Commentary: Evidence-based pharmacotherapy of bipolar disorder

Transcription

1 International Journal of Neuropsychopharmacology (2003), 6, Copyright f 2003 CINP DOI: /S Commentary: Evidence-based pharmacotherapy of bipolar disorder SPECIAL SERIES S. Nassir Ghaemi, Federico Soldani and Douglas J. Hsu Cambridge Health Alliance, Cambridge, MA, USA and Harvard Medical School, Boston, MA, USA Abstract This Commentary summarizes findings from three other papers in this issue with recommendations for evidence-based treatment with lithium, anticonvulsants, antipsychotics, and antidepressants in bipolar disorder. We will also provide a summary of levels of evidence and examine two important methodological issues in assessing drug-induced mania: reliance on significance testing for assessment of sideeffects, and limitations of randomized controlled trials (RCTs) for assessing frequency of side-effects. If a study is not specifically powered and designed to assess a side-effect, then no significance testing should be conducted, and side-effects should simply be reported as effect estimates and confidence intervals. Further, RCTs only establish a categorical response to a research question, i.e. whether or not something happens. The frequency of an event (treatment response, side-effects) is often more accurately assessed with observational studies. Received 11 June 2003; Reviewed 19 June 2003; Revised 23 June 2003; Accepted 25 June 2003 Key words: Antidepressants, Bipolar disorder, clinical trials, depression, drug-induced mania, drug therapy, epidemiology, evidence, evidence-based medicine, mania, methodology. Introduction Levels of evidence Any discussion of the evidence basis for any subject requires an initial definition of the meaning of evidence. A major contribution of the evidence-based medicine (EBM) movement is to highlight the concept of levels of evidence. In Table 1, we summarize five levels of evidence, as adapted by our group from the EBM literature (Sackett et al., 2000). Our adaptation relates these levels to the kind of data one observes in psychiatric studies (Ghaemi and Soldani, 2003). The recognition of levels of evidence allows one to have a guiding principle by which to assess the literature. Basic rules are: (1) All other things being equal, a study at a higher level of evidence provides more valid (or powerful) results than one at a lower level. (2) Base judgements as much as possible on the highest levels of evidence. (3) Levels II and III are often the highest levels of evidence attainable for complex conditions, and are to be valued in those circumstances. (4) Higher levels of evidence do not guarantee certainty; any one study can be wrong, thus look for replicability. (5) Within any level of evidence, studies may conflict based on other methodological issues not captured by the parameters used to provide the general outlines of levels of evidence. Much of the confusion in conflicting reviews of the literature and opinions regarding pharmacotherapy in bipolar disorder can be resolved by a general acceptance and application of the concept of levels of evidence. In this Commentary, we intend to apply this approach to the papers published in this issue in Part 2 of the Special Section on New Advances in the Understanding and Treatment of Bipolar Disorder, as well as to make other methodological points of importance, as we summarize our interpretation of the evidence basis for pharmacotherapy of bipolar disorder. Evidence-based recommendations Address for correspondence: Dr S. N. Ghaemi, Cambridge Health Alliance, Department of Psychiatry, 1493 Cambridge Street, Cambridge, MA 02139, USA. Tel.: Fax: ghaemi@hms.harvard.edu Acute and maintenance treatment with mood stabilizers (Bowden, 2003) Bowden seeks to summarize the literature on lithium and anticonvulsants in bipolar disorder, and along

2 304 S. N. Ghaemi et al. Table 1. Levels of evidence Level I. Double-blind randomized trials Ia. Placebo-controlled monotherapy Ib. Non placebo-controlled comparison trials, or placebo-controlled add-on therapy trials Level II. Open randomized trials Level III. Observational studies IIIa. Non-randomized, controlled studies IIIb. Large non-randomized, uncontrolled studies (n>100) IIIc. Medium-sized non-randomized, uncontrolled studies (100>n>50) Level IV. Small observational studies (non-randomized, uncontrolled (50>n>10) Level V. Case series (n<10), Case report (n=1), Expert opinion Reprinted with permission (Ghaemi and Soldani, 2003). the way makes important methodological points. He concludes that there is an excellent basis for the use of lithium, valproate, and lamotrigine for the treatment of bipolar disorder, with combinations of these and other anticonvulsants as secondary back-ups in case of failure. He concludes also that lithium and valproate are primarily useful for manic symptoms and lamotrigine for depressive symptoms in bipolar disorder, both acutely and in maintenance. The literature on lithium s efficacy in bipolar depression, especially in prophylaxis, is complex. Bowden refers to a few papers, some of which were level II, but there are also a number of level I randomized clinical trials (RCTs) with good methodologies (parallel designs, placebo groups, a-priori outcome measures) that support lithium s efficacy in the prevention of depressive episodes in bipolar disorder (Kane et al., 1982; Prien et al., 1973, 1984; Quitkin et al., 1981). Sometimes it is argued that in these studies lithium was not compared directly with a placebo-only group. While this is a relevant issue, it does not argue against the other methodological strengths of these studies, or against the advantage of multiple replication in different samples. Thus, we would agree with the American Psychiatric Association (APA) treatment guidelines (Hirschfeld et al., 2002) that lithium or lamotrigine are appropriate first-line agents for bipolar depression, especially in prophylaxis, and that lithium or valproate are the best first-line mood stabilizers for acute mania, with or without concomitant antipsychotic use. These agents should be used indefinitely for their preventive benefit. If ineffective, there is some evidence of utility in combination, as well as with adjunctive use of other novel anticonvulsants or atypical antipsychotics (Ghaemi, 2002). Antipsychotic drugs in bipolar disorder (Ertugrul and Meltzer, 2003) Ertugrul and Meltzer provide a snapshot of the rapidly progressing literature on antipsychotic use in bipolar disorder. They conclude that these agents have moderate effect sizes and that atypical antipsychotics are preferable to traditional agents due to enhanced efficacy and fewer side-effects, with little differentiation between them as to preferred agent of use. We would agree with this overall assessment. All these agents are probably effective, although more data are available with olanzapine. Clinicians should primarily choose among these agents based on sideeffects, moving from one to another as warranted. Duration of treatment should probably be long-term if the patient completely responds for the acute phase of mania with an atypical antipsychotic agent, and perhaps short-term in those with incomplete acute response (Tohen et al., 2003). Recent placebo-controlled acute trials in bipolar depression (Muzina and Calabrese, 2003) Muzina and Calabrese raise some important methodological issues on research in bipolar depression. In their interpretation of this literature, they conclude that lamotrigine is effective and safe, and that some antidepressants, like paroxetine and moclobemide, are also safe and effective when used with mood stabilizers. We agree with those statements, but we would add that bupropion should also be added to the list of antidepressants with apparent low risks of acute mania, based on the data included in the paper by Muzina and Calabrese (2003). However, we would like to emphasize more clearly the very limited evidence of efficacy with antidepressants in bipolar disorder, especially in the prevention of depressive episodes. There are well-designed RCTs (parallel design, a- priori hypotheses, placebo groups) from the 1970s and 1980s (Kane et al., 1982; Prien et al., 1973, 1984; Quitkin et al., 1981) that demonstrate that tricyclic antidepressants (TCAs) failed to show benefit in prevention of depression in bipolar disorder, either alone in comparison to lithium, or when added to lithium. While Muzina and Calabrese also allude to the potential risks of antidepressants, it is worth emphasizing the available level I and level III data on potential worsening of the long-term course of bipolar disorder

3 Commentary: Evidence-based pharmacotherapy of bipolar disorder 305 in approx. 25% of patients, with induction of rapidcycling or more mood episodes over time (Ghaemi et al., 2002). Hence, this literature suggests inefficacy of antidepressants in long-term treatment, not just absence of evidence of efficacy. Combined with the suggestion of some risk of worsening, the case for caution with antidepressant use in bipolar disorder would seem to follow. Some (Moller and Grunze, 2000) are critical of the change in the current APA guidelines (Hirschfeld et al., 2002), which recommend avoidance of antidepressant use in bipolar depression unless necessary for severe episodes, but we think the evidence is consistent with this recommendation. We view lithium and lamotrigine as important firstline choices for bipolar depression, acutely and longterm, followed by valproate, carbamazepine, and other anticonvulsants and atypical antipsychotics as adjuncts. The role for antidepressants is, based on the evidence, supported only for acute major depressive episodes, but probably not for long-term treatment, again unless proven necessary in the individual case. In our experience, only approx. 20% of patients with bipolar disorder need antidepressants for long-term treatment (Ghaemi and Goodwin, 2001). The vast majority can be managed with at least moderate improvement with mood stabilizers in various combinations. Thus, mood stabilizers, like lithium and lamotrigine, can be used acutely and should be used longterm, while antidepressants should be used acutely when needed, but long-term only in the minority of cases. Among antidepressants, paroxetine and bupropion, and perhaps also moclobemide and pramipexole (Goldberg et al., 2002), have been shown to have the lowest risk of acute manic switch in level I studies, and thus should be preferred agents. TCAs should be generally avoided. Methodological issues in assessing drug-induced mania The above articles also raise the importance of understanding the methodology of research in assessing and obtaining evidence. We want to focus here on two methodological issues related to drug-induced mania. The problem of side-effects and significance testing Statistics can be applied to clinical medicine in two ways, as significance or hypothesis testing (in statistical language the expression hypothesis testing refers to a specific technique, and should not be strictly identified with its more general scientific meaning), and as effect estimates and confidence intervals. Unfortunately, clinicians often engage solely in significance testing, overdoing it and thereby exposing their work to errors of both omission and commission. The importance of simply reporting effect estimates is often unrecognized. A common error in assessing the results of a clinical trial or of an observational study is to evaluate sideeffects across patient groups based on whether or not they differ statistically (e.g. drug vs. placebo group). However, as Muzina and Calabrese (2003) mention, most clinical studies are not powered to assess sideeffects, especially when side-effects are not frequent. Thus, significance testing is not appropriate, since the risk of a false negative finding using this technique in isolation is too high. In other words, when an investigation is designed, researchers have to choose one or a few primary outcome measures for which the study should be properly powered [an arbitrary level of 0.80 or 0.90 (power=1xtype II error) is a standard convention]. Usually, there is a main efficacy outcome measure, with one or two secondary efficacy outcome measures, as described well by Bowden (2003). An efficacy outcome measure or a side-effect to be tested can be established either a priori (always the case for primary and secondary outcomes) or post hoc. For instance, a study of acute bipolar depression with lamotrigine can have an acute efficacy measure as a primary outcome, and occurrence of acute manic switch as another outcome. Researchers should avoid too many a-priori outcome measures to avoid the type I error of false positive findings; that is, if they employ significance testing at a type I error level of 0.05 as their criterion of significance (against which the calculated p value will be compared), then for each outcome they assess, they need to apply a smaller type I error level (such as when adopting the Bonferroni correction) to avoid overestimating the positive results based on chance. Other analyses, such as subgroup effects (e.g. bipolar I vs. bipolar II), are usually underpowered and post hoc and can only be seen as exploratory. Significance (hypothesis) testing does not apply to them; they do not prove hypotheses, they only provide more suggestive evidence which then needs to be tested in future a-priori fashion by properly powered studies. Hence we do not think side-effects should be interpreted based on p values and significance testing because of the high false negative (type II) error risk. They are not hypotheses to be tested, but simply observations to be reported. The appropriate statistical

4 306 S. N. Ghaemi et al. Table 2. Treatment-emergent mood events: all controlled studies to date Lamotrigine a (n=379) Placebo b (n=314) Test statistic Relative risk 95% CI Hypomania 2.1% 1.9% x 2 =0.01, p= Mania 1.3% 0.3% x 2 =1.01, p= Mixed episode 0.3% 0.3% x 2 =0.33, p= All events 3.7% 2.5% x 2 =0.41, p= a Bipolar disorder (n=232); unipolar disorder (n=147). b Bipolar disorder (n=166); unipolar disorder (n=148). From Bowden (2003). approach is to report the effect size (e.g. percent) with 95% confidence intervals (the range of expected estimated observations based on repeated studies). These issues are directly relevant to the question of whether a drug has a risk of causing mania. In the case of lamotrigine, for instance, Bowden (2003) reviews the clinical trials, which fail to find a difference with placebo. However, it is not evident to us that those studies were designed to detect such a difference. In Table 2, we converted the data on lamotrigine and induced switch to relative risks and 95% confidence intervals, and also x 2 significance testing for comparison. While it may indeed be the case that lamotrigine does not carry a higher risk than placebo, it is of concern that the overall risk of pure manic episodes (1.3%) is 4-fold higher than placebo (0.3%; relative risk=4.14; 95% CI, ). These confidence intervals indicate that while there may be no increased risk with lamotrigine (i.e. on repeated testing the relative risk could be as low as 0.49, meaning approx. 2-fold less risk with lamotrigine than placebo), the likelihood is that there is some risk with lamotrigine of inducing pure mania (observed to be approx. 4-fold, but which could be as high as 35-fold increased risk in repeated studies). The sample size required to statistically detect (i.e. using significance hypothesis testing procedures) the observed 4-fold difference in pure mania would be achieved with a study comparing two arms of almost 1500 patients each (at a type II error level of 0.80, with the statistical assumptions of no drop-outs, perfect compliance, and equal-sized arms). To give another example, if we accept a spontaneous baseline manic-switch rate of approx. 5% over 2 months of observation, and further assume that the minimal clinically relevant difference to be detected is a doubling of all events at a 10% rate in the lamotrigine group, the required sample size of a study properly powered to statistically detect this clinically significant difference should be almost 1000 overall (assuming no drop-outs, perfect compliance and equal-sized arms). Only with such a sample could we be confident enough that a reported p value greater than 0.05 really does reflect a substantial, clinical equivalence of lamotrigine and placebo in causing acute mania. The current pooled data presented by Bowden (2003) involved 693 patients (Table 2), which is somewhat more than half the needed sample, but even larger samples would be needed due to the statistical assumptions requiring no drop-outs, full compliance, and equal sample size in both arms. In summary, one cannot assume no difference when studies are not designed to test a hypothesis. The problem of RCTs and frequency of events RCTs are designed to answer one or a few questions, as noted above. They provide categorical, yes/no, answers to those questions. When an RCT finds that drug X leads to a more rapid reduction in depression symptoms than placebo, one can conclude, perhaps, that drug X is better than placebo for depression. The next question, which clinicians need to know, is How much better? The RCT gives an answer, but the answer may or may not be accurate. In other words, in the standard epidemiological summary, the strength of RCTs is that they are the closest thing we have to demonstrating causation; i.e. they can tell us if something happens. Their weakness is generalizability, which stems from their strength. One can only establish something akin to causation with randomization, and usually also double-blinds, use of placebos, rigorous protocols, avoidance of dropouts and non-compliance, and so on. All these strictures of clinical trials result in the selection of highly qualified research subjects. As Bowden (2003) notes, this often excludes many, if not most, patients seen in real-world clinical practice. Hence, there is a potential lack of generalizability. Usually, RCTs

5 Commentary: Evidence-based pharmacotherapy of bipolar disorder 307 under-report side-effects due to the extreme homogeneity and medically healthy nature of research subjects. The classic example is sexual dysfunction with serotonin reuptake inhibitor (SRI) antidepressants, which was reported to occur infrequently in RCTs, but which was shown in real-world practice to occur in over 50% of patients. Studies of levels II IV provide what level I RCTs lack: generalizability; but their weakness is validity. In the absence of level I data, the experience of data from levels II IV is still subject to substantial doubt as to whether it actually is correct or not. However, if level I data demonstrate that something happens, then data from levels II IV, often observational, can tell us how frequently that thing happens. This issue is directly relevant also to the problem of frequency of antidepressant-induced mania. Muzina and Calabrese (2003) note differing numbers in various RCTs, ranging from 0 to 67%. But we would suggest that none of those figures are very informative. All we can conclude from the RCTs is that antidepressants do indeed cause mania, and that the relative frequency of some agents is less than others. But the exact frequencies are anyone s guess, if we restrict ourselves to RCTs. Many clinicians and experts interpret these studies, especially the more recent studies on new-generation antidepressants such as those emphasized in the paper by Muzina and Calabrese (2003), as demonstrating low rates of antidepressantinduced mania, often in the 0 10% range. However, as noted above, such side-effects are notoriously underreported in RCTs and cannot be generalized to the overall population of patients with an illness. Welldesigned level III studies demonstrate that the acute mania/hypomania risk with SRI antidepressants is probably around 20% (Henry et al., 2001), which may be less than the 30 60% reported in the past with TCAs (Goodwin and Jamison, 1990), but is certainly enough to continue to be a notable risk. Acknowledgements Dr Ghaemi was partially supported by NIMH grant no. MH and Dr Soldani by the PhD programme of the University of Pisa (Department of Psychiatry, School of Medicine). References Bowden C (2003). Acute and maintenance treatment with mood stabilizers. International Journal of Neuropsychopharmacology 6, Ertugrul A, Meltzer HY (2003). Antipsychotic drugs in bipolar disorder. International Journal of Neuropsychopharmacology 6, Ghaemi SN (Ed.) (2002). Polypharmacy of bipolar disorder. In: Polypharmacy in Psychiatry. New York: Marcel Dekker. Ghaemi SN, Goodwin FK (2001). Long-term naturalistic treatment of depressive symptoms in bipolar illness with divalproex vs. lithium in the setting of minimal antidepressant use. Journal of Affective Disorders 65, Ghaemi SN, Ko JY, Goodwin FK (2002). Cade s disease and beyond: misdiagnosis, antidepressant use, and a proposed definition for bipolar spectrum disorder. Canadian Journal of Psychiatry 47, Ghaemi SN, Soldani F (2003). Meta-analysis of observational studies: the case of rapid-cycling bipolar disorder. Acta Psychiatrica Scandinavica 108, 1 3. Goldberg JF, Burdick KE, Endick CJ (2002). A placebocontrolled trial of pramipexole for bipolar depression [Abstract]. 155th Annual Meeting of the American Psychiatric Association, Philadelphia, PA, May. Goodwin FK, Jamison KR (1990). Manic Depressive Illness. New York: Oxford University Press. Henry C, Sorbara F, Lacoste J, Gindre C, Leboyer M (2001). Antidepressant-induced mania in bipolar patients: identification of risk factors. Journal of Clinical Psychiatry 62, Hirschfeld RMA, Bowden CL, Gitlin MJ, Keck PE, Perlis RH, Suppes T, Thase ME (2002). American Psychiatric Association practice guideline for the treatment of patients with bipolar disorder (revision). American Journal of Psychiatry (Suppl.) 159, Kane JM, Quitkin FM, Rifkin A, Ramos-Lorenzi JR, Nayak DD, Howard A (1982). Lithium carbonate and imipramine in the prophylaxis of unipolar and bipolar II illness: a prospective, placebo-controlled comparison. Archives of General Psychiatry 39, Moller HJ, Grunze H (2000). Have some guidelines for the treatment of acute bipolar depression gone too far in the restriction of antidepressants? European Archives of Psychiatry and Clinical Neuroscience 250, Muzina DJ, Calabrese JR (2003). Recent placebo-controlled acute trials in bipolar depression: focus on methodology. International Journal of Neuropsychopharmacology 6, Prien RF, Klett CJ, Caffey EM (1973). Lithium carbonate and imipramine in prevention of affective episodes. Archives of General Psychiatry 29, Prien RF, Kupfer DJ, Mansky PA, Small JG, Tuason VB, Voss CB, Johnson WE (1984). Drug therapy in the prevention of recurrences in unipolar and bipolar affective disorders: a report of the NIMH Collaborative Study Group comparing lithium carbonate, imipramine, and a lithium carbonate-imipramine combination. Archives of General Psychiatry 41, Quitkin FM, Kane J, Rifkin A, Ramos-Lorenzi JR, Nayak DV (1981). Prophylactic lithium carbonate with and without

6 308 S. N. Ghaemi et al. imipramine for bipolar 1 patients. Archives of General Psychiatry 38, Sackett DL, Straus SE, Richardson WS, Rosenberg W, Haynes RB (2000). Evidence-based Medicine (2nd edn). London: Churchill Livingstone. Tohen MF, Bowden CL, Calabrese JR, Sachs GS, Jacobs T, Baker RW, Evans AR (2003). Olanzapine versus placebo for relapse prevention in bipolar disorder [Abstract]. 156th Annual Meeting of the American Psychiatric Association, San Francisco, CA, USA, May.