Published on 19 September 2026
Does bipolar disorder change the symptom profile of obsessive-compulsive disorder?
The essentials
Seventeen studies, 3,844 patients, including 682 with both obsessive-compulsive disorder and bipolar disorder, compared with 3,162 patients with obsessive-compulsive disorder alone. Fifteen studies could be meta-analyzed. Two differences emerge out of twenty-one comparisons: fewer contamination obsessions among comorbid patients (odds ratio 0.71, 95% CI 0.53 to 0.95, p = 0.021) and more sexual obsessions (1.77, 1.03 to 3.04, p = 0.04). The second finding survives neither a leave-one-out analysis nor restriction to the higher-quality studies. Overall severity does not differ, but this finding is itself fragile. Across the three severity scores, heterogeneity ranges from 88.9% to 93.7%, and an asymmetry test suggests publication bias for total symptom severity (p = 0.009). The article therefore describes a phenomenological lead, not a diagnostic marker.
Context
The co-occurrence of obsessive-compulsive disorder and bipolar disorder in the same patient is not rare, and it complicates two decisions: what to treat first, and with what. Earlier reviews focused mainly on the prevalence of this association and its predictors, without describing the clinical features that define it. The question asked here is different and more precise: does the presence of bipolar disorder come with a distinct symptom profile of obsessive-compulsive disorder?
The question has practical relevance. Certain obsessional themes are rarely reported spontaneously, sexual obsessions in particular, and they can be mistaken for other clinical phenomena, notably the manifestations of a manic or hypomanic episode. Knowing where to look changes how the interview is conducted.
The study at a glance
| Element | |
|---|---|
| Population | |
| Detail3,844 patients with obsessive-compulsive disorder, of whom 682 also have bipolar disorder. Seventeen studies included in the review, fifteen in the meta-analysis. Thirteen studies were in adults, three in children or adolescents, one in both: the population is not exclusively adult. | |
| Exposure studied | |
| DetailThe presence of bipolar comorbidity. The perspective is that of obsessive-compulsive disorder, not the reverse: the reference group consists of obsessive-compulsive patients without bipolar disorder. | |
| Comparator | |
| Detail3,162 patients with obsessive-compulsive disorder without bipolar comorbidity. | |
| Outcomes | |
| DetailLifetime frequency of obsessional and compulsive themes, number of obsessions and compulsions, overall severity of obsessive-compulsive disorder. The instruments used differ from one study to another, which is itself a source of heterogeneity. | |
| Design | |
| DetailSystematic review and meta-analysis of observational studies: fourteen cross-sectional studies and three prospective cohorts, no controlled trial. Registered protocol, compliance with PRISMA 2020 guidelines, random-effects model, quality assessment of the studies, meta-regressions and sensitivity analyses. |
Quality control
| Point checked | Judgment |
|---|---|
| Protocol registration | Present |
| FindingProtocol registered on PROSPERO under identifier CRD42023392296. The authors report three deviations from the initial protocol: the addition of a sensitivity analysis restricted to higher-quality studies, an extension of the meta-regressions beyond the originally planned rule of ten studies, and a change in the heterogeneity detection threshold from p = 0.05 to p = 0.10. | |
| Nature of the data | Observational |
| FindingFourteen cross-sectional studies and three prospective cohorts. A difference in profile between two groups does not mean the comorbidity produces that profile: it may reflect differences in recruitment, symptom recognition or illness duration. The authors acknowledge this explicitly. | |
| Heterogeneity | 88.9% to 93.7% |
| FindingAcross the three severity scores, the dispersion between studies is such that the pooled estimate loses clinical meaning. It is lower elsewhere: 86.6% for the number of obsessions, 61.6% for the number of compulsions, and null for several themes. The meta-regressions conducted to identify its source are exploratory, not planned in this form in the protocol, and not corrected for multiplicity. | |
| Publication bias | Reported, partial screening |
| FindingEgger’s test was conducted only on the three comparisons pooling at least ten studies. It is significant for total obsessive-compulsive symptomatology (z = −2.5854, p = 0.009) and negative for aggressive obsessions and washing compulsions. The two significant findings of the review, contamination and sexual themes, were not tested for lack of a sufficient number of studies: on these two points, there is no absence of bias, there is an absence of analysis. | |
| Robustness of the most discussed result | Not maintained |
| FindingThe excess of sexual obsessions disappears with the removal of almost any of the included studies, with two exceptions, and it also disappears when the analysis is restricted to methodologically higher-quality studies (odds ratio 2.17, 0.80 to 5.87, p = 0.127). Two tests, both negative. | |
| Funding and conflicts of interest | No funding, conflicts declared |
| FindingThe authors declare no funding source for this work. Two of them report extensive conflicts of interest with the pharmaceutical industry, as consultants, advisory board members or speakers, and one holds a grant from the La Caixa Foundation. The other nine authors declare no conflicts of interest to report. Details appear in the annotated bibliography. | |
Results
| Result | Value |
|---|---|
| Contamination obsessions | Odds ratio 0.71 (0.53 to 0.95), p = 0.021 |
| ReadingLess frequent among comorbid patients, across nine studies. The upper bound of the interval approaches 1, so the difference is modest and the estimate imprecise. The result stops being significant when either of two studies is removed, but it holds and strengthens when the analysis is restricted to higher-quality studies (0.66, 0.48 to 0.89, p = 0.007), where a lower frequency of washing compulsions is also seen (0.59, 0.37 to 0.93, p = 0.022). | |
| Sexual obsessions | Odds ratio 1.77 (1.03 to 3.04), p = 0.04 |
| ReadingThe finding that draws attention, and the one that does not hold up. The interval starts barely above 1, the significance threshold is crossed narrowly, heterogeneity reaches 57.6%, the effect disappears when almost any study is removed, and it vanishes when the weaker studies are excluded. It cannot be cited without this caveat. | |
| Other obsessional and compulsive themes | Nineteen non-significant comparisons |
| ReadingNo difference detected on the other dimensions. For several of them the intervals are very wide, for example 0.46 to 3.29 for repeating rituals: this is imprecision, not demonstrated equivalence. Given the heterogeneity and the size of the comorbid group, nearly five times smaller than the reference group, this does not permit the claim that these dimensions are distributed identically. | |
| Overall severity of obsessive-compulsive disorder | Standardized mean difference −0.305 (−0.663 to 0.053), p = 0.094 |
| ReadingNo significant difference between the two groups, across twelve studies. The finding is, however, less solid than it appears: heterogeneity reaches 90.4%, this is the outcome on which the asymmetry test is significant, and removing a single study is enough to make the difference significant (−0.375, −0.74 to −0.01, p = 0.044). What the comorbidity might change is the shape rather than the intensity of the symptomatology, but this reading remains a hypothesis. | |
Critical appraisal
| Domain | Judgment |
|---|---|
| Conduct of the review | Rigorous |
| FindingFour databases searched up to August 7, 2023, 10,393 references identified, selection and extraction performed in duplicate with arbitration, quality of the studies assessed with a validated scale, sensitivity analyses conducted and reported. The review does what is needed to show the weaknesses of its own results. | |
| Level of evidence of the material | Observational |
| FindingA meta-analysis cannot exceed the quality of what it pools. Stacking cross-sectional studies produces a more precise estimate, not evidence of a different order. The authors themselves state that the cross-sectional design of most studies rules out any causal conclusion. | |
| Definition of the groups | Variable |
| FindingDiagnoses were meant to rely on DSM or ICD criteria, but five of the included studies did not use a structured or semi-structured interview. Fourteen studies report the type of bipolar disorder, with 38.5% type I and considerable dispersion between studies, from 2% to 82%. Nine studies document mood state at the time of assessment: 88.9% of patients were euthymic. The results therefore mainly apply to patients in mood remission. | |
| Multiplicity of comparisons | No correction reported |
| FindingTwenty-one meta-analyses are reported, together with numerous meta-regressions, nine significant associations from which are highlighted. No correction for multiple comparisons is reported. Two significant results out of twenty-one, one of them at p equal to 0.04, should be read in this context. | |
| Adequacy of the conclusions | Late nuance |
| FindingThe discussion is cautious and the authors’ conclusion is measured, but the fragility of the sexual-obsessions finding deserves to appear as soon as the result is stated, otherwise it is the short version that circulates. | |
| Relevance to practice | Indicative |
| FindingNone of these differences has the precision required to guide an individual diagnosis. Their value is to suggest where to focus attention during the interview. | |
Level of evidence
PEB assessment: reasonable confidence in the conduct of the review, low confidence in each of its results. The review applies the rules of systematic reviewing, registers its protocol, discloses its deviations from that protocol, and publishes the analyses that weaken its own conclusions.
Confidence is low on each result. For the profile differences, heterogeneity makes the pooled estimates difficult to interpret, neither of the two significant differences could be tested for publication bias, and the most widely cited result disappears as soon as one study is removed or the analysis is restricted to the strongest studies. For the absence of a difference in overall severity, caution is equally warranted: heterogeneity reaches 90.4%, this is the only outcome for which publication bias is documented, and removing a single study is enough to produce a significant difference. What is suggested is that bipolar comorbidity comes with a somewhat different distribution of obsessional themes, the lower frequency of contamination obsessions being the most consistent signal. What is not established is the existence of an excess of sexual obsessions. What is not shown, and could not have been shown with this material, is that bipolar disorder causes these differences.
The colleague test
What an experienced colleague would say if shown this study in two minutes, between two consultations.
The result everyone will be talking about falls apart as soon as you remove one study or keep only the good ones, and heterogeneity exceeds 88% on the severity scores. I’m not going to change how I examine a patient because of this. What I take away is that you have to ask about sexual obsessions, something we forget to do, and that reason doesn’t need a meta-analysis.
Translation for practice: the article provides no usable criterion for distinguishing one patient from another. It does, however, serve as a reminder that part of the obsessional symptomatology stays unspoken unless it is explicitly asked about.
What you can do with this
- Systematically explore obsessional themes that are rarely reported spontaneously, including sexual or aggressive obsessions, in every patient with obsessive-compulsive disorder. The justification is clinical and does not depend on how solid this particular finding is.
- Do not use a symptom profile to infer bipolar comorbidity. None of the differences reported here has the precision needed for that purpose, and overall severity does not distinguish the two groups.
- Distinguish, during the interview, a sexual obsession experienced as intrusive and distressing from a change in sexual behavior occurring during a manic or hypomanic episode. The two can be confused, and their therapeutic implications differ.
- Keep in mind that the data come overwhelmingly from patients assessed while euthymic, and that they concern lifetime-reported symptoms, with no indication of their current presence or course.
- On the choice of treatment in an obsessive-compulsive patient with bipolar disorder, this study contributes nothing: it measures neither response nor tolerability. The usual precautions regarding the use of an antidepressant without a mood stabilizer come from other sources, to be checked and dated before informing a patient.
- In girls, adolescents and women of childbearing potential, valproate carries a well-documented teratogenic risk. In many countries, prescribing conditions for valproate in women of childbearing potential are strictly regulated and evolve over time. This point does not come from this publication: check the regulatory texts in force in your own country at the time of prescribing, since these rules differ between countries and change over time.
Frequently asked questions
Should we conclude that sexual obsessions are more frequent when bipolar disorder is also present?
Not as things stand. The observed difference is modest, it crosses the significance threshold only narrowly, it disappears when almost any of the included studies is removed, and it vanishes when the analysis is restricted to higher-quality studies. This is the typical behavior of a finding that still needs confirmation.
What does heterogeneity of 88.9% to 93.7% mean?
That nearly all of the variability observed between study results reflects real differences between those studies rather than sampling chance. The populations, instruments and clinical contexts differ too much for a single average to summarize them. The results are then read as an order of magnitude rather than a precise measurement.
Does a significant asymmetry test prove publication bias?
It strongly suggests it, without proving it. Asymmetry can also reflect genuine heterogeneity or quality differences between small and large studies. Its presence invites the assumption that the pooled effect is probably overestimated. It should be added that the test could only be conducted on three comparisons, those pooling at least ten studies: elsewhere, bias was not ruled out, it was simply not tested for.
Why doesn’t overall severity differ?
The pooled estimate does not cross the significance threshold, which suggests that comorbidity might change the shape of the symptomatology rather than its intensity. This finding, however, is the least stable in the whole study: heterogeneity reaches 90.4%, publication bias is documented for this outcome, and removing a single study produces a significant difference. It cannot be presented as a robust result.
Does this evidence change patient management?
No. Nothing in this study concerns treatment, therapeutic response or outcome. Its contribution is descriptive, and it can guide the clinical interview, nothing more.
Annotated bibliography
Source study. De Prisco M, Tapoi C, Oliva V, Possidente C, Strumila R, Takami Lageborn C, Bracco L, Girone N, Macellaro M, Vieta E, Fico G. Clinical features in co-occuring obsessive-compulsive disorder and bipolar disorder: A systematic review and meta-analysis. European Neuropsychopharmacology. 2024;80:14–24. DOI 10.1016/j.euroneuro.2023.11.006 · PMID 38128332. The title is reproduced as published, including the spelling “co-occuring”. Open-access publication under a Creative Commons license. Funding: the authors declare no funding source for this work. Conflicts of interest: E. Vieta reports grants and consulting, advisory board or speaking activities for AB-Biotics, AbbVie, Angelini, Biogen, Biohaven, Boehringer-Ingelheim, Celon Pharma, Compass, Dainippon Sumitomo Pharma, Ethypharm, Ferrer, Gedeon Richter, GH Research, GlaxoSmithKline, Idorsia, Janssen, Lundbeck, Medincell, Novartis, Orion Corporation, Organon, Otsuka, Rovi, Sage, Sanofi-Aventis, Sunovion, Takeda and Viatris, outside the submitted work; G. Fico reports continuing medical education or consulting honoraria from Angelini, Janssen-Cilag and Lundbeck, and holds a grant from the La Caixa Foundation; the other nine authors declare no conflicts of interest.
Protocol registration. PROSPERO registry, identifier CRD42023392296. Deviations from the initial protocol are described by the authors in Appendix I of the supplementary material.
Supplementary material. Search strategies, list of excluded studies with reasons, study-by-study quality assessment, detailed results of the meta-regressions and sensitivity analyses, analysis restricted to higher-quality studies, PRISMA 2020 checklist.
Editorial collections
Tags
Verified on September 2, 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 19, 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.