Nassir Ghaemi, professor of psychiatry at Tufts, recently responded to a systematic review we conducted on the evidence for the common claim that lithium prevents suicide.
The content and style of Ghaemi’s article, which is more of a rant than a scientific commentary, suggest he was extremely upset that this cherished belief had been challenged (Ghaemi, 2022). The manner of its publication in the Journal of Psychopharmacology implies he was not alone. As such, the piece provides an interesting insight into the importance of the medical or disease model of treatment to the identity of professional psychiatry.
Ghaemi, along with the reviewers and editors, seem to need to defend the reputation of psychiatric drugs as sophisticated and targeted agents and to shut down any notion that they might not be that special. The article illustrates the desperate measures some will take to defend this view, and the way in which a group of biological psychiatrists exert their influence over the scientific literature.
Our review of data from randomised trials was the largest to date and did not provide support for the claim that people treated with lithium have lower suicide rates or rates of suicide attempts than people treated with placebo (Nabi et al, 2022).
One of us (Joanna Moncrieff) had conceived the review because the belief that lithium prevents suicide is prevalent and influential. The main evidence cited to support this belief came from a meta-analysis published in 2013, which had excluded a large proportion of trials due to the fact that no suicides had occurred in them (Cipriani et al, 2013). The technique of excluding trials with ‘zero events’ is problematic, however, because it excludes much relevant data and makes suicide seem more common than it is. The technique was popular primarily because older statistical methods of meta-analysis could not incorporate such trials. Initially, therefore, we planned to do a simple analysis, combining the data from each trial as if it were from a single trial. Then Martin Plöderl joined the team and brought expertise in new statistical methods of meta-analysis that have been developed to manage ‘zero event’ trials. So we applied these too.
In a previous paper, Ghaemi declared that lithium ‘is the only drug in psychiatry which is proven to be disease-modifying,’ by which he meant that it affects the pathophysiology of the disease process and the course of the illness, including mortality due to suicide. In contrast, other psychiatric treatments are non-specific, ‘symptomatic’ treatments, according to Ghaemi, which have no effect on the underlying condition (Ghaemi, 2022).
Ghaemi’s categorisation of drugs in this way is misleading. Symptomatic treatments may nevertheless target physiological processes that produce symptoms. In fact, most medical treatments affect symptom mechanisms rather than the ultimate pathology of a disease. There is little evidence that psychiatric drugs do this, however. As one of us (Joanna Moncrieff) has explained in many publications and talks (including one at which Ghaemi was present as a co-presenter), psychiatric drugs can modify the manifestations of mental disorders by altering normal biological processes (the drug-centred model). There is little justification for supposing that they have any additional impact on the hypothesised mechanisms that produce the feelings and behaviours we call symptoms of mental illness (the disease-centred model) (e.g. Moncrieff, 2008; Moncrieff, 2018). These mechanisms are not established, and arguably never will be, since mental illness typically consists of complex situations that are unlikely to be explained by a deterministic, mechanical model of causation (Moncrieff, 2020).
So when Ghaemi claims that psychiatric drugs target symptoms rather than modify diseases he is not saying anything that is inconsistent with the conventional medical model of psychiatric treatment.
But Ghaemi wants to claim that lithium is special—that it does more than target symptom mechanisms, it modifies the disease process that underpins bipolar disorder.
Ghaemi is idiosyncratic in suggesting that only lithium affects the disease itself, but he joins the throng of psychiatrists who regularly and authoritatively proclaim that we know the biological basis of major mental conditions. With respect to bipolar disorder, Ghaemi claims that the ‘basic pathophysiology is known to involve biology of recurrence,’ which, he suggests, happens to involve systems that are affected by lithium (Ghaemi, 2022).
While, in contrast, most biological psychiatrists admit we do not understand the biological basis of bipolar disorder (Harrison et al, 2018) or the mechanism of action of lithium (Chokhawala et al, 2024), they regularly make similar arguments to justify the disease-modifying effects of other drugs. Those psychiatrists who protested about the umbrella review of serotonin and depression conducted by one of us (Moncrieff et al, 2022) insisted that there is some evidence of a link, despite the overall picture being inconsistent and unconvincing (Jauhar et al, 2023), and others resorted to alternative theories to argue for the disease-targeting effects of antidepressants and other drugs that are being introduced for depression (such as esketamine) (Belko, 2024). At the recent Royal College of Psychiatrists’ annual conference, it was firmly pronounced that schizophrenia is related to dopamine dysfunction, which would therefore respond to dopamine blocking drugs.
Ghaemi thinks glutamate is the culprit in schizophrenia, however, which is unaffected by antipsychotics (Ghaemi, 2022). Coupled with his propositions about the basis of bipolar disorder, this enables him to differentiate between lithium and antipsychotics in terms of their relationship to the hypothesised underlying disease.
But Ghaemi’s case for lithium’s special status as a curative agent also rests on his claim that lithium reduces mortality, including suicides. It is understandable, therefore, that he should want to challenge our systematic review. It is less clear why he felt the need to be quite so pejorative and unprofessional in his response. We will describe some of the derogatory comments he makes and then briefly set out a refutation of his substantive points—most of which had been made in another, more civilised response to our review (Bschor et al, 2022) to which we have replied (Moncrieff et al, 2022).
The title of Ghaemi’s recent paper, ‘The Pseudoscience of Lithium and Suicide: Reanalysis of a Misleading Meta-Analysis,’ gets the insults in before we even get to the paper. The opening sentence of the introduction repeats the allegation that our review is ‘pseudoscience’ and accuses us of spreading falsehoods and of using ‘metaanalysis as a tool to mislead oneself and others’ (Ghaemi, 2024).
Ghaemi then explains the meaning of pseudoscience, for those who might not know, and of the process through which ‘pseudoscientists’ deceive people:
‘Pseudoscientists deceive themselves, adhering to a set of unchanging beliefs. Then they can mislead honestly, based on their own self-deception. Self-deception is a precondition for deception.’
In contrast, Ghaemi seems to be setting himself up as the real scientist, arguing that ‘Science is a much harder task than pseudoscience, just as refutation of one’s beliefs is much harder than confirmation.’
In his conclusion he puts the boot in further: ‘This kind of article is not “research” in the sense of new knowledge: it produced not a single datum of new fact. It is social activism disguised as science. It uses scientific journals as a public relations tool, providing a patina of respectability for explicit opinion-based propaganda on the internet and in social media.’ The fact that we might disagree with his opinion is adduced as evidence that although we ‘believe’ we are engaging in science, we ‘are doing the exact opposite of science’. In the process, ‘Pseudoscientists deceive themselves first, then earnestly foist their false beliefs on others.’ Ghaemi is disabusing people of our misleading propaganda. ‘It takes some attention to understand why their meta-analysis was wrong,’ he explains, ‘but it is worth the effort if one seeks knowledge instead of self-deception.’
Ghaemi was not solely responsible for the tone of his article, however. One of us was asked to review the initial version that he submitted to the journal. Instead of reviewing it, we suggested that we be invited to provide a response to be published alongside the paper. This is also recommended by the Committee on Publication Ethics (COPE), which the Journal of Psychopharmacology is committed to. Although we were initially told that we would be invited to do this, in the end, no invitation was forthcoming. Instead, the editor and peer-reviewers not only facilitated the unscientific tone of the paper, but also failed to correct clear misrepresentations of our study.
In the initial version of the paper that was sent out for review, the title was ‘Lithium and Suicide: Critique and Reanalysis of a Recent Systematic Review.’ The article mentioned pseudoscience but quite briefly. In the published version, the title was changed to include the accusation of pseudoscience and two whole new sections on ‘pseudoscience’ were added to the text, one in the introduction, and one at the end. Most of the explanation about our supposed deceptive practices, criticism of our scientific credentials and pejorative language, such as references to ‘social activism,’ ‘opinion-based propaganda,’ ‘foist their beliefs’ were added subsequently. These changes presumably reflect the suggestions of referees or the journal’s editors. The main editors and editorial board happen to include several biological psychiatrists who have taken exception to other work that questions the biological narrative of mainstream psychiatry (Jauhar et al, 2023). One member of the editorial board, Sameer Jauhar, posted approvingly about the article on X: ‘Nassir writes elegantly and imo he is correct’ (Jauhar, 2024).
Ironically, the self-deception and the promotion of ideological views that Ghaemi accuses us of engaging in seems highly evident in his own article. The fact that he can conclude that there is not just evidence but ‘strong evidence’ for lithium’s preventive properties on the basis of a selective analysis based on a very small number of suicides is indicative of his presuppositions. Even other proponents of lithium, such as Baldessarini and Tondo (2022), have acknowledged the uncertainty of lithium’s anti-suicidal properties, describing how ‘recruiting participants to such trials [suicide prevention trials of lithium] may be made difficult by an evidently prevalent belief that the question of anti-suicidal effects of lithium is already settled, which it certainly is not.’
Ghaemi’s main criticism of our review is of our inclusion and exclusion criteria. He accuses us of ‘statistical alchemy’ because he thinks we should have excluded trials with zero suicides, which we included so as not to exclude a large amount of data, and included trials conducted before 2000, which we had excluded from our main analysis on the basis of unreliable reporting (there is evidence that suicides were not reported in at least one of these trials, as explained in our rebuttal to the earlier critique (Moncrieff et al, 2022). However, we had, in fact, performed a sensitivity analysis excluding zero event trials, and a subgroup analysis including the earlier trials. Neither detected a significantly lower suicide rate with lithium. Large parts of Ghaemi’s argument are built on the claim that we omitted earlier trials, yet we presented this analysis in Figure 3 in our paper and mentioned it in the abstract. It is curious that both Ghaemi and the reviewers seemed to miss this.
Despite Ghaemi’s idea of the clear-cut nature of science, every review involves making decisions about what you will include and what you won’t. We pre-registered our protocol outlining and justifying our eligibility criteria. Ghaemi, in contrast, appears to use a post-hoc selection strategy: selecting studies and statistical methods which lead to results that confirm his preconceptions, and then finding reasons for justification. Pre-registered systematic reviews were invented to prevent these biases.
Ghaemi cites Sweeting et al. (2004) and Diamond et al. (2007) to justify his criticism of our inclusion of zero event trials, references which are now up to 20 years old. In our original paper and also in our response to the previous critique of our review (Moncrieff et al., 2022), we carefully explained why these trials cannot simply be dismissed and how modern statistical research recommends they be included. Example quotations from relevant papers include:
“To utilize all available information and reduce research waste and avoid overestimating the effect, meta-analysts should incorporate DZS [double zero studies], rather than simply removing them” (Ren et al., 2019).
“Methods that ignore information from double-zero studies or use continuity corrections should no longer be used” (Kuss, 2015).
“Including double-zero studies in meta-analysis improved performance substantively when compared to excluding them, especially when the proportion of double-zero studies was large” (Xu et al., 2022).
The Cochrane Collaboration also now provides a tutorial on how to deal with such data and is critical about dismissing it (Cochrane Collaboration, 2024).
Ghaemi decided to exclude studies with zero suicides and to include trials published before the year 2000. This just failed to produce a statistically significant difference between lithium and control conditions (p=0.07). Then he adjusted the numbers of suicides in the recent, large VA trial by Katz et al. (2022). He included a death which was an overdose but not classified as a suicide, and a suicide that took place a month after the trial had ended. We excluded this since it is uncertain that other trials would have reliably reported events that occurred after the official end of the trial. By making these choices, Ghaemi managed to obtain a statistically significant difference in favour of lithium (p=0.02).
One can always argue about selection criteria in a systematic review, but we suggest that the alchemy may be on Ghaemi’s part, not ours.
One of Ghaemi’s arguments for excluding zero event trials is that they involve people at low risk of suicide. It is true that trials that are not aimed at suicide prevention often exclude people at risk of suicide, but this doesn’t necessarily mean there will be no suicides. The correct way to look at this issue is to look at trials which are specifically designed to test lithium’s suicide prevention properties, which include people at high risk of suicide. There are only four such trials, and we conducted a subgroup analysis including these which found no effect. Ghaemi fails to acknowledge this.
In his section about the three aspects of suicidality (suicide ideation, suicide attempts, suicides), Ghaemi rightly points out that suicide is difficult to study because it is so rare. Hence it is not surprising that lithium has not been shown to have effects on it with the available data. However, Ghaemi fails to mention that evidence for suicide attempts does not support a preventive effect of lithium in any recent meta-analyses (including that of Cipriani et al, 2013). Suicide attempts happen about 20 times more frequently than suicides and thus the trial data has greater power to detect a preventive effect for lithium. If lithium does prevent suicide, a preventive effect on suicide attempts should be seen, too, and it is not.
Ghaemi ends with a strong claim: “The clinical conclusion is clear and the opposite of the pseudoscience: The anti-suicide effect of lithium is supported strongly by randomized clinical trials.” This statement is clearly at odds when considering all the evidence, the findings for suicide attempts, the large and high-quality suicide prevention trial by Katz et al, (2022), and when appropriately taking into account the uncertainty, as we have outlined here.
Given that Ghaemi’s case against our meta-analysis rests principally on our failure to exclude zero-event trials, which is no longer recommended, and that we didn’t include trials conducted prior to the year 2000, which we did (in a sensitivity analysis that was prominently presented in our paper) it is difficult to understand how his article was published. The fact that it was modified to make it more accusatory and pejorative than it initially was suggests that the editors or reviewers involved in processing the paper were not primarily interested in a scientific debate about our meta-analysis, but wanted to use the paper as a vehicle to undermine the credibility of our research. Could the unscientific and aggressive tenor of the response indicate that some sections of the biological psychiatric establishment feel threatened?
On the one hand it feels as if they have never had it so good. More people than ever are using psychiatric drugs, such as antidepressants, and seeking psychiatric diagnoses including depression, bipolar disorder, ADHD and autism. On the other hand, people who have been harmed by psychiatric diagnosis and drugs are more connected and more powerful, and ordinary people have better access to alternative views about the nature of mental health problems and treatments. Doctors are no longer the only gatekeepers of medical knowledge and although this opens people up to the nefarious influence of the pharmaceutical industry or quackery, it also creates opportunities for people to inform themselves and each other outside of the medical system. Knowledge is power, and power means the ability to make truly informed choices about how to understand and manage your own problems. People who have manic depression, bipolar disorder or manic episodes no longer need to be misled about the miraculous anti-suicidal properties of lithium. They can see the evidence for what it is and make up their own minds.
