Predominant polarity in bipolar disorder and validation of the polarity index in a German sample

In spite of advances in evidence-based drug treatment, bipolar disorder (BD) remains an often recurrent illness causing severe impairments in patients' daily life. Thus, relapse prevention is the main objective in treatment strategies of BD. Regarding the drug regimen in BD, a patient's actual symptomatology plays a crucial role. However, for maintenance treatment the individual illness progress and family history (e.g. information about drug responders) must be considered. Residual symptoms, psychiatric comorbidity, and number of prior episodes have been identified as risk factors of recurrence [1]. Additionally, the polarity of the first mood episode seems to be a good predictor of the predominant polarity of subsequent episodes over time [2], i.e. an overall higher number of depressive episodes is observed in patients with first episode depression and vice versa, and thus should be taken into account in treatment planning.

The categorization of bipolar disorder patients (BP) according to predominant polarity might help to optimize relapse prevention. Patients with a depressive predominant polarity (DPP) have a higher risk of further depressive episodes in the future and therefore need a treatment tailored to preventing depressive relapses. In contrast, patients with manic or hypomanic predominant polarity (MPP) require a drug regimen specifically efficient in the prevention of mania. For example, antidepressants should be avoided in patients with MPP due to the risk of triggering manic symptoms [3]. However, not every BP has a clear tendency for one pole of the disorder. Recent research reported that a clear predominant polarity can be observed in roughly half of BP, and that DPP is more prevalent than MPP [4],[5].

Several disease specific variables have been associated with the concept of predominant polarity [5],[6]. For instance, it has been demonstrated that depressive polarity coincides with a worse prognosis and decreased treatment response compared to manic polarity [7],[8]. Furthermore, the later patients with DPP, often initially misdiagnosed as patients suffering from major depression ("hidden bipolars"), receive a correct diagnosis (and an adequate drug treatment, accordingly), the more severe the course of illness [5],[9], i.e. patients with DPP are at a higher risk for misdiagnosis and subsequent worse prognosis. Thus, subtyping of BD by using the concept of predominant polarity could improve the planning of clinical care, long-term prevention, and the development of new therapeutic strategies. Furthermore, more homogeneous subgroups might provide the potential to inform biological studies and the search for biomarkers. Therefore, mechanisms involved in the pathophysiology of BD could be studied in a more target-oriented fashion. However, up to now there are no genetic, familial or biomarker studies on predominant polarity in BD, and the biological underpinnings remain unknown. Treatment guidelines refer to the concept of predominant polarity [10], and some authors even suggest that diagnostic criteria of BD should include predominant polarity as a relevant course descriptor for the illness [11],[12]. However, the recently published DSM-5 [13] did not add predominant polarity as a course specifier.

Recently, the idea of predominant polarity in BD was introduced as a helpful tool for maintenance treatment. Popovic et al. [14] proposed the Polarity Index (PI) as a descriptor for categorizing drug profiles in the relapse prevention of BD. The Polarity Index is a numeric expression of the efficacy profile of a given drug, derived from the ratio of Numbers Needed to Treat (NNT) for the prevention of depression and NNT for the prevention of mania. The NNT expresses the reciprocal of the absolute risk reduction and is calculated based on the results of several randomized placebo-controlled trials [14]. According to the authors, a PI of 1 indicates equal efficacy of a drug in the prevention of manic and depressive episodes alike. Drugs with a PI >1 have a stronger anti-manic prophylactic effect, while drugs with an index <1 are more efficient in preventing depressive episodes. In a second study, Popovic et al. provided evidence for the validity of the index in a large sample of BP by calculating a PI value for each patient [15]. They demonstrated that the mean PI values in their bipolar sample were significantly higher in patients with MPP compared to patients with DPP, indicating a stronger antimanic treatment regimen in MPP. The treatment of patients with DPP was characterized by lower mean PI, indicating a better prevention of depressive episodes. Therefore, the authors suggested the PI as a useful numerical expression of evidence-based maintenance treatment in BD and recommended it as a simple concept for physicians, one that is easily remembered and translatable into clinical practice.

Up to now, most studies on predominant polarity were performed by the Barcelona Bipolar Disorder Group [14]-[17]. Therefore, the aim of the present study was to validate the PI in a clinical setting with a similar study design in a German sample. To do so, we aimed to replicate the results by Popovic et al. by examining a well-characterized bipolar sample in order to determine the role of PI in clinical decision-making regarding maintenance treatment in BD [15]. Furthermore, we were interested in associations between demographic or clinical characteristics and predominant polarity in BP.

In the present study we investigated a sample of 336 BP with regard to their predominant polarity, which is defined as twice the number of previous episodes of one pole over the other. We aimed to identify clinical characteristics associated with predominant polarity. Furthermore, we attempted to validate the recently published Polarity Index [14], a metric algorithm which expresses the efficacy profile of a given drug in the maintenance treatment of BD.

Approximately half of our sample was classified as predominantly depressive (50.3%), 13.7% of patients as predominantly manic, and 36% of patients had no defined polarity. We found that the diagnosis type BP-I was more prevalent in patients with MPP compared to patients with DPP. These findings are in line with previous studies about predominant polarity and polarity of first episode [4], given that patients with a manic episode at illness onset more often receive the diagnosis type BP-I [2]. A recent follow-up study by Gonzalez-Pinto et al. [7] demonstrated that patients with DPP overall had more episodes, more hospitalizations, and more suicide attempts. However, we found no significant differences in number of previous episodes and suicidality between the three groups. At onset of illness, manic episodes were more frequent in patients classified with MPP compared to patients with DPP. This result, again, is in line with previous work demonstrating that polarity of first episode predicts the polarity of subsequent episodes over time [21],[22]. In addition, patients with MPP were significantly younger when experiencing their first manic episode compared to patients classified as DPP. In line with previous studies, our results show that DPP was more prevalent in women, while MPP was equally distributed in women and men [5],[23]. Regarding family history of psychiatric disorders, the literature showed an association with DPP [7], however we could not confirm this finding in our sample.

Regarding the Polarity Index, we could not replicate the results by Popovic et al. [15]. In our sample, the PI of maintenance treatment did not differ significantly between patients with DPP and MPP. Therefore, we could not validate the Polarity Index as a numeric expression of the efficacy of maintenance treatment under naturalistic conditions. Only when mood stabilizers (lithium and lamotrigine) were considered separately were we able to detect a difference: Patients with DPP had a lower PI of mood stabilizers compared to patients with MPP. Furthermore, patients with DPP were treated significantly more often with lamotrigine and antidepressants than patients with MPP. In turn, lithium, valproate, carbamazepine, risperidone and FGAs were more frequently prescribed in MPP compared to DPP. Hence, our bipolar sample received an evidence-based treatment for relapse prevention, supported by pertinent guidelines such as those published by the German Association for Psychiatry, Psychotherapy and Psychosomatics (DGPPN) [24] or the Canadian Network for Mood and Anxiety Treatments [25]. It is therefore possible that the difference in results may be due to different prescription patterns in the present sample than in the Spanish sample. In our sample, valproate, quetiapine, and aripiprazole were prescribed considerably more often than in the study by Popovic et al. [15]. In contrast, risperidone, olanzapine, and carbamazepine, which are often used in maintenance treatment by the Spanish colleagues, were rarely prescribed in our sample. Furthermore, antidepressants as add-on therapy were administered more often in our sample. Due to the naturalistic study design, our study sample also differed in sociodemographic and clinical parameters from the sample of Popovic et al. [15]. For instance, diagnosis type BP-II was overrepresented in our sample. Additionally, the Spanish sample consisted of 44.4% of patients classified as MPP, compared to only 21.4% in our sample. Therefore, despite using a similar study design, different sample characteristics and prescription patterns may be attributable for the divergent results in our German sample.

A further factor hindering validation might be that no PI was assigned to valproate due to missing RCTs. Indeed, 30% of our patients were medicated with valproate, compared to only 17% in the sample of Popovic et al. [15]. The exclusion of a compound used in almost a third of the patients therefore might have distorted our results. In a similar vein, antidepressants do not have a PI either, despite playing an undervalued role in the maintenance treatment of BD [26], and consequently were not considered in the present analysis. Moreover, psychosocial interventions were also not considered for analysis because information was not available about how many patients of our sample received psychoeducation or psychotherapy in the past. As Popovic et al. [17] demonstrated, psychosocial interventions differ in their depression- or mania-preventive effects and can be evaluated by the PI. We therefore conclude that the unavailability of a PI for several frequently used compounds in maintenance treatment in BD might account for our non-validation. Since we still consider the PI a clinically useful measure, an effort should be made to assign a PI to all substances used in the treatment of BD. In the perspective of evidence-based medicine, the PI could act in addition to clinical interviews to provide a high standard of patient care.

However, the PI has some general limitations which should be considered. First of all, the index does not reflect pharmacodynamic interactions in the case of polypharmacy, as it is only the numeric mean of the PIs of the individual drugs, regardless of their dosage or possible interactions. Furthermore, the PIs of drugs used in maintenance treatment are calculated according to current available RCT studies. Up to now, well-designed RCTs are rare, especially for mood stabilizers like valproate and carbamazepine [16]. Therefore, the PI is subject to continuous change with each new study. Other open questions concern the maintenance treatment of patients without predominant polarity (in the present sample, 36%), treatment of first-episode patients, and how mixed states should be classified. Furthermore, sub-syndromal depressive symptoms, which are prevalent during remission in BD, are not taken into account in the classification of predominant polarity [6]. Despite these methodological shortcomings, the idea of assigning a PI to a given drug and following rule-based treatment upon the predominant polarity seems reasonable and worthwhile; although we assume that psychiatrists follow these rules intuitively, a formal description bears advantages especially for education purposes.

The present study design also has methodological limitations. Our data consisted of retrospective self-reports on previous episodes from BP, their relatives, and clinical data. These are partially subjective, thus reducing the reliability of the data. Especially hypomanic episodes are often not remembered well or tend to be misinterpreted by patients. Furthermore, the subgroup of patients with MPP was relatively small, which could explain the absence of statistical differences in PI values between patients with DPP and MPP.

To summarize, we found that predominant polarity was associated with several clinical variables like sex, diagnosis subtype, irritability during mania, polarity at illness onset and age of onset. However, we could not validate the Polarity Index suggested by Popovic et al. [15] in our bipolar sample. Despite evidence of different treatment strategies in patients with depressive or manic predominant polarity, we found no significant differences in PI values. Nevertheless, psychopharmacological strategies must be adapted to the most frequently observed polarity. The reason for the negative finding in this study might well be that for some frequently used drugs in maintenance treatment of BD no PI value is available. Furthermore, we had different sample characteristics and prescription patterns compared to Popovic et al. [15]. Therefore, despite the failed validation of the PI in our sample, we suggest the PI as a helpful orientation in decision-making of maintenance treatment.