Polypharmacy as maintenance treatment in bipolar illness: A systematic review

ABSTRACT Objectives Polypharmacy is common in maintenance treatment of bipolar illness, but proof of greater efficacy compared to monotherapy is assumed rather than well known. We systematically reviewed the evidence from the literature to provide recommendations for clinical management and future research. Method A systematic review was conducted on the use of polypharmacy in bipolar prophylaxis. Relevant papers published in English through 31 December 2019 were identified searching the electronic databases MEDLINE, Embase, PsycINFO, and the Cochrane Library. Results Twelve studies matched inclusion criteria, including 10 randomized controlled trials (RCTs). The best drug combination in prevention is represented by lithium + valproic acid which showed a significant effect on time to mood relapses (HR = 0.57) compared to valproic acid monotherapy, especially for manic episodes (HR = 0.51). The effect was significant in terms of time to new drug treatment (HR = 0.51) and time to hospitalization (HR = 0.57). A significant reduction in the frequency of mood relapses was also reported for lithium + valproic acid vs. lithium monotherapy (RR=0.12); however, the trial had a small sample size. Lamotrigine + valproic acid reported significant efficacy in prevention of depressive episodes compared to lamotrigine alone. Conclusions The literature to support a generally greater efficacy with polypharmacy in bipolar illness is scant and heterogeneous. Within that limited evidence base, the best drug combination in bipolar prevention is represented by lithium + valproic acid for manic, but not depressive episodes. Clinical practice should focus more on adequate monotherapy before considering polypharmacy.


| INTRODUCTION
Bipolar illness is a disabling mental illness with a high risk of relapse and recurrence affecting about 45 million people worldwide. 1 In most patients, it is a chronic condition, associated with clinical, psychosocial, and cognitive decline. 2 Sustained remission of mood episodes has been associated with better outcomes on cognitive performances and brain morphology, while affective recurrences have been shown to have an impact in terms of treatment resistance and disability. 3,4 Therefore, an effective maintenance treatment is needed to prevent relapses, hence minimize illness progression, reduce residual symptoms, restore functioning and quality of life.
Lithium, some antiepileptics, and second-generation antipsychotics are recommended and widely used as maintenance treatments. 5 Commonly it is stated that bipolar illness tends to require polypharmacy in most patients. 5 Experts and clinicians feel comfortable with this approach, assuming that this disease is too severe to respond to monotherapy in most cases. 6 In line with the Canadian Network for Mood and Anxiety Treatments (CANMAT) and International Society for Bipolar Disorders (ISBD) 2018 guidelines for the management of bipolar illness, "if therapy with one or a combination of the first-line agents at optimal doses is inadequate or not tolerated, the next step is to switch to or add on an alternate first-line agent." 5 Despite its wide use in clinical practice, evidence on polypharmacy in bipolar prophylaxis is still scant. Using (HR = 0.57) compared to valproic acid monotherapy, especially for manic episodes (HR = 0.51). The effect was significant in terms of time to new drug treatment (HR = 0.51) and time to hospitalization (HR = 0.57). A significant reduction in the frequency of mood relapses was also reported for lithium + valproic acid vs. lithium monotherapy (RR=0.12); however, the trial had a small sample size.
Lamotrigine + valproic acid reported significant efficacy in prevention of depressive episodes compared to lamotrigine alone. Conclusions: The literature to support a generally greater efficacy with polypharmacy in bipolar illness is scant and heterogeneous. Within that limited evidence base, the best drug combination in bipolar prevention is represented by lithium + valproic acid for manic, but not depressive episodes. Clinical practice should focus more on adequate monotherapy before considering polypharmacy.

Summations
• Polypharmacy is common in maintenance treatment of bipolar illness, but proof of greater efficacy compared to monotherapy is assumed rather than well known. • Within that limited evidence base, the best drug combination in bipolar prevention is represented by lithium + valproic acid for manic, but not depressive episodes. • Clinical practice should focus more on adequate monotherapy before considering polypharmacy.

Limitations
• The heterogeneity of selected studies and the low certainty of the outcomes represent the main limitations of this review. • The majority of the studies addressed the addition of a second drug to the maintenance monotherapy in case of failure. • To date, the field of polypharmacy in bipolar illness maintenance treatment lacks longitudinal RCT in which patients who failed drug A respond better to A+B than to B alone.
the possible combinations of couples of first-line maintenance drugs reported by the CANMAT and ISBD 2018 guidelines, we systematically reviewed the evidence from the literature on bipolar illness maintenance polypharmacy to provide recommendations for clinical management and future research.

| METHODS
This review was conducted according to methods recommended by the Cochrane Collaboration and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 7,8

| Study population and study design
We considered studies that included bipolar adolescents (from 13 to 18 years) and adults (older than 18 years) treated with combinations of couples of first-line maintenance drugs as suggested by the CANMAT and ISBD 2018 guidelines for the management of bipolar illness. Participants of both sexes were considered. All studies included subjects with a diagnosis of bipolar illness. Given the aim of this review, subjects had to be evaluated after the remission of an acute episode or, if remission was not an explicit inclusion criterion, the study had to include outcomes related to prophylaxis (frequency of relapse/hospitalization, time to relapse/hospitalization). No exclusion criteria were set for the recruitment setting or for the source of the clinical data. All experimental and observational study designs were included apart from case reports and case series. Narrative and systematic reviews, letters to the editor, and book chapters were excluded.

| Comparisons and outcomes
Only studies with two or more treatment groups were included to allow comparison of interventions. Studies were excluded if they provided aggregated efficacy data regarding different drug combinations and the considered polypharmacy was not a combination of CANMAT and ISBD firstline maintenance treatment options.
The primary outcome was the effectiveness of combinations of first-line bipolar illness maintenance drugs in terms of clinical relapses or hospitalizations. Secondary outcomes were (I) frequencies of response and remission, (II) time to discontinuation, (III) safety and tolerability of drug combinations.

| Study selection and data extraction
Identified studies were independently reviewed for eligibility by two authors (DR and NM) in a two-step process.
A first screening was performed based on title and abstract, and then full texts were retrieved for a second screening. At both stages disagreements by reviewers were resolved by consensus. Data were extracted by two authors (DR and NM) and supervised by a third author (AAm) using an ad hoc developed data extraction spreadsheet. The selected studies are presented on a spreadsheet piloted on ten randomly chosen papers and modified accordingly.
Efficacy data were presented at the level of single outcome on a modified version of the Evidence profile template obtained from the Grade Development Tool. Outcomes were ordered in different tables according to the intervention and comparison. All the relevant outcomes (see section "Comparisons and Outcomes") presented in the papers are reported in this review. Wherever possible, outcomes were reported with effect sizes (ES), confidence intervals (CI), and p-values.

| Quality assessments
The quality of individual outcome results was assessed according to the Grading of Recommendations Assessment, Development and Evaluation (GRADE) method with the following parameters: bias, inconsistency, indirectness, imprecision, publication bias, effect size, influence of confounding factors, dose response gradient. 9 Each outcome had a qualitative grade of certainty assigned from the four possible levels, ranging from very low to high. Outcomes of randomized controlled trials (RCTs) were rated high, and those from observational studies began the grading with moderate level of certainty. Afterward, outcomes were downgraded, respectively, by one or two levels for every parameter showing serious or very serious concerns. We generally followed the criteria from the GRADE handbook to downgrade the outcome certainty. An exception was imprecision which was downgraded because of small number of participants: by 1 level if the study had less than 30 participants per treatment arm, by 2 levels if the study had less than 10 participants per arm, in line with the thresholds used to evaluate the quality of evidence within the CANMAT and ISBD 2018 guidelines. We also downgraded for imprecision by 1 level if the CI of the ES included 0.75 and 1.25 and by 2 levels if it included 0.5 and 1.5, and if the study had lower power for the outcome than estimated in its power analyses (when the result was not significant).

| RESULTS
The literature search in the four databases yielded 1161 results, and 3 studies were retrieved outside the electronic search through citations in relevant literature in the field. One thousand and ninety-two studies remained after the removal of duplicates. One thousand and fifty-three results were excluded on the basis of title and abstract. The remaining 39 items were read in full text, and 27 of them were excluded. The total number of findings from the single databases and the reasons for exclusion of full texts are displayed in Figure 1.

| Included studies
Twelve studies matched the inclusion criteria for the present review (Table 1). Six out of ten possible combinations of CANMAT and ISBD 2018 first-line medications had at least one efficacy study as maintenance treatment.

| Combination of lithium and valproic acid
Only the BALANCE (Bipolar Affective disorder: Lithium Anticonvulsant Evaluation) study compared the preventive efficacy of valproic acid + lithium vs. valproic acid on mood episodes (Table 2), showing a significant effect on time to mood relapses (HR = 0.57), especially for the manic ones (HR = 0.51). 13 The effect was significant in terms of time to new drug treatment (HR = 0.51) and time to hospitalization (HR = 0.57).
Four studies allowed the comparison in efficacy as maintenance treatment of valproic acid + lithium versus lithium ( Table 2). The pilot trial conducted by Solomon and colleagues was the only study showing a significant effect on the frequency of mood relapses of the drug combination versus lithium monotherapy (relative risk, RR = 0.12), although several shortcomings as the small sample size (<10 patients per arm). 16 The other studies, 2 RCTs and 1 observational study, [12][13][14] consistently suggest some advantages in terms of time to mood relapse or treatment failure in the combination treatment group (hazard ratio, HR, between 0.72 and 0.80), although these results did not reach statistical significance. The BALANCE study also shows a non-significant delay in time to manic episodes with polypharmacy (HR = 0.66), although the same tendency does not apply to depressive episodes (HR = 1.06). 13 One observational study measured suicide attempts in populations exposed to valproic acid + lithium (6.3/10,000 months of exposure), lithium monotherapy (7.7/10,000 months of exposure), and valproic acid monotherapy (7/10,000 months of exposure) ( Table 2). 11 The study did not provide any statistical tests to compare these measures.
Finally, one RCT compared valproic acid + lithium vs. carbamazepine + lithium ( Table 2). The study states there were no significant differences between the time-to-relapse curves (no ES reported, p = 0.45). 15

| Combinations of lamotrigine and lithium or valproic acid
One RCT compared lamotrigine + divalproex with lamotrigine alone (Table 3). 19 The study failed to show statistically significant effect for time to depressive relapse (HR = 0.67) for the combination treatment. However, secondary outcomes such as the number of patients reaching the cutoff for depressive symptoms (RR = 0.66) and patients with discontinuation due to depressive symptoms (RR = 0.30), showed significant efficacy for the drug combination in prevention of depressive episodes. Statistical significance was not reached for the risk of manic relapses (RR = 0.61).
Another RCT studied the efficacy of the combination of lamotrigine + lithium vs. lithium (Table 3). 17 The study recruited patients who were depressed at the baseline and allowed the use of paroxetine to reach remission from the acute episode. Paroxetine was unequally used in the treatment groups (more in the monotherapy group), and results were reported narratively without statistical testing. Study results reported that more patients maintained the responder status in the combination treatment group than in the monotherapy group and the median time to depressive episode was higher with a combination treatment (10 months) compared to monotherapy (3.5 months).
Lamotrigine + lithium has been also compared to olanzapine + lithium combination for the treatment of comorbid anxiety in another RCT (Table 3). 18 The study reported a significant reduction of anxiety in both groups, although it lacks a control group with placebo or monotherapy. Therefore, a reduction of anxiety cannot be generalized as an effect of the addition of lamotrigine or olanzapine to lithium or vice versa. Moreover, the study did not find a statistically significant difference between the olanzapine and lamotrigine   combinations, although frequencies of response and remission of anxiety symptoms were higher with the olanzapine combination compared to the lamotrigine combination (RR = 0.61 for response; RR = 0.51 for remission).

| Combinations involving atypical antipsychotics
A single RCT compared aripiprazole + lamotrigine versus lamotrigine (Table 4) without reaching a statistical significance. 21 However, a trend toward significance was found in time to manic or mixed episode (HR = 0.6) and time to any episode (HR = 0.7), in favor of the combination treatment. Time to depressive episode (HR = 0.8) was not significantly different between the two groups. Quetiapine + lamotrigine combination was compared to lamotrigine alone in one RCT (Table 4) showing nonsignificant reductions in the risks of new treatments for depression (RR = 0.84) and mania (RR = 0.67). 22 The combination of asenapine + lithium or valproate vs. lithium or valproate was tested in a single RCT (Table 4). 20 Results narratively reported that time to first mood episode did not show significant differences between the two groups in the 52-week maintenance phase. Frequencies of manic or depressive symptoms, reported as adverse effects, were higher in the combination treatment group, although not significantly (RR = 1.58).

| Safety and tolerability
The combinations of couples of first-line maintenance drugs suggested by the CANMAT and ISBD 2018 guidelines demonstrated a safe and adequate tolerability profile (Table 5). Serious adverse events and discontinuations due to adverse events were low both in poly-and monotherapy treatment groups. No new or unexpected adverse events were reported that differed from the established profile for each individual medication as monotherapy in maintenance treatment. Three out of twelve studies (25%) did not mention side effects in the results. 11,12,19

| DISCUSSION
Studies from the literature on polypharmacy in bipolar illness maintenance treatment are sparse, heterogeneous, and often underpowered, so data interpretation requires caution. However, given the available scientific evidence supporting a statistically significant efficacy in bipolar prevention, some observations can be made (Table 6).
Findings from BALANCE reported that lithium + valproic acid is the only drug combination showing a statistically significant effect in terms of time to mood relapse compared to valproic acid monotherapy. 13 More than 40% relative benefit was irrespective of baseline severity of illness, maintained for up to 2 years, and most apparent in prevention of manic relapse. BALANCE was a randomized, open-label, threegroup trial, designed to imitate routine use of the agents in clinical practice relapse prevention, with up of 24 months of follow-up. This trial presented no serious methodological issues, except for it being an open-label study, designed mainly to compare combination therapy with monotherapy, with a relatively small sample (N = 110) per treatment arm. On the contrary, BALANCE could neither confirm nor refute a benefit of combination therapy compared with lithium monotherapy. In particular, for depression relapse, lithium monotherapy was as effective as combined therapy, and thus, the addition of valproate was not helpful for added efficacy.
Lithium + valproic acid might also offer some advantages on bipolar relapse or recurrence prevention compared to lithium alone, as reported by a clinical trial conducted in l997. 16 The study was a randomized pilot trial designed to compare the efficacy of a combination therapy (Lithium + divalproex sodium) with monotherapy (lithium), with up of 12 months of follow-up. The study had several shortcomings, including the small sample size (less than ten people per arm).
The second drug combination reaching a statistically significant efficacy in bipolar illness maintenance treatment is represented by lamotrigine + divalproex that reduced the risk for depressive episode and for trial discontinuation due to depressive symptoms compared to lamotrigine. 19 The study was a randomized, double-blind, parallel group trial designed to compare the efficacy of a combination therapy (lamotrigine + divalproex) with monotherapy (lamotrigine), with up of 8 months of follow-up. The use of a sample size with inadequate power for a strong test of several hypothesis and the enrollment of an insufficient number of bipolar patients to support separate analyses limited the generalization of the results. It could be also hypothesized that a longer duration of the study could allow the detection of a larger number of manic episodes, hence demonstrating an effect for adjunctive divalproex on manic episodes.
Moreover, evidence from the literature supporting an efficacy in bipolar prevention, although non-statistically significant, is available for the following drug combinations: lithium + valproic acid compared to lithium, 12-14 lamotrigine + lithium compared to lithium, 17 valproic acid + lithium compared to carbamazepine + lithium, 15 aripiprazole or quetiapine + lamotrigine compared to lamotrigine. 21,22 The impact of the duration of the follow-up on the observed effect sizes can be complex. While longer duration allows more chances for the outcome events to happen, it might also increase the number of drop-out events. Interestingly, the studies reaching a statistically significant result had a significant difference in duration, BALANCE lasted 24 months, 13  21,22 As noted previously, experts often comment that most patients with bipolar illness receive polypharmacy, and they endorse this practice on the grounds of the complexity of response in this condition. 23 However, the scientific evidence for such practice, as shown here, is still weak, and the belief that polypharmacy is more effective than the best monotherapy needs to be proven by more clinical studies. The complexity of the problem rises further questions about the clinical populations that might benefit from polytherapy and about the specific advantages or disadvantages of each drug combination.

T A B L E 4 Combinations involving atypical antipsychotics
A key caveat to most claims of efficacy of polypharmacy is that they tend to involve designs where patients are preselected to fail a single agent in monotherapy, and then are either continued in the failed monotherapy group or randomized to the addition of a second agent. Since the monotherapy group is already proven to be ineffective, such a design is biased in favor of polypharmacy. Instead, patients should be randomized from the beginning to monotherapy or combined therapy, to test the hypothesis whether two treatments are more effective inherently than one. Very few studies have this design.
If designs are limited to the inclusion to failed monotherapy, then the generalization of results would not involve inherent superiority of polypharmacy to monotherapy, but only added benefit after failed monotherapy. These distinctions are important to make since many clinicians do not provide adequate monotherapy doses and durations of treatment before rapidly moving to polypharmacy practice.
Furthermore, nonadherence seriously limits the effectiveness of any maintenance treatment. Thus, lack of adherence should be always considered in case of treatment failure, especially in the populations at risk, such as people with treatment-related side effects, comorbid personality, or substance use disorders. 24

| LIMITATIONS AND STRENGTHS
The main limitations of the review are the heterogeneity of the studies and the low certainty of the outcomes, ranging from very low to moderate in most instances. Quality issues of single outcomes in the studies are detailed in Tables 2, 3, and 4.
Studies with similar outcomes were limited to lithium and valproic acid combinations. A meta-analysis was not conducted given the paucity of these outcomes, the different study designs, follow-up times, confounding factors, and inclusion criteria. Therefore, a qualitative review was preferred to address the specific characteristics of the studies. Results are reported as they were available in the original manuscript, and only minor analyses have been done in some instances, such as the calculation of the relative risks and the absolute outcomes with the respective confidence intervals. No stratification of results across different bipolar subtypes was performed.
The main strength of this review is its being systematic and its including the entire scientific evidence published so far on the main medical databases. Moreover, the quality assessment process performed through the GRADE system, allows transparency and systematic assessment of critical issues within selected studies.

| CONCLUSIONS
There is minimal evidence to support the use of combinations of drugs for bipolar illness maintenance treatment. Keeping in mind scantiness and heterogeneity of the available literature, the best drug combination in bipolar prevention is represented by lithium + valproic acid which has been shown to be more effective in preventing manic relapse compared to valproic acid monotherapy, though not for depression relapse, where lithium monotherapy was very effective.
To date, the field of polypharmacy in bipolar illness maintenance treatment lacks longitudinal RCT in which patients who failed drug A respond better to A+B than to B alone. Progress in this area would serve to facilitate more rigorous prophylaxis against manic and depressive relapses. Future research efforts may lead to more grounded guidelines, which are greatly needed in a recurrent and disabling condition as bipolar illness. Sets