In their current meta-analysis, Nabi et al. (Reference Nabi, Stansfeld, Plöderl, Wood and Moncrieff2022) conclude that there is no convincing evidence that lithium reduces suicide or suicidal behaviour. This conclusion thus contradicts previous high-quality meta-analyses and previous reviews (e.g. Cipriani et al., Reference Cipriani, Hawton, Stockton and Geddes2013; Smith and Cipriani, Reference Smith and Cipriani2017; Baldessarini and Tondo Reference Baldessarini and Tondo2022).

Since the outcome of a meta-analysis depends on which studies are included (Walker et al., Reference Walker, Hernandez and Kattan2008), the reasons for the discrepancy between the findings of Nabi et al. and previous meta-analyses may depend on the seemingly arbitrary inclusion and exclusion of randomised controlled trials (RCTs) and on an unsatisfactory reinterpretation of data compared to the original publications.

The decision to exclude RCTs published prior to 2000 is puzzling. As a result, more RCTs were excluded (N = 15) from the primary analysis than those included (N = 12). Also questionable and not justified by the authors is the inclusion of only RCTs comparing lithium against placebo or treatment as usual, but not against an active comparator. At least three studies were excluded on this basis: (1) the randomised comparison of lithium v. lamotrigine in bipolar II disorder (Parker et al., Reference Parker, Ricciardi, Tavella and Hadzi-Pavlovic2021), (2) the comparison of lithium v. valproic acid or carbamazepine in the maintenance therapy of bipolar disorder (Peselow et al., Reference Peselow, Clevenger and IsHak2016) and (3) the comparison of lithium plus quetiapine v. quetiapine in the treatment of bipolar depression (AstraZeneca, 2009). Lamotrigine, valproic acid, carbamazepine, and quetiapine are usual treatments for bipolar disorder and we question why these trials were excluded.

Furthermore, the authors state that in a sensitivity analysis they included trials published before 2000, using data extracted in a previous meta-analysis by Cipriani et al. (Reference Cipriani, Pretty, Hawton and Geddes2005). This procedure does not seem accurate, since three key studies from Cipriani et al.'s meta-analysis are missing (Greil et al., Reference Greil, Ludwig-Mayerhofer, Erazo, Engel, Czernik, Giedke, Müller-Oerlinghausen, Osterheider, Rudolf, Sauer, Tegeler and Wetterling1996, Reference Greil, Ludwig-Mayerhofer, Erazo, Engel, Czernik, Giedke, Müller-Oerlinghausen, Osterheider, Rudolf, Sauer, Tegeler and Wetterling1997a, Reference Greil, Ludwig-Mayerhofer, Erazo, Schochlin, Schmidt, Engel, Czernik, Giedke, Müller-Oerlinghausen, Osterheider, Rudolf, Sauer, Tegeler and Wetterling1997b).

In addition, the authors' assumption that no suicides took place in studies with no information on suicide events is flawed; this compromises the validity of included data. It also leads to an inflation of zero studies, i.e. studies with no events in either treatment arm, potentially biasing the results towards false negative. Furthermore, the largest RCT included by the authors (Katz et al., Reference Katz, Rogers, Lew, Thwin, Doros, Ahearn, Ostacher, DeLisi, Smith, Ringer, Ferguson, Hoffman, Kaufman, Paik, Conrad, Holmberg, Boney, Huang and Liang2022) reports one death in the lithium group and three deaths in the placebo group. In their meta-analysis however, Nabi et al. considered only one suicide per study arm, although the other two deaths were due to an opioid overdose or were classified as suicide by the VA records and the National Death Index (Katz et al., Reference Katz, Rogers, Lew, Thwin, Doros, Ahearn, Ostacher, DeLisi, Smith, Ringer, Ferguson, Hoffman, Kaufman, Paik, Conrad, Holmberg, Boney, Huang and Liang2022), respectively. Accepting three suicides among the placebo-treated participants in Katz et al.'s RCT changes the results of the meta-analysis by Nabi et al. into a statistically significant anti-suicide effect of lithium (p = 0.034, Peto's method).

The analysis of rare events leads to statistical difficulties (Liu, Reference Liu2019), as discussed in detail in the publication by Nabi et al. Nevertheless, the assessment of suicide risk is of the highest clinical relevance. Beyond statistical significance, considering the absolute numbers can therefore help. If in Nabi et al.'s meta-analysis waiving the arbitrary criterion of excluding studies before 2000, the 27 RCTs found seven suicides among 1784 subjects in the control arms and only two suicides among the 1953 lithium-treated subjects, a crude reduction of two-thirds (Fig. 3 of Nabi et al.'s publication). For the analysis of rare events, it is also useful to use large open studies. In fact, in a meta-analysis of 31 studies with a total of over 85 000 person-years of risk exposure, Baldessarini et al. (Reference Baldessarini, Tondo, Davis, Pompili, Goodwin and Hennen2006) showed a reduction in suicides and suicide attempts during lithium treatment of four-fifths compared to no lithium therapy (RR = 4.91, 95% CI 3.82–6.31; p < 0.0001).

The adequate evaluation of RCTs also provides statistically reliable evidence of the preventive effect of lithium on suicidal acts: In an overview of RCTs, Baldessarini and Tondo (Reference Baldessarini and Tondo2022) recently found one suicidal act among 622 lithium-treated study participants compared to 14 suicidal acts among 828 study participants in the control groups, resulting in a pooled risk factor of 0.23 (95% CI 0.09–0.59; p = 0.002). The meta-analysis by Cipriani et al. (Reference Cipriani, Hawton, Stockton and Geddes2013) found six suicide-related deaths among 241 subjects in the control groups but none among the 244 subjects in the lithium groups (OR 0.13, 95% CI 0.03–0.66; p = 0.01). Remarkably, in this meta-analysis, Cipriani et al. also demonstrated a statistically significant reduction in all-cause mortality in the lithium groups (OR 0.38, 95% CI 0.15–0.95; p = 0.04), thereby confirming older studies (Ahrens et al., Reference Ahrens, Müller-Oerlinghausen, Schou, Wolf, Alda, Grof, Grof, Lenz, Simhandl, Thau, Vestergaard, Wolf and Möller1995).

For these reasons, we question the validity of Nabi et al.'s findings that are probably flawed and a misleading representation of the evidence base. Looking at the whole picture of all studies available demonstrates the well-established effect of lithium to prevent suicidal acts.