One‐year outcome with antidepressant treatment of bipolar depression – is the glass half empty or half full?

We as providers focus on what we are able to do for our patients, the extent to which we are able to help our patients recover from their serious mental illness. Our patients on the contrary, focus on what we are unable to do for them, the magnitude of their unmet clinical need. Whether or not the glass (the state of the art) is viewed, as being ‘half empty or half full’ very much depends on personal perspective. Consumers tend to focus on unmet need, whereas we as providers tend to focus on incremental improvement (frequently modest) as steps in the right direction. If the glass were ‘half full’, we would have a plethora of large acute and long-term therapeutic effect sizes. Unfortunately, the effect sizes of most of our clinical interventions for serious mental illness are modest in size. In point of fact, it is a rare drug development effort that leads to a new treatment for a serious mental illness that has a large effect size. Cohen (1) defines an effect size as improvement over placebo divided by pooled standard deviation, and notes that large effect sizes are at least 0.8. For example, an effect size of 1.0 would suggest that the treatment was twice as effective as placebo. At first glance this sounds like clinically impressive improvement. However, using the three acute antidepressant studies in bipolar depression as an example, a more detailed look at the data highlight how modest our effect sizes really are. For example, improvement in the total score on the Montgomery Depression Rating Scale over placebo (the delta) was as following in each of the three large-scale bipolar depression studies: 5.9, 6.5, 3.1, 6.6, 6.1, and 6.5, for lamotrigine 50 mg, lamotrigine 200 mg, olanzapine monotherapy, olanzapine plus fluoxetine, quetiapine 300 mg and quetiapine 600 mg daily respectively (2-4). Most often, this means patients with severe illness experience only moderate improvement during the first 2 months of treatment. With rare exception, the vast majority of our clinical interventions have only modest effect sizes. This observation has significant ramifications as clinical interventions that have large effects tend to be used as monotherapies, whereas treatments that possess only medium effect sizes tend to be used in combination with other treatments. With this in mind, it is not surprising to note that the majority of our patients receive multiple interventions or medications concurrently for their illnesses. Unfortunately, it is a rare patient with a serious mental illness who recovers completely on one medication. If the glass were ‘half full’, we would have definitions of a marked response that represented clear, compelling, and robust evidence of improvement. In mood disorders research, the definition of a ‘marked response’ is typically only described as being a 50% decrease in a baseline symptom severity scale. Worse yet, in schizophrenia research, the definition of a ‘marked response’ is even more modest, and described as being only a 30% decrease in a baseline symptom severity scale. However, in both instances, it is worth noting that routine outcome analyses of acute studies, more often than not, now include remission rate analyses, which are more patient (consumer)-friendly measures of true clinically relevant improvement. We must conclude that the glass is most often ‘half empty’, and that the state of the art in the treatment of serious mental illness is not that good. So why is this degree of scientific rigour so important? It is important for clinical investigators to recognize the magnitude of the unmet need in serious mental illness. It is particularly important for pharmaceutical companies to accurately characterize the magnitude of the clinical effect and the limitations of new medications. Such scientific candor drives insightful observations about where we need to go and what we need to do next. Unmet need should drive future drug development. There is emerging consensus that the greatest unmet need in the clinical management of bipolar disorder is the acute and long-term treatment of depressive symptoms. It is quite clear that patients with bipolar disorder live their lives ‘below baseline’, and that the majority of our new treatments, manage symptoms from ‘above baseline’. There are convincing data that suggest that the vast majority of time spent symptomatic for a patient with bipolar I disorder, and to a much greater extend, patients with bipolar II disorder, is spent in the depressed phase of the illness. For example, Judd et al. (5, 6) reported that the ratio of time spent depressed to manic/hypomanic is 3 : 1 in bipolar I and 39 : 1 in bipolar II disorder. Regarding the data published by Joffe et al. (7) in this issue, they should be congratulated for writing up data that targets an area of such great unmet need. These authors present an open, non-random case series of patients with bipolar disorder meeting inclusion criteria as specified in the manuscript. These Canadian data show that duration of antidepressant treatment of acute bipolar depression has a significant impact on the long-term outcome of bipolar disorder in terms of relapse into depression. Specifically, the longer the antidepressant is continued, the greater the risk of relapse into depression is reduced. Of particular note, patients still on the antidepressant at 1-year follow-up did better than those who received <6 months or between 6 and 12 months of antidepressant treatment. These open, non-random data sets, however, are quite troublesome because the reader is unable to ascertain the magnitude of the universe, the denominator. The reader is unable to answer the question of ‘generalizability’. The reader does not know how enriched the sample was as only responders were analysed. This data set is very interesting, but limited by this design feature. One wonders what percentage of patients had a response to antidepressants in the first place. These data are substantially less interpretable than the Altshuler data because the latter data set reported the denominator (8). The reader could surmise that the responders to antidepressants referred to a descrete small minority of the original sample (about 15%). These important data do not confirm or extend the findings of Altshuler because of this particular issue. In point of fact, these findings are more difficult to interpret. Of the entire cohort, 66.1% relapsed into a mood episode within 1 year of follow-up. Almost half of the sample (44.4%) relapsed into depression even while on antidepressant medication, as was the case for the Altshuler data set, which reported 41% relapse rates. The magnitude of the unmet need in both data sets is substantial. A 44% relapse rate in 12 months is poor, not good long-term outcome. Some significant emphasis should be placed on the observation that overall relapse rates were not good regardless of subgroup. Of those who stayed on antidepressants for 6 months or less, 90% relapsed. Of those who had more than 6 months of treatment, 54% relapsed. Although these differences are significantly different, the effect sizes here are modest and it is quite clear that both of the groups had high rates of relapse. There is no reference to and no attempt to distinguish those who had high rates of relapse from those who had moderate rates of relapse, which would have been helpful. These kinds of analyses help the clinician uncover the clinical characteristics of the small cohort of patients who have good acute responses to the conventional antidepressants that lead to good long-term responses. For example, one might hypothesize that the relapsers are those patients with rapid cycling, as was the case in the Altshuler data. None of the antidepressant responders in the Altshuler data set were rapid cyclers. The next big issue to consider is whether the patients who did not relapse into depression actually relapsed into hypomania, mania, or mixed states. In most studies of bipolar depression, drug-incuded switched into hypomania or mania are counted as evidence of efficacy, not adverse events. What proportion of the long-term antidepressant effect was mediated by relapse into hypomania or mania. Did relpase into mania count as evidence of antidepressant efficacy or did is count as an adverse event? If a patient does not get depressed, but instead relapses into mania, this is a bad outcome as well. Patients who experience switching into mania/hypomania should be counted as adverse events and subtracted out of ‘change from baseline’ and ‘responder analyses’ analyses. So if there is consensus that the greatest unmet need in bipolar disorder is in the clinical management of bipolar depression, why have the vast majority of our new treatments for bipolar disorder initially targeted the manic/hypomanic/mixed phase of the illness? With the exception of lamotrigine and the combination of olanzapine plus fluoxetine, one must conclude that ease of drug development, not unmet need, is driving drug development for bipolar disorder. The vast majority of regulatory approvals worldwide have focused on the treatment of mania. The glass remains ‘half empty’ and we would be doing our patients and their illness a service by focusing more of our attention on what we are unable to do (as Joffe and colleagues have done), which is to manage the depressed phase of bipolar disorder.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.