Number needed to treat analyses of drugs used for maintenance treatment of bipolar disorder
Due to the episodic and chronic nature of bipolar disorder (BD), maintenance therapy represents a critical part of treatment; however, there is a paucity of studies comparing effectiveness of available long-term treatments.
The aim of this study is to determine and compare the efficacy of pharmacological treatments for maintenance treatment of BD by means of the number needed to treat (NNT).
The efficacy of drugs used for maintenance treatment of BD, as emerging from the results of randomized controlled trials, was assessed using the size effect measure of NNT. PubMed searches were conducted on English-language articles published until May 2010 using the search terms “bipolar disorder,” “mania,” “mixed episode,” or “bipolar depression,” cross-referenced with trial characteristic search phrases and generic names of medications. The search was supplemented by manually reviewing reference lists from identified publications.
In 15 studies, aripiprazole, olanzapine, quetiapine, risperidone long-acting injection, lithium, lamotrigine, and divalproex proved effectiveness in terms of NNTs (≥10% advantage over placebo) for prevention of relapse into any mood episode. Quetiapine, lithium, risperidone long-acting injection, aripiprazole, and olanzapine are effective in manic recurrence prevention. Lamotrigine, quetiapine, and lithium present significant NNTs for prevention of depressive relapses.
All of the pharmacological agents assessed were effective in the prevention of any kind of mood episode; however, different efficacy profiles were found for prevention of manic and/or depressive relapses. The comparison of NNT values of the available agents may represent a useful tool in clinical settings, in order to facilitate implementation of long-term pharmacological interventions in patients with BD.
KeywordsBipolar disorderNNTTreatment efficacyMaintenance treatment
The ongoing research on bipolar disorder (BD) has highlighted its pervasive and debilitating nature, characterized by lifelong recurrent episodes and residual intraepisodic symptomatology (Keck et al. 2007). Recurrence rates can reach up to 49% within 2 years of recovery from an initial episode (Perlis et al. 2006), with the polarity of the index episode which tends to predict the polarity of relapse and the risk of recurrence that increments with the number of mood episodes (Calabrese et al. 2004).
Management of BD after acute treatment of mood episodes entails first continuation therapy aiming to prevent relapses, followed by maintenance therapy focusing on prevention of recurrences (Calabrese et al. 2006). The maintenance treatment of BD continues to represent a major clinical issue. The current first-line treatment strategies for long-term treatment of bipolar disorder are represented by lithium, lamotrigine, valproate, olanzapine, quetiapine in monotherapy and as adjunctive therapy, aripiprazole for the prevention of manic events, risperidone long-acting injection monotherapy and as adjunctive therapy, and adjunctive ziprasidone for the prevention of mood events (Yatham et al. 2009).
Monotherapy trials against placebo remain the gold standard design for determining efficacy in BD (Goodwin et al. 2008). A recommended tool for reporting results of clinical trials for bipolar disorder is represented by the number needed to treat (NNT) analysis (Martinez-Aran et al. 2008). NNT summarizes the effect of treatment in terms of the number of patients a clinician needs to treat with a particular therapy to expect to prevent one adverse event. NNT is a measure of effect size, and calculation of the NNT can quantify the clinical relevance of a statistically significant study result (Citrome 2008). Although NNT has been described as “the least misleading and most clinically useful measure of treatment effectiveness” (Gray 2004), it is considered likely to help translate efficacy-driven clinical data to information that will more readily guide clinicians on the benefits of specific interventions in BD (Martinez-Aran et al. 2008). As a matter of fact, most up-to-date clinical trials in bipolar disorder have not included NNT analyses.
The aim of the present study was to evaluate and compare the effects of the pharmacological treatments used for the long-term management of bipolar disorder by the means of NNT.
We systematically reviewed randomized controlled trials (RCTs) of medications used for the treatment of any phase of BD. A comprehensive PubMed search of all English-language articles published up to April 2010 was performed. Terms “bipolar disorder,” “mania,” “mixed episode,” or “bipolar depression” were cross-referenced with trial characteristics search phrases and generic names of medications used for maintenance treatment of BD. The search was supplemented by manually reviewing reference lists from the identified publications.
Included RCTs assessed the effectiveness of drugs in the treatment of BD compared to placebo with a minimal duration of 6 months and in patients aged over 18. Exclusion criteria were a small sample size (meaning median sample inferior to 16.5 subjects in each group as suggested by Richy et al. 2004), a study sample not exclusively composed of bipolar patients, the use of rating scales not validated in patients with bipolar disorder, and an absence of a placebo control group.
NNT for prevention of relapse into any episode, mania, and depression for the drugs used for long-term management of BD compared to placebo were calculated from the reports of all the studies that fulfilled the inclusion criteria.
NNT can be expressed as the reciprocal of the absolute risk reduction, and is calculated by taking the reciprocals of the differences between the rates of the outcomes for two interventions, and rounding upwards to the next whole number. Only NNTs <10 are considered clinically meaningful. Lower NNTs reflect larger clinical difference between the comparisons. NNTs of 3, 4 and 9, corresponding to values of Cohen’s d of 0.8, 0.5 and 0.2, represent “large,” “medium,” and “small” effect sizes (Kraemer and Kupfer 2006; Cook and Sackett 1995).
In addition, a 95% confidence interval (CI) for the NNT was calculated, constructed by inverting and exchanging the limits of a 95% CI for the absolute risk reduction (Cook and Sackett 1995). When the treatment effect is significant at the 5% level, the 95% CI for the absolute risk reduction will not include zero, and thus the 95% CI for the NNT will not include infinity (Altman 1998).
Two competing methods have been proposed in order to calculate NNT from results of systematic reviews; one method involves calculating the NNT from meta-analytical estimates, the other by treating the data as if it all arose from a single trial. Altman and Deeks (2002) found the “treat-as-one-trial” method to be susceptible to bias when there were imbalances between groups within one or more trials in the meta-analysis (Simpson’s paradox). In this paper, we have followed their suggestion not to use the treat-as-one-trial method of calculating NNTs, thus, when more RCTs regarding a single agent were available, we reported NNTs for data reported in each trial separately.
We have proceeded to calculate NNTs for the prevention of any kind of recurrences, followed by NNTs of manic and depressive episodes for each treatment assessed, as reported in the original studies.
Fifteen out of 67 available studies have met the inclusion criteria. Patient characteristics and primary outcomes of the selected studies have been reported in the original publications. All studies included an acute treatment phase, followed by a double-blind, relapse-prevention (maintenance) treatment phase. Some of the RCTs used a three-arm design thus could be used to make two comparisons each. In some cases, two or more articles/references provide data for the same RCT.
Characteristics of included randomized controlled studies
Trial (in order of appearance in text)
Patient inclusion criteria (maintenance phase)
Dosage (mg/day) or plasma levels/mean dosage
Keck et al. (2007)
Bipolar I ≥ 18 years
ARI = 78
ARI, 15–30 mg/day; mean, 23.8 mg/day
YMRS ≤ 10
PLA = 83
MADRS ≤ 13
No hospitalization in previous 3 months
Tohen et al. (2006)
OLZ = 225
OLZ, 5–20 mg/day
PLA = 136
YMRS ≤ 12
HAM-D ≤ 8
Two prior mixed or manic episodes in the past 6 years
Tohen et al. (2004)
LI/VPA + PLA = 48
OLZ, 5–20 mg/day; mean, 12.5 mg/day
LI/VPA + OLZ = 51
LI, 0.66–0.86 mEq/L
YMRS ≤ 12
VPA, 60.1–73.8 μg/mL
HRSD-21 ≤ 8
Vieta et al. (2008a)
QUE + LI/VPA = 336
QUE, 400–800 mg/day; mean, 497 mg/day
PLA + LI/VPA = 367
LI, 0.5–1.2 mEq/L
YMRS ≤ 12
VPA, 50–125 μg/mL
HAM-D ≤ 12
Suppes et al. (2009)
QUE + LI/VPA = 310
QUE, 400–800 mg/day; mean, 519 mg/day
PLA + LI/VPA = 313
LI, 0.5–1.2 mEq/L
YMRS ≤ 10
MADRS ≤ 13
Mean, 0.71–0.74 mEq/L
VPA, 50-125 μg/mL
Mean, 68.91–71.38 μg/mL
Weisler et al. (2008)
QUE = 404
QUE, 300–800 mg/day
LI = 364
Li, 0.6–1.2 mEq/L
YMRS ≤ 12
PLA = 404
MADRS ≤ 12 Acute current or recent (past 26 weeks) manic, depressive, or mixed index episode treated with QUE
Quiroz et al. (2010)
RLAI = 140
RIS, 12.5–50 mg i.m.; mean, 25 mg
PLA = 136
Recent manic/mixed episode or stable patients with ≥ 1 mood episode in past 4 months
Macfadden et al. (2009)
RLAI + TAU = 65
RLAT, 25–50 mg/2 weeks
PLA + TAU = 59
≥4 episodes in the past year
Bowden et al. (2010)
ZIP + LI/VPA = 127
ZIP, 80–160 mg/day
PLA + LI/VPA = 113
LI, 0.6–1.2 mEq/L
Current or recent manic/mixed episode
Mean, 0.7–0.9 mEq/L
VPA, 50–125 μg/mL; mean, 67.4–72.8
Bowden et al. (2003)
LAM = 59
LAM, 100–400 mg/day
LI = 46
LI, 0.8–1.1 mEq/L
Current or recent (hypo)mania ≥ 1 additional (hypo)manic and 1 depressive episode in the past 3 years
PLA = 70
Calabrese et al. (2003)
LAM = 221
LAM, 50–400 mg/day; mean, 200 mg/day
LI = 121
LI, 0.8–1.1 mEq/L; mean, 0.8 ± 0.3 mEq/L
Current or recent MDE
PLA = 121
≥1 additional (hypo)manic and 1 depressive episode in the past 3 years
Calabrese et al. (2000)
Bipolar I and II, rapid
LAM = 90
LAM, 100–300 mg/day
PLA = 87
<3 on item 3 HAM-D stable for 4 weeks
Prien et al. (1973)
Manic–depressive, manic type
LI = 101
LI, 0.5–1.4 mEq/L
PLA = 104
Bowden et al. (2000)
VPA = 187
VPA, 71–125 μg/mL
LI = 90
LI, 0.8–1.2 mmol/L
Manic episode ≤ 3 months before randomization
PLA = 92
MRS ≤ 11
DSS ≤ 13
GAS > 60, no serious suicidal risk
Vieta et al. (2008b)
Bipolar I or II
OXC + LI = 26
OXC, 1,200 mg/day
PLA + LI = 29
LI, 0.6 mEq/L
YMRS ≤ 12
MADRS ≤ 20
No acute phases in 6 months
Relapse rates reported in randomized controlled trials of pharmacological agents used for maintenance treatment of bipolar disorder
Any episode n (%)
Mania n (%)
Depression n (%)
(Keck et al. 2007)
(Tohen et al. 2006)
Olanzapine combined with
lithium/divalproex (Tohen et al. 2004)a
Quetiapine combined with
Quetiapine combined with
lithium/divalproex (Suppes et al. 2009)
(Weisler et al. 2008)
(Quiroz et al. 2010)
Risperidone LAI +
treatment as usual (Macfadden et al. 2009)
Ziprasidone combined with
lithium/divalproex (Bowden et al. 2010)
(Bowden et al. 2003)
(Calabrese et al. 2003)
(Calabrese et al. 2000)
(Goodwin et al. 2004)
(Prien et al. 1973)
(Goodwin et al. 2004)
(Bowden et al. 2003)
(Calabrese et al. 2003)
(Bowden et al. 2000)
(Weisler et al. 2008)
(Bowden et al. 2000)
Oxcarbazepine combined with lithium
NNT values for recurrence prevention of pharmacological agents used for maintenance treatment of bipolar disorder compared to placebo
NNT any episode (95% CI)
NNT mania (95% CI)
NNT depression (95% CI)
Aripiprazole (Keck et al. 2007)
7 (3.6, 24.9)
50.0 (7.7, infinity)
Olanzapine (Tohen et al. 2006)
3 (2.3, 4.2)
5 (3.4, 8.8)
12 (5.3, infinity)
Olanzapine combined with Lithium/Divalproex (Tohen et al. 2004)a
6 (2.4, infinity)
12 (3.4, infinity)
6 (2.6, infinity)
7 (4.8, 10.1)
7 (4.9, 10.0)
Quetiapine combined with Lithium/Divalproex (Suppes et al. 2009)
4 (2.6, 4.1)
9 (5.7, 14.0)
6 (3.9, 7.7)
Quetiapine (Weisler et al. 2008)
3 (2, 2.6)
3 (2, 2.8)
4 (2.8, 4.2)
Risperidone LAI (Quiroz et al. 2010)
4 (2.4, 5.6)
4 (2.5, 5.0)
26 (8.6, infinity)
Risperidone LAI+Treatment as Usual (Macfadden et al. 2009)
5 (2.6, 15.7)
8 (4.0, 198.7)
16 (5.2, infinity)
Ziprasidone combined with Lithium/Divalproex (Bowden et al. 2010)
8 (4.2, 61.5)
15 (6.9, infinity)
56 (9.5, infinity)
Lamotrigine (Bowden et al. 2003)
5 (2.6, 17.0)
24 (5.0, infinity)
7 (3.3, 38.5)
Lamotrigine (Calabrese et al. 2003)
51 (7.6, infinity)
59 (10.0, infinity)
28 (6.9, infinity)
Lamotrigine (Calabrese et al. 2003)
7 (3.5, 108.8)
Lamotrigine (Goodwin et al. 2004)
11 (5.5, 1764.7)
26 (8.6, infinity)
20 (7.2, infinity)
Lithium (Prien et al. 1973)
4 (2.2, 5.1)
4 (2.5, 6.4)
22 (7.6, infinity)
Lithium (Goodwin et al. 2004)
7 (3.8, 17.7)
8 (4.6, 16.3)
49 (8.4, infinity)
Lithium (Bowden et al. 2003)
4 (2.3, 20.5)
6 (3, 26.4)
13 (4.1, infinity)
Lithium (Calabrese et al. 2003)
12 (4.7, infinity)
14 (6.3, infinity)
86 (7.4, infinity)
Lithium (Bowden et al. 2000)
14 (4.7, infinity)
69 (7.5, infinity)
17 (6.4, infinity)
Lithium (Weisler et al. 2008)
3 (2, 2.6)
3 (2.2, 2.7)
4 (2.8, 4.4)
Valproate (Bowden et al. 2000)
7 (3.9, 37.7)
22 (6.8, infinity)
11 (5.6, 74.3)
5 (2.2, infinity)
9 (3.0, infinity)
6 (2.5, infinity)
Pharmacological agents used in maintenance treatment of bipolar disorder
The only relapse-prevention study conducted on aripiprazole in BD I patients showed that aripiprazole 15–30 mg/day was significantly superior to placebo in prevention of relapse into any mood episode during a 100-week-long double-blind treatment, with an NNT value of 6 (Keck et al. 2007).Aripiprazole was significantly superior to placebo in preventing relapse to mania, which translates into an NNT of 7. No difference in prevention of depressive relapses was noted between aripiprazole and placebo.
Given that efficacy was shown for the prevention of any mood episode and for mania in particular, aripiprazole represents a first-line option for treatment and prevention of mania in maintenance treatment of bipolar disorder.
When comparing olanzapine to placebo over 12 months (Tohen et al. 2006), the NNT was three for prevention of symptomatic relapse to any mood episode, indicating a large clinical maintenance effect size difference between olanzapine and placebo. When olanzapine plus lithium or valproate was compared with placebo plus lithium or valproate over an 18-month period, the NNT for the prevention of any episode was not significant.
Furthermore, olanzapine showed a significant difference in the risk of relapse into manic episodes over placebo for up to 12 months (Tohen et al. 2006), corresponding to an NNT of 5; however, no significant differences emerged from the study comparing olanzapine combined with lithium or valproate with placebo combined with lithium or valproate over 18 months (Tohen et al. 2004). Olanzapine in monotherapy or combined with lithium or divalproex was not significantly superior to placebo in preventing relapse into depressive episodes (Tohen et al. 2006).
Overall, olanzapine in monotherapy has proven efficacy for prevention of any mood episode as well as manic episodes. Although olanzapine combined with lithium or valproate (Tohen et al. 2004) had an NNT of six for prevention of any mood episode, as well as of depression, it cannot be considered significant since CI crossed infinity.
Two studies (Vieta et al. 2008a; Suppes et al. 2009) assessing effectiveness of quetiapine as add-on therapy during continuation treatment for up to 104 weeks demonstrated that quetiapine in combination with lithium or divalproex was significantly more effective than lithium or divalproex alone in the prevention of mood episodes, which translates into an NNT of 4. In a recent study by Weisler et al. (2008), evaluating quetiapine as monotherapy for up to 104 weeks, quetiapine was found effective in the prevention of any mood episode, with an NNT of 3.
Quetiapine in combination with lithium or divalproex was significantly more effective than the placebo in combination with lithium or divalproex in preventing recurrence of mania, assuming an NNT value of seven in the study by Vieta et al. (2008a) and nine in the study by Suppes et al. (2009). Likewise, quetiapine in monotherapy (Weisler et al. 2008) was significantly more effective than placebo in the prevention of mania, with an NNT of 3.
NNTs were significant for quetiapine plus lithium or divalproex in preventing depressive relapses in the two studies evaluating quetiapine as adjunctive therapy, with values of 7 (Vieta et al. 2008a) and 6 (Suppes et al. 2009). Quetiapine in monotherapy assumed an NNT value of 4 in the prevention of depressive episodes (Weisler et al. 2008).
Taken together, these data indicate that quetiapine alone and combined with lithium or valproate was effective in preventing any mood episode and has a similar efficacy for prevention of mania and depression.
Risperidone long-acting injection
Although the long-term efficacy of oral risperidone has not yet been assessed (Yatham et al. 2009), two RCTs have examined the efficacy of risperidone long-acting injection (RLAI) for maintenance treatment in BD (Quiroz et al. 2010; Macfadden et al. 2009).
A recent study examined the long-term efficacy of RLAI (Quiroz et al. 2010) in patients with recent manic or mixed episode followed up for up to 24 months.
RLAI monotherapy was shown to be superior to placebo in preventing any mood episode with an NNT of 4. A study conducted by Macfadden et al. (2009) assessed RLAI as an adjunct to treatment as usual for 52 weeks in 139 patients who had frequently relapsing BD. Significantly fewer patients in the adjunctive RLAI group relapsed into any mood episode compared with those in the placebo group, corresponding to an NNT of 5.
Overall, RLAI has proven efficacy for the prevention of any kind of mood episodes, as well as for the prevention of mania while it did not result effective for the prevention of depressive episodes.
Two RCTs compared lamotrigine, lithium, and placebo in maintenance treatment of recently manic (Bowden et al. 2003) and depressed (Calabrese et al. 2003) bipolar I patients. Since the studies were prospectively designed for combined analyses, a pooled analysis from the two abovementioned studies (Goodwin et al. 2004), allowing greater power with respect to the original studies, was available for NNT analysis and therefore included in the present paper. An additional double-blind, placebo-controlled study examining lamotrigine as maintenance monotherapy for rapid-cycling bipolar patients was available for NNT analysis exclusively for prevention of any mood episode (Calabrese et al. 2000).
When compared to placebo, lamotrigine in monotherapy was found to be effective in the prevention of any mood episode in recently manic patients (Bowden et al. 2003) with an NNT of 5, and showed no significant difference in recently depressed patients (Calabrese et al. 2003). Data emerging from the pooled analyses (Goodwin et al. 2004) evidenced an NNT of 11 for prevention of any mood episode.
Lamotrigine was found to be superior to placebo in preventing depressive episodes in recently manic patients (Bowden et al. 2003), with an NNT value of 7. In contrast, lamotrigine was not significantly superior to placebo in the prevention of depressive episode neither in recently depressed patients (Calabrese et al. 2003) nor in the pooled analyses (Goodwin et al. 2004).
Lamotrigine in monotherapy was found effective in preventing any kind of mood episode. Its effectiveness was also demonstrated for the prevention of depressive episodes in recently manic patients.
Although lithium has been considered to be the cornerstone of bipolar disorder maintenance treatment for many years, since the mid 1970s until 2000, no RCTs assessing lithium efficacy were published (Coryell 2009).
Only one of the studies on lithium conducted in the 1970s satisfied the inclusion criteria. A double-blind trial by Prien et al. (1973) assessed the efficacy of lithium over a 2-year period. NNT analyses relevant to a 1-year period evidenced effectiveness of lithium in preventing any mood episode, translating to an NNT of 4.
Four recent double-blind studies examined the efficacy of lithium as a maintenance treatment (Bowden et al. 2000, 2003; Calabrese et al. 2003; Weisler et al. 2008). In the study by Bowden et al. (2000), lithium was not significantly superior to placebo in preventing mood episodes over 1 year in a cohort of 372 patients with BD I. In two studies (Bowden et al. 2003; Calabrese et al. 2003), both previously described, lithium was studied as an active comparator medication compared to a lamotrigine-enriched sample of patients with BD I. In the study by Bowden et al. (2003), lithium was found to be significantly more effective than placebo in preventing any mood episode, with an NNT of 4 while in recently depressed BD I patients the difference was not significant. In a pooled analysis of the two studies (Goodwin et al. 2004), lithium was significantly more effective than placebo in preventing any mood episode, with an NNT value of 7. In a recent RCT conducted by Weisler et al. (2009) where bipolar I patients were followed up for up to 104 weeks, lithium, assessed as active comparator, has proven its effectiveness in preventing any kind of mood episode, with an NNT of 3.
Regarding the prevention of manic episodes, in the study by Prien et al. (1973) lithium was effective in the prevention of manic episodes, with an NNT of 4. Lithium was also found to be significantly more effective than placebo in preventing mania in recently manic patients (Bowden et al. 2003), with an NNT of 2, as well as in pooled analyses of the two studies (Goodwin et al. 2004). Likewise, in the study by Weisler et al. (2009), lithium assumed an NNT value of three for mania prevention; however, lithium was not significantly superior to placebo in preventing manic recurrences in the remaining two studies (Bowden et al. 2000; Calabrese et al. 2003).
Lithium was found significantly superior to placebo in preventing depressive episodes, with an NNT of 4, only in the study by Weisler et al. (2009) while it resulted not significantly superior to placebo in preventing depression in any other of the abovementioned studies (Prien et al. 1973; Bowden et al. 2000; Bowden et al. 2003; Calabrese et al. 2003; Goodwin et al. 2004).
In conclusion, since the efficacy of lithium was largely demonstrated for the prevention of any mood episode, in particular for mania, lithium continues to represent a first-line maintenance treatment for bipolar disorder for the treatment and prevention of mania.
Only one study assessing valproate in maintenance treatment of bipolar disorder was available for NNT analyses. A randomized, double-blind, parallel-group multicenter study of treatment outcomes was conducted over a 52-week maintenance period (Bowden et al. 2000). Patients with a previous manic episode were randomized to maintenance treatment with divalproex, lithium, or placebo. Although the authors concluded that the treatment arms did not differ significantly on time to recurrence of any mood episode during maintenance therapy, the NNT for prevention of any mood episode was 7; however, valproate did not result superior to placebo in preventing neither manic nor depressive episodes.
Due to the episodic and chronic nature of the illness, maintenance therapy is a critical part of treatment for bipolar disorder. The ultimate goal of treatment is the prevention of episode recurrence through the long-term management of the symptoms (Thase 2008).
NNT is a useful and meaningful concept for practicing clinicians which may help to translate the results of randomized controlled clinical trials to evidence-based medicine (Martinez-Aran et al. 2008). Nonetheless, most of the published studies have not reported NNT values for assessed medicaments. To the best of our knowledge this is the first paper providing NNTs for all available placebo-controlled trials for agents used in long-term treatment of BD.
NNT analysis evidenced differences in efficacy profiles among various agents used for the maintenance treatment for bipolar disorder. In general, our findings are in agreement with the recommendations of recent guidelines for the maintenance therapy of bipolar disorder (Yatham et al. 2009). Aripiprazole, olanzapine, quetiapine, risperidone LAI, ziprasidone, lithium, lamotrigine, and valproate have single-digit NNTs and are significantly effective for the prevention of any mood episode, but show substantial differences as to their ability to prevent mania and depression separately. Among the agents assessed, the only exception is given by oxcarbazepine combined with lithium that had a significant single-digit NNT for prevention of any episode, mania, and depression, but CI did not result significant, probably for the lack of power due to the relatively small sample size.
The NNT analyses confirmed the efficacy of quetiapine (in monotherapy and combined with mood stabilizer), lithium, risperidone LAI, aripiprazole, and olanzapine in preventing manic recurrences. Lamotrigine, lithium, and quetiapine alone and combined with lithium/valproate presented significant NNTs for the prevention of depressive relapses. The data emerging from our analysis provide utter evidence that the treatments assessed differentiated in terms of whether they primarily prevent mania or depression or have bidirectional effects. The clinical relevance of the directional efficacy of the various medications is heightened in the context of predominant polarity, a parameter correlated with treatment response and outcome of later acute episodes (Vieta et al. 2009; Colom et al. 2006); nevertheless, placebo-controlled randomized clinical trials assessing long-term effectiveness of mood stabilizers and antipsychotics are surprisingly sparse, and important questions remain unanswered.
In order to present the most complete data available, we have included both RCTs assessing drugs in monotherapy as well as combined with mood stabilizers such as lithium and valproate. The studies were not completely homogeneous with respect to clinical characteristics of the sample (rapid-cycling course, manic/mixed states or depression, refractory patients or unbiased samples), sample size, and rates of study completion, which may compromise to some extent the generalizability of reported NNTs.
In the interest of brevity, we have not addressed the number needed to harm analysis, although medication selection is based on tolerability as well as effectiveness. NNTs for the prevention of mixed episodes were omitted since most available RCTs have not proceeded to such assessment and the number of events in the trials that did look at it was extremely small.
An apparent paradox emerges from NNT values analyses. Namely, it may seem surprising that NNT for the prevention of any mood episode is often smaller than NNT for the prevention of manic or depressive relapses separately; however, one should bear in mind that the included studies were designed to assess efficacy in terms of prevention of any mood episode as primary aim. NNT for prevention of any mood episode reflects the sum of effects that a drug has upon any episode, manic, depressive, mixed, or unknown type, conferring it more power than any episode taken singularly; thus a drug may perform better overall in comparison to its best effect upon a single episode.
Successful long-term management often appears to require combination treatment, and although assessing combined treatments went beyond the aims of our study, lack of such trials, e.g., lamotrigine as add-on treatment, should be an object of future research. Further studies are needed in order to address the long-term effectiveness of agents such as carbamazepine, oxcarbazepine, and valproate. Some antipsychotics, such as amisulpride, asenapine, or paliperidone have not been assessed in long-term placebo-controlled studies.
The present review of clinical effectiveness by the means of NNT comparison aimed to investigate all pharmacological treatments approved as maintenance therapy in BD. Most of the pharmacological agents assessed were effective in the prevention of any kind of mood episode; however, different efficacy profiles were found for the prevention of manic and/or depressive relapses. The comparison of NNT values of the available medicaments may represent a useful adjuvant in clinical settings, in order to facilitate implementation of long-term pharmacological interventions in patients with BD.