Prevalence and correlates of manic/hypomanic and depressive predominant polarity in bipolar disorder: systematic review and meta-analysis

Background Identification of the predominant polarity, i.e. hypomanic/manic (mPP) or depressive predominant polarity (dPP), might help clinicians to improve personalised management of bipolar disorder. Aims We performed a systematic review and meta-analysis to estimate prevalence and correlates of mPP and dPP in bipolar disorder. Method The protocol was registered in the Open Science Framework Registries (https://doi.org/10.17605/OSF.IO/8S2HU). We searched main electronic databases up to December 2023 and performed random-effects meta-analyses of weighted prevalence of mPP and dPP. Odds ratios and weighted mean differences (WMDs) were used for relevant correlates. Results We included 28 studies, providing information on rates and/or correlates of mPP and dPP. We estimated similar rates of mPP (weighted prevalence = 30.0%, 95% CI: 23.1 to 37.4%) and dPP (weighted prevalence = 28.5%, 95% CI: 23.7 to 33.7%) in bipolar disorder. Younger age (WMD = −3.19, 95% CI: −5.30 to −1.08 years), male gender (odds ratio = 1.39, 95% CI: 1.10 to 1.76), bipolar-I disorder (odds ratio = 4.82, 95% CI: 2.27 to 10.24), psychotic features (odds ratio = 1.56, 95% CI: 1.01 to 2.41), earlier onset (WMD = −1.57, 95% CI: −2.88 to −0.26 years) and manic onset (odds ratio = 13.54, 95% CI: 5.83 to 31.46) were associated with mPP (P < 0.05). Depressive onset (odds ratio = 12.09, 95% CI: 6.38 to 22.90), number of mood episodes (WMD = 0.99, 95% CI: 0.28 to 1.70 episodes), history of suicide attempts (odds ratio = 2.09, 95% CI: 1.49 to 2.93) and being in a relationship (odds ratio = 1.98, 95% CI: 1.22 to 3.22) were associated with dPP (P < 0.05). No differences were estimated for other variables. Conclusions Despite some limitations, our findings support the hypothesis that predominant polarity might be a useful specifier of bipolar disorder. Evidence quality was mixed, considering effects magnitude, consistency, precision and publication bias. Different predominant polarities may identify subgroups of patients with specific clinical characteristics.

Bipolar disorder is a severe and chronic condition, affecting about 1-2% of the general population. 1,2Its clinical course is characterised by mood recurrencies, in which depressive episodes alternate with manic or hypomanic episodes, according to the conventional differentiation between bipolar-I disorder (BD-I) and bipolar-II disorder (BD-II). 3However, despite epidemiological assumptions that people with bipolar disorder spend more time affected by depression than by mania, 4,5 the clinical course and trajectories of bipolar disorder may be rather heterogeneous. 6,7In particular, it has been proposed that a more fine-grained classification of bipolar disorder should consider whether the clinical course is characterised by a depressive (dPP) or manic/hypomanic (mPP) predominant polarity. 8,9The concept of predominant polarity was first introduced by Jules Angst, 10 based on a study investigating 95 individuals with bipolar disorder.Participants were subdivided into three subtypes according to their mood recurrencies, i.e. the preponderantly manic, the preponderantly depressed and the nuclear type, in which there was a balanced proportion of depressive and manic episodes. 10Thereafter, Colom et al 11 provided a more detailed definition of predominant polarity in bipolar disorder, proposing that to define mPP, manic/hypomanic episodes should represent at least two-thirds of the overall number of lifetime mood episodes.On the other hand, dPP requires that among lifetime mood episodes, at least two-thirds are depressive.Finally, an undetermined predominant polarity should be considered if there is a sufficiently balanced proportion between manic and depressive episodes in the clinical course of bipolar disorder, without any clear mood episode predominance. 11Alternative definitions have been proposed, 12 together suggesting that the identification of the predominant polarity might help clinicians to improve the personalised management of bipolar disorder by making its clinical trajectories clearer. 9Indeed, available evidence suggests that mPP or dPP may influence individual response to acute and long-term treatment for bipolar disorder, as well as the effectiveness of psychopharmacological agents used for the stabilisation phase. 8,13Exploring the hypothesis of predominant polarity as a possible clinical specifier, previous reviews have suggested that mPP and dPP might involve approximately half of all people with bipolar disorder and might be associated with particular individual characteristics. 8,14owever, despite the growing scientific interest in this field, [15][16][17][18] no systematic analyses on rates and individual characteristics associated with mPP versus dPP are available so far.To shed light on this topic, we performed a systematic review and meta-analysis of observational studies aimed at identifying the prevalence and clinical correlates of different mood predominance types in bipolar disorder, as well as assessing the quality of evidence in terms of strength, precision, consistency and risk of publication bias.

Method Study design and protocol
This systematic review and meta-analysis is reported following the Meta-analysis Of Observational Studies in Epidemiology guidelines. 19The study protocol was registered on 27 November 2023 in Open Science Framework Registries (https://doi.org/10.17605/OSF.IO/8S2HU) and amended on 8 December 2023 because of changes in the search strategy.

Eligibility criteria
We included any observational studies (a) providing information on prevalence rates of mPP and dPP in people with bipolar disorder and (b) comparing them with respect to one or more sociodemographic or clinical characteristics.To be considered, studies had to include at least ten individuals in each group (mPP and dPP).Moreover, in order to improve consistency across studies and to reduce the risk of misclassification bias, we included only studies which used the recommended Colom's definition for predominant polarity. 11Based on this, mPP and dPP are defined as a lifetime ratio ≥2:1 of either hypomanic/manic episodes or depressive episodes, respectively.This restrictive definition, splitting patients in three categories (mPP, dPP and undetermined predominant polarity), is considered to be more stable and conservative over time than other definitions, 12 making patients less likely to be switched from one category to another across different episodes. 9We excluded studies (a) not providing information on predominant polarity, (b) not comparing mPP and dPP in terms of relevant sociodemographic or clinical characteristics, (c) including samples with a mean age <18 years, (d) using definitions of predominant polarity based on different criteria, and (e) published before the release date of DSM-IV. 20In order to avoid duplicate results, we excluded data on correlates derived from the same sample, including only the study that provided the larger amount of information.Finally, we excluded scientific reports not undergoing a peer-review process, such as conference abstracts, dissertations and grey literature.

Article screening
We searched the Embase, PubMed, APA PsycInfo (via ProQuest), and Emcare (via Ovid) databases for articles indexed up to 8 December 2023, without any language restrictions.We used the following search phrases adapted for each database: (a) Embase: 'bipolar disorder':ti,ab,kw AND 'predominant polarity':ti,ab,kw; (b) PubMed: bipolar [Title/Abstract] AND predominant polarity [Title/Abstract]; (c) PsycInfo: tiab(bipolar disorder) AND tiab(predominant polarity); (d) Ovid Emcare: (bipolar and (predominance or 'predominant polarity')).ti.or (bipolar and (predominance or 'predominant polarity')).ab.An additional manual search of studies included in two relevant reviews 8,14 was carried out to check for further potentially eligible studies.References were managed using EndNote web software.After the preliminary screening based on titles and abstracts had been completed, full texts were retrieved to assess the final eligibility of studies.These procedures were completed by three authors (C.B., M.G. and L.G.) independently, and reasons for exclusion after full-text review were recorded.Disagreements concerning suitability for inclusion were resolved by discussion and consensus involving all authors.

Data extraction
Data were extracted between 11 and 13 December 2023, using a standard template to collect key information from all eligible studies: year of publication; country; setting; inclusion and exclusion criteria; sample size, mean age, and sex proportion; methods used to define predominant polarity; prevalence rates of mPP and dPP; and sociodemographic and clinical correlates of mPP versus dPP.Four authors (F.B., C.B., M.G. and L.G.) independently extracted data and blindly cross-checked them for accuracy.

Data analysis
Meta-analyses of mPP and dPP prevalence in bipolar disorder were based on random-effects weighted proportions with 95% confidence intervals using arcsine-based transformation.Considering the expected low consistency of meta-analyses of prevalence rates, 21 subgroup analyses were run to test potential variations in mPP and dPP prevalence rates by the geographical area of included studies.An omnibus test from the random-effects meta-regression was performed to test the overall moderating effects of subgroups.Moreover, in order to deal with a skewed distribution of prevalence rates, we reported the overall median and interquartile range of the mPP and dPP point prevalences for descriptive purposes.Weighted differences in arcsine-transformed proportions (WPDs) were estimated for both overall and subgroup analyses by geographical area.
To compare mPP and dPP for relevant correlates, randomeffects meta-analyses were conducted for variables with data available from at least five different studies.P < 0.05 was used as the threshold for statistical significance.We used odds ratios and weighted mean differences (WMDs) with 95% confidence intervals for categorical and continuous variables, respectively.Heterogeneity across studies was evaluated according to standard cut-offs for I2 statistics to measure inconsistency of meta-analyses on correlates. 22ublication bias was assessed using Egger's test for meta-analyses with data available from at least ten studies. 23To evaluating the magnitude and precision of the effects (see 'Grading of the evidence' section), each WMD was converted into the equivalent effect size (standardised mean difference [SMD]), performing relevant metaanalysis, while each odds ratio was converted into an SMD by dividing the relevant ln(OR) by 1.81. 24Conventional cut-offs (0.2 small, 0.5 medium, 0.8 large) were used to interpret the magnitude of the effect. 25Data analyses were performed using Stata statistical software, release 17 (StataCorp LLC, 2021).OpenMeta[Analyst] software 26 was used to generate forest plots.

Grading of the evidence
Following a similar approach used in recent meta-analyses, 27,28 we used GRADE (Grading of Recommendations, Assessment, Development, and Evaluations) items, 29 adapted for non-interventional observational studies, to classify the quality of evidence as high, moderate, low or very low for each variable showing a statistically significant estimate (P < 0.05).
First, we assessed the consistency of findings according to the I 2 value.We downgraded by one level the quality of evidence if inconsistency was estimated (I 2 ≥ 50%).
Second, we evaluated the precision of findings by checking the width of the equivalent effect size 95% CI.We downgraded the quality of evidence by one level if (a) the 95% CI width of the equivalent effect size was ≥0.40 for meta-analyses showing small or medium effect sizes, or (b) the 95% lower and upper confidence limits of the equivalent effect size (SMD) were not both ≥0.80 for meta-analyses showing large effect sizes.
In addition, we assessed the risk of publication bias, downgrading the quality of evidence by one level if (a) meta-analyses included fewer than ten studies, or (b) the Egger's test P-value was <0.10 for meta-analyses including at least ten studies.
Finally, we evaluated the magnitude of the effect, upgrading the quality of evidence by one level if the magnitude of the equivalent effect size was large (SMD ≥ 0.80).

Study selection
The systematic search on relevant databases generated 374 records, namely 140 from Embase, 89 from PubMed, 80 from APA PsycInfo and 65 from Ovid Emcare.After deduplication, there were 192 articles left to be screened, including additional studies retrieved from the reference lists of the two reviews. 8,14After screening by titles and abstracts, 70 studies were identified as potentially eligible.Following the final screening based on full texts, 28 studies met the eligibility criteria and were included in the meta-analysis. 12,15-18,30-52A flowchart with details of screening, the study selection process and reasons for exclusion is presented in Fig. 1.

Study characteristics
Studies were published between 2009 40,50 and 2023. 31All studies, with the exception of one reported in Spanish, 43 were written in English.The sample sizes varied between 42 31 and 788. 50The majority of studies (k = 16) were conducted in Europe, i.e. six were from Italy, 30,[38][39][40][41]45 three from Spain, 15,46,49 two from France 32,42 and Germany, 51,52 and one each from Belgium, 18 Finland 17 and Greece. 31Five studies were conducted in Asia: four in India 16,36,44,47 and one in Singapore; 37 and five studies in South America, i.e. three were from Brazil 33,34,48 and two from Colombia.35,43 Two studies were based on data from multiple countries of different geographical areas. 12,50 Stdy characteristics are reported in Table 1. Some studies were lily to have a partial overlap between included samples, i.e.(a) Belizario et al (2019) 33 and Belizario et al (2018); 34 (b) Fico et al (2022) 15 and Popovic et al (2014); 46 and (c) Pacchiarotti et al (2011) 45 and Mazzarini et al (2009).40 We prioritised data from Belizario et al (2019), 33 Fico et al (2022) 15 and Pacchiarotti et al (2011), 45 respectively, as these studies were all based on larger sample sizes. For meta-anlyses, we used data from smaller studies 34,40,46 only if these were not provided by the main study.

Selection of variables
We extracted data for 22 correlates from at least five studies based on unique samples.Twenty-four studies 12,[15][16][17][18][30][31][32][33][34][35][36][38][39][40]43,44,[46][47][48][49][50][51][52] had data suitable for at least one correlate. We did not consider man and depressive symptoms or the lifetime number of manic/ hypomanic and depressive episodes, even if these variables were based on more than five studies, owing to the inherent association with the corresponding predominant polarity.Moreover, we did not consider pharmacological treatments because of the extreme variability across studies in terms of current and lifetime treatments.We distinguished variables included for meta-analyses as 'variables associated with mPP', 'variables associated with dPP' and 'variables not associated with any predominant polarity'.
Variables associated with hypomanic/manic predominant polarity A summary of findings and related details on quality of evidence are reported in Tables 3 and 4, respectively.

Gender
Based on a meta-analysis of 19 studies with 3335 total participants, we found that people with mPP were more likely to be male (odds ratio = 1.39, 95% CI: 1.10 to 1.76; P = 0.005; Supplementary Fig. 5).

Manic polarity of first episode
We found that people with mPP were more likely to report a first mood episode characterised by manic polarity (k = 8, N = 1557; odds ratio = 13.54,95% CI: 5.83 to 31.46;P < 0.001; Supplementary Fig. 7).The overall quality of evidence was moderate; despite the inconsistency of findings (I 2 = 84.8%)and the unclear risk of publication bias, the magnitude of the effect and the lower and upper confidence limits (equivalent effect size: SMD = 1.44, 95% CI: 0.97 to 1.90) were large.

Psychotic features
Based on ten studies including 1970 participants with bipolar disorder, those with mPP were more likely to have psychotic features than those with dPP (odds ratio = 1.56, 95% CI: 1.01 to 2.41; P = 0.047; Supplementary Fig. 9).However, the overall evidence was

Variables associated with depressive predominant polarity
A summary of findings is provided in Table 3, and related details on the quality of evidence are given in Table 4.

Depressive polarity of first episode
We found that a depressive polarity of the first mood episode was associated with dPP (k = 9, N = 1655; odds ratio = 12.09, 95% CI: 6.38 to 22.90; P < 0.001; Supplementary Fig. 11).The evidence was of moderate quality, considering the poor consistency across studies (I 2 = 78.3%)and the unclear risk of publication bias, despite the large effect (equivalent effect size: SMD = 1.37, 95% CI: 1.02 to 1.73).Factors associated with hypomanic/manic predominant polarity (versus depressive predominant polarity) Age (years)

Number of mood episodes
Individuals with dPP had more mood episodes than people with mPP (k = 9; N = 1773; WMD = 0.99 mood episodes; 95% CI: 0.28 to 1.70 mood episodes; P = 0.006; Supplementary Fig. 12).Despite the consistency across studies (I 2 = 0%) and the precision of findings (equivalent effect size: SMD = 0.12, 95% CI: 0.02 to 0.22), the evidence was of moderate quality, being downgraded by one level because of uncertainty about publication bias.

Being in a relationship
Meta-analysis based on six studies and 856 participants showed that people with dPP were more often married or in a relationship than those with mPP (odds ratio = 1.98, 95% CI: 1.22 to 3.22; P = 0.006; Supplementary Fig. 13).However, the quality of evidence was low, owing to the poor precision of findings (equivalent effect size: SMD = 0.38, 95% CI: 0.11 to 0.65) and the unclear risk of publication bias, despite the low between-study heterogeneity (I 2 = 28.9%).

Variables not associated with any predominant polarity
We found no differences between mPP and dPP as regards other variables, including years of education, unemployment, duration of illness, mixed polarity of first episode, rapid cycling course, number of hospital admissions, number of suicide attempts, comorbid alcohol and substance use disorders, and family history of bipolar disorder, any affective disorders or suicide.A summary of these findings is provided in Supplementary Table 1, and relevant forest plots are shown in Supplementary Figs.14-25.

Summary and interpretation of findings
To our knowledge, this is the first systematic review and meta-analysis investigating prevalence rates and possible correlates of mPP and dPP in bipolar disorder.Based on 28 studies, conducted in 12 countries across Europe, Asia and South America, our work provides several important findings.First, we found no differences in prevalence rates between predominant polarities, as around a third of subjects with bipolar disorder had mPP, and a similar proportion of individuals had dPP.Thus, a predominant polarity, as defined by Colom's criteria, 11 seems to affect approximately two-thirds of patients with bipolar disorder.Overall rates of both mPP and dPP as estimated in our work appeared higher than those reported in a previous review on this topic, which found that around half of people with bipolar disorder might have a well-defined predominant polarity. 8econd, rates of mPP and dPP might vary according to different geographical areas.Whereas dPP was more frequent than mPP in European countries, opposite estimates were found in studies conducted in Asia, in which prevalence rates of mPP were almost double those of dPP.This was not surprising, as the clinical trajectory of bipolar disorder might be influenced by genetic, cultural and environmental factors that are likely to vary by country, 53 shaping the expression of manic and depressive symptoms.Moreover, differences in predominant polarity might be explained also by cross-national variations in prevalence rates of BD-I and BD-II, 54 which are associated with mPP and dPP, respectively.Finally, it should be considered that even the heterogeneity of mental healthcare delivery systems worldwide may influence the probability of access to care for patients with bipolar disorder, especially during depressive episodes. 55,56Indeed, fewer than half of people with bipolar disorder receive mental health treatment, particularly in low-income countries, and only a quarter report contacts with the mental health system. 57However, those with manic episodes and related behavioural abnormalities may be less likely to avoid detection and treatment.
Third, we uncovered several clinically meaningful correlates associated with a distinct predominant polarity.Individuals with mPP were younger, more often male and more likely to be affected by BD-I, as well as being more likely to show psychotic features and have an earlier onset characterised by manic symptoms.These findings were consistent with evidence from a recently published metaanalysis involving participants whose manic predominance was set by definition, i.e. those with unipolar mania, who were again more often males, with younger age at onset and higher rates of psychotic features. 28In addition, our findings, showing an association between mPP and BD-I, seem consistent with the natural history of BD-I, 58 which is typically characterised by more psychotic features and an earlier onset than BD-II.On the other hand, people with dPP were more likely to have a depressive polarity at onset, a higher number of mood episodes and a history of suicide attempts, and were more often in a relationship than people with mPP.Our findings seem to be supported by a relatively recent body of evidence in the field. 1,59 balanced interpretation of our findings needs to consider that for some explored variables, the quality of evidence was low (i.e.psychotic features and being in a relationship) or moderate (i.e.age, polarity of first episode, BD-I and number of mood episodes).This was influenced by several issues, including the imprecision and inconsistency of overall estimates, as well as the uncertain probability of publication bias.Additional research is needed to substantiate our meta-analytic evidence for these variables.Nonetheless, our findings reasonably support the hypothesis that predominant polarity might identify specific subgroups of people with bipolar disorder; in particular, the associations with male gender and earlier bipolar disorder onset in mPP and with higher rates of suicide attempts in dPP were based on evidence of high quality.

Clinical implications
The assessment and definition of predominant polarity, because of its potential as a long-term specifier, may represent a valuable tool in clinical practice to help choose appropriate treatments and predict probable outcomes for individuals with bipolar disorder.In particular, the predominant polarity conceptualisation might represent a useful alternative to the traditional and yet complex 60,61 distinction between BD-I and BD-II. 9In view of the existing scientific evidence in this field, the inclusion of predominant polarity in the treatment decision-making process for bipolar disorder has been already hypothesised, with the concept of the polarity index. 9,13This defines the ratio between antimanic and antidepressant properties of single pharmacological and nonpharmacological approaches used for maintenance treatment of bipolar disorder. 62,63More generally, establishing clinical features associated with mPP or dPP may guide clinicians in the early identification of possible trajectories of bipolar disorder and in the selection of the most appropriate interventions for the prevention of mood relapses.In addition, the conceptualisation of predominant polarity, involving a fine-grained assessment of the behaviour, cognition, emotions and social interactions of individuals with bipolar disorder, might clash with emerging perspectives towards an unified and transdiagnostic view of affective disorders. 64nother important clinical implication of this study is related to the strong concordance between the long-term predominant polarity and the polarity of the first mood episode: our meta-analyses showed large effects for the relationships between a manic onset and mPP, as well as between a depressive onset and dPP, ruling out any association between mixed symptoms at onset and subsequent predominant polarity.While stressing the clinical heterogeneity of mixed features possibly occurring in both manic and depressive episodes, 65 our findings make clearer the predictive role of the first mood episode on predominant polarity.This has also been suggested by some pioneering studies in the field, 66,67 which showed that affective polarity at onset was associated with the polarity of the following episodes.As the main critical element of the current conceptualisation of predominant polarity involves the large amount of time that elapses between onset of bipolar disorder and the subsequent assessment of mPP, dPP or undefined polarity, across at least three mood episodes,8 the first manic or depressive episode might be considered to be a useful clinical proxy for the polarity of subsequent recurrences.
Finally, from a completely different perspective, the definition of potential neurobiological correlates of mPP and dPP might represent an important area for future research.Recent studies have reported, for example, that individuals with predominant polarityespecially those with mPPmay show differences in cortical thickness in several areas of the central nervous system. 31,35urther evidence highlighted that biomarkers of bipolar disorder may be more influenced by its specific clinical features, e.g.9][70] Potential neurobiological underpinnings of predominant polarity in bipolar disorder are also a matter for future research.

Limitations
The findings of the current systematic review and meta-analysis should be interpreted with caution, considering some limitations.First, as our work investigated cross-sectional differences between mPP and dPP, we cannot draw any conclusions about causal inference.Second, we need to consider some methodological variability across studies in terms of objectives, sample size, inclusion criteria and methods used to assess clinical variables.For instance, the retrospective nature of data should be considered as a possible source of misclassification of predominant polarity.Third, owing to the observational nature of the included studies, without any predefined protocol, we should consider that the possible effects of unpublished data and the likelihood of selective reporting bias may have at least partially influenced our meta-analytic estimates.Moreover, considering the lack of valid and recommended tools for meta-analyses based on non-intervention and non-randomised studies, 71 we did not formally assess the risk of bias of included studies.Although we included only studies with consistent definitions of predominant polarity, the poor representativeness of some studies (e.g.those including only in-patients) and some risk of recall bias (due to the nature of the tested variables) may have affected the results of our meta-analysis.In addition, the clinical course of bipolar disorder may be influenced by the different effectiveness of treatments for preventing relapses; however, insufficient data were available from the included studies for us to explore the role of this confounder on polarity predominance.For example, lithium, the gold-standard treatment for bipolar disorder, 72 has shown to be more effective in preventing mania than depression, 73 and, in general, there are fewer evidence-based treatments for bipolar depression than for mania. 74imilarly, data on other important correlates, such as seasonality, 75 affective temperaments 32 and comorbid anxiety disorders, 17 were available only from a limited number of studies, suggesting a need for additional research.

Future perspectives
The findings of this systematic review and meta-analysis support the hypothesis that predominant polarity might be a useful specifier of bipolar disorder, identifying subgroups of individuals with different clinical characteristics.High quality of evidence shows an association of mPP with male gender and an earlier onset of bipolar disorder, whereas dPP is more often associated with a history of suicide attempts.Different predominant polarities in bipolar disorder may represent particular targets for more appropriate care programmes and effective approaches to personalised treatment.

Fig. 1
Fig.1Flow diagram of included and excluded studies.

Table 1
Characteristics of included studies

Table 2
Meta-analyses of prevalence rates of hypomanic/manic and depressive predominant polarity in people with bipolar disorder by geographical area of low quality, because of the moderate-to-high heterogeneity (I 2 = 66.1%) and imprecision (equivalent effect size: SMD = 0.25, 95% CI: 0.01 to 0.49), regardless of the low risk of publication bias (Egger's coefficient: −0.56, P = 0.71).

Table 3
Factors associated with hypomanic/manic or depressive predominant polarity: summary of findings k, number of included studies; N, number of study participants; N/A, not applicable; OR, odds ratio; SMD, standardised mean difference; WMD, weighted mean difference.

Table 4
Factors associated with hypomanic/manic or depressive predominant polarity: quality of evidence