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,^{Reference Colom, Vieta, Daban, Pacchiarotti and Sánchez-Moreno11} 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.^{Reference Carvalho, McIntyre, Dimelis, Gonda, Berk and Nunes-Neto8}

Second, 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,^{Reference Rowland and Marwaha53} 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,^{Reference Merikangas and Lamers54} 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.^{Reference Bartoli, Bachi, Calabrese, Cioni, Guzzi and Nasti55,Reference Bartoli, Cavaleri, Nasti, Palpella, Guzzi and Riboldi56} Indeed, 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.^{Reference Merikangas, Jin, He, Kessler, Lee and Sampson57} However, 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 meta-analysis 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.^{Reference Bartoli, Nasti, Palpella, Piacenti, Di Lella and Mauro28} In addition, our findings, showing an association between mPP and BD-I, seem consistent with the natural history of BD-I,^{Reference Baldessarini, Bolzani, Cruz, Jones, Lai and Lepri58} 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.^{Reference McIntyre, Berk, Brietzke, Goldstein, López-Jaramillo and Kessing1,Reference Grande, Berk, Birmaher and Vieta59}

A 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^{Reference Kogan, Maj, Rebello, Keeley, Kulygina and Matsumoto60,Reference McIntyre, Alda, Baldessarini, Bauer, Berk and Correll61} distinction between BD-I and BD-II.^{Reference Colom, Vieta and Suppes9} In 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.^{Reference Colom, Vieta and Suppes9,Reference Carvalho, Quevedo, McIntyre, Soeiro-de-Souza, Fountoulakis and Berk13} This defines the ratio between antimanic and antidepressant properties of single pharmacological and non-pharmacological approaches used for maintenance treatment of bipolar disorder.^{Reference Popovic, Reinares, Goikolea, Bonnin, Gonzalez-Pinto and Vieta62,Reference Popovic, Reinares, Scott, Nivoli, Murru and Pacchiarotti63} More 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.^{Reference Norton and Paulus64}

Another 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,^{Reference Bartoli, Crocamo and Carrà65} 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,^{Reference Daban, Colom, Sanchez-Moreno, García-Amador and Vieta66,Reference Turvey, Coryell, Arndt, Solomon, Leon and Endicott67} 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,^{Reference Carvalho, McIntyre, Dimelis, Gonda, Berk and Nunes-Neto8} 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 polarity – especially those with mPP – may show differences in cortical thickness in several areas of the central nervous system.^{Reference Argyropoulos, Christidi, Karavasilis, Bede, Antoniou and Velonakis31,Reference Ruiz G, Ospina JP, Vargas, Aguirre Acevedo and López-Jaramillo35} Further evidence highlighted that biomarkers of bipolar disorder may be more influenced by its specific clinical features, e.g. depression and mania, than by the disease itself.^{Reference Bartoli, Cioni, Cavaleri, Callovini, Crocamo and Misiak68–Reference Zhang, Xiao, Wang, Gao, Su and Lu70} 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,^{Reference Stang71} 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,^{Reference Bartoli72} has shown to be more effective in preventing mania than depression,^{Reference Fountoulakis, Tohen and Zarate73} and, in general, there are fewer evidence-based treatments for bipolar depression than for mania.^{Reference Ketter, Miller, Dell'Osso, Calabrese, Frye and Citrome74} Similarly, data on other important correlates, such as seasonality,^{Reference Fico, de Toffol, Anmella, Sagué-Vilavella, Dellink and Verdolini75} affective temperaments^{Reference Azorin, Adida and Belzeaux32} and comorbid anxiety disorders,^{Reference Pallaskorpi, Suominen, Rosenström, Mantere, Arvilommi and Valtonen17} 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.