Reconsidering Lithium as the Gold Standard for Bipolar Disorder: A Personal History
“Lithium is one of our tools but has no special status.”
Dr. Robert Haim Belmaker is an Israeli psychiatrist and was the Hoffer-Vickar Professor of Psychiatry at Ben-Gurion University of the Negev, Beersheva Israel (holding the first named Chair of Psychiatry in Israel) until his retirement and is now Emeritus. He opened a public Bipolar Disorders clinic 40 years ago and continues to see patients at his private practice in Modiin, Israel. He has received numerous awards and honors over the course of his career, including the National Alliance for Research on Schizophrenia and Depression Lifetime Achievement Falcone Award for research in mood disorders (2000). He was the President of the CINP 2008-2010 (International College of Neuropsychopharmacology). He is the author, with Pesach Lichtenberg, of the 2023 volume “Psychopharmacology Reconsidered - A Concise Guide Exploring the Limits of Diagnosis and Treatment” (Springer).
A wave of critical psychiatry in its moderate and thoughtful form, epitomized by Awais Aftab’s column, has engaged with many entrenched myths of the profession, whether these are ideas of neurotransmitter deficiencies or simplistic notions that decoding the human genome would uncover the essences of psychiatric illness. One of the still-remaining and rarely touched psychiatric shibboleths has been my personal anguish in recent years, the idea that lithium is the “gold standard” treatment for bipolar disorder. In 2025, I wrote about this issue in detail in an academic paper for a special issue of Pharmaceuticals celebrating the 75th anniversary of lithium as psychiatric therapy, and I write now to share my perspective with the readers of Psychiatry at the Margins.
I participated in one of the first groups to use lithium in the United States in 1971 while a medical student at Duke University Medical School, where my teachers had an Investigational New Drug (IND) application for the as-yet unapproved use of lithium. In 1972 I went to NIMH as a psychiatric clinical associate. The environment was full of excitement about the promise of neurochemistry in psychiatry and I rapidly became enthusiastic and optimistic. I acquired training in clinical trials and the belief that controlled clinical trials were the only key to any real knowledge of psychopharmacologic treatment. I also acquired neurochemical laboratory skills in a lab affiliated with Nobel Prize-winning Julius Axelrod who made famous discoveries about monoamines. Lithium and bipolar disorder were one of the key interests of the enthusiastic NIMH psychiatrists because it is a simple ion that has a profound effect in bipolar disorder. The ethos was that if we could understand the neurochemistry of lithium we would unravel the neurochemistry of bipolar disorder. Bipolar disorder at the time had been diagnosed in the US much less frequently than schizophrenia whereas in Kraepelinian Europe the ratio was the other way around. Fred Goodwin at NIMH did studies of lithium in highly psychotic manic patients and found that many of them responded to lithium. He promoted the concept that affective symptoms, when present with a remitting clinical course, were a reliable predictor of lithium response, even in highly psychotic patients. His work influenced subsequent revision of DSM to widen the diagnosis of bipolar disorder and to increase enthusiasm for the new drug lithium that was approved for use in the US in the ensuing years.
On coming to Jerusalem in 1974, I fully embraced the NIMH model of working both in a clinical research unit and an adjacent psychopharmacology lab. My collaborator Richard Ebstein, a Yale-trained neurochemist, gave our work considerable gravitas. I inherited a lithium clinic that had been founded by Elliot Gershon who later went onto a career in American psychiatric genetics. For ten years in Jerusalem and for twenty-five years thereafter in Beersheba, I maintained a personal weekly Lithium Clinic maintaining hundreds of patients. We came upon an idea that lithium might work by inhibiting adenylate cyclase, the enzyme that converts ATP into cyclic AMP. Cyclic AMP had recently been found to be the second messenger system for many neurotransmitters. Lithium has been reported obscurely to inhibit this enzyme and we thought that this could fit a monoamine-based theory of mania. It worked and we followed up the leads from this exciting finding for almost a decade.
Meanwhile, in our clinical research I was eager to test Fred Goodwin’s idea that lithium response could define manic depressive illness (which is what we then called bipolar disorder). Collaborating with Joeseph Biederman who went on to become a distinguished Harvard professor of child psychiatry in later years in Boston, we treated all acute psychiatric admissions with standard antipsychotic (haloperidol) treatment. In addition, they received in a randomized fashion either placebo tablets or lithium tablets. Blood lithium was measured weekly and the results reported to a technician who made up false lithium levels for those who were taking placebo. The results were contrary to our hypothesis and were published in the Archives of General Psychiatry in 1979. Lithium added benefit across the whole range of acute psychoses as an additive to dopamine blocking antipsychotic medication. There was no clear boundary between excited schizophrenia, schizoaffective mania, psychotic manias with mood incongruent symptoms, psychotic manias with mood congruent symptoms, and non-psychotic manias. The first seeds of doubt for me were planted about the idea that lithium was specific to bipolar disorder.
In the lab doubts also arose about whether lithium could really be acting on adenylate cyclase as its mechanism of action. Adenylate cyclase is ubiquitous in the body and its inhibition should affect almost every conceivable neuroendocrine and neurotransmitter process. We tried to find whether some adenylate cyclases were more sensitive to lithium but made little progress. In the early 1980s Sir Michael Berridge from Oxford received the Wolf Prize in Jerusalem for his findings that lithium inhibits an entirely different enzyme, inositol monophosphatase and lowers the concentration of an important second messenger precursor in the brain, inositol. I was invited to the ceremony at the Knesset and was deeply impressed. By coincidence, an important contributor to the inositol and lithium field was Bill Sherman, Professor in the Psychiatry Department at Washington University in St. Louis, where my wife’s parents lived. I was able to do a two-month sabbatical with Bill Sherman and tooled myself up for studying this new theory of lithium action, the inositol depletion hypothesis.
By this time I had become Professor and Chairman of Psychiatry at Ben Gurion University of the Negev in Beersheva. I continued to maintain a weekly personal lithium clinic seeing patients who had been hospitalized at least once for bipolar disorder and who were being followed up for lithium prophylaxis. When I first started in Jerusalem in 1974 I had the flame spectrometer in its most primitive version in my own office and I took blood and measured lithium levels myself getting a result with the patient still in the room. My patients were almost all on lithium monotherapy. By 1982 I organized the International College of Neuropharmacology (CINP) meetings in Jerusalem. I met there Hinderk Emerich, a German psychiatrist who had serendipitously noticed that valproate, which had been used as an antiepileptic and sedative, was also effective in bipolar disorder. Charles Bowden in the USA took up this idea after Abbott Pharmaceuticals achieved a patent for the derivative divalproex sodium. (Divalproex sodium dissolves in the stomach into valproate and in blood is measured as valproate and has no conceivable scientific advantage over valproate). Bowden conducted a large commercial multi-center clinical controlled trial with a lithium arm, a valproate arm and a placebo arm. He did studies both in acute mania and in prophylaxis. Valproate was at least as good as lithium. It was then reported that Okuma in Japan (in Japanese) that a controlled trial of carbamazepine, another antiepileptic, was effective in the prophylaxis of bipolar disorder. His results were rapidly taken up by Bob Post at NIMH and carbamazepine was studied and introduced for bipolar disorder in most countries. While I was reluctant at first to abandon my belief in lithium as monotherapy, clinical needs pushed me to reality: Many of my patients had relapses on lithium and did better when valproate or carbamazepine was added. Some patients developed intolerable lithium side effects and we found that we could manage their bipolar disorder by switching to valproate or carbamazepine. Some patients did not want to start on lithium treatment because they did not want frequent blood testing and we started them on valproate or carbamazepine instead because the anti-epileptics required less frequent blood testing. Many patients seemed to do as well as the patients on lithium or as well as they themselves had done on lithium. In 2010 we reluctantly published our results, called the New Lithium Clinic, where only about a third of patients were taking lithium monotherapy. Many were taking valproate or carbamazepine monotherapy, but even more were taking lithium/valproate combination, lithium/carbamazepine combination therapy or even lithium/carbamazepine/valproate triple therapy. Bob Post dignified this by calling it rational polypharmacy. This gave us some comfort as we did not want to be confused with the sloppy polypharmacy that we had criticized so much at the beginning of the psychopharmacological revolution.
Back to the lab we rushed to see if we could find any biochemical common denominators of lithium, carbamazepine and valproate. If they were all mood stabilizers in bipolar disorder, Occam’s Razor would suggest that we could find their mechanism of action by finding some common mechanism. Inspired by Husseini Manji, whom I met at NIMH on a visit in 1983 and who led this field, we spent some effort on enzymes like protein kinase C which might have been affected by all three compounds. However, nothing specific emerged. Valproate and carbamazepine had no effects of the kind that lithium had on the phosphatidylinositol system. In 1995 Manji participated in a conference I organized in Jerusalem on the mechanism of antibipolar drugs. Husseini went on to become Global Head of Neuroscience Research at Johnson & Johnson. A conclusive common mechanism of action of these three mood stabilizers has yet to emerge.
In the clinical treatment of bipolar patients, the rug was pulled out from under the lithium monotherapy clinic by the development of second-generation antipsychotics. I had become active in the International College of Neuropsychopharmacology beginning in 1978 and joined the council after the 1982 meetings in Jerusalem and was President 2008-2010. This position allowed me to visit many countries outside of the US and Israel. Lithium was rarely available in Asian, Latin American, African countries or Soviet bloc countries because of the absence of access to frequent blood monitoring. First generation antipsychotics, especially in injectable long-lasting form, were the mainline treatment for all recurrent psychoses including those who were clearly affective and even bipolar. I assumed (mea culpa) that results were poor or that side effects were intolerable and that Western standards of course were superior. No textbook of psychiatry mentioned first generation antipsychotics as options for bipolar disorder prophylaxis. When I corresponded about this with several textbook editors, I was told that the question was uninteresting because tardive dyskinesia would in any case prevent a trial of such compounds. When second generation antipsychotics became available after clozapine showed the way, our center participated in some of the early commercially sponsored multi-center trials of risperidone and olanzapine in one or more than one of the phases of bipolar disorder. Second generation antipsychotics rapidly proved themselves to be effective treatments for all phases of bipolar disorder. While all these compounds called second generation antipsychotics have complex pharmacologies, all block dopamine receptors. Moreover, a few small but convincing studies have gone back to show that first generation injectable long-acting antipsychotics are also prophylactic in bipolar disorder as well as being clearly anti-manic as monotherapy. Second generation antipsychotics were rapidly added to our armamentarium in the “New Lithium Clinic” and are now routinely used either on a continuous basis together with lithium, valproate and carbamazepine for prophylaxis for bipolar disorder or as immediate supplementations upon the appearance of manic or depressive symptomatology in a bipolar or schizoaffective patient taking lithium or an anti-epileptic mood stabilizer.
In the lab we had spent an exhilarating decade studying the inositol depletion hypothesis but came to a dead end. In light of this failure, we shifted to a new theory of lithium action first proposed by Klein and Melton in 1996 and called the glycogen synthase kinase 3 beta theory of lithium action. Like adenylate cyclase and inositol monophosphatase, GSK3b is a ubiquitous enzyme in the body but its inhibition by lithium seemed exciting and promising. We shifted our resources in the lab to studying this new enzyme. Some of the published findings we simply could not replicate. This could be part of a well-known phenomenon and the subject of many recent books and articles dealing with the difficulty of non-replication in biological sciences, psychological sciences and particularly in neurochemistry and neuropharmacological systems. More relevant to our story here is the fact that lithium inhibition of GSK3b rapidly became just one of the effects of psychiatric drugs on GSK3b. GSK3b is also affected by antipsychotic drugs and many drugs in different ways, places and time courses. It is not, to my knowledge, affected by valproate and carbamazepine; so it is not the Occam’s Razor that we were looking for that could be a common mechanism of action for all mood stabilizers. (GSK-3beta is also important in cancer research).
As the complexities of treatment of bipolar disorder increased, I feel more and more effective as a clinical psychiatrist in treating bipolar disorder. I have many tools, not just one: lithium. I could offer patients choices. I could offer patients who failed lithium many choices. I could offer patients who respond to lithium but who had milder relapses than they had before lithium a polypharmacy approach. I began to have experiences with patients who were referred to me after the treatment of a manic episode in hospital with risperidone or olanzapine and who did not want to go onto lithium with its blood tests. These patients have often done very well even on small maintenance doses of 2.5 mg olanzapine or 1-3 mg of risperidone. I also noticed that the exact DSM diagnosis did not predict response to lithium vs. carbamazepine vs. valproate vs. second generation antipsychotic in my practice. Some patients reached me with a diagnosis of schizoaffective disorder or SSRI-resistant depression and responded well to lithium.
I was thus taken aback to see in 2023 an editorial in Bipolar Disorders declaring that lithium is still the gold standard in the treatment of bipolar disorder. It encouraged me to reread the literature that I thought I had been reading. I could not find a single study or meta-analysis finding lithium superior to valproate or carbamazepine in bipolar disorder. One could evaluate quality of evidence, or perhaps decide that there are more studies of lithium than the others, but that is an odd kind of epistemology in my opinion. The strongest statement of our literature, in my opinion, are the network meta-analyses of Kishi et al (2021; 2022). A meta-analysis allows researchers to combine the results of many controlled trials, say of lithium vs. valproate, that use different measuring instruments and different sample sizes into one overall measure. In our field there are meta-analyses that compare lithium to valproate, and those that compare lithium to second generation antipsychotics etc. However, a network meta-analysis allows the researcher to combine studies that compared A to B with those that compared B to C with those that compared C to D with those that compared D to F. The meta-analysis then allows the researcher to compare A to F, even if there were few or no studies directly comparing A to F head-to-head. These meta-analyses show absolutely no overall superiority of lithium to any of the other treatments of bipolar disorder aforementioned. This does not rule out the possibility that there are some individuals who only respond to lithium. This is clearly true. However, there are some bipolar disorder patients who respond only to valproate or only to carbamazepine and, heaven forbid, there are some clearly bipolar disorder patients who do absolutely best on olanzapine, risperidone or another second-generation antipsychotics.
Several large observational and nationwide cohort studies have found that people with bipolar disorder taking lithium have lower rates of rehospitalization. However, the picture isn’t entirely clear-cut in my assessment. The advantage over other medications tends to be weak and of uncertain clinical significance, and in many cases, lithium doesn’t statistically separate from other mood stabilizers or antipsychotics when you look at the data. Plus, this benefit isn’t unique to lithium, we see similar results with other treatments, like long-acting injectable antipsychotics.
There’s also ongoing debate about lithium’s relationship with suicide risk. Some research suggests fewer suicide attempts and deaths among lithium users, but it’s not certain this is directly due to the medication itself. It’s possible that what we’re seeing is partly a side effect of the regular blood tests lithium requires, these frequent check-ins mean patients on lithium get more consistent medical monitoring, which could be protective in its own right, or that the blood test requirements self-select for certain sorts of patients. I have no problem with suicidality risk being a positive consideration in the choice of lithium treatment in patients if clinical use of lithium is otherwise indicated. But the hype around this has led to the misguided use of lithium in the population of chronically suicidal patients and patients whose suicidality is related to personality disorders, substance use, and chronic pain.
Why is it then that there seems to be a movement in academic psychiatry to reaffirm the concept, in my own view long dead, that lithium is the gold standard specific treatment for bipolar disorder?
The answer is not pharmaceutical company backing. Lithium is not patentable and no company has funded large clinical trials or paid for promotion of lithium as a treatment (an exception might be some delayed release formulations of lithium). Pharmaceutical companies have sometimes benefited indirectly from lithium’s academic halo by funding symposia at academic psychiatry meetings focused on their own new antibipolar compounds, be it Abbott’s Divalproex or Lilly’s olanzapine. They include a lecturer on lithium as an academic proof of the seriousness of their compound even if the symposium is completely oriented in a commercial way to the new compound. What then could be the reason why so many papers have been published in the last decade lamenting the declining use of lithium and claiming that lithium is still the gold standard for the treatment of a DSM-definable illness called “bipolar disorder”? It seems that these authors have not realized that bipolar disorder itself is a highly heterogenous concept whether looked at genetically, phenomenologically or prognostically. The effect of these declarations of the gold standard is to make the clinical psychiatrist in the field and the resident in the emergency room confused and feel inferior. It is a kind of town-gown conflict of the old times when I studied at Harvard College 1963-1967 where the students of the elite university looked down on the residents of the city. Academics who deal with data analysis, laboratory research summaries and meta-analyses love lithium. The resident in the emergency room has a much more difficult choice to refer a patient to lithium without taking many economic and logistical issues into account. Highly psychotic, violent, overactive manic patients in the inpatient service are almost never treatable with lithium in today’s short stay reality and short-staffed psychiatry wards. Injectable antipsychotics are a frequent necessity. When the patient is discharged and prophylaxis is considered, sometimes it is easiest to remain on the same second-generation antipsychotic that the patient started in his inpatient mania. Lithium should certainly be considered in some patients who have poor prophylaxis or side effects from antipsychotics: valproate and carbamazepine should also be considered.
The academic hubris mentioned above is complicated in those academic psychiatrists who also do biochemical research. We have held our flag up for 75 years that we have a simple compound that can fix a complicated illness. We have erred. This may be a simple chemical element but it is hugely complicated in its biochemical effects, as complicated as valproate, carbamazepine and olanzapine. Moreover, it has serious long term side effects on the kidney and the parathyroid, side effects that are no less serious than the tardive dyskinesia of the first-generation antipsychotics or the metabolic side effects of second-generation antipsychotics. Lithium is one of our tools but has no special status. The search for a specific biochemical effect of lithium that could then unravel the cause of bipolar disorder has been a fruitless search. So far, the Occam’s Razor approach of finding a common denominator of the antibipolar armamentarium has also led us nowhere. Apparently, mental illness is highly heterogenous, highly complex and we need a more trial-and-error empirical relationship to find the right treatment for the right patient. This approach de-emphasizes the role of specific clinical trials that recruit specific DSM diagnoses with a monotherapy treatment. Such trials are demanded by the FDA but they produce knowledge that is not relevant for most clinical work or at least not entirely adequate for clinical work. The true clinical work goes on in a doctor/patient relationship where a doctor sees a patient over time and can compare the clinical results with one treatment and then another treatment and then a combination of treatments and find thereby the right treatment for the right patient. Precision medicine, where perfect treatments can be predicted in advance, would be wonderful but so far there are far many more articles written on the subject than patients who have been helped by it. We must accept that there is an imperfect relationship between diagnosis and treatment in psychiatry. Examples: antidepressants are useful for anxiety and OCD as well as depression and can sometimes help depressive symptoms in schizophrenia. Lithium in bipolar disorder is no exception to this heterogeneity and non-specific approach. The continued halo has misled our field into a great excess of articles with titles like “Lithium in immortalized white blood cells affects gene transcription of gene x.” Such papers make big waves but should never be accepted without a comparison group of valproate, carbamazepine and olanzapine which may do the same thing in at least one or two other genes which may be experiencing the same effect. In 2004 in the New England Journal of Medicine I proposed a theory of the history of lithium research. I found that in 1970s when monoamines were all the rage in neurochemistry, scientists reported lithium effects on monoamines; on the 1980s when second messengers were all the rage, scientists reported lithium effects on second messengers; in the 1990s when third messengers were all the rage, scientists reported lithium effects on third messengers; in 2000s when neuroprotection was all the rage, scientists reported lithium effects on neuroprotection. These reports were barely replicable by other laboratories. If they were replicated, they never led to further mechanistic understanding. Some of the effects were never disproved but merely abandoned when scientists went on to the new framework of the day.
Lithium will always be an important tool in the armamentarium of the clinical psychiatrist. The percentage of patients taking lithium is not and should not be equal to the percentage of the population diagnosed with bipolar disorder but will vary according to the alternative medicines available, the socioeconomic framework of medical care in particular countries, cities and settings and the preferences of individual patients for different potential side effects of the different drugs. It is time for academic leadership in psychiatry to learn from what’s going on in the field rather than only criticizing it.
See also:
Fascinating piece. Would love to see a dialogue between Drs. Belmaker and Ghaemi on their contrasting positions on lithium and psychiatric nosology.
I would love to hear more about how thoughtful psychiatrists approach prescribing for patients who have done well on lithium for years after initial manic psychosis and severe depressions, are stable and get regular lithium level tests, and are experiencing signs of renal failure.