Bipolar disorder presents a unique challenge for patients and clinicians in medicine, as it involves managing two completely different presentations of the same illness, both in terms of treatment and preventing relapse. Different treatment options are also not equally effective in either treating or preventing episodes of mania or depression.
Therefore, in the first episode of mania in bipolar disorder, the question clinicians face is not only “how can I treat this person and get them better as quickly as possible?”, but also how to prevent future relapses. Guidelines such as NICE (NICE, 2014) and British Association of Psychopharmacology (Goodwin et al 2016) suggest using a second-generation antipsychotic (SGA) as a first option and adding a mood stabiliser if this isn’t effective. For patients who do respond well with SGAs and recover, clinicians and patients are faced with a decision regarding long-term treatment for relapse prevention: do I stick with the current antipsychotic medication which has led to recovery, or do I switch or add a mood stabiliser for relapse prevention?
As SGAs are also an option for relapse prevention, it seems many are choosing to ‘stick’, rather than risk a switch with new medication with unknown efficacy and side effects from an individual perspective. As previous elves have noted, the use of lithium is declining (Edward & Ahmed 2019) though it remains the most effective medication in preventing depressive relapses in bipolar disorder (Swiffen 2025).
The recent paper by Szmulewicz et al (2026) used observational data to compare the risk of relapse between those on mood stabiliser monotherapy, SGA monotherapy and combination therapy of a mood stabiliser and SGA.
Methods
The researchers used data from electronic health records of patients treated within first episode psychosis services and explicitly emulated a ‘target trial’, i.e. stated how a hypothetical randomised controlled trial (RCT) would answer their research question.
In this case, the target trial was assessing whether mood stabilisers alone, SGAs alone or a combination of both were better at preventing relapse (defined as hospitalisation or emergency department attendance due to any psychiatric reason) in patients with bipolar disorder who had been treated for mania with psychotic symptoms. Participants were aged over 16 years were included from several first episode psychosis cohorts in North America, Chile and Spain. Patients were included if they hadn’t been taking antipsychotics or mood stabilisers over the previous 3 months and were classified as being started on either SGAs (Olanzapine, Risperidone, Aripiprazole or Quetiapine), mood stabilisers (Lithium or Valproic Acid) or combination therapy.
The study followed participants up to 2 years, or until relapse or loss to follow up. As well as overall relapse they looked at manic/mixed relapse, and depressive relapse/suicide-related event as secondary outcomes. Over the two-year period, individual participants could be eligible for multiple “trials” e.g. if they were on a mood stabiliser, then discontinued for more than 3 months, and then commenced an antipsychotic.
The study used both intention-to-treat (according to assigned treatment) and per protocol (according to adherence to treatment) analyses. Sensitivity analyses were conducted to analyse a range of different possible scenarios that could have affected the results.
Results
There were 371 eligible participants with first episode psychotic mania that corresponded to 465 treatment initiations (as some participants contributed more than one “trial” during the follow up). There were 112 mood stabiliser monotherapy initiations (in 79 individual patients), 167 second-generation antipsychotic monotherapy initiations (in 108 individual patients), and 186 combination therapy initiations (in 184 individual patients).
The median follow up was 19.8 months and across the whole cohort 42% had a relapse, 35% were lost to follow up and 4% had a change of diagnosis.
After adjusting for baseline differences (including demographics, baseline severity, substance use disorder, previous episodes or admissions, previous medication use) the 2-year risk of relapse was:
- 42.3% for mood stabiliser monotherapy,
- 49.5% for SGA monotherapy and
- 42.2% for combination therapy.
Head-to-head comparisons of showed a risk reduction of -7.3% (-15·8 to -1·4) for mood stabilisers compared to SGA monotherapy, and -7.4% (-16·0 to -2·1) for combination therapy compared with SGA monotherapy. Essentially this means that the risk of relapse was very similar between mood stabiliser monotherapy and combination groups, which were both about 7 percentage points lower than the SGA monotherapy group.
The results were similar in the per-protocol analysis which took into account treatment adherence. After adjusting for relevant confounders, the 2-year risk of relapse was 38·9% for mood stabiliser monotherapy, 46·8% for SGA monotherapy and 38·4% for combination therapy. Head-to-head comparisons showed a -7·9% (-13·9 to 0·5) risk reduction for mood stabilisers compared to SGA monotherapy, and -8·4% (-15·4 to -0·8) for combination therapy compared with SGA monotherapy. These results were similar to the intention-to-treat analysis, though the confidence intervals comparing mood stabilisers to SGA monotherapy included zero, which suggests a less significant difference between these groups.
In secondary analysis the risk of manic or mixed relapses was very similar across groups, but the risk of depressive relapse was lower in either the mood stabiliser monotherapy -5·2% (-11·1 to -0·1) or combination therapy -7·0% (-12·1 to -1·9) groups compared to SGA monotherapy.
Conclusions
Therefore, relapse rates were lower in those receiving mood stabilisers alone or in combination with antipsychotics, compared to SGAs alone. This suggests that in first episode mania with psychosis, a treatment strategy should include a mood stabiliser to prevent relapse, either alone or in combination with an SGA.
Strengths and limitations
This cohort study which emulated a ‘target trial’ is a well-designed, important study which contributes to our understanding of how treatments could reduce relapse following first episode psychotic mania. The authors acknowledge that the ideal study to answer this question would be an actual RCT (as opposed to an emulation of one), but there are practical difficulties in terms of the set-up, retention and funding for these. The authors have explicitly defined the target trial, used robust analysis to account for a variety of potential confounders and sensitivity analysis of a range of alternative scenarios, which adds further weight to their findings.
Nonetheless, a challenge with all observational studies is the risk of uncontrolled confounding when a variable that influences both the exposure and the outcome is left unmeasured or unadjusted. Here there may have been factors that could have affected the choice of treatment and were linked to the outcome, despite adjustments in the analysis. There was higher baseline severity in the group started on SGA monotherapy, and also a higher proportion of antidepressants prescribed in this group, though these were adjusted for in the analysis. The authors acknowledge other potential unmeasured confounders, such as personality traits, family history, as well as risk profile which could have affected treatment decisions and outcomes. That said, the study design comparing three groups gives some further weight to their findings, as logically you may expect those with greater risk of relapse or poor initial response to be on combination therapy, which had similar outcomes to the mood stabiliser group, but significantly lower relapse rates than the SGA monotherapy group.
It is worth considering that medications used in the study were grouped as ‘mood stabilisers’ or ‘SGAs’ due to sample size considerations. Lithium was the most commonly used mood stabiliser and the SGAs used were Olanzapine, Risperidone, Aripiprazole and Quetiapine, with Olanzapine and Risperidone used most frequently. There is evidence of differences among antipsychotics in preventing depressive relapses with Quetiapine showing better evidence compared to Risperidone and Aripiprazole (Kishi et al 2021), which could have affected the results.
Finally, the study did not assess adverse effects of medications or explicitly report discontinuation due to adverse effects. Side effects are a major consideration in the choice of prescribing long-term treatments for relapse prevention and therefore make up a considerable part of the consultation and decision-making process in clinical practice.
Implications for practice
This study demonstrates in a real world setting that mood stabilisers, either alone or in combination with SGAs, may reduce relapse rates following first episode psychotic mania compared to SGAs alone. Though the study was not randomised, it was a well-designed cohort study which emulated a target trial. Moreover, in conjunction with existing evidence it gives further weight to the finding that mood stabilisers, in particular lithium, are preferable for preventing relapse in patients with bipolar disorder.
Future studies should confirm these findings with well-designed randomised studies. Ideally different mood stabilisers (e.g. lithium and valproate), and different SGAs should be compared to determine if there are differences with specific medications. Studies that examine adverse effects and discontinuation in a real-world setting would also support clinicians and patients in their decision making.
The study helps to add further evidence to answer the clinical question for many patients with bipolar disorder: when a manic episode has been successfully treated with antipsychotics, in order to prevent relapse, should I stick or twist? Many patients and clinicians appear reluctant to switch, which may have led to the declining use of lithium and increase of SGA use in bipolar disorder. This may be due to concerns about side effects, monitoring requirements, or simply uncertainty as to what extra benefit would be achieved by switching. Patients may often feel they are managing well with their initial SGA, and therefore are reluctant to ‘rock the boat’ or change something that appears to be working.
Szmulewicz et al’s findings suggest that starting with a mood stabiliser may offer a better alternative, particularly for reducing the risk of depressive relapse, which is often the most challenging and disabling mood phase of bipolar disorder for patients. Being able to provide clinicians with this information will hopefully lead to a more informed discussion with patients about whether to ‘stick or twist’.
Statement of interests
Tobias Rowland has no conflicts of interest to declare. All intelligence used in the writing of this blog was strictly non-artificial.
Editor
Simon Bradstreet.
Links
Primary paper
Alejandro Szmulewicz, Yu-Hsin Liang, Gonzalo Martínez-Alés, Ryan O’Dea, Lorna Thorpe, Sarah Conderino, José Rubio, Kerry Smith, Deepak Sarpal, Covadonga Díaz-Caneja, Celso Arango, Paul Vöhringer, Sophia Aldunate, Ridha Joober, Jai Shah, Ashok Malla, Diane Fredrikson, Lakshmi Yatham, Miguel Hernán, Dost Öngür (on behalf of the BD-CAUSAL Collaboration) (2026). Comparative effectiveness of mood stabiliser monotherapy versus second-generation antipsychotic monotherapy versus combination therapy for continuation and maintenance treatment after a first episode of psychotic mania in North America, Chile, and Spain: a target trial emulation using data from the BD-CAUSAL Collaboration. The Lancet Psychiatry, 13, 861-869
Other references
Goodwin GM., Haddad PM., Ferrier IN. et al (2016). Evidence-based guidelines for treating bipolar disorder: Revised third edition recommendations from the British Association for Psychopharmacology. Journal of Psychopharmacology 2016, Vol. 30(6) 495 –553.
Kishi, T., Ikuta, T., Matsuda, Y. et al (2021). Mood stabilizers and/or antipsychotics for bipolar disorder in the maintenance phase: a systematic review and network meta-analysis of randomized controlled trials. Molecular Psychiatry 26, 4146–4157 (2021).
NICE (2014). Bipolar disorder: Assessment and Management | Guidance | NICE. National Institute for Health and Care Excellence.
