Tuesday, August 22, 2006

Efficacy of some pharmacological treatments for bipolar disorder   posted by Darth Quixote @ 8/22/2006 11:48:00 PM

A few months ago Agnostic alerted us to an issue of PLoS Medicine devoted to skeptical appraisal of "disease mongering." In this issue the psychiatrist David Healy claimed that the evidence supporting the efficacy of medications commonly prescribed for the prophylactic treatment of bipolar disorder is either weak or nil.

As bipolar disorder is the most painful and debilitating non-terminal disease that I have witnessed so far during my brief time on earth (although mild forms may be conducive to creativity), I was sufficiently provoked by this claim to take a look at the literature in this area. In this post I give a brief report of what I have found. Please note that I have no particular expertise in neurochemistry, pharmacology, or psychiatry, and in any case I make no warrant for the comprehensiveness of this post. The findings in this field are unfortunately quite confusing and contradictory; authorities going over the same evidence sometimes come to opposite conclusions in their reviews, and there is no particular reason to trust my arbitration of this matter. So read at your own risk. However, I think this post is worthwhile if it reminds us that an incidental benefit of the pursuit of knowledge is the alleviation of human suffering. (Note: I personally believe that this is true of all knowledge, including the "forbidden" knowledge deemed by the skeptics of GNXP's more controversial posts to be wicked and irresponsible.) As always, of course, I welcome criticism and pointers from informed readers.

Related: Psychotic mania in bipolar disorder resembles schizophrenia in many respects; common symptoms include hallucinations (primarily auditory), delusions, disordered thinking, and loss of contact with reality. Moreover, a curious risk factor that the two diseases share in common is a seasonality effect; sufferers from bipolar disorder show an excess of winter/spring births. Greg Cochran and his colleages have written two papers that I know of addressing the likely infectious causation of schizophrenia (Cochran et al., 2000; Ledgerwood et al., 2003). Much of their arguments, including evolutionary considerations, are just as applicable to bipolar disorder. I hope that work on pathogenic hypotheses will become a top priority for bipolar-disorder researchers.

My impression is that valproic acid (also known as divalproex and by its trade name Depakote) is rapidly becoming the preferred treatment for bipolar disorder. Its reputation among clinicians is growing as a result of their experience with patients, and its undesirable side effects are demonstrably less numerous and severe than those of lithium.

Although there exist at present fairly detailed mechanistic hypotheses regarding how valproic acid, lithium, and other drugs might dampen epilepsy and the mood oscillations characteristic of bipolar disorder, my fallible impression is that none of these commands anything approaching a consensus that it is the correct explanation. In fact, I suppose that physicists and other real scientists would complain that explanations on the table regarding the mechanistic action of all mood-affecting drugs are rather handwavy. This uncertainty in our knowledge severely limits the kinds of inferences that can be made. For example, if the efficacy of a drug in treating symptoms at the height of their acuity is securely established, can we generalize this to its effectiveness as a prophylactic? It is hard to say without a more detailed model of the disease etiology and the mechanistic action of the drug. And so we resort to statistical black-box evidence.

There are actually two things that are wanted from valproic acid or any treatment for bipolar disorder: (1) efficacy in treating patients displaying acute symptoms; and (2) efficacy as a prophylactic treatment preventing the occurrence of future episodes. The evidence that valproic acid does in fact accomplish (1) looks fairly good to me. Many studies have found valproic acid to be effective in diminishing the symptoms of acute mania, and it is on the basis of these studies that the FDA has approved valproic acid for this use. The most-cited of these studies that I have found is by Bowden et al. (1994). The abstract:

OBJECTIVE--To compare the effectiveness of divalproex sodium with that of lithium and placebo in patients with acute mania. DESIGN--Randomized, double-blind, parallel-group study of treatment outcomes in patients with manic-depressive illness.... RESULTS--Intent-to-treat analysis for efficacy was based on data from 68, 35, and 73 patients in the divalproex, lithium, and placebo groups, respectively. Groups were initially comparable except that all eight patients with four or more manic episodes in the previous year were in the divalproex group. In 30%, 33%, and 51% of the above groups, treatment was prematurely terminated due to lack of efficacy, with fewer premature terminations from divalproex than placebo (P = .017). The proportions of patients improving at least 50% were higher for divalproex and lithium groups than for the placebo group: 48% for divalproex (P = .004) and 49% for lithium (P = .025) vs 25% for placebo. Divalproex was as effective in rapid-cycling manic patients as in other patients. CONCLUSIONS--Both divalproex and lithium were significantly more effective than placebo in reducing the symptoms of acute mania. The efficacy of divalproex appears to be independent of prior responsiveness to lithium.

For references to more studies in this vein, the interested reader may consult Torrey and Knable (2002), an accessible guidebook aimed at non-academics. In fact, Torrey and Knable is a good background source for everything covered in this post; its descriptions of relevant studies are telegraphic but fair. (Incidentally, Fuller Torrey was an early adopter of pathogenic causation of mental illness.)

Now, the evidence that valproic acid works as a prophylactic is considerably more iffy. The gold standard for multi-center, placebo-controlleded, double-blind, randomized studies in this regard is Bowden et al. (2000), and its results on balance favor the effectiveness of valproic acid but are still somewhat ambiguous.

If we look at just the raw percentages of patients terminated from the one-year trial phase of the study within each treatment group, we seem to have some clear evidence that divalproex is superior to placebo. The percentage of patients on divalproex removed from the trial phase for depressive symptoms severe enough to warrant hospitalization (or for severe bipolar symptoms of any kind) was significantly smaller than the percentage of patients on placebo removed for the same reasons. However, it turns out that the sheer number or percentage of patients removed from a treatment group because of bipolar symptoms was not the a priori primary outcome measure. Rather, the authors employed the time elapsed until the occurrence of a bipolar episode (i.e., any episode severe enough to warrant treatment discontinuation or the prescribing of antidepressants). It is not clear to me why the latter outcome should be preferred over the former rather than both being seen as complementary, but apparently such is the convention in this research community. Now, none of the differences in "survival times" among the treatment groups were significant at the Bonferroni-corrected 0.05 level given the large confidence interval of this statistic. On the other hand, the numbers strike me as noticeably tending toward the expected direction. For example, the longer time to development of any bipolar episode by the divalproex group relative to placebo was marginally significant (p < 0.06). Also,

[a]mong patients taking divalproex at the end of the open phase, those randomized to divalproex had a 46% longer duration of prophylaxis in the maintenance phase (p = 0.03) than those randomized to placebo; they were also 42% less likely to be prematurely dropped from the study for any major affective episode (p = 0.04) and 32% less likely to be prematurely terminated for any reason (p = 0.002).

I gather that the hypotheses tested in the above quote were not formulated a priori but were a result of post hoc trawling through the data. For this reason we should be suspicious of the results. But if taken at face value, they suggest that the patients deemed by their doctors to be suited best for divalproex as a treatment for acute mania during the open phase (and in fact treated successfully for acute mania during the open phase with divalproex) were biased toward responding favorably to divalproex as a prophylaxis during the trial phase. This strikes me as an eminently reasonable explanation and should be tested in future studies. In a later follow-up, the authors present results of more post hoc analyses of these data with a focus on depressive symptoms (Gyulai et al., 2003). The results are promising, although again they should be taken with a grain of salt.

The overall marginality of the primary outcome measures may well be an artifact of the stringent criteria that patients had to meet before they could be randomized into the trial phase (full recovery from the index manic episode, two consecutive GAS score of 60+, etc.). These requirements were initially thought to make for a cleaner experiment, but in hindsight they may have unduly compromised its power by biasing the intent-to-treat sample toward patients predisposed to good outcomes. On the basis of previous studies the authors expected over half of the placebo patients relapse into mania, but only 22% actually did so. It is of course fortunate that the majority of the patients on placebo experienced such favorable outcomes, but strictly for purposes of the study the wellness of the placebo group is undesirable in that it may have created too high a baseline; if everyone is going to do relatively well anyway, then there is little room for a treatment drug to give a boost.

Another surprising result of the study was the relatively weak performance of lithium on a number of indices. But these indices suffered from the power problem noted above, especially because lithium was not the primary study drug and thus had a smaller treatment group than divalproex. As far as I can tell, these findings are atypical and not accepted as disproving the efficacy of lithium as a prophylactic treatment. See Muzina and Calabrese (2005) for a review of older studies supporting the effectiveness of lithium prophylaxis and also two recent placebo-controlled, double-blind, randomized studies conducted over a longer time period that found patients on lithium faring significantly better than those on placebo with respect to both manic and depressive symptoms.

The atypical (or second-generation) antipsychotics are also rising in popularity because of the mildness of their side effects relative to typical (or first-generation) antipsychotics. My subjective impression is that olanzapine (also known by its trade name Zyprexa) is becoming the atypical antipsychotic of choice. It has also, as far as I know, the only antipsychotic subjected to a placebo-controlled, double-blind, randomized study of its efficacy as a prophylactic (Tohen et al., 2006). Fortunately, the results of this study are much more clear-cut than those of the similar study of valproic acid--perhaps because this time the authors included in the intent-to-treat sample only those patients who had responded well to the study drug (olanzapine) during the open phase. From the abstract:

Time to symptomatic relapse into any mood episode was significantly longer among patients receiving olanzapine (a median of 174 days, compared with a median of 22 days in patients receiving placebo). Times to symptomatic relapse into manic, depressive, and mixed episodes were all significantly longer among patients receiving olanzapine than among patients receiving placebo. The relapse rate was significantly lower in the olanzapine group (46.7%) than in the placebo group (80.1%).

See also Perlis et al. (2006) for a quantitative meta-analysis demonstrating the efficacy of atypical antipsychotics in the treatment of acute mania.

For the most part I will not discuss the undesirable side effects of these drugs, although they must certainly be given some weight in any decision as to whether they should be prescribed. The one side effect that I will briefly comment on is the risk of suicide. In his commentary Healy present a compelling analysis in favor of the contrarian conclusion that prophylactic treatment for bipolar disorder actually increases the probability of suicide. I have found one quantitative meta-analysis of the antisuicidal effect of lithium that is contra Healy (Tondo, Hennen & Baldessarini, 2001):

METHOD: Broad searching yielded 22 studies providing suicide rates during lithium maintenance; 13 also provide rates without such treatment. Study quality was scored, between-study variance tested, and suicide rates on vs. off lithium examined by meta-analyses using random-effects regression methods to model risk ratios.
RESULTS: Among 5647 patients (33 473 patient-years of risk) in 22 studies, suicide was 82% less frequent during lithium-treatment (0.159 vs. 0.875 deaths/100 patient-years). The computed risk-ratio in studies with rates on/off lithium was 8.85 (95% CI, 4.1-19.1; P<0.0001).

But even if we are to split the difference and say that prophylactic treatment for bipolar disorder does increase the risk of suicide in some samples, it is not clear that this is evidence of an undesirable side effect. In a perverse way, it may be evidence the treatment is working as intended. It is part of the clinical folklore that suicide risk among bipolar patients is greatest when they are recovering from an acute episode, not while in the midst of one. Patients suffering from severe depression may not have the energy to carry out a suicide attempt; often they simply lie in bed around the clock, lacking the will to face their troubles, hoping that the next time they open their eyes the bad dream of life will have passed away. If a prophylactic treatment succeeds in keeping a patient out of this kind of crushing depressive state, then given the influence of moderating variables not identified by the above authors, the patient may actually be more likely to carry out a suicide atttempt because of his increased energy and willpower. Now, I admit that this speculation on my part is completely non-rigorous. My hope is that results of relevant studies continue to be tallied; if Healy's effect turns out to be robust now that we are looking for it, then continued evidence for the efficacy of prophylaxis in preventing bipolar symptoms per se may force us to the entertain the kind of mechanism that I have just sketched.

So, in conclusion. I think that the tenor of Healy's commentary should be tempered somewhat with respect to the narrow arena that I have briefly examind here. Although I would like the evidentiary foundation of valproic acid's prophylactic value to be firmer, given its demonstrated efficacy in the treatment of acute symptoms I'm willing to give it an informative prior. More research is needed (as always, I suppose). With respect to olanzapine, I am thoroughly impressed by the results of the recent gold-standard study (robust across several post hoc analyses not discussed here) addressing its prophylactic efficacy.

There is always the temptation to leave problems to the experts charged with taking care of them, to trust that they are doing their jobs and are on top of things. We must all succumb to this tempation at least occasionally, as our resources of time, energy, and conscientiousness are finite. I hope that I am justified in feeling that our trust in this particular instance is justified. I also hope that readers who feel differently or want to add further nuance to the overall picture will leave comments. This is probably too much to ask, but I'd also appreciate it if readers aware of new developments in this area would email me relevant links or PDFs. This request extends indefinitely into the future. Thanks in advance.