Do you like chocolate cake?
Suppose that a pharmaceutical company decided to market chocolate cake. It wanted a Food and Drug Administration (FDA) indication for maintenance treatment with chocolate cake. It consulted with the FDA and got approval to design the pivotal research study this way: First, a sample of research subjects would eat chocolate cake and vanilla cake. If they liked vanilla cake only they would be excluded from further study. If they liked chocolate cake only, they would now be entered into a research study where they would be randomized to continue eating chocolate cake or eat vanilla cake instead.
What do you think the results of the “randomized” clinical trial would show? Have we now proven that chocolate cake is better than vanilla cake?
This is how the FDA approves maintenance treatment for antipsychotics for bipolar illness (and also for antidepressants in major depressive disorder).
The key issue in adjudicating the question of whether antipsychotics are mood stabilizers involves how the research studies test that question. The maintenance studies of dopamine blockers all use a similar basic design: they are “enriched” studies, also called randomized discontinuation trials (RDTs). The term “enriched” refers to the fact that the sample is enhanced in such a way that it is more likely to show a treatment effect. Put another way, the sample is preselected for subjects who have already responded to the medication being studied. The problem, in the PL viewpoint, is that those studies test treatment response sample already preselected for treatment response. This is obviously a tautology, if true.
As described in the Classic Article of the Month, the basic enriched design of a RDT is as follows: a dopamine blocker is given to a group of patients with acute mania; those who don’t respond or don’t tolerate medication are then excluded; the remaining patients who have responded to the dopamine blocker for acute mania are then entered into the randomized maintenance trial, in which they either stay on the dopamine blocker or come off it (receive placebo). They are followed for up to one year to see if new mood episode relapse occurs. If they relapse quickly on placebo, it is inferred that the dopamine blocker prevented the mood episode from occurring as quickly. As described in the Article of the Month, the PL critique is that most relapses occur within the first few months of the maintenance trial, meaning usually within 3 months after the dopamine blocker was stopped. These studies aren’t really showing benefit for prevention of mood episodes 6 months to one year or longer after the acute phase, but rather they are showing benefit during and soon after improvement from the acute phase.
All this research design lingo can be translated into clinical experience as follows: Suppose you have an acutely manic patient in hospital. You give that patient a dopamine blocker. The patient improves. You then stop the medication one month later. The patient relapses two months afterward. Have you just proven that this medication is a long-term preventive agent for mood episodes that would happen years into the future?
In other words, patients are preselected for short term acute efficacy. In the supposed maintenance trial, one is still assessing short term acute efficacy rather than true long-term prophylaxis. This is the main critique made by PL.
The main response from the supporters of the enriched design is that the maintenance trial is capturing a different phase of illness, namely new mood episodes occurring outside of the acute phase of treatment. Let’s examine if this is correct.
Another way to answer this question can be found in an analysis conducted by members of the PL editorial board. In that paper, we looked at data from the enriched maintenance trials of lamotrigine in bipolar illness. The analysis was applied to antipsychotics for their enriched maintenance trials in bipolar illness, because the research design issues were the same. We did not have access to data from dopamine blocker trials, because the pharmaceutical companies who conducted those trials will not make those data publicly accessible to research scholars. However GlaxoSmithKline, the maker of lamotrigine, agreed to give its data to some researchers for analysis after its medication had gone into generic use. Thus we were able to reanalyze their data to address the questions being raised in this special article.
In that analysis, we assessed relapse in the maintenance phase in two ways: a) Did it occur before or after 6 months; b) What was its polarity?
One enriched maintenance trial preselected patients who had acutely responded to lamotrigine for a depressive episode and then they were randomized to either continue or stop it. Another study preselected lamotrigine response for an acute manic episode, and then they were randomized to stay on or come off. Six months or longer was chosen to represent the maintenance phase of relapse, based on natural history research which shows that the untreated average acute manic or depressive episode in bipolar illness lasts about 3 to 6 months (see By the Numbers). Polarity of relapse gave an indication of whether the supposedly new mood episode was indeed different from the acute mood episode. If the polarity of the relapse was the same as the acute episode, and it occurred very quickly after the acute episode was treated, the apparent relapse was likely to reflect the same recent mood episode, rather than representing a completely new mood episode. Indeed, most relapses occurred in less than six months after the maintenance trial began. Furthermore, all the mood episodes that occurred in the first 6 months were of the exact same polarity as the acute mood episode that had been used for preselection before the study began. In contrast, almost all the mood episodes that occurred after 6 months were of the opposite polarity of the original index mood episode before the study began. This finding is consistent with a century of natural history research indicating that mood episodes in bipolar illness cycle from one phase to the other. In other words, mania typically is followed by depression, and depression is followed by mania. If you don't prevent the opposite phase, it’s probably because you aren’t preventing the next mood episode.
An example is the aripiprazole maintenance study, which received an FDA indication. Acutely manic patients were preselected for aripiprazole response, and then randomized for up to 6 months to stay on or come off that agent. Aripiprazole only prevented manic, not depressive, episodes. Why? Is it simply because it works well for mania, not depression? This is highly unlikely given that aripiprazole has been proven effective for acute depressive episodes in major depressive disorder. In fact, it was ineffective in one placebo-controlled trial of acute mania. In contrast, the results would make sense if they simply reflect the fact that the study was not long enough, in the PL view, to assess new mood episodes, and that its lack of efficacy in the opposite phase of illness simply indicates that it wasn't preventing new mood episodes.
Few realize that there has never been a negative enriched RDT of any medication in any psychiatric illness. Either all medications are immensely effective for everything, or this design is guaranteed not to fail, which is great for pharmaceutical industry profits, but which directs clinicians and patients into fools’ errands.
In sum, these enriched studies aren’t preventing new episodes, but simply retesting acute efficacy after already proving acute efficacy.
This controversy can be addressed also as a matter of the English language. In a prior issue, PL expressed the view that the term “antipsychotic” is not a scientifically and clinically accurate term. PL prefers the term “dopamine blockers.” These agents often are effective in non-psychotic conditions, like depression and mania. The term “dopamine blocker” was suggested it because it is true, although it is not reflective of all of the biological effects of these agents, and it is neutral as to clinical effect. The term “mood stabilizer” also is meaningless scientifically. These medications do not “stabilize” mood, but rather they treat both acute depressive and manic episodes, and they prevent those mood episodes. What distinguishes these agents clinically to the greatest degree from other classes of medications is the long-term prevention of mood episodes. So the term mood stabilizer really does not capture what these medications do. A more neutral term that reflects their biological mechanism is “second messenger modifier.”
So the question - ‘Are antipsychotics mood stabilizers?’ - can be translated more neutrally and scientifically to the question - ‘Are dopamine blockers second messenger modifiers?’ When restated this way, the answer is obvious. Dopamine blockade is not the same thing as second messenger modification. Just to clarify: dopamine blockade happens at the synapse and is associated with improvement in psychotic or manic symptoms primarily. It’s an acute effect. Second messenger modification happens postsynaptically and it relates mainly to long-term changes in neurons and their connections with other neurons. Those biological effects would correlate more with long-term clinical effects of prophylaxis of mood episodes. In other words, the basic biological effects of these two different classes of medications differ and correlate with differences in clinical strengths, with acute symptom benefit for dopamine blockers and long-term prophylaxis for second messenger modifiers.
Therefore at one level we can assert that antipsychotics aren’t mood stabilizers simply because dopamine blockers aren’t second messenger modifiers.
In sum, the PL view is that antipsychotics are not mood stabilizers because “enriched” maintenance randomized discontinuation trials are inherently invalid based on their preselection of acute treatment efficacy, followed by reassessing acute treatment efficacy. They don’t assess new mood episodes in the maintenance treatment of bipolar illness. Because they involve a tautology, they never fail. The FDA approves this design, but doesn’t appreciate, in the PL view, the invalidity of how this design is used in psychiatric maintenance studies. This error results in major profits for the pharmaceutical industry, but misleads clinicians and patients into major errors in long-term treatment of bipolar illness.
Given these considerations, PL strongly recommends that clinicians shouldn’t consider dopamine blockers as equal to or better than standard second messenger modifiers, like lithium. These maintenance studies don’t justify replacing lithium with quetiapine, or replacing valproate with olanzapine. Rather PL recommends that the latter agents be used short term, preferably, and if used long-term, they should be seen as adjuncts only to second messenger modifiers, not as “mood stabilizers” by themselves, that could be effective for bipolar illness in place of standard second messenger modifiers like lithium or valproate or lamotrigine or carbamazepine.