ² = 98.7%) (Fig. 2; Supplementary Fig. 6). The proportion of individuals with schizophreniform disorder was 14.28% (k = 34; 95% CI 10.12–19.00%; I ² = 96.8%), with substantial between-study variability (Fig. 2; Supplementary Fig. 7). The proportion of delusional disorder was estimated at 4.22% (k = 27; 95% CI 2.79–5.90%; I ² = 92.2%), with high between-study heterogeneity (Fig. 2; Supplementary Fig. 8).

Psychosis not otherwise specified had an estimated proportion of 17.86% (k = 45; 95% CI 13.49–22.69%; I ² = 98.3%) (Fig. 2; Supplementary Fig. 9). The proportion of substance-induced psychosis was 5.42% (k = 25; 95% CI 3.85–7.23%; I ² = 93.4%) (Fig. 2; Supplementary Fig. 10).

Finally, the proportion of other psychotic disorders, defined in the studies as brief psychotic disorders or other unspecified psychotic conditions, was estimated at 11.43% (k = 49; 95% CI 8.35–14.83%; I ² = 98.2%) (Fig. 2; Supplementary Fig. 11).

For further details on the proportion estimates, see Supplementary Results 1.

Subgroup analyses and meta-regression

Regarding sociodemographic variables, a higher percentage of females was associated with a slight but significant decrease in the proportion of affective psychotic disorder (β = −0.0050; p = 0.0041). However, age, percentage of White participants and year of publication were not significant moderators (Supplementary Table 6).

In the subgroup analysis by continent, substantial heterogeneity was observed (Q = 1575.2; d.f. = 69; p < 0.0001; I ² > 89% across subgroups). Under the random-effects model, the pooled prevalence of affective psychosis was 16.8% (95% CI 14.7–19.0%) in Europe (k = 38), 6.9% (95% CI 3.1–11.9%) in Asia (k = 7), 19.4% (95% CI 12.0–28.1%) in Oceania (k = 9) and 26.2% (95% CI 18.3–34.9%) in North America (k = 14). Only one study each was available for South America (29.4%; 95% CI 15.1–46.0%) and for multi-continental samples (27.7%; 95% CI 25.1–30.4%). The test for subgroup differences was significant (Q = 64.1; d.f. = 5; p < 0.0001), indicating that prevalence estimates varied significantly across continents (Supplementary Table 6). In the meta-regression analysis with continent as a moderator, Europe was used as the reference group. Compared with Europe, studies conducted in Asia showed a significantly lower prevalence of affective psychosis (β = –0.16; p = 0.002), whereas studies from North America showed a significantly higher prevalence (β = 0.11; p = 0.005). No significant differences were observed for Oceania or South America, although the latter was represented by only one study.

In the subgroup analysis by NIL, pooled prevalence of affective psychotic disorder was highest in the low NIL group (k = 1; 21.9%; 95% CI 16.1–28.3%) and in the low–medium NIL group (k = 1; 19.3%; 95% CI 14.4–24.7%). The high NIL group (k = 59) showed a pooled prevalence of 18.9% (95% CI 16.3–21.8%), whereas the lowest prevalence was observed in the high–medium NIL group (k = 7; 9.7%; 95% CI 4.0–17.5%).

When stratified by urbanicity, pooled proportion of affective psychotic disorders was higher in urban settings (k = 49; 19.9%; 95% CI 16.7–23.2%) compared with rural settings (k = 21; 13.9%; 95% CI 10.5–17.7%). The difference between subgroups was statistically significant (p = 0.0268).

Among clinical characteristics, neither positive, negative, general nor total symptom scores were significantly associated with proportion of affective psychotic disorders. Similarly, GAF scores and NOS scores did not influence proportion estimates (Supplementary Table 6). For complete meta-regression estimates please see the Supplementary Material.

Heterogeneity and quality of included studies

Heterogeneity varied across analyses (I ² ranged from 0 to 98.3%) but was very high (>90%) for almost all pooled estimates. The scale ranged from 5 to 9, with a mean of 7.37 (s.d. = 1.19) (Supplementary Table 7). Visual inspection of funnel plots and Egger’s regression did not suggest marked small-study effects (Supplementary Fig. 12). However, given the very high between-study heterogeneity observed across analyses, these assessments should be interpreted cautiously and considered exploratory rather than definitive indicators of absence of bias.

Summary of findings

In this systematic review and meta-analysis, we examined the diagnostic composition of FEP cases treated within EIS, focusing on the proportion of affective and non-affective psychotic disorders across services and contexts. To the best of our knowledge, this is the first systematic review and meta-analysis to examine the relative proportion of affective and non-affective psychotic disorders in treatment-seeking FEP samples, i.e. people attending EIS. We found a pooled proportion of 18% for affective psychotic disorder across included studies, with substantial variability between studies. Given the substantial between-service heterogeneity, the pooled estimate should be interpreted as an average across heterogeneous EIS rather than as a single stable parameter; accordingly, the wide prediction interval indicates that the proportion of affective psychoses may vary markedly across services depending on local eligibility criteria and diagnostic practices. Bipolar disorder was the most prevalent form of affective psychotic disorder at 12%, followed by MDD-PF at 7.19%. Schizophrenia accounted for nearly 45.48% of cases, plus an additional 14.28% classified as schizophreniform disorder.

EIS have been developed and implemented within heterogeneous service models, with some programmes – such as those reported in the USA ^{Reference Mueser, Penn, Addington, Brunette, Gingerich and Glynn68} and Spain ^{Reference Mayoral-van Son, Juncal-Ruiz, Ortiz-García de la Foz, Vázquez-Bourgon, Setién-Suero and Tordesillas-Gutiérrez69} – explicitly restricting inclusion to non-affective psychoses, whereas others, particularly within the UK National Health Service framework, ^{Reference NHS70} adopt a fully transdiagnostic approach that includes both affective and non-affective psychotic disorders. Consistent with this heterogeneity, our findings show substantial between-study variability in the proportion of affective psychotic disorders across EIS cohorts, while nonetheless indicating that affective psychoses constitute a meaningful minority of presentations within services that accept them.

Female gender was associated with a slight but significant decrease in the proportion of affective psychotic disorder. This is somewhat counterintuitive, given that MDD is approximately twice as prevalent in females ^{Reference Kuehner71} and bipolar disorder has an equal prevalence across gender. ^{Reference Merikangas, Akiskal, Angst, Greenberg, Hirschfeld and Petukhova72} However, these general prevalence patterns may not directly translate to the subset of affective disorders with psychotic features. A possible explanation is differential patterns of help-seeking and access to services, where females with affective psychotic symptoms may be more likely to receive care in primary or general psychiatric settings rather than within specialised EIS. A qualitative study of pathways to care in early psychosis found that women’s help-seeking was often dismissed or downplayed by both family members and clinicians, partly because of stereotypes portraying them as ‘overdramatic’ or ‘too functional’ to warrant specialised care. ^{Reference Ferrari, Flora, Anderson, Haughton, Tuck and Archie73} These gendered dynamics may contribute to their underrepresentation in early intervention programmes.

Other factors such as age, percentage of White participants and year of publication did not significantly influence proportion estimates. However, the proportion of affective psychotic disorders in FEP varied significantly geographically. Compared with Europe, the proportion of affective psychosis was significantly lower in Asia, whereas in North America, proportion was higher. These variations may reflect regional differences in diagnostic practice, cultural perceptions of affective symptoms (and therefore access to care) or structural factors influencing access to EIS. ^{Reference Richardson, Hameed, Perez, Jones and Kirkbride74–Reference Omer, Finnegan, Pringle, Kinsella, Fearon and Russell76} Differences in study settings, including sample population, geographic region and healthcare access, may explain variations in findings. ^{Reference Jongsma, Turner, Kirkbride and Jones77–Reference Lasalvia, Tosato, Brambilla, Bertani, Bonetto and Cristofalo79}

In our analysis, affective psychotic disorders appear less common in rural areas than urban settings, consistent with prior research. ^{Reference Szöke, Charpeaud, Galliot, Vilain, Richard and Leboyer80} Urban environments often present higher levels of social fragmentation, socioeconomic deprivation and environmental stressors, all of which have been associated with increased risk of psychotic disorders. ^{Reference Fett, Lemmers-Jansen and Krabbendam81} Although affective psychotic disorders appeared less common in rural areas in our sample, it is important to note most existing evidence on urbanicity and psychosis risk relates to non-affective psychoses. ^{Reference Fett, Lemmers-Jansen and Krabbendam81} Although the link between urbanicity and schizophrenia is well established, evidence for an association with affective psychotic disorders is far more limited and inconsistent. ^{Reference Fett, Lemmers-Jansen and Krabbendam81,Reference Schofield, Thisted Horsdal, Das-Munshi, Thygesen, Pedersen and Morgan82} Individuals in urban areas had more frequent contact with medical services (77%) compared with those in rural areas (52%), ^{Reference Dahlberg, Forsell, Damstrom-Thakker and Runeson83} despite similar rates of anxiety and depression symptoms. These findings suggest availability and accessibility of mental health services in urban settings may contribute to the higher proportion of affective psychotic disorders observed in our review.

On the other hand, the proportion of affective psychotic disorder was significantly lower in those with high-medium NIL compared with those in the high NIL group. In contrast, only single studies contributed data for the low and low–medium NIL groups, limiting the generalisability of findings in these categories. Differences in healthcare infrastructure and access across NIL levels may substantially influence the detection and reporting of affective psychotic disorders. This occurs despite the well-established association between low NIL and higher prevalence of affective psychotic disorders. ^{Reference Hakulinen, Komulainen, Suokas, Pirkola, Pulkki-Råback and Lumme84}

In contrast, clinical characteristics such as psychotic symptom severity and global functioning were not significantly associated with the proportion of affective psychotic disorder in our sample. This differs from previous studies showing that patients later diagnosed with bipolar disorder tend to have better baseline functioning and fewer negative symptoms. ^{Reference Salagre, Grande, Vieta, Mezquida, Cuesta and Moreno85} A likely explanation is that inclusion in EIS requires marked, pervasive psychotic symptoms and impairment, potentially excluding milder affective cases with better prognosis. Notably, such differences have been observed outside of EIS settings, following longer-term diagnostic follow-up. ^{Reference Peralta, Janda, de Jalon, Moreno-Izco, Sánchez-Torres and Cuesta86} These results indicate that severity of psychotic symptoms, as measured by clinical scales, may not be the primary determinant in distinguishing affective psychotic disorder from other psychotic disorders in FEP populations at baseline.

Importantly, the moderators examined in this review – such as gender, age, geographical region, ethnicity, NIL and urbanicity – should be interpreted as indexing contextual and study- or service-level characteristics, rather than individual-level clinical or biological risk factors.

Strengths and limitations

This study provides the most comprehensive synthesis to date on the proportion of affective psychotic disorders in FEP cohorts. Its clinical relevance is strengthened by the focus on individuals treated within EIS, which provides more standardised and comparable data than studies conducted in heterogeneous, non-EIS settings. This enhances the clinical meaningfulness of the results, as they reflect real-world service use and diagnostic practices.

Limitations include lack of data from some geographical regions, such as South America and Africa. We acknowledge that diagnoses in FEP cohorts may change over time, although meta-analysis suggests reasonable diagnostic stability for schizophrenia (90%) and affective psychotic disorders (84%), ^{Reference Fusar-Poli, Cappucciati, Rutigliano, Heslin, Stahl and Brittenden19} with better diagnostic stability for bipolar disorder. ^{Reference Morgan, Lappin, Heslin, Donoghue, Lomas and Reininghaus87} Because most diagnoses in the included studies were assigned at first presentation, early diagnostic reclassification could not be examined and may contribute to discrepancies between service eligibility criteria and observed diagnostic distributions.

Although schizoaffective disorder occupies a diagnostically hybrid position, with criteria that span both primary psychotic symptoms and major mood episodes, we classified schizoaffective disorder a priori as a non-affective psychotic disorder, in line with its placement within the primary psychotic disorders grouping in the DSM-5-TR ⁵⁶ and ICD-11 ⁵⁸ and its general management within schizophrenia spectrum pathways in many EIS. This decision inevitably simplifies an underlying clinical continuum, but provides a nosologically consistent and service-relevant practice, allowing us to focus the affective group specifically on bipolar disorder with psychotic features and MDD-PF.

Additionally, findings relating to NIL should be interpreted with caution. The small number of studies in two subgroups and the unbalanced NIL distribution required broad categorisation and limited the use of more nuanced measures (e.g. relative deprivation). Consequently, these analyses are exploratory and highlight the need for further research in diverse socioeconomic contexts. In the same way, alternative urbanicity classifications (e.g. Organisation for Economic Co-operation and Development frameworks) could not be systematically applied because granular location-level data were not consistently reported across studies.

Our findings should be interpreted in light of the fact that EIS vary widely in their organisational structure and recruitment pathways. A further limitation is the substantial heterogeneity observed across studies, which likely reflects genuine differences in diagnostic practices, sampling methodology and EIS eligibility criteria. Some recruit on a catchment basis, whereas others rely on referrals from traditional care pathways or even convenience samples; others report on research subsamples rather than consecutive clinical cohorts. Each of these dimensions profoundly shapes the diagnostic composition of EIS populations. Because such information was not consistently reported across primary studies, we were unable to include it in moderator analyses. Our estimates therefore reflect pooled proportions across heterogeneous contexts, which are likely to mirror the reach and implementation of services, rather than any universal distribution of affective versus non-affective psychoses. Accordingly, our pooled proportions should be interpreted as describing diagnostic composition within existing EIS, rather than population-level prevalence at first episode. Future studies should systematically document and report service-level features (e.g. inclusion criteria, catchment design, referral sources) to enable a more nuanced understanding of heterogeneity.

Implications

Notwithstanding the above limitations, our findings raise important questions about the extent to which affective psychoses – particularly bipolar disorder with psychotic features and MDD-PF – are recognised. In our meta-analysis, affective psychoses accounted for approximately 18% of all cases, highlighting a clear subgroup of people within EIS for whom care pathways may be less explicitly differentiated, particularly in services where protocols are primarily structured around non-affective psychoses. Our results suggest that EIS models may benefit from more systematic detection and management of affective disorder presentations, through routine use of structured mood-oriented screening at initial presentation (e.g. the Mood Disorders Questionnaire, Montgomery-Åsberg Depression Rating Scale and Calgary Depression Rating Scale), ^{Reference Wang, Woo, Ahn, Ahn, Kim and Bahk88} and dedicated clinical pathways that integrate disorder-specific pharmacological and psychosocial interventions. This emphasis on diagnosis-sensitive pathways within early intervention frameworks is consistent with prior calls for improved recognition and targeted management of first-episode mania and bipolar disorder within EIS settings. ^{Reference Chia, Cotton, Filia, Phelan, Conus and Jauhar44,Reference Jauhar, Ratheesh, Davey, Yatham, McGorry and McGuire89} Strengthening these elements within EIS could facilitate earlier, more accurate diagnosis, and support more appropriate treatment allocation for this important subgroup.