Illness stage and predominant polarity in bipolar disorder: Correlation with burden of illness and moderation of treatment outcome

Abstract

Bipolar disorder often follows a set progression best described in stages where advanced stages are associated with poorer outcomes. Bipolar disorder is also often characterized by a predominance of episode polarity, where some individuals experience more depressive episodes (termed predominant depressive polarity) while others experience more hypo/manic episodes (termed predominant hypo/manic polarity). We examined the associations between staging and predominant polarity with measures of illness burden and treatment outcome utilizing data from a six-month comparative effectiveness trial of lithium and quetiapine in bipolar disorder (Bipolar CHOICE). We used number of self-reported lifetime mood (depressive and hypo/manic) episodes as a proxy for staging and ratio of depressive to manic episodes to define predominant polarity. Polarity and staging were correlated with several measures of burden of illness. Childhood abuse was correlated with more lifetime mood episodes, while more depressive episodes and depressive polarity were correlated with more anxiety disorder comorbidity. Depressive polarity was also correlated with more past trials of psychotropics, particularly antidepressants. However, neither staging nor predominant polarity moderated the randomized treatment effect of lithium vs. quetiapine. Number of depressive episodes in the past year was identified as a potential predictor of overall worse treatment outcome, regardless of medication condition. In conclusion, though staging and predominant episode polarity correlated with several measures of illness burden, they were not associated with differential treatment outcomes. This could be because many of our patients presented for treatment at advanced stages of illness and further highlights the need for early intervention in bipolar disorder.

Introduction

Bipolar disorder is a highly disabling disorder in which symptoms often present on a spectrum from depression to mania or mixed episodes (Colom et al., 2006; González-Pinto et al., 2010). During clinical encounters, healthcare providers routinely assess not only the current mood episode, but also the number and type of past mood episodes the patient has experienced (Judd et al., 2003; Judd et al., 2002; Shah et al., 2017). Bipolar disorder can be classified based on the predominant polarity of previous episodes (i.e., more depressive episodes vs. manic/hypomanic episodes; Colom et al., 2006) and the number of past mood episodes (i.e., is the patient early in their illness course with only few past mood episodes vs. later with many past episodes?; Kapczinski et al., 2014). There is evidence that these clinical factors can assist clinicians in making treatment choices and determining prognosis and treatment outcome (Baldessarini et al., 2012; Berk et al., 2007b; Colom et al., 2006; González-Pinto et al., 2010; Ortiz et al., 2017; Popovic et al., 2014; for reviews see Cosci and Fava, 2012; Kapczinski et al., 2014; Mcgorry et al., 2014).

Some medical conditions have a set pattern of progression (often called staging), that assists clinicians with treatment decisions and prognosis determination. The best examples of staging are in cancer (Sobin, 2003). There has been an effort to identify similar stages of illness in psychiatric disorders (Kapczinski et al., 2014), including bipolar disorder (Berk et al., 2007a; Berk et al., 2017c). Some staging models for bipolar disorder have proposed using neuroimaging and biomarkers to define stages (Frank et al., 2015; Siwek et al., 2017). Others have used the number of lifetime episodes as a proxy for staging (Grande et al., 2014; Magalhães et al., 2012; Rosa et al., 2012). Most postulated models consistently acknowledge that there are at least two definable stages, specifically an “early” or “at-risk” stage and a “late” or “advanced” stage, and patients in the earliest stages typically respond better to treatment (Berk et al., 2011; Kapczinski et al., 2014).

Psychotherapy and psychosocial treatments also have shown effects based on staging. For example, family psychoeducation was found to be more effective in preventing recurrence in early stages of bipolar disorder compared to later stages (Reinares et al., 2010), and psychotherapy was more effective in bipolar patients in the STEP-BD trial with fewer past depressive episodes (Peters et al., 2014). Magalhães and colleagues (2012) found that depressed patients with bipolar disorder from the STEP-BD trial who had multiple previous episodes had consistently poorer cross-sectional and prospective outcomes. However, there was no significant effect for staging on antidepressant response in this randomized trial (Magalhães et al., 2012).

Nonetheless, patients’ response to medications may differ based on stage of illness. In a recent study, patients experiencing their first episode of mania were controlled with a combination of lithium and quetiapine and then randomized to maintenance therapy with only one agent (Berk et al., 2017a). Those randomized to lithium had superior outcomes during the yearlong follow up phase (Berk et al., 2017a) with potential neuroprotective effects (Berk et al., 2017b).

Bipolar disorder can also be classified as predominantly depressed or hypo/manic based on the ratio of previous mood episode types. Predominant depression has been associated with more suicide attempts, antidepressant and lamotrigine use, Axis-II disorder comorbidity, longer time to diagnose with bipolar disorder, and bipolar II diagnosis (González-Pinto et al., 2010; Popovic et al., 2014). Predominant hypo/manic polarity has been associated with more drug abuse, psychotic symptoms, use of atypical antipsychotics and conventional neuroleptics, and bipolar I diagnosis (Baldessarini et al., 2012; Popovic et al., 2014). Previous studies have also found an association between type of mood episode at illness onset and future predominant polarity, with depressive onset patients presenting as predominantly depressive (Baldessarini et al., 2012; Colom et al., 2006; Popovic et al., 2014) and manic, mixed, or psychotic onset patients presenting as predominantly manic (Baldessarini et al., 2012). Overall, depression has been described as the dominant polarity for both bipolar I and bipolar II patients (Colom et al., 2006; González-Pinto et al., 2010; Judd et al., 2003; Judd et al., 2002). However, there have been variable results when it comes to the correlation between predominant polarity and other important factors, such as age of onset and number of hospitalizations. It is suggested that depressive polarity has a poorer prognosis (González-Pinto et al., 2010), but there is limited information regarding the effects of predominant polarity on treatment outcome.

The present study used the self-reported number and type of past episodes to classify patients based on predominant polarity (i.e., ratio of previous depressed to manic episodes) and stage of illness (i.e., using number of past mood episodes). We aimed to examine how predominant polarity and the stage of illness affected outcomes in a six-month comparative effectiveness trial of lithium vs. quetiapine in individuals with bipolar disorder and to identify if lithium or quetiapine had different clinical effects depending on the polarity or stage. Additionally, we examined whether these variables could help predict sustained remission above and beyond variables previously identified as significant predictors of the overall treatment outcome (Nierenberg et al., 2016). We used both lifetime episodes and number of episodes reported in the past year in our analyses. We looked at the correlations between lifetime and past year episodes with outcomes and with each other. We also asked if the self-reported number of mood episodes or the predominant polarity was in fact a meaningful clinical feature by exploring the correlation with other measures of burden of illness. We examined clinical features that have been previously associated with staging and predominant polarity (e.g., age of onset, type of mood episode at the onset of illness, use of antidepressants and antipsychotics) and measures of burden of illness (e.g., suicide attempts, substance use, and anxiety comorbidity). We examined childhood abuse, as childhood adversity has been associated with a bipolar diagnosis (Rodriguez et al., 2021). We hypothesized that: 1) bipolar patients with depressive polarity and those with more previous episodes of illness would have poorer response to either treatment and higher burden of illness; and 2) the number of mood episodes and a depressive predominant polarity in the year prior to study entry would also be positively correlated with burden of illness.

Material and Methods

Participants

This was a secondary analysis of the Bipolar CHOICE (Clinical and Health Outcomes Initiative in Comparative Effectiveness for Bipolar Disorder) study, a comparative effectiveness trial of lithium vs. quetiapine in bipolar I and II (Nierenberg et al., 2016; Nierenberg et al., 2014). The study enrolled 482 adult participants across 11 sites and compared treatment outcome between the lithium and quetiapine arms. A secondary analysis looked at predictors of sustained remission (Clinical Global Impressions Scale for Bipolar Illness (CGI-BP) ≤ 2 for 8 weeks; Spearing et al., 1997) in the whole sample. Participants were mainly female (59%, n = 283) and diagnosed with bipolar I disorder (68%, n = 329). Individuals were excluded from the study if they demonstrated any contraindication to lithium or quetiapine. The study was approved by the Institutional Review Board (IRB) at each of the 11 study sites. See Table 1 for additional participant demographics.

Table 1.

Participant demographics.

Variable	N	Mean (SD)
Duration of illness (Years)	482	23.20 (12.45)
Age of first symptoms
Manic	478	19.82 (9.46)
Depressive	480	16.45 (7.99)
Age of first episode	482	15.68 (7.79)
Number of psychiatric hospitalizations	481	1.56 (3.96)
Number of medications
Psychotropic medications (at Baseline)	482	1.09 (1.31)
Antidepressant (Lifetime)	482	2.49 (2.27)
Antimanic agents (Lifetime)	482	1.77 (1.93)
Lifetime number of episodes
Hypo/manic	453	37.69 (47.45)
Depressive	459	39.46 (41.29)
Quality of life (Q-LES-Q)	478	44.29 (17.78)
LIFE-RIFT	476	14.22 (3.38)
	N	(%)
Gender
Female	283	59
Male	199	41
Employee/student	219	45
Married/living as married	150	31
Primary diagnosis
Bipolar I	329	68
Bipolar II	153	32
History of childhood abuse	257	53
Comorbid anxiety	277	57
Type of first episode
Manic	72	15
Depressive	281	58
Both episodes in first year of illness	123	26
Substance use disorder (lifetime)	296	61
Suicide attempt history (lifetime)	173	36
Lithium + adjunctive treatments	240	50
Quetiapine + adjunctive treatment	242	50
Sustained Remission	103	21
Mood episode at study entry
Major Depressive Episode	269	56
Hypo/manic episode	56	12
Mixed	82	17
Neither (not meeting full criteria)	75	16
Number of depressive episodes in past year
0–2	248	51
3–5	120	25
6–9	48	10
10+	61	13
Number of hypo/manic episodes in past year
0–2	237	49
3–5	124	26
6–9	48	10
10+	68	14
Percentage of last year spent depressed
0–20%	52	11
21–40%	115	24
41–60%	138	29
61–80%	127	26
81–100%	50	10
Percentage of last year spent hypo/manic
0–20%	222	46
21–40%	158	33
41–60%	61	13
61–80%	28	6
81–100%	11	2

Note: Some percentages may not sum to exactly 100% due to rounding

Materials and Procedure

Participants were randomized to either lithium or quetiapine plus adjunctive personalized treatment and followed for 6 months. Co-primary outcome measures included Clinical Global Impressions-Efficacy Index (CGI-EI; Feighner et al., 1983; Guy, 1976) and necessary clinical adjustments, which measured number of changes in adjunctive personalized treatment. This analysis examined the predominant polarity (lifetime and in the year prior to study entry) and number of self-reported past mood episodes (lifetime and in the year prior to study entry) of the study participants, and their relationship to treatment outcome and burden of illness. To determine predominant polarity, participants provided self-reported total number of manic/hypomanic and depressive episodes experienced over their lifetime and in the year prior to study entry. Participants reported actual numbers for total lifetime episodes. For the number of episodes in the past year, the participants were given ranges to choose from (0–2, 3–5, 6–9, 10+). They also reported an estimate of what percentage of time they spent depressed or manic in the past year (0–20%, 21–40%, 41–60%, 61–80%, 81–100%). We calculated two ratios for the number of depressive episodes over the number of manic/hypomanic episodes experienced, one for lifetime and one for the past year. Specifically, we took the log of the total number of lifetime depressive episodes plus one, over the total number of lifetime manic episodes plus one. A similar log ratio was developed for episodes in the past year. We assigned a numerical value to each category (0–2=1, 3–5=2, 6–9=3 and 10+=4). The log of the ratio was taken to transform these data such that negative numbers reflected a more manic polarity, zero reflected no polarity, and positive numbers reflected a more depressed polarity. The addition of one in the numerator and denominator of the ratio was used to avoid undefined ratios. As a proxy for stage of illness, we used the self-reported number of lifetime mood episodes.

Participants completed several other measures during the initial clinical assessment and during follow-up visits that were used to represent the burden of illness. History of suicide attempts, comorbid anxiety disorder diagnoses, lifetime history of substance use disorders, and type of mood episode the participant was experiencing at study entry (MDD, Mania/Hypomania, Mixed or neither when they had mood symptoms but did not meet full DSM-IV criteria) were assessed using the electronic Extended Mini-International Neuropsychiatric Interview (MINI; Sheehan et al., 1998) at the baseline visit. History of childhood abuse, age at onset of illness, age of first depressive symptoms that interfered with functioning, age of first manic symptoms that interfered with functioning, number of past psychiatric admissions, percentage of last year spent depressed, percentage of last year spent manic/hypomanic, number of psychiatric medications prescribed immediately before study entry, and number of lifetime antidepressants and antimanic agents prescribed were gathered at baseline as part of the self-report medical history form. Based on the age of onset of depression and mania, the group was divided into those whose illness started with depression, those who first had hypo/mania, and those who had both depression and mania in the first year of their illness. Life functioning in areas such as in work, with family and friends, and activities was gathered using the Longitudinal Interval Follow-up Evaluation Range of Impaired Functioning Tool (LIFE-RIFT; Leon et al., 2000) at the baseline, week 12, and week 24 (or end of treatment if the participant terminated treatment early) visits. Subjective quality of life in areas such as physical health, recreational activities, and social relationships was assessed by the Quality of Life, Enjoyment, and Satisfaction Questionnaire (Q-LES-Q; Endicott et al., 1993) at the baseline, week 12, and week 24 (or end of treatment if the participant terminated treatment early) visits.

Severity of psychiatric symptoms was assessed with the Bipolar Inventory of Symptoms Scale (BISS; Bowden et al., 2007; Gonzalez et al., 2008) at baseline and weeks 2, 4, 6, 8, 12, 16, 20, and 24 (or end of treatment if the participant terminated treatment early), and the BISS score was used to calculate depressive and manic symptoms according to the Montgomery Asberg Depression Rating Scale (MADRS; Montgomery and Asberg, 1979) and the Young Mania Rating Scale (YMRS; Young et al., 1978), respectively. The BISS is correlated with the MADRS and the YMRS (Gonzalez et al., 2008). The full list of assessments administered in the Bipolar CHOICE study is reported elsewhere (Nierenberg et al., 2016).

Statistical Analyses

All analyses were conducted using SAS 9.2 statistical software (SAS Institute, Inc., 1994). Two-tailed hypothesis tests were used throughout; null hypotheses were rejected using a threshold of p = 0.05. We did not correct for multiple comparisons, as our analyses were exploratory in nature.

To examine relationships between number of depressive and hypo/manic episodes and polarity over the lifetime and in the past year, Spearman’s rank-order correlations were calculated. We conducted mixed-effects linear regression analyses to examine associations between the burden of illness measures and 1) the number of self-reported episodes of depression, 2) the number of self-reported episodes of mania, 3) the total number of mood episodes, and 4) the predominant polarity ratio. These analyses were conducted for lifetime and past year episodes. To investigate any moderating effect of number of mood episodes or predominant polarity on outcome when randomized to lithium or quetiapine, we fit mixed-effects linear regression models including an interaction between polarity and staging moderators and treatment assignment.

Stepwise logistic regression was conducted to examine the effects of episode number and predominant polarity on treatment outcome, above and beyond what was reported in the CHOICE study (Nierenberg et al., 2016). In the original paper, a stepwise procedure was used to select a subset of variables that predicted sustained response as defined by a CGI-BP ≤ 2 for at least 8 weeks. In the present analysis, we included these six variables in our model (i.e., current comorbid anxiety diagnosis, MADRS current depression scores, YMRS current mania scores, employment status, MINI suicide risk status, and bipolar I or II disorder subtype (Nierenberg et al., 2016), but then additionally considered inclusion of variables related to predominant polarity and number of episodes using an entry criterion of p = 0.10.

Results

The number of lifetime and past year mood episodes were all highly correlated with each other (all p < 0.001). The highest correlation was between number of lifetime depressive episodes and number of lifetime hypo/manic episodes (r = .60, p < 0.001). The lowest correlation was between number of lifetime depressive episodes and number of past-year hypo/manic episodes (r = .28, p < 0.001).

Tables 2 and 3 describe associations between burden of illness measures and lifetime and past year mood episodes and depressive polarity. Individuals with the greatest number of lifetime mood episodes (either depressed, hypo/manic, or total) had been ill for more years and had an earlier age of onset of illness. Those with the most total lifetime mood episodes were more likely to report a history of childhood abuse, had spent a higher percentage of time hypo/manic in the year before study entry, and had reported taking fewer psychotropic medications during their lifetime. They were also more likely to report that their illness started with a hypo/manic episode or that they had experienced both types of mood episodes in the first year of their illness. Looking at lifetime and past year mood episodes based on episode type (Table 3), we found that those with more lifetime depressive episodes were more likely to be married, report a history of childhood abuse, have comorbid anxiety, and have more past trials of antidepressants. Participants reporting more lifetime hypo/manic episodes spent a larger percentage of time in the last year hypo/manic, were more likely to enter the study with hypo/manic or mixed symptoms, were less likely to report that their first lifetime episode was depressed and had fewer lifetime psychotropic medication trials.

Table 2.

Correlations between burden of illness measures and total episodes and depressive polarity (lifetime).

Measures	Total Episodes (Lifetime)		Depressive Predominant Polarity (Lifetime)
	Mean [95% CI]	p	Mean [95% CI]	p
Years ill	1.66 [1.11, 2.21]	<0.001	0.01 [−0.00, 0.02]	0.051
Gender (F)	−3.24 [−17.63, 11.15]	0.658	0.05 [−0.16, 0.25]	0.662
Employment status	−5.58 [−19.83, 8.66]	0.442	−0.13 [−0.34, 0.07]	0.193
Age of first hypo/manic episode	−2.17 [−2.89, −1.45]	<0.001	0.01 [−0.00, 0.02]	0.194
Age of first depressive episode	−2.17 [−3.04, −1.30]	<0.001	−0.01 [−0.02, 0.00]	0.061
Married (vs not)	9.37 [−6.02, 24.76]	0.232	0.22 [0.00, 0.44]	0.048
Child abuse (vs not)	15.20 [1.02, 29.37]	0.036	0.01 [−0.19, 0.21]	0.926
Primary diagnosis (Bipolar I vs. II)	9.48 [−5.70, 24.66]	0.220	−0.02 [−0.24, 0.19]	0.833
History of suicide attempts	1.75 [−12.46, 15.97]a,b,c	0.809	0.09 [−0.12, 0.30]	0.418
Substance use disorder (lifetime)	4.86 [−8.97, 18.69]a,b,c	0.490	0.02 [−0.19, 0.23]	0.850
Current mood stated
Hypo/manic episode	19.61 [−3.12, 42.34]	0.091	−0.33 [−0.65, −0.00]	0.048
Mixed episode	15.24 [−4.35, 34.83]	0.127	−0.12 [−0.39, 0.16]	0.415
Not meeting full criteria for either	0.78 [−19.55, 21.10]	0.940	0.05 [−0.24, 0.33]	0.757
Type of first episoded
Hypo/manic	30.07 [9.93, 50.21]	0.004	−0.45 [−0.73, −0.16]	0.002
Same age of onset for hypo/manic and depressive episodes	17.69 [0.85, 34.54]	0.040	−0.24 [−0.48, −0.00]	0.049
Any anxiety disorder (lifetime)	8.16 [−5.53, 21.85]a,b,c	0.242	0.23 [0.02, 0.43]	0.030
LIFE-RIFT	0.77 [−1.22, 2.76]a,b,c	0.448	0.00 [−0.03, 0.03]	0.811
Q-LES-Q	−0.13 [−0.52, 0.25]a,b,c	0.498	−0.00 [−0.01, 0.00]	0.103
Percent depressed in previous year	0.80 [−5.00, 6.59]a,b,c	0.787	0.10 [0.02, 0.19]	0.017
Percent hypo/manic in previous year	15.32 [8.77, 21.86]a,b,c	<0.001	−0.33 [−0.42, −0.23]	<0.001
Number of antidepressant medications (lifetime)	1.34 [−1.65, 4.34]a,b,c	0.378	0.07 [0.02, 0.11]	0.003
Number of antimanic medications (lifetime)	3.20 [−0.33, 6.74]a,b,c	0.076	0.05 [−0.01, 0.10]	0.086
Number of psychotropic medications (lifetime)	−6.32 [−11.42, −1.21]a,b,c	0.016	0.10 [0.03, 0.18]	0.007
Number of psychiatric hospitalizations (lifetime)	0.68 [−0.96, 2.33]a,b,c	0.415	0.00 [−0.02, 0.03]	0.726

Note:

^aAdjusted for years ill,

^bAdjusted for age of first episode,

^cAdjusted for type of first episode;

^dReference category is Depressed mood;

LIFE-RIFT = Longitudinal Interval Follow-up Evaluation-Range of Impaired Functioning Tool; Q-LES-Q = Quality of Life Enjoyment and Satisfaction Questionnaire

Table 3.

Correlations between burden of illness measures and depressive and manic episodes (past year and lifetime) and polarity (past year).

Measures	Depressive Episodes (Past Year)		Depressive Episode (Lifetime)		Hypo/manic Episodes (Past Year)		Hypo/manic Episodes (Lifetime)		Depressive Predominant Polarity (Past Year)
	Mean [95% CI]	p	Mean [95% CI]	p	Mean [95% CI]	p	Mean [95% CI]	p	Mean [95% CI]	p
Years ill	0.01 [−0.00, 0.01]	0.105	1.01 [0.72, 1.30]	<0.001	−0.00 [−0.01, 0.01]	0.781	0.64 [0.30, 0.99]	<0.001	0.00 [0.00, 0.01]	0.032
Gender (F)	0.01 [−0.18, 0.21]	0.886	−0.54 [−8.21, 7.12]	0.890	−0.05 [−0.25, 0.14]	0.605	−2.98 [−11.87, 5.90]	0.510	0.02 [−0.07, 0.11]	0.599
Employment status	−0.02 [−0.21, 0.17]	0.815	−4.15 [−11.75, 3.46]	0.284	−0.02 [−0.22, 0.17]	0.805	−1.10 [−9.89, 7.70]	0.806	−0.01 [−0.10, 0.08]	0.819
Age of first hypo/manic episode	−0.02 [−0.03, −0.01]	<0.001	−0.87 [−1.26, −0.48]	<0.001	−0.02 [−0.03, −0.01]	<0.001	−1.30 [−1.75, −0.86]	<0.001	0.00 [−0.00, 0.01]	0.624
Age of first depressive episode	−0.02 [−0.03, −0.01]	<0.001	−1.33 [−1.79, −0.87]	<0.001	−0.01 [−0.03, −0.00]	0.022	−0.81 [−1.36, −0.27]	0.004	−0.00 [−0.01, 0.00]	0.229
Married (vs not)	0.01 [−0.19, 0.22]	0.898	9.49 [1.33, 17.65]	0.023	−0.04 [−0.25, 0.17]	0.684	−0.45 [−9.94, 9.04]	0.926	0.01 [−0.08, 0.11]	0.790
Child abuse (vs not)	0.06 [−0.13, 0.25]	0.546	7.58 [0.01, 15.15]	0.050	0.02 [−0.17, 0.22]	0.829	6.91 [−1.87, 15.68]	0.123	0.01 [−0.07, 0.10]	0.747
Primary diagnosis (Bipolar I vs. II)	0.07 [−0.13, 0.27]	0.495	3.21 [−4.90, 11.32]	0.437	0.06 [−0.15, 0.27]	0.577	6.39 [−2.97, 15.76]	0.181	0.01 [−0.09, 0.10]	0.889
History of suicide attempts	0.05 [−0.15, 0.25]b	0.632	0.79 [−6.82, 8.40]a,b	0.838	−0.04 [−0.24, 0.17]b,c	0.730	1.57 [−7.41, 10.55]a,b,c	0.731	0.05 [−0.04, 0.15]	0.263
Substance use disorder (lifetime)	0.06 [−0.14, 0.25]b	0.574	3.19 [−4.17, 10.55]a,b	0.394	0.19 [−0.01, 0.38]b,c	0.064	2.57 [−6.15, 11.28]a,b,c	0.563	−0.03 [−0.12, 0.06]	0.560
Current mood stated
Hypo/manic episode	−0.03 [−0.33, 0.28]	0.856	1.37 [−10.91, 13.65]	0.827	0.47 [0.16, 0.78]	0.003	18.01 [4.12, 31.90]	0.011	−0.23 [−0.38, −0.09]	0.001
Mixed episode	0.00 [−0.26, 0.27]	0.976	2.34 [−8.14, 12.81]	0.662	0.16 [−0.10, 0.43]	0.229	12.73 [0.67, 24.79]	0.039	−0.09 [−0.21, 0.04]	0.166
Not meeting full criteria for either	−0.07 [−0.34, 0.21]	0.622	−1.89 [−12.75, 8.97]	0.732	0.03 [−0.24, 0.31]	0.805	2.29 [−10.22, 14.80]	0.719	−0.06 [−0.19, 0.06]	0.323
Type of first episoded
Hypo/manic	0.15 [−0.12, 0.43]	0.269	1.55 [−9.34, 12.44]	0.780	0.32 [0.04, 0.59]	0.026	28.52 [16.20, 40.84]	<0.001	−0.05 [−0.18, 0.07]	0.415
Same age of onset for hypo/manic and depressive episodes	0.32 [0.10, 0.55]	0.005	5.13 [−3.92, 14.17]	0.266	0.35 [0.12, 0.58]	0.003	12.91 [2.61, 23.21]	0.014	−0.02 [−0.13, 0.08]	0.667
Any anxiety disorder (lifetime)	0.31 [0.12, 0.49]b	0.002	9.40 [2.23, 16.57]a,b	0.010	0.13 [−0.06, 0.33]b,c	0.180	−1.34 [−10.00, 7.32]a,b,c	0.761	0.10 [0.01, 0.19]	0.036
LIFE-RIFT	0.02 [−0.01, 0.04]b	0.229	0.41 [−0.65, 1.47]a,b	0.451	−0.00 [−0.03, 0.02]b,c	0.761	0.38 [−0.88, 1.63]a,b,c	0.553	0.01 [−0.01, 0.02]	0.237
Q-LES-Q	−0.00 [−0.01, 0.00]b	0.327	−0.12 [−0.33, 0.08]a,b	0.230	0.00 [−0.00, 0.01]b,c	0.114	−0.01 [−0.25, 0.23]a,b,c	0.938	−0.00 [−0.01, −0.00]	0.016
Percent depressed in previous year	0.12 [0.04, 0.20]b	0.003	2.36 [−0.71, 5.43]a,b	0.132	−0.04 [−0.12, 0.05]b,c	0.389	−1.48 [−5.13, 2.17]a,b,c	0.427	0.07 [0.03, 0.11]	<0.001
Percent hypo/manic in previous year	0.13 [0.03, 0.22]b	0.007	2.81 [−0.74, 6.37]a,b	0.121	0.43 [0.34, 0.52]b,c	<0.001	12.85 [8.79, 16.91]a,b,c	<0.001	−0.16 [−0.20, −0.11]	<0.001
Number of antidepressant medications (lifetime)	0.01 [−0.03, 0.05]b	0.708	1.68 [0.09, 3.28]a,b	0.038	−0.02 [−0.06, 0.02]b,c	0.379	−0.61 [−2.50, 1.28]a,b,c	0.527	0.02 [−0.00, 0.04]	0.064
Number of antimanic medications (lifetime)	0.01 [−0.04, 0.06]b	0.613	1.77 [−0.11, 3.65]a,b	0.065	0.01 [−0.04, 0.06]b,c	0.719	1.14 [−1.09, 3.37]a,b,c	0.315	0.00 [−0.02, 0.03]	0.673
Number of psychotropic medications (lifetime)	0.03 [−0.04, 0.10]b	0.392	−1.78 [−4.51, 0.95]a,b	0.201	−0.05 [−0.12, 0.02]b,c	0.194	−4.47 [−7.68, −1.25]a,b,c	0.007	0.04 [0.01, 0.08]	0.014
Number of psychiatric hospitalizations (lifetime)	0.02 [−0.01, 0.04]b	0.134	0.40 [−0.48, 1.29]a,b	0.370	0.02 [−0.00, 0.05]b,c	0.053	0.29 [−0.75, 1.33]a,b,c	0.588	−0.00 [−0.01, 0.01]	0.846

Note:

^aAdjusted for years ill,

^bAdjusted for age of first episode,

^cAdjusted for type of first episode;

^dReference category is Depressed mood;

LIFE-RIFT = Longitudinal Interval Follow-up Evaluation-Range of Impaired Functioning Tool; Q-LES-Q = Quality of Life Enjoyment and Satisfaction Questionnaire

The younger the age of onset, the greater the number of hypo/manic and depressive episodes reported in the year prior to study entry by the participant. Additionally, participants with more depressive episodes in the year prior to study entry were more likely to have similar ages of onset for their first hypo/manic or depressive episodes, have comorbid anxiety, and spend a higher percentage of time both depressed and hypo/manic in the year prior to entering the study. More hypo/manic episodes in the year prior to study entry was associated with a hypo/manic mood state at study entry, a smaller likelihood that the first lifetime mood episode was depression, and a higher percentage of time hypo/manic in the year prior to study entry. Comorbid substance use disorders and more lifetime hospitalizations may have been associated with the number of hypo/manic episodes in the past year but did not reach the usual 0.05 threshold (p = 0.064 and 0.053, respectively).

Although lifetime depressive polarity was not significantly correlated with years of illness (p = 0.051) or age at onset (p = 0.061), it was significantly associated with depressed mood at illness onset, comorbid anxiety, more total lifetime psychotropic medications and antidepressants, and spending more time depressed and less time manic in the year prior to study entry. Those with lifetime depressive polarity were also less likely to enter the study hypo/manic.

Depressive polarity in the year prior to study entry was correlated with a greater duration of illness, comorbid anxiety, lower quality of life (as measured by the Q-LES-Q), and more psychotropic medications at baseline, and was inversely correlated with being hypo/manic at study entry. In addition to reporting more frequent depressive episodes in the past year, depressive polarity patients reported more time depressed and less time manic in the past year.

Several clinical features were notably not correlated with our measures of polarity and staging of illness: type of bipolar disorder, sex, history of suicide attempts, employment status, number of antimanic medications (lifetime), and LIFE-RIFT (functioning) scores. Furthermore, the results of the mixed-effects linear regression analyses suggested that the number of depressive and hypo/manic episodes, total number of episodes, and polarity of episodes over the past year and over the lifetime had no moderating effect on outcome in lithium and quetiapine groups (ps > 0.260).

In the stepwise logistic regression model, we included the six variables that were previously examined as significant predictors of overall outcome in the Bipolar CHOICE study (i.e., current comorbid anxiety diagnosis, MADRS current depression scores, YMRS current mania scores, employment status, MINI suicide risk status, and bipolar I or II disorder subtype; Nierenberg et al., 2016). The only measure that may help predict sustained remission (i.e., a score on the CGI-BP ≤ 2 for at least 8 weeks) above and beyond the original six covariates was the number of depressive episodes within the past year (p = 0.056). Adjusting for the other six covariates, for each increase in category of the number of depressive episodes within the past year (0–2=1, 3–5=2, 6–9=3 and 10+=4), the odds of sustained remission decreased by about 27%.

Discussion

The present study examined how staging (based on self-reported number of lifetime and past-year mood episodes) and predominant polarity (based on ratio of lifetime depressive to lifetime hypo/manic episodes) affected outcomes in a six-month comparative effectiveness trial of lithium vs. quetiapine in individuals with bipolar disorder. We found that staging and predominant polarity were correlated with one another and with different burden of illness variables. Many findings matched our expectations or had previously been reported in the literature, although we found notable exceptions. Mainly, we found that participants with predominant depressive polarity or more lifetime or past-year depressive episodes had more comorbid anxiety disorders. Predominant depressive polarity was associated with taking more psychotropic agents, particularly more antidepressant over their lifetime and participants with a history of childhood abuse reported more lifetime mood episodes, mostly depressive type. However, neither staging nor predominant polarity moderated the effect of the two medication conditions on outcome. In addition to the six covariates considered in the primary study manuscript (Nierenberg et al., 2016), only number of depressive episodes in the past year was found to be a potentially useful predictor of overall treatment outcome, regardless of medication condition.

Although self-reported number of mood episodes may not have accurately reflected actual number of episodes due to its reliance on event memory, particularly when number of episodes was high, we found reported number of episodes over one’s lifetime and the past year correlated with one another, and we saw several similar findings between polarity in the past year and lifetime. In support of previous findings in the literature, individuals with depressive predominant polarity (lifetime or past year) tended to be more depressed and less manic in the previous year, less likely to enter the study in a hypo/manic state, and had greater antidepressant and total psychotropic medication use over their lifetime (González-Pinto et al., 2010; Popovic et al., 2014). Depressive polarity was also associated with higher rates of anxiety disorder diagnoses, but only depressive polarity in the past year showed an association with lower quality of life scores. Significantly, our sample did not show correlations between depressive polarity and a history of suicide attempts, childhood abuse or substance use, measures of functioning (LIFE-RIFT), number of hospitalizations, employment status, or sex.

Previous studies have reported that predominant depressive polarity is associated with more antidepressant and lamotrigine use (González-Pinto et al., 2010; Popovic et al., 2014), whereas predominant manic polarity is associated with atypical antipsychotic and conventional neuroleptics use (Popovic et al., 2014). In our study, individuals with depressive polarity had more trials of psychotropics and antidepressants (although this did not reach statistical significance for past year polarity) over their lifetime. This could reflect the relative paucity of pharmacological treatment options for bipolar depression compared to bipolar mania (Goodwin et al., 2016; Yatham et al., 2018), necessitating more frequent medication trials with agents with mixed effectiveness, such as antidepressants.

Unlike previous studies that found patients in earlier stages of illness to be more responsive to treatment and responding better to lithium (Berk et al., 2011; Berk et al., 2017a; Swann et al., 1999), we found no difference in response to lithium or quetiapine, and they did not significantly impact overall treatment outcome other than the number of depressive episodes in the past year. Magalhães and colleagues also found no significant effect of staging on response to antidepressants in bipolar disorder (Magalhães et al., 2012).

One explanation for this finding could be that our patient population was in advanced stages of illness. The average duration of illness in the present sample was 23 years and the average number of episodes experienced over the lifetime was high (hypo/manic: M = 37; depressive: M = 39). The number of participants with 5 or fewer episodes in our sample was only 12 (2.4% of the sample). According to the staging model proposed by Berk et al. (2007b), most of our participants would be in stage 3c or 4 (late-stage) with recurrent episodes of illness. This may explain why we did not find significant differences in treatment outcome or certain burden of illness variables such as suicide attempts, functionality, and substance use. By the time our patients presented for this study, the illness was already in the later stages with its associated morbidities and treatment resistance. For better precision medicine and risk stratification utilizing staging and predominant polarity, we need deep and accurate phenotyping, a more consistent definition of stages and predominant polarity, and consideration of the longitudinal trajectory of disease. Some strategies to achieve this include larger longitudinal naturalistic follow up studies and international consortia which can lead to greater sample size and statistical power and allows for the use of machine learning predictive strategies (Manchia et al., 2020).

There are limitations to the present study worth noting. This is a secondary analysis of the Bipolar CHOICE study. Because examining staging and polarity were not the primary aims of the Bipolar CHOICE study, polarity and staging variables were not considered in inclusion and exclusion criteria. For this reason, we had very few participants in early stages of bipolar disorder. Also, the number of episodes were self-reported, which allows for misreporting due to recall problems or bias. Lithium and Quetiapine are effective for both manic and depressed phases of the illness. This is one possible explanation for lack of differentiation between these agents based on polarity and is a limitation of this study. Staging and polarity are complex clinical descriptions, and the field has not yet come to a consensus on how to define them. This creates a limitation for our study. Using self-reported number of episodes to define staging or polarity, although utilized clinically, may not be able to predict treatment response when patients have had many previous episodes. However, this finding cannot be fully generalized to all other definitions of staging or polarity.

Despite these limitations, there are strengths worth noting. The present study included a large, relatively representative clinical sample from across the United States. We were able to obtain variables that approximated polarity and stages (i.e., the total number of episodes over the lifetime and past year), and we were also able to examine the effects of these variables on treatment outcomes.

Our findings suggest that although self-reported mood episodes are correlated with some measures of burden of illness, once patients have had multiple episodes, it has less value for staging and clinical prediction. Kessing et al. (2021) discuss how the change from DSM-IV to DSM-5 and the inclusion of activation to the diagnostic criteria, while increasing specificity, leads to less sensitivity, and this in turn could lead to further delays in diagnosis. Patients newly diagnosed with bipolar disorder according to ICD-10 were only diagnosed with bipolar disorder 60% of the time when using DSM-5 criteria. This number increased to 94% for chronic cases. This could potentially lead to difficulty with diagnosing bipolar disorder early in the course of the illness. Treatment options may be less effective as treatment progresses to later stages of the illness, emphasizing the need for early diagnosis and intervention (Vieta et al., 2018) and suggests that future studies of staging in bipolar disorder should be conducted with a specific focus on early stages of the disorder and in individuals with few mood episodes. By understanding how polarity and staging may impact outcomes of bipolar disorder, we may be able to better match patients to treatments with increased precision and help in determining prognosis.