Published on 24 September 2026
Bipolar disorder with comorbid obsessive-compulsive disorder: a more severe clinical profile?
European Psychiatry · 2025; 68(1): e142, pp. 1-13 · De Prisco et al.
DOI 10.1192/j.eurpsy.2025.10087
PMID 40855518
Scientific 68
Editorial 78
The essentials
A systematic review with meta-analysis pooled 26 observational studies covering 6,678 patients with bipolar disorder, of whom 1,095 had comorbid obsessive-compulsive disorder and 5,583 did not. Twenty-two of these studies could be aggregated, and all the principal estimates concern adult populations. In adults, patients with comorbid OCD report more lifetime suicide attempts (OR 1.85, 95% CI 1.21 to 2.84, p = 0.005, 12 studies) and show more impaired global functioning, measured with the GAF (SMD -0.42, 95% CI -0.59 to -0.24, p < 0.001, 4 studies). The other differences point in the same direction: chronic episodes, rapid cycling, panic disorder, eating disorder, substance use disorder, earlier age of onset. Mood symptom severity, by contrast, does not differ. In children and adolescents, no significant difference emerges.
What this work cannot say matters just as much. Twenty-two of the twenty-six studies are cross-sectional, and thirteen are judged to be of poor methodological quality. Above all, the sensitivity analysis restricted to good-quality studies alone drops most results below the significance threshold, including suicide attempts. What is established is an association between two clinical pictures, measured most often at the same time in the same patients. Nothing in this material says whether obsessive comorbidity worsens bipolar disorder, whether it signals a form that is more severe from the outset, or whether the two are mostly found together in patients who seek care more often and are therefore better identified.
Context
The comorbidity between bipolar disorder and obsessive-compulsive disorder is a long-standing clinical difficulty, and it is first of all a diagnostic one. A depressive rumination and an obsession resemble each other from a distance. Repeated checking can pass for anticipatory anxiety, or disappear behind a louder manic picture. When the patient already carries a bipolar label, the second condition is often read as a mode of expression of the first, rather than as an entity with its own criteria, its own scale, and its own management.
The stakes are not only nosological. A bipolar patient whose obsessive-compulsive disorder has gone unrecognized is a patient whose true symptomatic burden is not assessed, whose avoidance behaviors go unexplained, and who is sometimes offered a mood-focused intensification where the complaint stems from something else. Conversely, treating the obsessive dimension in a bipolar patient immediately raises the question of mood safety, which makes recognition useful before any prescribing decision.
Hence the interest of the question raised here, which is not one of prevalence but of prognosis: when the two disorders coexist, is the clinical picture different, and along which dimensions? Several studies have addressed the topic, with often modest sample sizes and scattered results. Pooling this data quantitatively was therefore legitimate, provided one does not forget what this data actually is: observations, not experiments.
The study at a glance
Population, comparison, criteria
| Population | |
| Patients with bipolar disorder, with or without comorbid obsessive-compulsive disorder, diagnosed under DSM or ICD criteria. No age or language restriction was applied to the search. Of the 26 studies, 20 concern adults, 4 concern children and adolescents, and 2 concern mixed samples. Analyses were conducted separately for adults and for children or adolescents. Patients with the comorbidity averaged 34.65 years (SD 9.6) and were 54.8% women, versus 40.31 years (SD 7.84) and 57.8% women in the group without the comorbidity. | |
| Exposure studied | |
| Presence of comorbid obsessive-compulsive disorder, not an intervention. No treatment is assigned or compared in this work. The authors further indicate they were unable to control their results for treatments received, for lack of sufficient data in the included studies. | |
| Comparator | |
| Patients with bipolar disorder without obsessive-compulsive disorder, N = 5,583, study sizes ranging from 15 to 1,613. | |
| Criteria analyzed | |
| Course of bipolar disorder (chronic episodes, rapid cycling, polarity of the first episode, psychotic features), number and duration of episodes, number of hospitalizations, age of onset and duration of bipolar disorder, associated psychiatric comorbidities, suicidality, symptom severity, and global functioning. | |
| Design | |
| Systematic review with meta-analysis of observational studies: 22 cross-sectional and 4 prospective cohorts. Search conducted in PubMed/MEDLINE, Scopus, PsycINFO, and Web of Science through April 15, 2024, across 11,959 initial records. Protocol registered with PROSPERO under number CRD42024536387, report compliant with PRISMA. Quality assessed with the Newcastle-Ottawa scale, whose scores were then converted into Agency for Healthcare Research and Quality categories. Random-effects models, restricted maximum likelihood estimator, metafor package under R 4.3.1. | |
| Sample sizes | |
| 26 studies included in the qualitative synthesis, 6,678 patients in total, including 1,095 with comorbid obsessive-compulsive disorder, study sizes in that group ranging from 6 to 201. 22 studies provide sufficient data for the meta-analysis, that is, 19 of the 22 cross-sectional studies and 3 of the 4 longitudinal studies. The four studies excluded from the pooled analyses are excluded for sample overlap with a more recent and larger study, or for non-poolable measures. | |
| Measurement instruments | |
| The only instruments named in the publication and its supplementary material are the GAF for global functioning and the CGI for overall clinical severity. The scales used for depressive severity, manic severity, and obsessive symptomatology are not detailed study by study, which leaves this part of the heterogeneity undocumented. |
Quality control
What was done, and what remains open
| Point checked | Verdict |
|---|---|
| Prospective protocol registration | Verified, with a declared deviation |
| FindingPROSPERO number CRD42024536387. A deviation is declared in the appendices: meta-regressions initially planned for any heterogeneous comparison were ultimately restricted to comparisons with at least ten studies, in keeping with the Cochrane handbook. Declaring this deviation is a positive point; it does mean fewer exploratory analyses are available. | |
| PRISMA reporting compliance | Declared |
| FindingThe declaration concerns how the review is reported, not the quality of the studies gathered. The PRISMA checklist is provided as an appendix. | |
| Quality assessment tool | Newcastle-Ottawa converted to AHRQ |
| FindingThe Newcastle-Ottawa scale served as the measure, with two independent raters and a third in case of disagreement, and the scores were then converted into Agency for Healthcare Research and Quality categories. This conversion is common practice; it does not make the tool more discriminating than it already is. | |
| Quality of the included studies | Thirteen of twenty-six studies of poor quality |
| FindingTwelve studies are rated good quality, one intermediate, and thirteen poor quality. Half of the pooled material is therefore fragile, which weighs on every pooled estimate. Item-level detail appears in the appendices: no study earns the sample-size point, and between-group comparability is the domain most often failed. | |
| Design of the pooled studies | Cross-sectional for twenty-two of twenty-six studies |
| FindingOnly four studies are prospective cohorts, and for these the authors preferentially retained baseline data rather than follow-up. In other words, even the longitudinal material is exploited cross-sectionally. This does not change the nature of the comparison available. | |
| Meta-regressions | Four models, no significant predictor |
| FindingOnly two criteria gathered the ten studies required: age of onset and suicide attempts in adults. The two predictors tested are mean age and proportion of women. Neither reaches the threshold, the closest being proportion of women for age of onset (beta -1.29, 95% CI -2.61 to 0.02, p = 0.05). Type of bipolar disorder, mood state, and treatments could not be tested. | |
| Sensitivity analyses | Performed, fragile results |
| FindingTwo approaches, leave-one-out and restriction to good-quality studies only. The second is the most informative and the most severe: restricted to good studies, the association with suicide attempts is no longer significant (OR 1.33, 95% CI 0.74 to 2.39, p = 0.34, 5 studies), nor are rapid cycling, panic disorder, substance use disorder, or age of onset. Only global functioning and chronic episodes survive it. Eating disorder cannot even be re-estimated, for lack of a good-quality study contributing to that criterion. | |
| Heterogeneity across studies | High on several criteria |
| FindingIt varies markedly by criterion: I² near zero for global functioning, chronic episodes, rapid cycling, and eating disorder, 11% for substance use disorder, 42% for panic disorder, 70% for suicide attempts, and 83% for age of onset. Prediction intervals, more honest than confidence intervals for anticipating a future study, include no effect for suicide attempts (0.55 to 6.21), substance use disorder (0.89 to 2.16), and age of onset (-1.08 to 0.54). | |
| Publication bias | Explored on two comparisons only |
| FindingVisual inspection of funnel plots and Egger’s test, applied only to comparisons pooling at least ten studies, that is, two of them. A significant bias is detected for age of onset of bipolar disorder in adults (z = -1.964, p = 0.049), not for suicide attempts (z = 0.923, p = 0.36). The authors nonetheless conclude to an overall minimal publication bias, which is an overreach, since the vast majority of comparisons were never tested. | |
| Certainty level of the evidence | Not assessed |
| FindingThe PRISMA checklist provided by the authors marks the two items on assessment of certainty of evidence as not applicable. No GRADE-type rating was therefore conducted, either for the method or for the results. The reader must judge for themselves how much confidence to place in each estimate. | |
| Declared use of an artificial intelligence tool | Declared by the authors |
| FindingThe authors state they used ChatGPT to improve the readability and language of the manuscript, that they then reviewed and corrected the content, and that they take full responsibility for the publication. This is reported here because it falls under the transparency expected, not because it says anything about the validity of the analyses. | |
| Funding and conflicts of interest | No dedicated funding, relationships disclosed |
| FindingThe research received no dedicated grant, whether commercial, institutional, or charitable. Individual institutional support, all public or academic, is mentioned in the acknowledgments. On conflicts of interest, two authors disclose relationships with the pharmaceutical industry, including the corresponding author, who lists about thirty companies for grants, consulting, or speaking, outside the submitted work. The other authors disclose no relationships. | |
Results
One estimate that holds, several signals that do not survive testing
Global functioning measured with the GAF in bipolar adults with comorbid obsessive-compulsive disorder, compared with bipolar adults without this comorbidity, 95% CI -0.59 to -0.24, p < 0.001, 4 studies, heterogeneity near zero.
Only one result comes through every robustness check unscathed: the impairment of global functioning. The standardized mean difference is -0.42, of moderate size, with the narrowest interval in the series, zero heterogeneity across the four contributing studies, and a value that holds when only good-quality studies are retained (SMD -0.46, 95% CI -0.64 to -0.27).
The most discussed result, the excess of suicide attempts, is more fragile than it appears. The odds ratio of 1.85 withstands the removal of any single one of the twelve studies, but its heterogeneity is high, its prediction interval includes no effect, and the association disappears when the analysis is restricted to the five good-quality studies. The other estimates should be read with the same reservation, compounded by the small number of studies behind them. The values below all concern adult populations.
| Criterion | Value reported |
|---|---|
| Global functioning, GAF | SMD -0.42, 95% CI -0.59 to -0.24, p < 0.001, 4 studies, 170 versus 749 patients. |
| ReadingModerate effect size, zero heterogeneity, result retained in the analysis restricted to good-quality studies. This is the most solid estimate in the whole set. Functioning is still measured by a global scale, sensitive to the timing of assessment. | |
| Lifetime suicide attempts | OR 1.85, 95% CI 1.21 to 2.84, p = 0.005, 12 studies, 568 versus 3,494 patients. |
| ReadingThe clinically most striking result, but not the most robust. I² of 70%, prediction interval of 0.55 to 6.21, and loss of significance when only the five good-quality studies are retained (OR 1.33, 95% CI 0.74 to 2.39). It describes a difference in frequency between two observed groups, not an excess risk produced by the comorbidity. | |
| Rapid cycling | OR 1.92, 95% CI 1.04 to 3.53, p = 0.037, 4 studies. |
| ReadingDirection consistent with the rest, but a lower bound sitting right against unity. Removing either one of two of the four studies is enough to erase significance, and the analysis restricted to good-quality studies gives a nonsignificant OR of 1.62. | |
| Associated panic disorder | OR 3.30, 95% CI 2.11 to 5.16, p < 0.001, 8 studies. |
| ReadingEmotional comorbidity, an estimate resting on the largest number of studies after suicidality and age of onset, intermediate heterogeneity (I² 42%). It drops, however, to 2.03, 95% CI 0.98 to 4.20, when only the four good-quality studies are retained. | |
| Associated eating disorder | OR 3.37, 95% CI 1.99 to 5.71, p < 0.001, 2 studies. |
| ReadingApparently large magnitude, but only two studies, neither of good quality, so this criterion is absent from the sensitivity analysis. The authors also note that eating disorders are treated as a single category, which rules out any reading by diagnosis. | |
| Substance use disorder | OR 1.39, 95% CI 1.02 to 1.89, p = 0.038, 9 studies. |
| ReadingA more modest gap than the preceding ones, lower bound at 1.02, prediction interval of 0.89 to 2.16 that includes no effect. The association does not hold in the analysis restricted to good-quality studies. | |
| Age of onset of bipolar disorder | SMD -0.27, 95% CI -0.52 to -0.01, p = 0.042, 12 studies, favoring earlier onset. |
| ReadingThe most fragile criterion on the list despite its large number of studies: upper bound at -0.01, heterogeneity of 83%, prediction interval crossing zero, publication bias detected, loss of significance on removal of several studies taken individually and in the analysis restricted to good-quality studies. To be treated as a lead, not as an established finding. | |
| Chronic mood episodes | OR 9.42, 95% CI 2.23 to 39.89, p = 0.002, 2 studies, 50 versus 50 patients. |
| ReadingThe interval spans nearly a twentyfold factor, across a hundred patients in total. The two contributing studies defined a chronic episode as criteria for a major mood episode met continuously for at least two years. An estimate drawn from two studies with this much dispersion does not transfer to the bedside, and the point value in particular should not be retained on its own. | |
| Symptom severity and episode polarity | No significant difference: CGI, depressive severity, manic severity, psychotic features, first-episode polarity. |
| ReadingThese comparisons were indeed conducted, on three to six studies each, and show no gap. The absence of a reported difference is not proof of equivalence: the intervals remain wide and heterogeneity is high on several of these criteria, notably the CGI (I² 88%) and depressive severity (I² 80%). | |
| Children and adolescents | No significant difference on eight comparisons, each based on 2 studies. |
| ReadingThe pediatric stratum rests on 105 patients with the comorbidity and 237 without. The estimate closest to threshold concerns tics (OR 3.32, 95% CI 1.00 to 11.03, p = 0.05). Here, the absence of a signal reflects above all a lack of data, not a demonstrated equivalence between the two groups. | |
Critical appraisal
The point that governs all the others
Design is the central limitation, and it cannot be worked around. When twenty-two of twenty-six studies measure exposure and outcome at the same time, no temporal sequence is available. Obsessive comorbidity may precede, accompany, or follow the worsening of bipolar disorder, and the data cannot settle the question. The four longitudinal studies do not fill this gap, especially since their baseline data were favored over their follow-up. The appendices also show that the sequence of onset is reported in only twelve of the twenty-six studies, and in detail in only seven, with opposite patterns from one study to the next, bipolar disorder always first for some authors, obsessive-compulsive disorder always first for others. Added to this limitation is the quality of the material, judged poor for thirteen studies, and above all the behavior of the estimates once those thirteen studies are removed.
| Domain | Risk of bias |
|---|---|
| Causal inference | Impossible by design |
| FindingThe direction of the relationship between obsessive comorbidity and bipolar severity is not established, and cannot be established with this type of data. The authors acknowledge this explicitly in their limitations. | |
| Methodological quality of the material | Poor for 13 of 26 studies |
| FindingThe pooled analyses rest for half on work whose conduct is insufficient, and most significant associations do not survive their removal. This calls for retaining the direction of the gaps, not their exact value. | |
| Measurement of exposure and outcomes | Poorly documented instruments |
| FindingOnly the GAF and the CGI are named. The scales used for mood severity and for obsessive symptomatology are not detailed study by study, even though the studies span nearly thirty years of practice and five successive versions of the diagnostic criteria, from DSM-III to DSM-5. The authors themselves note that OCD’s move out of the anxiety disorder category in DSM-5 may have changed how the diagnosis was made. | |
| Precision and robustness of the estimates | Only one criterion withstands the sensitivity analyses |
| FindingGlobal functioning holds up on every table. Chronic episodes retain significance but with an unusable interval. Suicide attempts, rapid cycling, panic disorder, substance use disorder, and age of onset lose significance as soon as only good-quality studies are retained. Eating disorder cannot be tested at all. | |
| Publication bias | Tested on two criteria, positive on one |
| FindingPositive for age of onset, negative for suicide attempts, not tested elsewhere. Such a bias usually pushes the pooled estimate away from no effect, which suggests treating the reported gap for age of onset as an upper bound. Concluding to an overall minimal bias, as the authors do, goes beyond what two tests allow. | |
| Fit between data and conclusions | Conclusion firmer than the data |
| FindingThe authors are clear about the impossibility of drawing causal conclusions, and they distinguish adults from children. But the abstract announces a more severe and complex clinical profile without mentioning that the sensitivity analyses erase most of the associations, and the discussion uses action-oriented language, of the kind synergistic impact on functioning, that the design does not support. No assessment of certainty of evidence accompanies these conclusions. | |
| Transparency of conduct | Protocol registered, data accessible |
| FindingPROSPERO registration with a declared deviation, PRISMA report, absence of dedicated funding, disclosed conflicts of interest and use of an assisted-writing tool, forest and funnel plots deposited on an open archive, dataset available on request. This is the strong point of the work. | |
| External validity | No French cohort |
| FindingThe studies come from thirteen countries, mainly the United States, Italy, Turkey, India, and Brazil, with one German-Dutch-American multicenter study. No French cohort. The recruitment setting, inpatient or outpatient, is not reported in the descriptive table, which prevents assessing the degree of sample selection. | |
Level of evidence
The work falls under level 2a of the Oxford Centre for Evidence-Based Medicine classification, that is, a systematic review of non-randomized studies. The authors conducted no certainty-of-evidence rating, so the appraisal below is the review’s own. Confidence is reasonable on one specific point: among the bipolar adults followed in these studies, the presence of obsessive-compulsive disorder is accompanied by more impaired global functioning. This is a descriptive finding, resting on four concordant studies with no heterogeneity, and it holds when the weakest studies are set aside.
Confidence is low on everything else. First, causality, ruled out by design. Then the excess of suicide attempts, whose direction is consistent from study to study but whose statistical significance disappears when only good-quality studies are considered: the signal deserves to be taken seriously in consultation, but it cannot be presented as demonstrated. Then the magnitude of the gaps, which the uneven quality of the included studies does not allow to be pinned down. Finally, the criteria carried by only two studies, chronic episodes and eating disorder, whose point values carry no practical weight. For mood symptom severity and for the pediatric population, the observed absence of difference above all reflects the scarcity of data; it does not demonstrate that the two groups behave the same way.
The colleague test
What an experienced colleague might say about this study in two minutes, between two consultations.
“What I take from this is that if a bipolar patient also has OCD, I really assess their day-to-day functioning, not just their episodes, and I look more closely at suicide risk. That said, these are cross-sectional studies, I don’t know what comes first, half of the studies don’t hold up methodologically, and once you remove those, almost nothing significant is left apart from functioning. That doesn’t stop me from looking for the OCD. It stops me from turning it into an explanation. And if I need to treat the obsessive side, I make sure the mood is covered first.”
Translated for practice: obsessive comorbidity serves as a signal of complexity, not an explanatory factor. It justifies intensifying the assessment, not reattributing the severity of the bipolar disorder to a new cause.
What you can do with this
- Actively screen for obsessive-compulsive disorder in the workup of a bipolar patient, even when the mood picture dominates the whole clinical field. An open question about checking, ritualized behaviors, and the time they consume costs only a few minutes.
- When the comorbidity is present, document global functioning, the only dimension for which the observed gap withstands every sensitivity analysis in this review. A global functioning scale is enough to provide a measurable starting point rather than an impression.
- Maintain careful assessment of suicide risk, keeping in mind that the signal is consistent in direction but does not reach statistical significance when the analysis is restricted to good-quality studies. This is a reason for clinical vigilance, not an established fact.
- Do not transpose as-is the values carried by only two studies, chronic episodes and eating disorder. Their direction is informative, their magnitude is not.
- Before introducing a serotonin reuptake inhibitor for the obsessive dimension in a bipolar patient, make sure mood coverage is in place first. This is a standard precaution in practice, and it does not follow from this publication, which evaluated no treatment. As of September 1, 2026, adult obsessive-compulsive disorder carries, in France, marketing authorization for clomipramine, fluvoxamine, fluoxetine, paroxetine, sertraline, and escitalopram, with sertraline alone carrying an explicit extension to ages 6 to 17; readers practicing under another jurisdiction should check their own regulator’s authorized indications, since these lists commonly differ from one country to the next. What no such authorization addresses, anywhere, is the safety of these agents on the mood axis in a bipolar patient: an approval granted for OCD alone says nothing about the risk of a manic or hypomanic switch in someone who also carries a bipolar diagnosis, and that is a separate question requiring its own guidance. No dedicated national guideline covers it in France. The closest reference remains the report of the International Society for Bipolar Disorders task force, which recommends avoiding antidepressant monotherapy in type I bipolar disorder and, when an antidepressant is used, combining it with a mood stabilizer, while stating plainly that the protective effect of that combination is not itself demonstrated. The British NICE CG185 guideline does not impose this same restriction, and limits itself to a discontinuation rule in case of manic emergence. This divergence between two major reference frameworks is worth knowing rather than smoothing over.
- Know how to answer a patient who asks whether their OCD is worsening their bipolar disorder: the two conditions often occur together, the combined picture looks on average heavier, and the direction of causality is not known. This answer is more accurate than either of the two clear-cut answers on offer.
Frequently asked questions
Does this meta-analysis show that OCD worsens bipolar disorder?
No. Twenty-two of the twenty-six studies are cross-sectional, meaning the comorbidity and the outcome are measured at the same time. The work establishes an association, it does not fix a direction for the relationship. A bipolar disorder that is more severe from the outset could just as well favor the expression of obsessive symptoms, or the two conditions could share common determinants. The authors themselves raise a third reading: a patient already followed for OCD is in contact with care, and therefore more likely to be diagnosed early for their bipolar disorder.
Is the odds ratio of 1.85 for suicide attempts usable in consultation?
It is usable as a signal of vigilance, not as an individual risk calculation, and it is less solid than its presentation suggests. It rests on twelve studies, seven of which are not of good quality, its heterogeneity is high, and the analysis restricted to the five good-quality studies brings the odds ratio down to 1.33 with a confidence interval that includes unity. The direction remains consistent from study to study, which justifies raising the level of suicide risk assessment by a notch, without allowing a figure to be announced to any given patient.
Why isn’t the odds ratio of 9.42 for chronic episodes emphasized?
Because it rests on two studies, fifty patients in each group, and its confidence interval runs from 2.23 to 39.89. An estimate whose upper bound is nearly twenty times its lower bound tells us almost nothing about the true magnitude. It indicates a direction, nothing more, and that is exactly the kind of figure a hurried reader latches onto, when they shouldn’t.
What should we make of the authors’ declared use of ChatGPT?
The authors state they used it to improve the readability and language of the manuscript, then reviewed and corrected the content, and take full responsibility for the publication. This is a transparency disclosure, now required by most journals, and it concerns the writing, not the literature search, the data extraction, or the statistical analyses. It does not change the assessment of the work’s validity, which rests on the design and on the quality of the studies gathered.
Should OCD be treated in a stabilized bipolar patient?
This publication does not settle the question: it evaluated no treatment and could not even control its results for treatments received, for lack of data. It is decided case by case, based on the functional impact of the obsessions and compulsions, the patient’s current mood stability, and the coverage already in place. The usual precautionary rule is that mood stability should be secured before introducing a serotonergic antidepressant in this context.
Annotated bibliography
Source study. De Prisco M, Tapoi C, Oliva V, Strumila R, Takami C, Girone N, Macellaro M, de Salles Andrade JB, Schmitz CN, Vieta E, Fico G. Clinical impact of obsessive-compulsive disorder comorbidity in bipolar disorder: a systematic review and meta-analysis. European Psychiatry, 2025, volume 68, issue 1, article e142, pages 1 to 13. DOI 10.1192/j.eurpsy.2025.10087. PMID 40855518. Received June 4, 2025, revised July 16, 2025, accepted July 19, 2025. Open-access article under a Creative Commons Attribution 4.0 license. Systematic review with meta-analysis of 26 observational studies published between 1995 and 2023, 6,678 patients, search through April 15, 2024, protocol registered with PROSPERO under number CRD42024536387, with a deviation declared in an appendix. Funding: the research received no dedicated grant from a funding agency, the commercial sector, or the charitable sector. Individual institutional support, all public or academic, is mentioned in the acknowledgments. Conflicts of interest: the corresponding author discloses grants and consulting, advisory, or speaking activities for about thirty pharmaceutical companies, outside the submitted work, and another author discloses training or consulting fees for three of them. The other authors disclose no relationships. Use of ChatGPT to improve the readability and language of the manuscript is declared. The authors present this work as the first systematic review with meta-analysis devoted to the clinical impact of OCD comorbidity in bipolar disorder, to their knowledge; their earlier meta-analysis, published in 2024, compared the comorbid picture to isolated obsessive-compulsive disorder rather than to isolated bipolar disorder. Its limitation lies in the nature of the pooled material, predominantly cross-sectional and of uneven quality, and in the fact that most significant associations do not survive restriction to good-quality studies. No supplementary material was missing: the nine appendix tables, including the sensitivity and publication-bias analyses, were consulted.
Context reference. Pacchiarotti I, Bond DJ, Baldessarini RJ, et al. The International Society for Bipolar Disorders (ISBD) Task Force Report on Antidepressant Use in Bipolar Disorders. American Journal of Psychiatry, 2013, volume 170, issue 11, pages 1249 to 1262. DOI 10.1176/appi.ajp.2013.13020185. PMID 24030475. Consensus report from an international task force, cited here for the antidepressant prescribing framework in bipolar patients, a question the analyzed publication does not address. This text is an expert consensus, not a quantitative synthesis of the evidence, and it states itself that the protective effect of the mood stabilizer against switch is not demonstrated. It has not been updated since 2013.
Editorial collections
Tags
Verified on September 1, 2026 against the full text of the publication and its supplementary material where available. This analysis underwent an independent double reading. The English version was checked for conformity on September 23, 2026, against the figures of the French version and against the source. How we verify what we publish
This analysis is intended for healthcare professionals. It does not constitute a prescribing recommendation and does not replace individual clinical judgment.
Analysis from Psychiatry Evidence Base, evidence-based psychiatry, explained with rigor.