Extending the specificity of mood stabilizers from clinical response to mortality reduction

Mood stabilizers are a heterogeneous class of drugs including pharmacological agents with diverse mechanisms of actions (anticonvulsants, second generation antipsychotics, and lithium) but sharing one fundamental clinical property: the ability to treat and prevent episode recurrences in patients affected by major affective disorders. An attempt to formulate some evidence-based taxonomic criteria suggested that an ideal mood stabilizer should be effective in: (a) treating acute manic symptoms; (b) treating acute depressive symptoms; (c) preventing manic symptoms; and (d) preventing depressive symptoms.¹ Lithium is probably the only mood stabilizer satisfying all these criteria, although it is less effective in treating bipolar depression.² Similarly, the clinical profile of anticonvulsants appears to be specific depending on the illness phase (acute and/or continuation/maintenance) and on the mood polarity. Indeed, lamotrigine is more effective in preventing depression, but not mania, and possibly in treating acute bipolar depression and rapid cycling.³ Conversely, carbamazepine and valproic acid are effective in the treatment of acute mania, as well as in maintenance.^{3, 4} The latter also shows more effectiveness in certain subgroups of patients presenting with mixed mania or mania with irritability, compared with other treatments.^{4, 5} Finally, antipsychotics also show effectiveness in preventing mood relapses,⁶ with possibly higher efficacy in treating acute mixed episodes in bipolar disorder.⁷ These patterns of efficacy/effectiveness appear to be associated with distinct clinical presentations that also show a degree of predictive power. This has indeed been demonstrated for lithium, where specific clinical characteristics, namely an episodic (mania-depression-interval) clinical course sequence, absence of rapid cycling, absence of psychotic symptoms, family history of bipolar disorder, shorter pre-lithium illness duration and later age of onset,⁸ predict a good clinical response to lithium. Interestingly, a recent machine learning study, evaluated whether lithium responsiveness was predictable using clinical markers in a large multicenter sample of patients with bipolar disorder.⁹ The authors showed that lithium responsiveness was predictable in the pooled sample with good accuracy area under the receiver operating characteristic curve of 0.80 and a particularly low false-positive rate (0.91).⁹ More importantly, features related to clinical course and the absence of rapid cycling appeared consistently informative in predicting lithium responsiveness.⁹ Although identified with less robust analytical methodology, similar patterns of clinical features have been associated to valproic acid (mixed or dysphoric mania, early age at onset, rapid-cycling, concurrent substance abuse)¹⁰ or lamotrigine (earlier onset of symptoms, non-episodic course of illness, rapid cycling, comorbidity with panic or substance use disorder, fewer hospitalizations, fewer prior medication trials and male gender) responsiveness.¹¹ Of note, these clinical (phenotypic) patterns of response appear to be underlined by specific genetic architectures as shown for lithium¹² and, more recently, for lamotrigine and valproic acid.¹³ In this context, it is plausible to ask whether these agent-specific clinical properties (possibly regulated by distinct neurobiological mechanisms) extend beyond the resolution of acute symptoms and/or the control of mood recurrences. In this issue of the Journal, Chen and co-authors explored the hypothesis that the use of mood stabilizers could modulate the risks of all-cause mortality, suicide, and natural mortality in individuals affected by bipolar disorder.¹⁴ To this end, they studied a cohort of 25,787 patients with bipolar disorder whose data were available in the Taiwan's National Health Insurance Research Database (NHIRD). Of these, 4,000 died after the index admission between February 1, 2001, and December 31, 2016, with suicide accounting for 760 cases (19.0%), and natural causes accounting for 2,947 cases (73.7%).¹⁴ Importantly, the authors showed that mood stabilizers significantly decreased the risks of all-cause mortality, suicide and natural mortality in the initial 5-year period. This protective effect was particularly high for lithium, and to a lesser extent for valproic acid, while lamotrigine and carbamazepine did not influence the risk of mortality. In addition, lithium exerted a dose-dependent effect on mortality risk reduction, which was detected in patients with a longer duration of lithium use as well as in those using a higher cumulative dose of the drug. This dose dependent risk reduction was also observable with valproic acid use. Importantly, the sensitivity analysis accounted for the immortal time bias, which can influence the results of this type of epidemiological cohort studies.

These findings are really timely as they come out during a period of ongoing debate on the effectiveness of mood stabilizers, and lithium above all, in reducing mortality, particularly when suicide related. Indeed, a 52-week recent double-blind, placebo-controlled randomized control trial (RCT) of add-on lithium in veterans with bipolar disorder or major depressive disorder and substantial comorbidities (substance use disorders, post-traumatic stress disorders) was stopped for futility being unable to detect a clear antisuicidal effect in this complex population of patients. Although affected by significant limitations, particularly the low level of adherence with lithium,¹⁵ there is a question on whether RCT might be the most appropriate design to detect its mortality-reducing effect that may take up to 2 year to establish.¹⁶ In this context, well-conducted and adequately powered epidemiological studies, such as Chen's work, provide a sufficient temporal window to observe the possible effects of treatments on suicide outcome, and in general on the reduction of mortality. Notably, the dose-dependent analysis between exposure to mood stabilizers and risk of mortality showed that a longer duration of lithium use was associated with lower risks of all-cause mortality, including suicide.¹⁴

Another important point is that the all-cause, suicide and natural mortality reduction, was detected also in patients treated with valproic acid. This is in agreement with other cohort studies¹⁷ and RCT¹⁸ indicating that valproic acid might be effective in reducing suicide mortality. A point of novelty is, however, the association of use of this anticonvulsant with the reduction in the risk of natural mortality. This finding, observed also in lithium-treated patients, suggests that biological pathways perturbed by both lithium and valproic acid might underlie this clinical effect. Indeed, if, as observed previously, lithium and valproic acid responders have very distinct clinical profiles, biologically these drugs share some common molecular effects such as the inhibition of glycogen synthase kinase 3 beta (GSK-3β).¹⁹ Further exploration of the role of this pathway should be purported, since GSK-3β is a key regulator of neuronal survival. In addition, there is paucity of data on the possible effects of valproic acid on the telomere shortening, a marker of biological aging, that has been shown to be reduced by lithium treatment.²⁰

The observation of a dose dependent effect (in terms of duration of exposure) is of great relevance, although limited by the lack of measurement of serum levels. The latter aspect is certainly of primary importance for ensuring adequate treatment adherence and, consequently, to establish whether the exposure to the drug is uniform during the period of observation. But the analysis of serum levels can also allow to detect therapeutic windows specific for a certain clinical effect. For instance, neuroprotective effects of lithium might be exerted at low doses,²¹ while its antiviral properties show greater potency at higher serum levels.²²

A final question concerns the possibility that the all-cause mortality reduction effect is coupled to the clinical effectiveness exerted by mood stabilizers. Although there is notion that the anti-suicidal effects of lithium might be to some extent independent from its mood stabilizing properties,²³ it is not clear if this is true also for mortality from natural causes. This area of research is of great interest and needs to be developed in future studies. Mood stabilizers play a fundamental role in the pharmacological armamentarium of the clinicians. Their clinical effectiveness appear to extend beyond the area of mood disorders,²⁴ and is observable also in pediatric population.²⁵ The presence of such a rich spectrum of effects motivates further clinical and translational research to clarify biological mechanisms, identify more accurately patients that can benefit from the use of these drugs as well as to develop pharmacological analogues that could be tested in clinical trials.

The author declares no conflict of interest.