Bipolar disorder is a complex and recurrent condition characterised by episodes of mania or hypomania which can alternate or co-occur with episodes of depression. In between episodes, people with the disorder may experience recovery and a return to usual social and occupational functioning, however sub-threshold symptoms are common and rates of recurrence are high (Oswald et al., 2007). For example, in a naturalistic study, 68% of participants with bipolar disorder were found to relapse within 4 years of discharge from a psychiatric community hospital (Simhandl et al., 2014). As such, long-term treatment is usually required to prevent future episodes and minimise sub-threshold symptoms.

Lithium has traditionally been the most common treatment for the long-term management of bipolar disorder, having been used for over 40 years. Yet a wide range of other drugs are often used in clnical practice. These include antipsychotics, anticonvulsants and antidepressants such as olanzapine, risperidone, carbamazepine and fluoxetine. Polypharmacy is also common, as clinicians strive to prevent both types of episodes.

Due to this wide ranging pool of treatment options, Miura and colleagues (2014) conducted a network meta-analysis to investigate the comparative efficacy and tolerability of various drugs (and their combination) for the long-term maintenance treatment of bipolar disorder.

Methods

The authors used a network meta-analysis to combine study outcomes. Whereas usual meta-analyses can only compare two interventions, network meta-analyses can use direct and indirect associations between interventions to estimate the relative efficacy between multiple interventions, as long as there is a common comparator in the network.

To be eligible for inclusion the studies had to:

• Be randomised controlled trials
• Compare any mood stabiliser, antipsychotic, anticonvulsant or antidepressant with placebo or an active comparator
• Have at least 12 weeks follow-up data
• Include participants who were not in an acute episode (prophylaxis design), or include participants who had responded to the investigational drug during an acute episode and were subsequently randomised to continue with the same treatment or switch to placebo/other comparator (an enrichment design)

Trials in which participants were randomised to long-term treatment during an acute phase were excluded (continuation design).

The primary outcomes were:

• The number of participants who relapsed (any type) as described by the study investigators (treatment efficacy)
• The number of participants who discontinued treatment due to side effects (treatment tolerability)

The secondary outcomes were:

• The number of participants who had a depressive episode
• The number of participants who had a manic, hypomanic or mixed episode
• The number of participants who discontinued treatment for any reason including relapse (treatment acceptability)
• The number of participants who killed themselves
• Social functioning

Risk of bias was assessed using the Cochrane Collaboration risk of bias tool, and the quality of each outcome was evaluated using the GRADE framework.

The authors carried out pairwise meta-analyses of direct evidence and a network meta-analysis, both using random effects models, and calculated relative risks (RR). Sensitivity analyses were also included to investigate the effect of publication year, subtypes of bipolar disorder, rapid-cycling course of illness, enrichment design, sponsorship bias, duration of follow-up and blinding of treatment group.

The network meta-analysis included 33 trials of 6,846 participants. 17 different interventions were included in the network, these were:

1. Aripiprazole
2. Carbamazepine
3. Fluoxetine
4. Imipramine
5. Lithium
6. Lithium & Imipramine
7. Lithium & Oxcarbazepine
8. Lithium & Valproate
9. Lamotrigine
10. Aripiprazole & Lamotrigine
11. Valproate & Lamotrigine
12. Olanzapine
13. Paliperidone
14. Quetiapine
15. Risperidone long-acting injection (LAI)
16. Valproate
17. Valproate & Aripiprazole

Results

Primary outcomes

• For any relapse, all interventions were found to be better than placebo except for Aripiprazole, Carbamazepine, Imipramine and Paliperidone.
• For discontinuation due to side effects, placebo and Lamotrigine were significantly better than Carbamazepine, Lithium and Lithium & Valproate.

The quality of comparisons was assessed for the outcome ‘any mood relapse’ using the GRADE framework. Out of 13 comparisons, two were rated as moderate (Lithium and Olanzapine), nine as low and two as very low.

Secondary outcomes

• When looking at depressive and manic relapse separately, only Lithium and Quetiapine were found to be better than placebo for both types.
• Olanzapine was significantly better than placebo for manic relapse but not for depressive relapse.
• The relative risks for completed suicide and social functioning were not calculated. For suicide this was because the rates were too low. For social functioning this was because only 5 studies reported the outcome.

Sensitivity analysis

There was little or no change in estimates when less weight was given to studies with an enrichment design, sponsorship from pharmaceutical companies. When studies with a follow-up of less than 52 weeks and open-label studies were removed from the analysis there was little or no change to estimates.

Conclusions

The authors concluded:

The results of our systematic review show the superiority of Lithium in all three efficacy outcomes. Our analysis seems to have had higher statistical power than previous analyses, because several new trials have been published since the previous reviews were done and because we used the network meta-analytical method.

Second, we were able to delineate the efficacy profiles of some newly examined compounds including Quetiapine, Olanzapine, Risperidone long-acting injection, or Lamotrigine, for which the previous reviews did not have enough randomised evidence.

Strengths and limitations

This was a good quality review which followed PRISMA reporting guidelines and assessed the quality of included evidence using the GRADE framework.

The main limitation lies in the quality of evidence. Out of 13 comparisons, 11 were rated as low or very low, indicating poor confidence in the estimates.

Another limitation is the inclusion of RCTs with different conceptual designs. The first type were trials in which participants were recruited and randomised to interventions whilst they were euthymic (symptoms were stable and they were not in an acute episode). In the other type of trial (enrichment design), participants were recruited to the study whilst in an acute episode and put on the investigational drug. Subsequently, only participants who responded to this drug (i.e. showed a reduction in symptoms in a particular time-frame) were then randomised to continue receiving the drug or to switch to placebo/other comparator. This type of research design is often used in studies sponsored by pharmaceutical studies as they tend to favour outcomes for the investigational drug. To overcome this methodological limitation, Miura and colleagues (2014), carried out a sensitivity analysis where they gave half the weight to studies with an enrichment design and found no changes in estimates. However, as these trials made up 58% of included studies, it would have been of interest to see what effect their removal would have had to the network results.

One other limitation relates to the definition of relapse/recurrence in included studies. There was a great variation in definitions across studies ranging from ‘meeting DSM criteria for a mood episode’, to ‘requiring extra medication’, to scoring above a certain threshold depression and mania rating scales (which again varied from study to study). Whilst this is mostly a conceptual limitation which is difficult to overcome, it would have been interesting to include this factor in the sensitivity analysis.

Summary

Overall, this was a good quality review which attempted to provide clarity to a very fragmented area of research. The findings generally support the use of Lithium as a first-line treatment for the prevention of manic and depressive episodes in people with bipolar disorder. Whilst other drugs, such as Quetiapine and Olanzapine, were found to be effective compared with placebo, the evidence for lithium was the most robust and least biased. These findings are in line with the recent NICE Bipolar guideline (NICE, 2014) which also recommended Lithium as the first-line maintenance treatment for bipolar disorder.

This review highlights the need for more consistency in the design of maintenance trials including definitions of relapse and recurrence. In addition, out of 33 included studies only five reported social functioning outcomes. As recovery encompasses improvements in social lives, employment and quality of life in addition to the prevention of relapse, there is a greater need for such outcomes to be included in future maintenance trials.

Links

Miura T, Noma H, Furukawa TA, Mitsuyasu H, Tanaka S, Stockton S, Salanti G, Motomura K, Shimano-Katsuki S, Leucht S, Cipriani A, Geddes JR, Kanba S. Comparative efficacy and tolerability of pharmacological treatments in the maintenance treatment of bipolar disorder: a systematic review and network meta-analysis. The Lancet Psychiatry – 1 October 2014 (Vol. 1, Issue 5, Pages 351-359 ) DOI: 10.1016/S2215-0366(14)70314-1. [Abstract]

Simhandl, C., Konig, B., Amann, B.L. A prospective 4-year naturalistic follow-up of treatment and outcome of 300 bipolar I and II patients (PDF). Journal of Clinical Psychiatry, 2014;75(3):254-263.

Oswald, P., Souery, D., Kasper, S., Lecrubier, Y., Montgomery, S., Wyckaert, S., Zohar, J., Mendlewicz, K. Current issues in bipolar disorder: a critical review. European Neuropsychopharmacology, 2007;17(11):687-95. [Abstract]

Bipolar disorder: the assessment and management of bipolar disorder in adults, children and young people in primary and secondary care. NICE clinical guideline 185, Sep 2014.