4. Discussion
4.1. Main findings
We identified 123 cases of neurological sequelae associated with lithium exposure published from 1965 to 2019, including 23 additional cases not identified in prior reviews. Cerebellar sequelae were observed in an overwhelming proportion of cases (79%). Nearly two out of three cases (63%) had a maximal lithium plasma level <2.5 mEq/l indicative of low/mild toxicity [19]. Fever and/or infection were reported in nearly half of the patients (48%). The likelihood of presenting with cerebellar vs. other neurological sequelae was independently increased by elevated plasma lithium level (≥ 2.5 mEq/l) and by a history of fever and/or infection. Stratified analyses showed that the association between fever/infection and occurrence of cerebellar sequelae was restricted to cases with lithium plasma level <2.5 mEq/l.
4.2. Interpretation of findings
While the incidence of SILENT in the population of lithium users is unknown, this complication may be considered as exceptional considering that the 123 cases identified in the present review were published over a 55-year period. The decreasing frequency of published cases after the ‘90s may be explained by a lower publishing rate of new cases, as is often the case when a drug adverse effect is considered as already well-documented in the literature. This decrease might also be related to changes in prescribing practice over the last decades, with a wider use of anticonvulsants and second-generation antipsychotics in bipolar disorder, leading to low rates of lithium exposure in the general population [17, 48, 49]. Due to the severity of this complication, it is important to keep in mind that SILENT cases are still occurring despite these changes in prescribing practices.
The clinical characteristics of the SILENT cases identified in the present review are comparable to those reported in prior reviews [9, 11-15]. These severe complications of lithium treatment are most frequently observed during routine lithium treatment, with fewer than 10% of cases occurring after accidental or intentional overdoses. Consistently with this low rate of overdose, plasma lithium levels were within the therapeutic range in a large proportion of cases. In cases with toxic plasma lithium levels, several factors known to increase plasma levels were noted such as anorexia [8], dehydration [50], chronic kidney disease [51] or coprescription of diuretics [13, 38, 52, 53].
Typically, the inaugural clinical picture is an episode of acute lithium poisoning with little symptomatic specificity. Neurological symptoms are frequently observed during these episodes, such as tremor, rigidity, hyperreflexia, myoclonia, disorientation, drowsiness or seizures, and transient cerebellar symptoms (dysarthria, ataxia, nystagmus) may also be present [6, 13]. Cases of SILENT without an acute lithium poisoning phase are rare [54].
The questions raised since the observation of the first cases of SILENT are why these neurological symptoms exceptionally persist after the resolution of the acute episode, and why cerebellar sequelae are so frequent in SILENT cases. The neurotoxicity of antipsychotic-lithium polytherapy in general and of haloperidol-lithium was long considered as a plausible hypothesis explaining the occurrence of these cases [12, 16] and is still mentioned as a key factor in the recent literature on drug-induced cerebellar syndromes [37, 55]. However, the existence of a causal link between exposure to haloperidol and the occurrence of SILENT was soon questioned as the frequency of haloperidol-lithium polytherapy was comparable in persons with and without neurological sequelae [9, 56]. In the present review, half of the cases were prescribed antipsychotics, a prescription pattern comparable to that observed in persons with bipolar disorders treated in the last decades of the 20th century [57]. No association was found between exposure to antipsychotics and the occurrence of cerebellar sequalae after adjusting for the other variables.
The striking elevated frequency of cases of SILENT with a history of fever or infection, first noticed by Schou [9], is confirmed in the present study. What the latter adds to prior narrative reviews is to show that fever is associated with an increased risk of cerebellar sequelae, independently from other characteristics, and that this increased risk is restricted to cases with plasma lithium levels <2.5 mEq/l (no/mild toxicity)[19]. In the group with no/mild toxicity, more than two thirds (71%) of patients with cerebellar sequelae had a history of fever/infection vs. 17% of patients with other sequelae. Conversely, the presence of fever/infection was not associated with the type of neurological sequelae in the group of cases with lithium levels ≥ 2.5 mEq/l (moderate/severe toxicity).
These findings suggest that two distinct pathophysiological pathways may lead to neurological sequelae induced by lithium. This hypothesis is speculative and has to be regarded as an oversimplification of the complex relationships between cerebellar sensitivity to both lithium neurotoxicity and fever. Schematically, the first pathway is characterized by an episode of lithium poisoning with elevated lithium levels leading to a wide range of neurological sequelae, including cerebellar ones, perhaps because the lithium neurotoxic threshold is reached everywhere in the nervous system. The second pathway is marked by the occurrence of cerebellar sequelae after a febrile intercurrent episode in patients with lithium levels within the therapeutic range or moderately elevated, perhaps because the lithium neurotoxic threshold is lower in this brain area in the event of fever. The biological plausibility of a causal link between fever and risk of cerebellar sequelae in lithium users is supported by the long-known sensitivity of the cerebellar cortex to heat and hyperthermia [13, 37]. The most consistent neuroanatomical findings found by brain imaging and post-mortem studies are cerebellar atrophy and loss of Purkinje cells [43, 46, 54, 58-62].
4.3. Implications for clinical practice
Decades after the event, prescribers who have encountered such cases keep a vivid memory of patients confined to a wheelchair [9, 46]. As therapeutic resources are limited, prevention of SILENT is crucial and requires prescribers’ awareness about the following points, which synthesize the key findings with implications for clinical practice drawn from current and prior reviews [9, 11-15]:
  1. SILENT may occur at any time during lithium treatment, from a few days after treatment to decades after.
  2. SILENT may occur even when lithium plasma levels are within the therapeutic range. Hence, the clinical symptoms of neurotoxicity are more important than the lithium level for deciding modifications of lithium treatment.
  3. Fever dramatically increases the risk of cerebral sequelae. The literature demonstrates that all causes of fever (infection, neuroleptic malignant syndrome, heat stroke, etc.) appear to be associated with this increased risk, and it is possible that no minimal hyperthermia can be considered as safe [27].
  4. Based upon the literature and the present findings, we recommend the interruption of lithium or at least a dose reduction in the event of fever. As no guidelines are currently available regarding this point, it might be recommended to halve the dosage of lithium. Even if there is a risk of mood recurrence due to lithium withdrawal [63, 64], this risk may be considered as low when the treatment is interrupted only over a few days, and it is negligible compared to the risk of neurological sequelae. Hence, the strategy recommended for lithium in the event of fever or infection should be very close to that recommended for clozapine, the other psychotropic drug with a narrow therapeutic index [65-68].
  5. When lithium users benefit from therapeutic education about the symptoms of neurotoxicity and conditions promoting the occurrence of these symptoms (dehydration, coprescription of diuretics or NSAID, etc), they should receive special warning about the fact that fever, irrespective of its cause, may induce lithium toxicity and that they must consult in the event of fever in order to adapt their dosage.
It is beyond the scope of the present paper to review the management of lithium poisoning: the cornerstones are the early identification of neurotoxic symptoms and the use of aggressive lithium removal methods [5, 6, 19, 69]. Owing to the low incidence of SILENT, no study has examined which strategy may help to prevent neurological sequelae. However, neurological symptoms are known to persist more frequently at intensive care unit discharge in patients with lithium poisoning not treated by extracorporeal toxin removal [6].
4.5. Limitations
The present review is limited by the fact that data were extracted from published case reports who may not be representative of all cases of neurological sequelae occurring in persons exposed to lithium. For instance, we cannot exclude that a publication bias may exist for cases presenting with cerebellar symptoms, leading to an overrepresentation of such cases in the literature. However, such a bias, if any, should not have impacted the direction and strength of the association between fever and occurrence of cerebellar sequelae, as a systematic publication bias of cases with fever and cerebellar symptoms is unlikely. Furthermore, we may have missed some published cases, but we have little reason to suspect that we selectively missed cases presenting with features different than those identified in the present review. Lastly, a third unknown factor and therefore not adjusted for in the present analysis, and independently associated with increased risk of fever and increased risk of cerebellar sequelae, may confound this association.
4.6. Conclusion
During the SARS-CoV-2 pandemic, prescribers have to be aware of the risks of cerebral sequelae associated with infection and fever in lithium users, and should warn them of the need to consult in the event of fever to adjust their lithium dosage. As the occurrence of SILENT is exceptional, there is no need to preventively modify lithium treatment because of the pandemic as its benefit/risk balance remains largely positive, including with respect to neuroprotection [70, 71].