Introduction
Bipolar disorder (BD) is a severe and persistent mood disorder characterised by cyclic alterations in mood, energy levels, and psychomotor activity [Reference Grande, Berk, Birmaher and Vieta1, Reference Vieta, Berk, Schulze, Carvalho, Suppes and Calabrese2]. The estimated prevalence is approximately 1–2% in the general population, with a significant functional impact and a high social and healthcare burden [Reference McIntyre, Berk, Brietzke, Goldstein, López-Jaramillo and Kessing3, Reference Goes4]. The episodic course and clinical variability contribute to the complexity of early recognition and therapeutic management [Reference Smedler, Bergen, Song and Landén5, Reference Nierenberg, Agustini, Köhler-Forsberg, Cusin, Katz and Sylvia6].
The aetiopathogenesis of BD is multifactorial, involving genetic vulnerabilities and neurobiological alterations, including neurotransmitter systems and circadian rhythms, as well as environmental factors [Reference Harrison, Geddes and Tunbridge7–Reference O’Connell, Koromina, van der Veen, Boltz, David and Yang9]. Clinical presentation is highly heterogeneous, with individual differences in the pattern of manic, hypomanic, depressive, and mixed episodes, as well as in relapse frequency and longitudinal course [Reference Phillips and Swartz10, Reference Bardi, Pinto, Simonetti, Di Nicola, Mazza and Manfredi11].
The course of BD is characterised by an alternation of acute episodes and phases of complete or partial remission. This pattern has led to the development of several staging models aimed at more accurately describing disease progression and guiding phase-specific therapeutic interventions [Reference Kapczinski, Dias, Kauer-Sant’Anna, Frey, Grassi-Oliveira and Colom12–Reference Berk, Post, Ratheesh, Gliddon, Singh and Vieta14].
The treatment of BD is particularly complex. Although mood stabilisers remain the first-line therapy, rates of incomplete response and relapse remain significant [Reference Nierenberg, Agustini, Köhler-Forsberg, Cusin, Katz and Sylvia6, Reference Cipriani, Barbui, Salanti, Rendell, Brown and Stockton15–Reference Fiorillo, Sampogna, Albert, Maina, Perugi and Pompili17]. Alongside pharmacotherapy, the effectiveness of psychotherapeutic interventions, such as psychoeducation, cognitive behavioural therapy (CBT), and therapies based on the regulation of interpersonal and social rhythms, is well documented [Reference Miklowitz, Efthimiou, Furukawa, Scott, McLaren and Geddes18–Reference Singh, Swartz, Cuellar-Barboza, Schaffer, Kato and Dols20].
International guidelines have synthesised the available scientific evidence and formulated diagnostic and therapeutic recommendations; however, their implementation in clinical practice remains variable. This heterogeneity reflects the complexity of the recommendations, differences between healthcare systems and cultural contexts, as well as the diverse training backgrounds and individual clinical beliefs [Reference Goodwin, Haddad, Ferrier, Aronson, Barnes and Cipriani21–Reference Fountoulakis, Grunze, Vieta, Young, Yatham and Blier24]. Across Europe, these differences contribute to a gap between evidence-based recommendations and real-world clinical decision-making [Reference McIntyre25–Reference Forsner, Hansson, Brommels, Wistedt and Forsell27]. Understanding these differences is essential to identify areas for improvement and potential strategies for harmonising care.
In this context, the structured network of the European Psychiatric Association (EPA) Ambassador Programme provides a unique opportunity to investigate clinical decision-making across different European regions. By using standardised clinical case vignettes within an established transnational professional network, it is possible to compare real-world therapeutic choices among psychiatrists. The present study was therefore conducted to explore treatment preferences in BD across Europe.
Methods
Study design, setting, and participants
The study followed the framework of previous research conducted within the EPA Ambassador Programme [Reference Rojnic Kuzman, Slade, Puschner, Scanferla, Bajic and Courtet28–Reference Rojnic Kuzman, Nordentoft, Raballo, Mohr, Fiorillo and Dom30]. Specifically, the survey consisted of three sections: Section A focused on general information about the participant; Section B addressed treatment choice preferences in the management of BD; and Section C presented a clinical case study, asking participants to read the clinical vignette and answer the related questions based on their own experience in clinical practice at their formal workplace. The case vignette read as follows: “CASE. A 30-year-old clerk officer has been treated with risperidone 3 mg daily for three months after a manic psychotic episode. While he has no more manic symptoms, he feels empty, with low mood, with no motivation, tiredness, trouble concentrating, and trouble following conversations with friends.”
Case vignettes were used as a tool to examine clinical judgements by health professionals, which can be highly generalisable to “real-life” behaviour [Reference Evans, Roberts, Keeley, Blossom, Amaro and Garcia31].
The case was constructed using experimental design. First, one expert designed initial cases following a set of recommendations (short case derived from clinical experience neutral with respect to cultural and socio-economic factors, following a narrative story-like progression) [Reference Falkai, Wagner, John, Yakimov, Galderisi and Bitter32]. The case was then reviewed and revised by three independent experts in the field of BD, including the chair of the respective EPA Sections (Section of Psychopharmacology) who read the vignette, provided a diagnosis, and confirmed the presence of key diagnostic features to improve the vignettes’ clarity, cultural neutrality, and validity. In the next step, the draft questionnaire containing revised case vignette was submitted to a pilot study among the remaining 17 members of the EPA Board, including representatives of GAMIAN [Reference Grol and Grimshaw26] and EUFAMI [Reference McIntyre25] to address issues of internal and external validity. They were asked to respond to the questions and to give separate feedback on the difficulty, clarity, and length of the survey.
The final questionnaire was reviewed and revised according to the pilot study analysis. All versions, as well as the final questionnaire, were developed and administered in the English language. Psychiatrists working in Europe and affiliated with the EPA community were contacted, including individual EPA members, member associations, and attendees of EPA congresses. In 2020, they were offered the opportunity to become “EPA Ambassadors” and to take part in EPA surveys, and the same recruitment procedure was applied for each subsequent Ambassador study. The full questionnaire is provided as Supplementary Material 4.
The survey was launched during the EPA Forum at the EPA Congress 2024, where the invitation to participate was open to Congress attendees. In the second phase, all representatives of the EPA Council of National Psychiatric Associations (Council of NPAs, comprising 44 NPAs) were invited to recruit members of their respective national associations. In the event of non-response from NPA representatives, a third recruitment phase involved members of the Board and representatives of the EPA Sections, who were asked to disseminate the invitations to their members.
Responses were collected from March 2024 to January 31 2025 using an online questionnaire, which took approximately 20–25 minutes to complete. The study was open to psychiatrists and psychiatry residents/trainees working in Europe. The authors affirm that the study procedures adhered to the ethical principles established by the appropriate national and institutional review bodies for research involving human participants, as well as to the Declaration of Helsinki (1975) and its subsequent revisions in 2008 and 2013 [33]. Ethical approval for the study was granted by the Ethics Committee of the Zagreb University Hospital Centre (approval number 02/013AG).
Statistical analysis
We collected and analysed sociodemographic data (age, gender, and country of employment), as well as information on the expertise of clinicians, training, and professional practice (time since qualification as consultant psychiatrists, specialty, work position, type of practice, clinical setting, and use of clinical guidelines in routine practice). Countries of employment were grouped into four regions according to EuroVoc (Supplementary Table S1): Central and Eastern Europe, Northern Europe, Southern Europe, and Western Europe.
We used descriptive statistics to summarise the sample characteristics, patient and clinical practice profiles, and treatment preferences. Descriptive statistics were also applied to summarise the use of assessment measures for BD, the frequency of patients in different phases or types of disorders, and adherence to clinical guidelines. Additionally, we applied latent class analysis (LCA) to classify participants into distinct latent groups based on their responses to various treatment options and characteristics. For the Case (initial treatment; n = 555) data subset, we conducted a LCA to classify participants into distinct groups based on their reported treatment choices. Because some treatment options were rarely selected (fewer than 10% of participants), we used eight combined treatment categories as binary input variables: antipsychotics, antidepressants, mood stabilisers, CBT, individual psychoeducation, lifestyle recommendations, other socio- or psychotherapy, and no psychopharmacological treatment/no therapy. The optimal number of latent classes was determined using several goodness-of-fit indices, including the log-likelihood, Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample-size adjusted BIC (SABIC). Models with lower values of AIC, BIC, and SABIC were considered to have better fit. Entropy and average posterior probabilities (APP) were examined as indicators of classification quality. After selecting the optimal model, participants were assigned to the class for which they had the highest posterior probability. Descriptive statistics were then used to characterise each latent class in terms of treatment choices. Differences in categorical variables across latent classes were assessed using chi-squared tests, while continuous variables were compared using analysis of variance (ANOVA). All LCA analyses were conducted using Stata’s gsem command (StataCorp, Stata Statistical Software: Release 17). This manuscript adheres to the STROBE guidelines for reporting cross-sectional studies [Reference Vandenbroucke, von Elm, Altman, Gøtzsche, Mulrow and Pocock34].
Results
Participant characteristics
The online survey yielded 687 responses from mental health professionals. After limiting participants only to countries defined by the EuroVoc regions, the final sample for this analysis consisted of 654 participants, as described in detail in Table 1. Concerning the clinical presentation of patients with BD, participants reported frequently encountering bipolar depression (n = 470, 72.5%), mania (n = 239, 36.9%), hypomania (n = 243, 37.7%), bipolar II disorder (n = 253, 40.0%), BD with psychotic symptoms (n = 240, 37.3%), mixed episodes (n = 214, 33.2%), cyclothymia (n = 102, 16.1%), and rapid cycling (n = 66, 10.5%).
Sample description (n = 654)
Table 1. Long description
The table is divided into two main sections based on the type of data reported.
Categorical variables reported as N and percentage:
* EuroVoc regions: Central and Eastern Europe 391 (59.8 percent), Northern Europe 51 (7.8 percent), Southern Europe 126 (19.3 percent), and Western Europe 86 (13.1 percent).
* Gender: Male 386 (59.0 percent), Female 266 (40.7 percent), and Other 2 (0.3 percent).
* Profession: Psychiatrist 539 (82.4 percent), Psychiatric trainee 102 (15.6 percent), and Other mental health care professional 13 (2.0 percent).
* Professional setting: University hospital 272 (41.6 percent), General hospital 57 (8.7 percent), Specialised psychiatric hospital 176 (26.9 percent), Primary care setting 31 (4.7 percent), Community mental health 58 (8.9 percent), and Other 60 (9.2 percent).
* Certified psychotherapist: Yes 263 (40.2 percent), No 391 (59.8 percent).
Continuous variables reported as Mean and S D:
* Age: 43.6 (12.7).
* Work experience in years: 15.8 (11.9).
* Percentage of work time with Outpatients: 56.4 (35.7).
* Percentage of work time with Inpatients: 44.0 (35.2).
* Percentage of work time with Other: 12.9 (19.2).
Abbreviations: M, mean; N, number of observations; SD, standard deviation.
Patient characteristics and clinical practice patterns
The clinical practices and patient characteristics of the 654 participants are summarised in Supplementary Table S2. Regarding the use of assessment measures for BD over the past 6 months, 192 participants (29.4%) reported not using any scales, while 36 (5.5%) used other unspecified measures. Diagnostic interviews such as Mini International Neuropsychiatric Interview (MINI) or Structured Clinical Interview for DSM (SCID) were applied by 208 participants (31.8%), at-risk assessments (Structured Interview for Psychosis-Risk Syndromes [SIPS], Comprehensive Assessment of At-Risk Mental States [CAARMS], Schizophrenia Proneness Instrument [SPI]) by 31 (4.7%), and neurocognitive assessments by 111 (17.3%). Functionality scales (Functioning Assessment Short Test [FAST], Social and Occupational Functioning Assessment Scale [SOFAS], and World Health Organization Disability Assessment Schedule [WHODAS]) were used by 73 participants (11.2%). Scales for psychoses (Brief Psychiatric Rating Scale [BPRS], Positive and Negative Syndrome Scale [PANSS]) and for specific psychotic symptoms (Scale for the Assessment of Positive Symptoms [SANS], Scale for the Assessment of Negative Symptoms [SAPS], Clinical Assessment Interview for Negative Symptoms [CAINS]) were applied by 203 (31.0%) and 55 (8.4%) participants, respectively. Scales for mood and anxiety disorders (e.g., Hamilton Depression Rating Scale [HAMD], Montgomery–Åsberg Depression Rating Scale [MADRS], Young Mania Rating Scale [YMRS], Hamilton Anxiety Rating Scale [HAMA], Quick Inventory of Depressive Symptomatology [QIDS], and Beck Depression Inventory [BDI]) were the most frequently used, with.
357 participants (54.6%) reporting regular use. Temperament assessment tools (Temperament Evaluation of Memphis, Pisa, Paris, and San Diego – Autoquestionnaire/Interview (TEMPS-I, TEMPS-A)) were used by 26 participants (4.0%), and personality assessment tests by 124 (19.0%). Non-use of clinical guidelines varied across different guidelines: 13.8% for national guidelines, 13.1% for National Institute for Health and Care Excellence [NICE], 28.0% for EPA, 36.2% for American Psychiatric Association (APA), 50.6% for Canadian Network for Mood and Anxiety Treatments (CANMAT), and 58.1% for World Federation of Societies of Biological Psychiatry (WFSBP) guidelines. On average, participants managed 16.8 patients with BD per month (SD = 31.5).
Clinician treatment choices and preferences
The treatment preferences of the 654 participants are summarised in Table 2. The majority of participants reported having experience in treating specific phases of BD, with 86.5% managing manic episodes, 64.1% rapid cycling, 74.8% mixed episodes, and 78.6% hypomanic episodes. Across all phases, pharmacological treatment was predominant, with only a small proportion of respondents reporting no medication use (2.3–4.3%), for the last patient they treated. Treatment strategies varied according to the phase of illness. Monotherapy was chosen by 22–38% of participants, being most frequently reported for hypomanic episodes, whereas two-drug combinations were selected by 24–28%. The use of three or more therapies was common, particularly in manic and mixed episodes, ranging from 34 to 48% of respondents.
Clinicians’ treatment preferences
Table 2. Long description
A table with seven columns and forty-six rows. The columns are: Treatment Category, Manic episode, Rapid cycling, Mixed episode, Hypomanic episode, C A S E initial treatment or depression, and C A S E maintenance phase or depression.
Key data points include:
* Treated B D-specific phase: Yes responses are highest for Manic episode at 566 (86.5 percent) and lowest for C A S E maintenance at 224 (34.3 percent).
* Medication count: Combination therapy (three or more medications) is most common for Mixed episodes at 48.3 percent, while monotherapy is most common for C A S E initial treatment at 49.0 percent.
* Antipsychotics: Olanzapine and quetiapine are the most preferred, peaking at 68.6 percent for Manic episodes.
* Mood stabilizers: Lithium is preferred by 41.3 percent for Manic episodes and 42.5 percent for Rapid cycling. Sodium valproate is preferred by 36.8 percent for Rapid cycling.
* Socio and psychotherapy: Individual psychoeducation is a leading choice across all episodes, ranging from 28.1 percent to 33.0 percent. Cognitive behavioral therapy is highly preferred for C A S E initial treatment at 44.0 percent.
* Other therapies: Hospitalization is highly preferred for Manic episodes at 89.2 percent, while Electroconvulsive therapy is preferred by 27.9 percent for the same phase.
Note: Data are presented as number (percentage) of participants. The full-sample denominator (n = 654) was used only to assess whether participants had treated or would treat a particular episode or case. All other percentages were calculated using the number of participants who reported having treated or being willing to treat the respective episode or the case.
“–” indicates that non-pharmacological options were not assessed in the treatment-switch/maintenance item (CASE – switch medication), as only pharmacological switch was queried.
* Non-pharmacological methods for BD-specific phases/types were assessed using a single-choice question, whereas for the medications and case they were assessed using a multiple-choice question.
Mood stabilisers and antipsychotics were the most frequently selected pharmacological approaches. Lithium and lamotrigine were widely used, with lithium preferred for manic (41.3%) and rapid cycling episodes (42.5%) and lamotrigine for rapid cycling (19.6%) and mixed episodes (21.3%). Sodium valproate was selected in 24.7–36.8% of cases, while carbamazepine was consistently used across phases (10.3–11.5%). Among antipsychotics, second-generation agents, grouped as amisulpride/aripiprazole/brexpiprazole/cariprazine, were reported by 25–33% of participants, whereas olanzapine/quetiapine were the most frequently selected antipsychotics across phases (54.5–68.6%). Risperidone/lurasidone/ziprasidone/sertindole/paliperidone were chosen in 11.2–24.6% of cases, and typical antipsychotics (chlorpromazine, haloperidol, fluphenazine, promazine, zuclopenthixol) were used less frequently, primarily in manic episodes (1.1–21.6%).
Antidepressants were selected in 3.5–19.3% of cases and hypnotics were selected by 5.0–27.0% of participants. Other medications, including atomoxetine, dextroamphetamine, methylphenidate, thyroid hormone, or supplements (omega-3, vitamin D) were rarely applied (<8%). A substantial proportion of participants reported that they would not apply any psychosocial intervention, particularly in manic episodes (18.1%) and mixed episodes (13.7%). When interventions were applied, CBT was recommended by 12.3–21.7% of respondents across phases and individual psychoeducation by 28.1–33.0%. Lifestyle recommendations, including advice on sleep, diet, physical activity, and substance use, were endorsed by roughly one-fifth of participants (17.3–22.4%). Other psychotherapeutic approaches, including psychodynamic, systemic/family, trauma-focused therapy, acceptance and commitment therapy, cognitive remediation, social skills training, creative therapies, body/movement therapy, mindfulness, and structured physical health programmes were less commonly applied (≤10%).
Electroconvulsive therapy was selected by 22.7–27.9% of participants for manic, rapid cycling, and mixed episodes, and less frequently in depressive cases (4.9%). Transcranial magnetic stimulation was reported in 6.0–9.1% of manic episodes and more often during treatment of depressive states (13.7%). Long-acting injectable antipsychotics were commonly considered for manic (54.9%) and rapid cycling episodes (43.0%), with lower use in hypomanic episodes (14.2%) and limited use in depressive cases (11.6%). Hospitalisation was frequently indicated for manic (89.2%) and mixed episodes (66.2%), while less commonly considered for hypomanic episodes (17.0%).
At the initial assessment of the clinical case vignette describing the depressive phase, the majority of participants reported that they would initiate pharmacological treatment (84.9%). Monotherapy was the most frequently selected strategy (49.0%), followed by two-drug combinations (29.4%) and three or more therapies (18.7%). Mood stabilisers and second-generation antipsychotics represented the core pharmacological options. Lithium was selected by 29.2%, lamotrigine by 20.5%, while quetiapine (24.5%) and aripiprazole (18.4%) were the most frequently preferred antipsychotics. Olanzapine was less frequently selected (13.2%). Antidepressants are reported in 19.3% of cases, while sedatives/hypnotics were reported in a minority of cases (<6%). CBT (44.0%), individual psychoeducation (40.9%), and lifestyle recommendations (32.3%) were the most frequently recommended psychosocial interventions. Other psychotherapeutic modalities were applied less often. Electroconvulsive therapy, transcranial magnetic stimulation, long-acting injectables, and hospitalisation were rarely indicated at this initial stage.
After resolution of the depressive phase in the clinical vignette, 34.3% of participants reported that they would switch pharmacological treatment for maintenance. Among those opting for a switch, mood stabilisers were the most frequently selected medications, with lithium (38.8%) and lamotrigine (27.7%) preferred. Quetiapine (26.8%) and aripiprazole (21.9%) were the most frequently used antipsychotics in the switch scenario, whereas antidepressants were selected in 16.1% of cases, and other antipsychotics and sedatives/hypnotics were applied in <6% of cases. Other interventions, including long-acting injectable antipsychotics (11.6%), hospitalisation (20.4%), and transcranial magnetic stimulation (13.7%), were selected by some participants, while electroconvulsive therapy was rarely used (4.9%).
Latent classes based on treatment choices of clinicians
The model with four latent classes showed the best fit to the empirical data (Supplementary Table S3). Participants were assigned to latent classes according to the highest posterior probability (Table 3). Class 1, labelled Preferring Antidepressants, included participants that preferred antidepressants (94.2%), followed by CBT (51.4%) and other socio- or psychotherapeutic interventions (48.5%). Class 2, labelled Preferring Mood Stabilisers, comprised participants all of whom preferred mood stabilisers (100%), with frequent use of antipsychotics (55.0%), individual psychoeducation (49.0%), CBT (48.6%), and other socio- or psychotherapeutic interventions (48.3%). Class 3, labelled Preferring Minimal/Other Treatment, included participants who reported no use of psychopharmacological treatment or therapy (100%), followed by antipsychotics (51.7%) and mood stabilisers (36.2%). Finally, Class 4, labelled Preferring Antipsychotics, comprised participants with antipsychotics being the most chosen treatment (61.7%), followed by individual psychoeducation (28.1%), lifestyle recommendations (25.8%), and CBT (25.8%).
Clinicians’ treatment choices in different latent classes
Table 3. Long description
The table consists of five columns. The first column lists treatment types, and the subsequent four columns represent Class 1, Class 2, Class 3, and Class 4. Data is presented as number and percentage in the format: n (percentage).
* Antipsychotics: Class 1: 12 (17.1); Class 2: 147 (55.0); Class 3: 30 (51.7); Class 4: 160 (61.7).
* Antidepressants: Class 1: 66 (94.2); Class 2: 30 (11.2); Class 3: 11 (18.9); Class 4: 0 (0.0).
* Mood stabilisers: Class 1: 18 (25.7); Class 2: 267 (100.0); Class 3: 21 (36.2); Class 4: 0 (0.0).
* Cognitive behavioural therapy: Class 1: 36 (51.4); Class 2: 130 (48.6); Class 3: 11 (18.9); Class 4: 67 (25.8).
* Individual psychoeducation: Class 1: 18 (25.7); Class 2: 131 (49.0); Class 3: 5 (8.6); Class 4: 73 (28.1).
* Lifestyle recommendations: Class 1: 22 (31.4); Class 2: 93 (34.8); Class 3: 9 (15.5); Class 4: 55 (21.2).
* Other socio- or psychotherapy: Class 1: 34 (48.5); Class 2: 192 (48.3); Class 3: 11 (18.9); Class 4: 62 (23.9).
* Other, no psychopharmacs, no therapy: Class 1: 2 (2.8); Class 2: 4 (1.4); Class 3: 58 (100.0); Class 4: 1 (0.3).
Note: Class 1 prefers Antidepressants. Class 2 prefers Mood Stabilisers. Class 3 prefers Minimal or Other Treatment. Class 4 prefers Antipsychotics.
Note: Data are presented as number (percentage) of participants. Class 1 – Preferring Antidepressants; Class 2 – Preferring Mood Stabilisers; Class 3 – Preferring Minimal/Other Treatment; Class 4 – Preferring Antipsychotics.
* Atomoxetine, dextroamphetamine, methylphenidate, thyroid hormone, Omega-3, vitamin D, supplements, etc.
These latent classes highlight distinct patterns of treatment choices among clinicians that were associated with different clinical and demographic characteristics (Table 4).
Description of the latent classes (n = 654)
Table 4. Long description
The table presents data for Class 1 n equals 70, Class 2 n equals 267, Class 3 n equals 58, and Class 4 n equals 259.
EuroVoc regions: Central and Eastern Europe is the most represented across all classes, peaking in Class 1 at 82.9 percent. Northern, Southern, and Western Europe follow with lower percentages. The p-value for regional distribution is less than 0.001.
Gender: Males represent the majority in all classes, ranging from 71.4 percent in Class 1 to 52.7 percent in Class 4. Females range from 28.6 percent to 47.3 percent. p equals 0.019.
Profession: Psychiatrists make up the vast majority (77.9 percent to 91.4 percent) compared to psychiatric trainees. p is less than 0.001.
Professional setting: University hospitals are a primary setting for Class 3 (62.1 percent) and Class 1 (50.0 percent), while specialized psychiatric hospitals are common for Class 2 (30.3 percent) and Class 4 (26.9 percent). p is less than 0.001.
Certified psychotherapist: Percentages of ‘Yes’ range from 28.6 percent in Class 1 to 45.9 percent in Class 4. p equals 0.039.
Continuous variables (Mean and S D):
- Age: Class 4 is the oldest (Mean 45.7) and Class 2 is the youngest (Mean 41.3). p is less than 0.001.
- Work experience: Class 3 has the most experience (Mean 17.6 years) while Class 2 has the least (Mean 13.9 years). p equals 0.005.
- Work time distribution: No significant difference (p greater than 0.05) was found for time spent with outpatients (ranging 49.4 to 58.7 percent) or inpatients (ranging 42.2 to 48.7 percent).
Abbreviations: M, mean; N, number of observations; SD, standard deviation.
Class 1 “Preferring Antidepressants” included 70 clinicians (10.7% of the sample) and was strongly represented by clinicians from Central and Eastern Europe (82.9%), while representation from other regions was minimal. Participants were predominantly male (71.4%) and included a mix of psychiatrists (80.0%) and psychiatric trainees (20.0%). Half of the clinicians in this class worked in university hospitals (50.0%), whereas smaller proportions were employed in specialised psychiatric hospitals (14.3%) or general hospitals (14.3%). Only 28.6% were certified psychotherapists. Clinicians in this class had a mean age of 43.7 years (SD = 12.4) and an average of 15.3 years of professional experience (SD = 10.9).
Class 2 “Preferring Mood Stabilisers” was the largest class, comprising 267 clinicians (40.8%). Compared with other classes, this group showed a more heterogeneous geographic distribution, although Central and Eastern Europe remained the most represented region (48.7%), followed by Southern Europe (25.5%), Western Europe (17.2%), and Northern Europe (8.6%). Male clinicians were again more prevalent (62.4%). Most participants were psychiatrists (77.9%), while 22.1% were trainees. In terms of professional setting, clinicians in this class were most commonly employed in university hospitals (41.6%) and specialised psychiatric hospitals (30.3%). The proportion of certified psychotherapists was 37.5%. This class had the lowest mean age (41.3 years, SD = 12.7) and the least professional experience (13.9 years, SD = 11.6). Class 3 “Preferring Minimal/Other Treatment” was the smallest class, with only 58 clinicians (8.9% of the sample). Similar to Class 1, this group was largely composed of clinicians from Central and Eastern Europe (77.6%). A large majority were psychiatrists (91.4%) and only a small proportion were trainees (8.6%). This class showed the highest proportion of clinicians working in university hospitals (62.1%). Compared with Classes 1 and 2, a higher proportion of clinicians were certified psychotherapists (41.4%). Clinicians in this group had a mean age of 44.5 years (SD = 11.8) and a mean of 17.6 years of professional experience (SD = 11.7).
Class 4 “Preferring Antipsychotics” included 259 of clinicians (39.6%) who came from a broader mix of European regions, although Central and Eastern Europe remained the largest subgroup (61.1%). The gender distribution was more balanced compared with other classes, with 52.7% male clinicians. This class consisted predominantly of psychiatrists (90.7%), with relatively few trainees (9.3%). Clinicians were most commonly employed in specialised psychiatric hospitals (26.9%) and university hospitals (41.6%), with smaller proportions working in community mental health services (8.9%) or general hospitals (8.7%). This class had the highest proportion of certified psychotherapists (45.9%). Participants in this group had a mean age of 45.7 years (SD = 12.4) and a mean of 17.3 years of professional experience (SD = 12.0). Post hoc comparisons indicated that clinicians in Class 4 had a significantly higher mean age than those in Class 2 and greater professional experience than those in Class 2.
Discussion
The survey of 654 psychiatrists and psychiatry trainees revealed clear patterns in the management of bipolar affective disorder across Europe. Across all illness phases, pharmacological treatment predominated, typically with mood stabilisers or second-generation antipsychotics, while antidepressant use remained more selective. This is consistent with current trends towards individualising antidepressant prescription in BD [Reference Fico and Vieta35].
Similar patterns emerged in the case vignette, where most participants favoured pharmacological monotherapy at initiation and a minority reported switching treatment for maintenance. Although self-reported preferences may not fully reflect prescribing behaviour, our lithium figures fall within the wide range reported for real-world European prescribing (22.0–70.6%). Documented rates in Scotland (22–26%) and in Swedish National Registry data (41%) are similar to or lower than reported here, whereas Dutch data show a substantially higher rate (70.6%), underscoring the considerable cross-country variability in lithium prescribing that likely also underlies our findings [Reference Karanti, Kardell, Lundberg and Landén36–Reference Lyall, Penades and Smith38]. This suggests that our findings, while based on stated preference, are not markedly discordant with documented prescribing patterns.
In line with previous literature [Reference Vieta, Berk, Schulze, Carvalho, Suppes and Calabrese2, Reference Goes4, Reference Nierenberg, Agustini, Köhler-Forsberg, Cusin, Katz and Sylvia6], the results highlight substantial heterogeneity in both pharmacological strategies and the integration of psychosocial interventions. Across all phases of the disorder, pharmacological treatment represented the primary choice, with monotherapy and combination therapies adapted to the specific clinical presentation [Reference Kishi, Ikuta, Matsuda, Sakuma, Okuya and Nomura39–Reference Yildiz, Siafis, Mavridis, Vieta and Leucht41]. When compared with current international pharmacological guidelines, our findings show substantial concordance alongside some notable discrepancies. For acute mania, the preference for lithium, valproate, and second-generation antipsychotics such as quetiapine and aripiprazole aligns with the first-line recommendations of both CANMAT/ISBD and WFSBP guidelines [Reference Keramatian, Chithra and Yatham42, Reference Grunze, Vieta, Goodwin, Bowden, Licht and Moller43]. However, the continued use of first-generation antipsychotics (up to 21.6%) and carbamazepine (10.3–11.5%), neither of which are first-line agents in these guidelines, suggests areas of divergence from evidence-based recommendations. For bipolar depression, the predominance of quetiapine, lithium, and lamotrigine mirrors the first-line options endorsed by both guidelines [Reference Keramatian, Chithra and Yatham42, Reference Grunze, Vieta, Goodwin, Bowden, Licht and Möller44]; however, the use of aripiprazole (18.4%), explicitly listed as not recommended for acute bipolar depression by CANMAT/ISBD [Reference Keramatian, Chithra and Yatham42], indicates a degree of discordance with current evidence, while the moderate use of antidepressants (19.3%) may be less concerning if used adjunctively rather than as monotherapy, which our survey did not distinguish. For maintenance treatment, the selected agents (lithium, lamotrigine, quetiapine, and aripiprazole) closely match the first-line maintenance options recommended by CANMAT/ISBD [Reference Keramatian, Chithra and Yatham42], reflecting strong guideline concordance in this phase. For mixed episodes, valproate use is supported by WFSBP evidence (RG 4 for acute treatment) [Reference Grunze, Vieta, Goodwin, Bowden, Licht and Azorin45], whereas the relatively frequent use of lamotrigine, for which the WFSBP mixed-states guideline reports no supporting evidence for acute presentations, appears not guideline-concordant in this context; notably, CANMAT/ISBD does not provide phase-specific recommendations for mixed states, so this comparison rests on WFSBP evidence alone.
Interestingly, the number of participants who would offer long-acting injectables is around 50% in the manic phase, which seems more compared to the results of our Ambassador study in schizophrenia [Reference Rojnic Kuzman, Nordentoft, Raballo, Mohr, Fiorillo and Dom30]. This may be particularly surprising, as in many countries, LAIs may not even be approved nor reimbursed for patients with BD. This appears particularly favourable considering the relatively high rates of non-adherence in patients with BD, which represents a critical factor in disease management. In light of international guidelines and recent evidence, LAIs may represent an early and strategic therapeutic option for patients with predominantly manic polarity [Reference Verdolini, Hidalgo-Mazzei, Del Matto, Muscas, Pacchiarotti and Murru46, Reference Pacchiarotti, Tiihonen, Kotzalidis, Verdolini, Murru and Goikolea47]. Recent studies have shown that early use of LAIs, such as risperidone or injectable aripiprazole,
significantly reduce the risk of hospitalisations and manic relapses without worsening depressive recurrences. Moreover, LAIs improve treatment adherence compared to oral antipsychotics, with positive effects on long-term outcomes, including reduced hospitalisations and improved quality of life [Reference Vieta, Tohen, McIntosh, Kessing, Sajatovic and McIntyre48]. On the other hand, such a high rate of LAIs in manic phase, at least partly, might be explained with several other possible factors such as limited human power resources (e.g., number of psychiatrists), pressure of time reserved for patients, and as a consequence, attitudes of clinicians to stay on the secure side for possible risky behaviours of patients in manic phase. In addition to evidence and guidelines, individual needs and shared decision-making should guide clinicians in their practice.
Psychosocial interventions [Reference de Barros Pellegrinelli, de O Costa, Silval, Dias, Roso and Bandeira49–Reference Cuijpers, Quero, Noma, Ciharova, Miguel and Karyotaki51] were not systematically implemented, and a notable proportion of clinicians reported that they would not use them, particularly during manic and mixed-episodes. When recommended, the most common approaches included psychoeducation, CBT in up to 40% in the depressive phase, and lifestyle recommendations, while other psychotherapeutic approaches and structured physical health programs were rarely used.
In the case scenario, psychosocial interventions were recommended more often than in routine practice, although they remained underutilised overall. Clinicians selectively recommended psychosocial or psychotherapeutic interventions that are less commonly applied in everyday care but receive stronger emphasis in structured assessments such as case vignettes.
This pattern aligns with international guidelines, which consistently endorse psychosocial interventions as adjunctive to pharmacological treatment, emphasising their role in relapse prevention, functional recovery, and patient empowerment, particularly during depressive and maintenance phases of BD [Reference Ventriglio, Ricci, Giallanella, Gaggiano, Torales and Castaldelli-Maia52]. Similarly, structured physical training is used in less than 10%, despite the well-known effect of antipsychotics and mood stabilisers on weight gain and lipid profiles [Reference Pillinger, McCutcheon, Vano, Mizuno, Arumuham and Hindley53], again contrary to existing guidelines [Reference Maurus, Wagner, Spaeth, Vogel, Muenz and Seitz54].
Electroconvulsive therapy(ECT) was selected by 22.7–27.9% of participants for manic, rapid cycling, and mixed episodes, and less frequently in depressive cases (4.9%). Transcranial magnetic stimulation(TMS) was reported in 6.0–9.1% of manic-spectrum episodes and more often during treatment of depressive states (13.7%). These preferences appear generally consistent with the APA guidelines [55], which recommend ECT primarily for severe or treatment-resistant cases. Moreover, recent evidence on TMS in BD, although limited, suggests that it may be a relatively safe and well-tolerated intervention, particularly during depressive episodes, with potential benefits for mood stabilisation and as an adjunct to pharmacotherapy. Small sample sizes and heterogeneity in study designs highlight the need for larger sham-controlled trials to clarify efficacy across manic, mixed, and depressive states, but current findings support the selective use of TMS in depressive phases of BD [Reference Konstantinou, Hui, Ortiz, Kaster, Downar and Blumberger56, Reference Mutz57].
LCA identified four distinct patterns of clinical practice among European clinicians. The first group represented 10.7% of the sample and was characterised by greater use of antidepressants combined with moderate use of psychosocial interventions and included more clinicians from Central and Eastern Europe as well as a higher proportion of trainees. The second and largest group (40.8%) relied primarily on mood stabilisers, often combined with antipsychotics and psychosocial support, and consisted of relatively younger and less-experienced clinicians. The third group, including 8.9% of the sample, rarely reported using pharmacological or psychotherapeutic interventions and included older, more experienced clinicians with a higher proportion of certified psychiatrists. The fourth group included 39.6% of the sample and was characterised by frequent use of antipsychotics alongside psychoeducation, CBT, and lifestyle recommendations, and consisted mainly of more experienced psychiatrists often working in specialised psychiatric hospitals.
Overall, the differences observed across latent classes indicate that treatment patterns are influenced not only by clinical considerations but also by clinician-related factors, such as geographic context, professional training, career stage, and clinical setting. These latent classes highlight that treatment choices are not uniform but vary according to demographic and professional characteristics of the clinicians and their geographic region. The findings underscore the importance of considering these clinician-level factors when promoting the implementation of evidence-based guidelines across European psychiatric services. Consequently, targeted dissemination and training strategies may be necessary to reduce variation in clinical practice and support more consistent adoption of evidence-based treatment approaches. The observed association between clinician experience, professional context, and treatment orientation underscores the potential impact of training pathways and institutional context on clinical decision-making [Reference Fisher, Manicavasagar, Sharpe, Laidsaar-Powell and Juraskova58–Reference Hidalgo-Mazzei, Mantingh, Pérez de Mendiola, Samalin, Undurraga and Strejilevich60]. This perspective is further supported by evidence highlighting the importance of collaborative decision-making processes in BD care. For example, a large observational study [Reference Renes, Metz, Nolen, Hoogendoorn, Kupka and Regeer61] has shown that key components of shared decision-making are positively associated with greater guideline concordance, higher patient satisfaction with care, and improved medication adherence. These findings suggest that clinical decision-making is shaped not only by clinician-level factors and institutional context but also by the extent to which patients are actively involved in the treatment process.
Although the study benefits from a relatively large sample of psychiatrists and psychiatry trainees from multiple European regions and the use of LCA to identify naturally occurring practice patterns, some limitations should be acknowledged.
First, participants were recruited through the EPA Ambassador initiative, including EPA members, member associations, and attendees of EPA congresses. This recruitment strategy may have introduced a selection bias towards clinicians who are more engaged with professional networks and academic activities, potentially limiting the representativeness of the sample with respect to the broader population of European psychiatrists. As the overall response rate could not be calculated, findings should be interpreted as reflecting the practices of an engaged subset of European psychiatrists rather than a representative sample of the profession. For example, real-world studies such as those based on National registries may better reflect the clinical landscape [Reference Jorgensen, Sloth, Larsen, Osler and Kessing62]. Second, the geographical distribution of respondents was uneven, with a higher proportion of participants from Central and Eastern Europe compared with other regions due to the higher total number of countries included in the CEE region compared to other regions (21 of the 24 countries in the CEE region, 6 of 8 countries in the NE region, 6 of 10 countries in the SE region, and 8 of 14 countries in the WE region according to EuroVoc). In addition, a substantial proportion of clinicians worked in university hospitals or specialised psychiatric settings, which may reflect a greater exposure to academic environments and guideline-based practice than in other clinical contexts. These factors may influence treatment preferences and limit the generalisability of the findings to all European mental health services. Third, although the choice of the final class solution was based on information criteria, classification quality, class sizes, and substantive interpretability, the Lo–Mendell–Rubin likelihood ratio test (LMR) and bootstrap likelihood ratio test (BLRT) were not reported because these tests are not available in the Stata 17 gsem implementation used for the present latent class analyses. In addition, Stata 17 does not provide built-in options for accounting for classification uncertainty when examining associations between latent class membership and sample characteristics. The study relied on self-reported treatment preferences, which may be subject to reporting or social desirability bias and may not fully reflect actual clinical practice. Furthermore, the cross-sectional design prevents causal inference and does not allow examination of how treatment approaches may evolve over time or in response to clinical outcomes [Reference Vieta and De Prisco63]. Fifth, maintenance-phase preferences were captured only through the case-vignette switch item, which queried pharmacological choice exclusively. Finally, although most participants reported using clinical guidelines in practice, variability was observed in adherence to different international recommendations, with a notable proportion reporting non-adherence to certain guideline frameworks [Reference Goodwin, Haddad, Ferrier, Aronson, Barnes and Cipriani21–Reference Fountoulakis, Grunze, Vieta, Young, Yatham and Blier24]. This variability may reflect differences in regional training, availability of guidelines, or institutional practices, and should be considered when interpreting the observed treatment patterns.
From a clinical perspective, these findings emphasise the importance of tailoring guideline dissemination, continuing education, and institutional support to clinician characteristics and professional context. Recognising that treatment choices vary according to both patient presentation, and clinician-level determinants may help reduce variability in practice, enhance adherence to evidence-based recommendations, and promote more integrated and consistent management of BD across European and International psychiatric services [Reference Goodwin, Haddad, Ferrier, Aronson, Barnes and Cipriani21, Reference Yatham, Chakrabarty, Bond, Schaffer, Beaulieu and Parikh23, Reference Keramatian, Chithra and Yatham42, 64].
Overall, our findings highlight the predominance of pharmacological interventions in the management of BD in Europe, with mood stabilisers and second-generation antipsychotics as the mainstay of treatment. Psychosocial interventions, although recommended by guidelines, are applied less consistently, particularly outside standardised clinical case assessments. Variation in treatment preferences is evident, with clear latent class patterns suggesting that clinician characteristics, regional context, and training influence therapeutic choices.
Study implications
The study has several important implications. First, it suggests a general convergence across Europe towards a preference for pharmacotherapy in the treatment of BD, but also underutilisation of non-pharmacological methods. This tendency is not fully aligned with current clinical guidelines and may instead reflect differences in resource availability and service organisation, which highlight the need for greater dialogue within NPAs and the EPA community in order to promote more consistent standards of best practice.
Notably, the findings also indicate that a substantial proportion of clinicians do not rely on any specific clinical guidelines when recommending treatment for BD, similarly to what was observed in other Ambassador studies [Reference Rojnic Kuzman, Padberg, Amann, Schouler-Ocak, Bajic and Melartin29]. This suggests that additional efforts are needed to improve guideline dissemination and implementation in routine clinical practice. The EPA can play an important role in identifying strategies that facilitate the adoption of evidence-based recommendations among clinicians. These results may therefore support further initiatives aimed at strengthening good clinical practice across Europe among mental health professionals, organisations, and policy makers.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1192/j.eurpsy.2026.12265.
Data availability statement
The data that support the findings of this study are available as open source.
Acknowledgements
We thank Thibaut Le Forsonney, EPA employee, for efficient and kind technical support of the study.
Financial support
This research received no grant funding from any external agency, commercial or not-for-profit sector, but was financially supported by the European Psychiatry Association (EPA).
Competing interests
Celso Arango has been a consultant to or has received honoraria or grants from Abbot, Acadia, Ambrosetti, Angelini, Biogen, BMS, Boehringer, Carnot, Gedeon Richter, Janssen Cilag, Lundbeck, Medscape, Menarini, Minerva, Otsuka, Pfizer, Roche, Rovi, Sage, Servier, Shire, Schering Plough, Sumitomo Dainippon Pharma, Sunovion, Takeda, and Teva. The other authors declare no conflict of interest.
Open access
Comments
No Comments have been published for this article.