Hippocratic psychopharmacology for Bipolar Disorder—An Expert's Opinion

Abstract

In the case of bipolar disorder, my experience is that differences of opinion among clinicians and researchers often have little to do with data and much more to do with underlying beliefs about psychopharmacology. In this review, I will inevitably discuss data, but I also will examine differing philosophies of psychopharmacology that influence how we interpret and use those data.

Introduction

Reviews of bipolar disorder usually involve presenting and interpreting data. The problem with that approach, as is well known, is that data can be selectively presented and interpreted to make different points.¹ This is especially the case when original articles are few and reviews are many, as with bipolar disorder.² On the other hand, systematic reviews (like meta-analyses) can also be abused; depending on inclusion and exclusion criteria and differences in design between studies, such reviews can also be tilted toward different conclusions based on the biases of the reviewers.¹

My main theme in this article is that most clinicians and researchers practice an anti-Hippocratic psychopharmacology and that we need to reverse this practice.

What does “Hippocratic” Mean?

There is a general misunderstanding of the term Hippocratic. Most physicians have never read any of the Hippocratic corpus and know little about the man or his times. All they know, usually, is that there was a Hippocratic oath that consisted of certain ethical maxims, the main one being “First do no harm.” Thus the Hippocratic tradition in medicine has come to be seen simply as a conservative cautious approach to treatment. This is partly true, but it is not the real meaning of the term Hippocratic.

Consequently, many physicians might be inclined to view practice as “more” or “less” Hippocratic, using Hippocratic interchangeably with cautious. I consciously use the terms Hippocratic and anti-Hippocratic to make the point that we have a choice to make: one or the other. There is no such thing as being more or less Hippocratic; it is all or nothing.

This is because the Hippocratic school or tradition involved far more than ethical maxims. It was an entire philosophy of medicine, and the ethical results grew out of that philosophy. In this article, I want to emphasize the overall Hippocratic philosophy (what one might call its metaphysics of disease) rather than just view the Hippocratic tradition simply as ethical rules.

Differing Philosophies of Psychiatry

Psychiatry has gone from being biological to rejecting biology and back again. In the process, the original biological psychiatry—that of Kraepelin and the turn of the 20th century German tradition—has been resurrected yet modified.^3,4 It has been modified, at least in the US, by the American pragmatic tradition of doing and acting. Kraepelin's school was therapeutically nihilistic and biologically deterministic: it saw biological roots to mental illness, mostly in heredity, but did not feel we could effectively intervene with nature. Freud's school was therapeutically optimistic and psychologically deterministic: it saw psychic, not physical, roots to mental illness, mostly in trauma, and felt that the truth would set us free.³

American psychiatry in the last few decades has combined these two traditions. We are now therapeutically optimistic and biologically deterministic.⁴ Yes, the roots to mental illness are in the brain, but we now have tools to fight back.

Perhaps bipolar disorder can be viewed as the greatest success story in this evolution. John Cade, a lone researcher in Australia, discovered the antimanic effects of lithium in 1949,⁵ in the same year in which the Nobel Prize was awarded to Egas Moniz for frontal lobotomy, a surgical procedure that rendered humans “improved” by altering their personalities.⁶ The premise of frontal lobotomy was that the sickest patients who could not be cured would at least be rendered placid. It is ironic that in that same year, lithium, the closest thing to a “cure” in psychiatry, was discovered for bipolar disorder, and that the man who perhaps should have received a Nobel Prize for the most effective treatment in psychiatry, John Cade, has never been adequately recognized.

Today, the psychopharmacology of bipolar disorder is indeed therapeutically optimistic and biologically based. The question I address in this review is whether our current practice is balanced or overweening, and if it is the latter, what approaches best lead to the most safe and effective psychopharmacology.

Two Basic Approaches to Psychopharmacology

Before we can look at the data, we need to acknowledge and assess our assumptions about how to do psychopharmacology. It is often accepted that psychotherapies have underlying theories and assumptions, yet psychopharmacology is viewed as straightforward and simple. Yet, as the great German psychiatrist Karl Jaspers repeatedly emphasized, we always have conceptual assumptions to everything we do.⁷ In the case of psychopharmacology, I believe that there are two basic approaches: Hippocratic and anti-Hippocratic. My colleague Ronald Pies, MD, has emphasized the importance of a Hippocratic approach to psychopharmacology (Ronald Pies MD, personal communication, 2005). I think it is the case that most clinicians today practice anti-Hippocratic psychopharmacology without realizing it.

Now Hippocrates is a venerated figure, and most physicians would not seek to overtly be critical of the Hippocratic approach in medicine, yet most physicians also have only a passing familiarity with Hippocratic ideas⁸—usually only some of the ethical maxims of the Hippocratic oath, such as Primum non Nocere (“First Do No Harm”). The conceptual basis for these ethical maxims is little known.

Basically, the Hippocratic notion in medicine is that nature is the source of healing, and the job of the physician is to aid nature in the process of healing. In contrast, the anti-Hippocratic approach, which has always been quite prevalent, is that nature is the source of disease, and the physician (and surgeon) needs to fight nature to effect a cure.⁸ Even in ancient Greece, physicians had many potions and pills to cure any ailment; Hippocrates resisted that interventionistic medicine, and most of his treatment recommendations involved things like diet, exercise, and wine—all designed to strengthen natural forces in recovery from illness. It is striking how those same factors are being rediscovered as key to prevention and treatment of many chronic illnesses. The Hippocratic ethical maxims stem from this philosophy of illness. If nature will cure, then the job of the physician is to hasten nature's work carefully, but at all costs to avoid adding to the burden of illness. Hippocratic medicine is at root anti-interventionistic and highly conservative.^8,9

Thus, Hippocratic psychopharmacology would seek to avoid treating patients with medications as much as possible, except where we can clearly help the natural process of healing with great attention to side effects. A Hippocratic psychopharmacologist would be highly aware at all times of the natural history of mental illnesses, knowing that many conditions will resolve spontaneously at some point, and would always seeking to intervene less with medications where such spontaneous recovery is more likely.¹⁰ In many cases, the Hippocratic psychopharmacologist would refrain from prescribing any medications at all, and might instead emphasize psychosocial interventions, such as psychotherapies or alterations in lifestyle (moving, changing jobs, exercise), among other approaches, that might spur the natural process of healing. This tradition is best exemplified by William Osler in modern times.⁹

In contrast, anti-Hippocratic psychopharmacology would consist of prescribing medications aggressively for all kinds of conditions, with limited attention to side effects and with the conviction that if such medications were not given, the illnesses would not abate. Thus all things considered, the anti-Hippocratic psychopharmacologist would view treatment, with all its side effects, as better than no treatment in almost all cases. This approach was most exemplified by the 19th century leader of American medicine, and founder of modern American psychiatry, Benjamin Rush.^10,11

In fact, I do not believe that either approach is definitively correct, for indeed nature appears to be both the cause of disease and the source of healing. And certainly, with some diseases, one simply has to rely on a surgeon to cure the disease by cutting it out. Yet, all in all, it seems that a Hippocratic approach to medicine causes more good than harm.

The application of Hippocratic psychopharmacology involves two rules that I have derived from the modern history of medicine (Table 1). The first, named after Oliver Wendell Holmes,¹⁰ requires proof of efficacy before we prescribe medications. In the 19th century, the materia medica (the equivalent of today's Physician's Desk Reference [PDR]) was much larger than today's PDR. Physicians used pills and potions for everything. Following Rush's advice, they were quite interventionistic. Holmes argued that most of those treatments did not work and simply caused harm. He argued legalistically. In the law, a person is innocent until proven guilty. In medicine, drugs are guilty until proven innocent. There is a presumption that they are harmful and most do have side effects; therefore, we need to start on the benefit side of the risk-benefit ledger. Since we presume all drugs to be harmful, none should be used until there is some proof of benefit (the more valid scientific proof, the better).¹⁰ With this approach, gabapentin would not have been prescribed widely for every ailment,¹² and antidepressants would not have been prescribed extensively for bipolar disorder, as they are, in the absence of preventive benefit.¹³

Table 1.

The two rules of Hippocratic psychopharmacology

HOLMES'S RULE—All medications are presumed harmful. Have proof of benefit before prescribing.

OSLER'S RULE—Treat diseases, not symptoms.

For elaboration, see References 10 and 42.

Many clinicians argue that the lack of adequate treatments for depression in bipolar disorder may underlie the “drive” to use antidepressants in the absence of scientific evidence to support their use. This perspective is emblematic of what I mean by “anti-Hippocratic” psychopharmacology. In the Hippocratic approach, if one does not have an adequate treatment, then one does not treat. It is no justification of unscientific treatments to state that we do not have scientific evidence. In such cases, one should refrain from treatment because, on the whole, one will produce more harm than good. Hippocrates's school divided diseases into the treatable and the untreatable and it left the untreatable illnesses alone. In my interpretation, I am not arguing that diseases are inherently untreatable (though some may be), but rather I am substituting Holmes's notion that we either have scientific evidence or we do not. Where we do not, we should refrain from unscientific, guess-based treatment.

The second rule, named after William Osler, emphasizes his view that 19th century medicine was not scientific because it was symptom-oriented, rather than disease-oriented.¹⁰ Physicians then used all kinds of medications for all kinds of symptoms. Osler argued that we needed to study the diseases that underlie the symptoms, and then the treatments would be clear.¹⁰ Scientific medicine is the treatment of diseases, not symptoms. Yet today many psychiatrists practice non-scientific, symptom-oriented treatment, giving sedatives for insomnia, stimulants for fatigue, anxiolytics for tension, antidepressants for depressive symptoms, and mood stabilizers for lability—leading to an excessive and ineffective polypharmacy.⁹

Critics may argue for a practical need for dimensional and even symptom-oriented approaches when there is a lack of certainty regarding the discreteness of putative diseases and a lack of treatments for specific diseases with multiple symptom constellations and comorbidities (such as borderline personality disorder). My own view is that such complex conditions are not discrete, valid disease entities, but are either conglomerations of multiple diseases or not diseases at all (they are, rather, “problems of living”). As for the practical need to treat symptoms, all physicians do so, and certainly it is humane to relieve suffering where this can be done symptomatically, such as using analgesics for pain. However, the Hippocratic approach would emphasize that this approach to pharmacology should be the exception, not the rule, and that we should not be happy about this state of affairs. The common use of drugs for symptoms reflects our ignorance about disease and is another reflection of the primitive state of our science. We should lose sleep over this reality, even though practically, in some cases (though not most, if we are Hippocratic), we provide symptomatic treatment.

Hippocrates Again

Being Hippocratic does not mean that one is ethical and being anti-Hippocratic does not mean that one is unethical. That implication, which is assumed by most physicians who have limited real knowledge about Hippocrates, is based on the notion, which I opposed above, that the Hippocratic tradition is simply one of ethics. I argue, again, that it involves a philosophy of health and disease, which we either need to accept or reject. It is not unethical to be anti-Hippocratic—it just involves a different conception of medicine.

Rush was quite overt about this when he wrote, “It is impossible to calculate the mischief which Hippocrates has done, by first marking Nature with his name, and afterwards letting her loose on sick people. Millions have perished by her hands.”¹⁴

Most psychiatrists practice anti-Hippocratically; this is not to say that they are unethical or even wrong, but simply that they believe that Nature is the enemy and they want to intervene aggressively to cure patients. My point is to emphasize that this is what most practitioners do and that it conflicts with the Hippocratic tradition that so many of them profess.

Application to Bipolar Disorder

Assuming we were to accept a Hippocratic approach to treating bipolar disorder, how would we interpret the available data? I focus my discussion of studies on maintenance efficacy.

Recurrence: The hallmark of bipolar disorder. We would initially want to know what characterizes the disease we are treating. Is bipolar disorder predominantly an illness of recurrent mood episodes, as Kraepelin argued, where recurrence is the key feature? If so, then treatment should focus on reducing recurrence, i.e., prevention or prophylaxis, with mood stabilizers. If we see bipolar disorder as an illness of acute alternating mood episodes, then treatment would focus on treating those acute episodes.¹⁵ Again, it is not completely an either/or matter, because bipolar disorder is characterized by both acute episodes and recurrence. It is not unreasonable, however, to emphasize recurrence as the main problem, because if one can prevent episodes, then one does not need to treat them acutely.

What is a mood stabilizer? I would argue that the first step in treatment is to recognize that bipolar disorder is an illness of recurrence, and that the key ingredient to its treatment is to prevent future episodes. Mood stabilizers may be defined as agents that have this effect: They are effective in prophylaxis or prevention of mood episodes. Here is another important conceptual problem: We need to define what we mean by mood stabilizers, by agents that treat this illness as a whole. This is a key point.

Historically, the term mood stabilizer was first used in the 1950s to refer to the combination of an amphetamine and a barbiturate.¹⁶ It has a symptomatic connotation, literally lifting mood and suppressing it at the same time. In the 1960s, Mogens Schou introduced the term mood normalizer to connote how lithium prevented future episodes of depression or mania in bipolar disorder.¹⁶ In Schou's conception, the effect was not symptomatic, but rather curative of disease.¹⁶ The term mood stabilizer later was applied to lithium in Schou's sense.

We can use medications symptomatically, as Band-Aids, in which case it may not matter what we call them. Or we can use medications to treat diseases, more or less as cures when they are fully effective for a disease. The older symptomatic approach is 19th century medicine.¹⁰ The disease-oriented approach is 20th century medicine, as inaugurated by Osler.¹⁰ On that definition, then, we would emphasize Schou's that a mood stabilizer is a drug that treats the disease of bipolar disorder, not just mood symptoms.

In the past, various definitions have been offered.^17–19 Some have suggested that this matter is too complex or that the pharmaceutical industry has exploited the vagueness of the term and that we should stop using the term mood stabilizer and simply refer to proven effects, such as acute antimanic efficacy or delay of time to relapse.²⁰ Yet the mere fact that a phrase is wrongly used or abused does not mean that we should avoid trying to give it a valid definition. The fact that it is used is a testament to its utility and suggests we should try to better define it. I have suggested previously that acute efficacy in one phase of bipolar illness along with prophylaxis for that phase might be a sufficient definition.¹⁷ I am changing my view here to state that I think that prophylaxis, pure and simple, should identify a mood stabilizer. Mood stabilizers should be agents that treat recurrence (the hallmark of bipolar disorder) and therefore prevent new episodes. Their acute effects, while possibly beneficial, are not necessary to their definition.

Defining maintenance efficacy in bipolar disorder. So if the disease of bipolar disorder is characterized primarily by recurrence, the Hippocratic approach would be to use mood stabilizers to alter the biological processes that affect recurrence. The clinical test of efficacy would be that the number of mood episodes would decrease in frequency and severity over time with treatment.

We are next left with the need to define what we mean by prophylaxis. This too is a conceptual problem that has not been addressed in the world of bipolar research, unlike work in unipolar depression. The acute phase of illness depends on the natural history of the untreated mood episode. In bipolar disorder, acute major depressive episodes last 3 to 6 months untreated (compared to 6–12 months for unipolar depression), and acute manic episodes last 2 to 4 months untreated.²¹ Thus the acute phase in bipolar disorder is probably in the 2 to 6-month range. Continuation treatment, when medication is continued so as to avoid relapse into the same continuing acute episode, would last in that time frame as well. Thus, maintenance treatment would consist of six months or longer after the acute phase began.

A mood stabilizer could then be defined as a drug that prevents mood episodes in bipolar disorder as shown in studies that assess prevention of future episodes six months or longer after the acute phase.

With that definition, we can assess which agents appear to have evidence of such efficacy. Table 3 demonstrates my interpretation of the evidence for efficacy with putative mood stabilizers in the treatment of bipolar disorder.

Table 3.

Two types of research design for maintenance studies of mood stabilizers

Design	Advantages	Disadvantages	Drugs Shown Effective vs Placebo	Comments
Prophylaxis	Generalizable to all patients	Smaller drug vs. placebo effect size	Lithium
Relapse Prevention	Larger drug vs. placebo effect size	Generalizable only to those with acute benefit; has risk of withdrawal syndrome; biased against active controls	Lithium, Divalproex, Lamotrigine, Olanzapine, Aripiprazole	Olanzapine study showed evidence of withdrawal syndrome. Aripiprazole study was only six months. Divalproex analysis was a secondary outcome.

Lithium

Lithium has been shown to be effective in over 20 randomized clinical trials (RCTs) of prophylaxis, most of which are one year or longer in duration, some of which last up to three years.²² Although some of these studies are small and use research designs that are not optimal (such as crossover designs as opposed to parallel assignment of treatments), the replicability and consistency of these results greatly strengthens proof of lithium's efficacy. Further, newer studies, designed to show the efficacy of other agents, are mostly confirming that lithium is also effective in prophylaxis of bipolar disorder.²²

There is also increasing evidence that lithium prevents suicide, reduces lifetime mortality, and lengthens lifespan in bipolar disorder.²³ The decreased mortality is primarily from decrease in death due to cardiovascular disease, showing apparent medical benefits to lithium.²⁴ Lithium has also been shown to improve neuronal viability in animals and enhance in vitro the effect of neuroprotective factors, and this effect in humans also appears to prevent or reduce long-term hippocampal atrophy and cognitive decline in bipolar disorder.²⁵ Thus, the grounds for John Cade's almost-cure of bipolar disorder have only strengthened with time.

There are, of course, problems with lithium in terms of side effects, tolerability, and medical risks. These problems, in my view, however, have been exaggerated in the public mind and in the psychiatric world over the past two decades, partly due to the marketing efforts of pharmaceutical companies promoting new mood stabilizers and partly due to the fact that lithium was likely overused in the 1970s and 1980s since it was the only mood stabilizer available at that time. Given to many patients who did not respond to or tolerate it, clinicians became overly skeptical about lithium's effects. In fact, many of the concerns about lithium's risks appear to have been overblown. For instance, it has two major medical risks if used appropriately: hypothyroidism and long-term chronic renal insufficiency (CRI). Both are preventable with blood monitoring, and hypothyroidism is treatable. As for CRI, prospective long-term follow-up data suggest that it is associated with acute lithium toxicity (which is also preventable in most cases) and with multiple daily dosing (lithium should instead be given only once daily). Duration of treatment is not clearly associated with CRI, and, with appropriate monitoring, the frequency of irreversible CRI is rather small.²⁶ Nuisance side effects, such as increased urination, cognitive impairment, and acne, often are more of a practical problem, and some persons simply cannot take lithium. (It is notable in my experience that different formulations often have less side effects: For instance, lithium citrate has less gastrointestinal side effects than lithium carbonate, and slow-release lithium has less polyuria than immediate release). Thus, while lithium can have intolerability, as with any drug, it is less than is commonly assumed and can be managed with adequate medical care.

Carbamazepine

This anticonvulsant has been compared head-to-head with lithium in a large, multicenter, open, randomized, German clinical trial with about equal benefit in both treatments over two years of follow-up.²⁷ In those with mixed episodes and less classical bipolar symptoms, carbamazepine appeared more effective than lithium, while in those with pure manic episodes and longer periods of euthymia by natural history, lithium appeared more effective.²⁷ The absence of a placebo group decreases the validity of these analyses, but nonetheless these randomized data provide some long-term evidence of benefit with carbamazepine. Use of this drug has been limited by side effects more than anything else over the years, particularly risk of hepatitis, leucopenia, hyponatremia, aplastic anemia, Stevens-Johnson syndrome (about 1:10,000), and drug interactions.²⁸ Its analogue oxcarbazepine has much fewer side effect risks (with the exception of hyponatremia) and fewer drug interactions (an exception is a reduction of efficacy of oral contraceptives). Yet oxcarbazepine is also less studied in RCTs and appears less effective in clinical practice for maintenance treatment.²⁹ A slow-release version of carbamazepine, Carbatrol^® or Equetro^®, appears to have fewer nuisance side effects and may be better tolerated.³⁰ A major benefit of carbamazepine is that it does not lead to weight gain.

Divalproex

This agent has been studied in one maintenance, open, randomized, clinical trial in which it was similar to lithium in efficacy, and in one double-blind RCT in which it was the same as lithium and placebo in a primary analysis of all patients who entered the study, including those who had not received divalproex previously (i.e., they were treated with lithium, antipsychotics, or antidepressants).³¹ In a secondary analysis limited to those who had received divalproex before the study, divalproex was more effective than placebo and lithium in prevention of mood episodes in one year follow-up, especially depressive episodes.³² This led the researchers involved with that study to conclude that this kind of design, the “relapse prevention” design, would be preferable to the standard “prophylaxis” design (Table 3).³³ In other words, in the relapse prevention design, only those patients who respond to the drug initially for an acute episode (of mania or depression) are allowed into the maintenance study (to prevent future mania or depression). This contrasts with the prophylaxis design where patients can enter the maintenance study after having responded to any drug for an acute episode. As seen with the divalproex study, the relapse prevention design provides a larger drug versus placebo effect size than the prophylaxis design. However, this comes at the price of generalizability. One can only then conclude that Depakote®, for instance, is shown effective in maintenance treatment for people who initially respond to divalproex acutely for mania. This contrasts with lithium, which is the only drug shown effective in a prophylaxis design for bipolar disorder,²² and thus one can say that it should be effective in maintenance treatment in general for anyone with bipolar disorder (not limited to those who might respond to lithium for an acute mood episode).

Lamotrigine

The lessons learned with the divalproex study were applied to two maintenance studies with lamotrigine, both of which proved its effectiveness and led to its FDA indication in maintenance treatment of bipolar disorder.^34,35 These studies both initially required that patients either respond to or tolerate lamotrigine for 6 to 12 weeks for an acute episode of depression or mania before entering the maintenance study. Once the maintenance study began, the patients were randomized to stay on lamotrigine, switch to lithium, or switch to placebo. Lamotrigine and lithium were both more effective than placebo in prevention of all mood episodes in the primary analyses. In secondary analyses, lithium was more effective than lamotrigine in mania prevention, and lamotrigine was more effective than lithium in depression prevention. Much has been made of the latter secondary analysis, with the common conception that these studies prove that lamotrigine is more effective than lithium for bipolar depression, but it is only limited to initial lamotrigine responders and is not a fair comparison with lithium. Such a comparison is one of apples and oranges: A relapse prevention design with lamotrigine versus a prophylaxis design with lithium. The fact that lithium was still more effective than lamotrigine for mania prevention, despite the disadvantage of this design comparison, is remarkable and valid.

Also, it is commonly believed that lamotrigine shows evidence of maintenance efficacy while divalproex does not. But this also compares apples and oranges: A relapse prevention design with lamotrigine versus a prophylaxis design with divalproex. Comparing apples to apples, both lamotrigine and divalproex have been shown to be effective in relapse prevention designs for maintenance treatment of bipolar disorder.

The existence of a 1:6000 risk of Stevens Johnson Syndrome (in adults with a slow titration of 25mg per week or less) with lamotrigine, which can be potentially fatal, must be taken into account when using this drug.³⁶ Otherwise, it is well tolerated in terms of weight gain and nuisance side effects.

Atypical Antipsychotics

Olanzapine and aripiprazole now have FDA indications for maintenance treatment of bipolar disorder, although the data on which those indications are based do not, in my interpretation, supply notable evidence of such maintenance benefit.

With olanzapine, the only placebo-controlled study at least partly represents a discontinuation (or withdrawal) syndrome.³⁷ In that relapse prevention design, all patients initially had to respond openly to olanzapine for acute mania for a minimum of two weeks, and then they were randomized to staying on olanzapine or stopping it (placebo). About 80 percent of the placebo group relapsed within two months of the study. Thus the study mainly shows that if patients respond acutely to olanzapine and the drug is then stopped in two weeks, 80 percent will relapse in two months. This is not necessarily evidence of long-term prevention of new episodes, if one views maintenance efficacy as requiring benefit shown six months or later after the acute phase. However, in a secondary analysis, limited to those who stayed in the study longer than two months, there appeared to be continued benefit with olanzapine versus placebo. The study suggests efficacy in this paradigm, but the majority of patients appeared to experience an acute discontinuation syndrome rather than long-term prophylaxis.

The aripiprazole study³⁸ did not have such evidence of discontinuation relapse in the placebo group, but it was only six months long in total, and thus it simply did not study the maintenance phase. Rather, like the olanzapine study, it can best be seen as suggesting short-term continuation benefit for up to six months after recovery with the antipsychotic for acute mania. These designs do not assess prevention of new episodes in the maintenance phase of six months or one year or longer after the acute episode (unlike the lamotrigine, lithium, carbamazepine, and divalproex studies).

There are three other randomized studies directly comparing olanzapine to divalproex or lithium, respectively, or combined with divalproex or lithium versus either mood stabilizer alone. In the two monotherapy comparisons, olanzapine appeared similar to divalproex (in one study)³⁹ and lithium (in another study),⁴⁰ but the absence of a placebo group does not allow one to conclude proof of efficacy. The combination prophylaxis study did not show any benefit with addition of olanzapine for the primary outcome of prevention of new mood episodes.⁴¹ A secondary analysis suggested some benefit for mood morbidity with continued olanzapine usage, but the clinical effect size of that benefit does not appear to be robust given the absence of utility in prevention of new full mood episodes (unlike the lithium studies and the secondary analysis of the divalproex study).

Summary of the Data

As seen in Table 2, the number of replications and the consistency of the lithium literature set it apart from all other potential mood stabilizers in terms of evidence of prophylaxis. After lithium, similar kinds of evidence of likely benefit exist for lamotrigine, divalproex, and carbamazepine, with some differences (a single replication with lamotrigine, reliance on secondary analysis with divalproex, absence of placebo arms with carbamazepine). The antipsychotics do not truly assess or provide evidence of benefit in the maintenance phase of treatment, given the interpretation and rationale above.

Table 2.

Mood stabilizer efficacy: An interpretation of the randomized evidence

Drug	Depression	Mania	Prophylaxis	Relapse Prevention
Lithium	++	++	+ + + +	+ + + +
Valproate	+	+++	±	+
Carbamazepine	±	++	±	+
Lamotrigine	±	--	?	++
Olanzapine	--	+++	?	±
Aripiprazole	No data	++	?	±

Many clinicians and researchers will not agree with the logic or content of some of this discussion, and they might consider this review to be based on opinion. My interpretation is not the only one that could be made of these data or is definitely the correct one. But all data require interpretation—data never speak for themselves—and therefore all one can do is provide one's interpretation and the rationale for that interpretation, and then leave it to readers to make their own judgments.

Conclusions

A Hippocratic psychopharmacology of bipolar disorder would emphasize the use of the only proven agent in the prophylaxis design, lithium, followed by those agents with the strongest evidence of efficacy in long-term treatment with the relapse prevention design, lamotrigine and divalproex. Carbamazepine is likely also effective, though not proven in placebo-controlled designs. Atypical antipsychotics are not yet proven effective in maintenance designs, though they have continuation phase benefit and likely are useful as adjuncts (but not as mood stabilizers used in monotherapy).

A Hippocratic approach would emphasize agents with the most proven benefit (Holmes's rule) and focus on using mood stabilizers to treat the whole disease of bipolar disorder as opposed to using agents for individual symptoms (Osler's rule). Following these guidelines, we will do the most good and the least harm.