What is the optimal serum level for lithium in the maintenance treatment of bipolar disorder? A systematic review and recommendations from the ISBD/IGSLI Task Force on treatment with lithium

Abstract Aims To systematically review the existing trials on optimal serum levels for lithium for maintenance treatment of bipolar disorder and to develop clinical recommendations. Methods Systematic literature search. Discussion of major characteristics, limitations, methodological quality, and results of selected trials. Delphi survey consisting of clinical questions and corresponding statements. For statements endorsed by at least 80% of the members, consensus was considered as having been achieved. Results With strict inclusion criteria no studies could be selected, making it difficult to formulate evidence‐based recommendations. After loosening the inclusion criteria 7 trials were selected addressing our aims at least to some extent. Four of these studies suggest better efficacy being associated with lithium serum levels in a range above a lower threshold around 0.45/0.60 and up to 0.80/1.00 mmol/L. These findings support the outcome of the Delphi survey. Conclusions For adults with bipolar disorder there was consensus that the standard lithium serum level should be 0.60‐0.80 mmol/L with the option to reduce it to 0.40‐0.60 mmol/L in case of good response but poor tolerance or to increase it to 0.80‐1.00 mmol/L in case of insufficient response and good tolerance. For children and adolescents there was no consensus, but the majority of the members endorsed the same recommendation. For the elderly there was also no consensus, but the majority of the members endorsed a more conservative approach: usually 0.40‐0.60 mmol/L, with the option to go to maximally 0.70 or 0.80 mmol/L at ages 65‐79 years, and to maximally 0.70 mmol/L over age 80 years.


| INTRODUC TI ON
Lithium has been licensed in Europe and North America for the longterm maintenance treatment of bipolar disorder (BD) for more than 45 years and is justifiably considered one of the first line options in major global treatment guidelines, such as the WFSBP guideline, 1 the NICE guideline, 2 the RANZCP guideline, 3 the Dutch guideline, 4 the BAP guideline, 5 the CINP guideline 6 and the CANMAT/ISBD guideline. 7 However, lithium has a very low therapeutic index, that is, a low ratio between the dose (or serum level) that is associated with toxicity (mainly CNS and renal toxicity) and the dose associated with therapeutic effect. The therapeutic index for lithium is approximately 2. 8 For maintenance treatment of BD several reviews recommend differing minimum effective lithium serum levels ranging from 0.40 mmol/L 9-11 and 0.50 mmol/L, 12 or 0.60 mmol/L 13 to as high as 0.80 mmol/L. 14,15 In addition, some major clinical practice guidelines recommend optimal ranges, but these also lack consistency. For instance, the NICE guideline, 2 the RANZCP guideline 3,16 and the BAP guideline 5 recommend 0.60-0.80 mmol, the CANMAT/ISBD guideline 7 0.60-1.00 mmol/L, and the CINP guideline 6 0.60-1.20 mmol/L, whereas other guidelines, for example, the WFSBP guideline, 1 do not provide specific recommendations. Finally, some reviews more specifically recommend targeting serum levels for individual patients "on the basis of efficacy and tolerability" without further specification 15 or "primarily at 0.60-0.75/0.80 guided by clinical response and tolerability" 16 within a broad range of 0.40-1.20 mmol/L, 10 which is also recommended in the Dutch guideline. 4 The major problem underlying these different recommendations is that large studies, well designed to assess the optimal serum level in maintenance treatment, specifically the serum level that best balances efficacy and tolerability for most patients, are lacking.
In clinical pharmacology recommendations for the optimal dose (or serum level) of a drug should be based on studies in which patients are randomized to different groups with various a priori defined fixed doses/serum levels (or narrow dose/serum level ranges) of the drug in question. This is necessary as in studies with flexible doses/serum levels or with wide dose/serum level ranges, prescribers are inclined to increase the dose/serum level especially in those patients who are not responding to (and are tolerating) the original dose/serum level, resulting in a so called channeling effect: nonresponding patients are treated with higher doses/serum levels than patients who have already responded to lower doses/serum levels 17 or dosing following outcome rather than outcome following dose. 18 To our knowledge, the most recent systematic review on optimal serum levels for lithium in the maintenance treatment of BD by Severus et al 10  Based on the need for an updated and more critical systematic review of the literature on optimal serum levels for lithium in the maintenance treatment of BD, a joint International Society for Bipolar Disorders (ISBD) and International Study Group on Lithium (IGSLI) Task Force on the Treatment with Lithium was charged to address this topic (and other lithium related topics). In addition, it was decided to develop a consensus on recommendations for optimal serum levels that also attempted to answer key clinical questions such as those outlined above, for which there apparently was no direct evidence available, using the Delphi method. above a lower threshold around 0.45/0.60 and up to 0.80/1.00 mmol/L. These findings support the outcome of the Delphi survey.
Conclusions: For adults with bipolar disorder there was consensus that the standard lithium serum level should be 0.60-0.80 mmol/L with the option to reduce it to 0.40-0.60 mmol/L in case of good response but poor tolerance or to increase it to 0.80-1.00 mmol/L in case of insufficient response and good tolerance. For children and adolescents there was no consensus, but the majority of the members endorsed the same recommendation. For the elderly there was also no consensus, but the majority of the members endorsed a more conservative approach: usually 0.40-0.60 mmol/L, with the option to go to maximally 0.70 or 0.80 mmol/L at ages 65-79 years, and to maximally 0.70 mmol/L over age 80 years.

K E Y W O R D S
bipolar disorder, Lithium, maintenance treatment, serum level 2 | ME THOD

| Composition of the task force
The ISBD/IGSLI Task Force on Lithium (further Task Force) was   comprised of global experts active in ISBD or IGSLI and furthermore selected on their contributions to the literature on lithium. The Task Force started at the annual conference of the ISBD in Seoul (2014) while a subgroup of the Task Force (WN, ES, RL, GM, AY, MT) initiated this project on optimal serum levels for lithium at the annual ISBD conference in Washington (2017) and further discussed it at the annual ISBD conference in Mexico City (2018).

| Selection of literature
As a first step 2 members of the Task Force (WN and ES on August 28, 2017) initiated a systematic literature search in PubMed with the following search terms limited to "clinical trials" and "reviews": "lithium" and "bipolar disorder" and ["serum level" or "serum concentration" or "plasma level" or "plasma concentration" or "blood level" or "blood concentration" or "lithium level" or "lithium concentration"].
From the retrieved trials we initially intended to examine and discuss only those which fulfilled a priori defined inclusion criteria (see Table 1). The included trials were rated for their methodological quality regarding randomization, blinding and description of withdrawals, and dropouts ensuring intent to treat analysis, according to the Jadad Scale 19 as poor (score 0-2), moderate (score 3-4), or good (score 5).

| Delphi method
Anticipating that our literature review would fall short of providing clear evidence-based conclusions and definite guidance regarding optimal lithium serum levels, we concurrently developed a survey consisting of clinical questions and corresponding statements using the Delphi method. 20 Informed by the findings of clinical trials that have been conducted and by existing reviews and guidelines we formulated 32 statements on 16 clinical questions (see Supplemental   Table S1) presented to all 50 Task Force members, representing 15 countries. They were asked to score all statements as "Agree" or "Not agree", with the option in several statements to mark a preferred additional more specified response. For statements endorsed by at least 80% of the participating members consensus was considered as having been achieved.
Two survey rounds were initially planned. In the first survey round we tested the statements and besides asking members to choose among the additional responses, we also asked for comments on the statements and on the additional responses. After having received the responses and comments it was concluded that a second survey round was not needed as the comments received did not reveal a need to change statements or responses.

| Further procedure
The subgroup that initiated this project consisted of WN, ES, RL, GM, AY, and MT. Selection of studies and data extraction was conducted by WN and ES. The first drafts of the paper and the Delphi statements were written by WN in collaboration with ES and RL and

| Selection of literature
The results of our literature search and the further selection is presented in Figure 1.
The literature search in PubMed resulted in 128 trials and 204 reviews, of which we based on the abstracts selected 7 trials and 5 reviews for full-text reading. In addition, we added another 12 papers from reference lists of the selected papers and already known to us: 9 trials and 3 reviews.
In the first instance we excluded all 16 selected trials, as none of the trials fulfilled all 6 inclusion criteria (see above). However, we then decided to modify the inclusion criteria by loosening them (see again Papers excluded on abstract: n = 320 -Other aim of the trial: n = 121 -Other topic of the review: n = 199 Additionally identified papers: n = 12 -Trials: n = 9 -Reviews: n = 3 Full-text papers assessed for eligibility: n = 24 -Trials: n = 16 -Reviews: n = 8 In first instance excluded trials: n = 16 -Not fullfilling inclusion criteria: n = 13 -Duplicates: n = 3 Included trials: n = 7 Included reviews: n = 8 In second instance included trials: n = 7 -Fullfilling loosened criteria: n = 7 Included papers: n = 8 -Trials: n = 0 -Reviews: n = 8

IdenƟficaƟon
Identified papers from PubMed search: n = 332 -Trials: n = 128 -Reviews: n = 204 had not required hospitalization" in the past 2 years, indicating that this study was not a pure recurrence prevention study. At baseline randomization all were using lithium while potential use of other psychotropic medication is not reported. Patients were randomized to 2 different a priori defined lithium level ranges: 0.30-0.80 and 0.80-1.40 mmol/L. Endpoint was "change in mood of a sufficient degree to require clinically additional pharmacological intervention […]".
Thus, this study fulfilled only 3 of the original inclusion criteria (1, 5, and 6). Further limitations are that the timing of blood collection was not reported and that the procedure used for randomization is unclear.
Coppen 23  Thus, this study only fulfilled 2 of the original inclusion criteria (1 and 6). Another limitation is that the study was not blind. Thus, this study fulfilled 5 of the original inclusion criteria (1, 2, 3, 5, and 6). A source of bias could be the abrupt lowering of the lithium levels at randomization in patients previously responding to high levels resulting in a risk of drug-discontinuation related morbidity. 34 Another limitation is that only the patients but not the investigators/ clinicians were blind to the dosage-assignment.

Vestergaard 26 included 57 BP and 34 UP patients in a 2-year
follow-up trial. At baseline all patients were still hospitalized (indicating that they were not in remission and implying that it was not a real recurrence prevention study) and about half of them were still using antipsychotics or antidepressants. Whether (some) patients were also already using lithium remains unclear in the report. At baseline patients were randomized double blind to 2 different a priori defined lithium level ranges of 0.50-0.80 and 0.80-1.00 mmol/L. Endpoint was recurrence (or relapse) defined as "re-emergence of affective disorder of a severity that rendered rehospitalization necessary".
Thus, this study fulfilled only 2 of the original inclusion criteria (5 and 6). Further limitations are that the study was not blind and that only the number of BP patients who were included is presented, leaving unclear how many BP patients completed the study and how many dropped out.
Nolen 27 published a post-hoc analysis of a long-term study intended to test the efficacy of quetiapine. 39 In this study patients with BP-1 disorder with a recent manic of depressive episode who had Thus, this study fulfilled 4 of the original inclusion criteria (1, 2, 3, and 4). Besides that the composition of both lithium arms was not the result of randomization and the levels were not a priori defined, another major limitation of this study is that it was discontinued prematurely by the sponsor after an interim analysis after 300 mood events had revealed superiority of quetiapine over placebo, which resulted in loss of probably 15%-20% of patients before completing the 2-year follow-up. Finally, although called a "trough" level, it is unclear form the paper whether blood was sampled after 12 ± 1 hour after the last evening intake.
In summary, none of the 7 trial studies described above fulfilled all original inclusion criteria (  (Waters), who were hospitalized (Vestergaard) or included patients with morbidity (Coppen, Maj) and thus also examined the direct/acute effects. Only 1 study (Nolen)

| Quality assessment of the included studies
Regarding JADAD criteria to rate methodological quality, 5 of the 7 studies involved randomized assignment to particular serum concentrations of lithium. However, the randomization was described adequately in only studies (Gelenberg, Vestergaard) and not in the remaining 3 studies (Jerram, Waters, Coppen); 2 studies were not randomized (Maj, Nolen). Three studies had a double-blind design, which was well described in 1 study (Nolen)   The findings of the 7 studies are summarized in Table 4. Taken

| Delphi survey
The Delphi survey was completed by 33 of the 50 (66.6%) Task Force members. Many statements received additional comments, most frequently that evidence regarding a statement or a specified response was lacking.
There was consensus on recommendations regarding timing of blood sampling to measure lithium serum levels during the day and on whether recommendations for optimal serum lithium levels should be the same with once daily dosing compared to twice daily dosing, as well as with different formulations of lithium medication (immediate release vs extended release).
There was a lack of consensus on the overall statement "The higher the lithium serum level, the higher the likelihood of a preventative response" (endorsed by only 54.5%) while several members commented that serum levels in the lower range (below 0.60 mmol/L) are associated with efficacy but in the higher range (above 1.00 mmol/L) with more adverse effects and toxicity. In addition, there was no con- The major recommendations including 7 statements endorsed by at least 80% of the members (consensus) are summarized in Table 5.
Regarding recommended standard lithium serum levels in the maintenance treatment there was only consensus on the recommendation for adults, but not for children <12 years, adolescents 12-17 years, the elderly 65-79 and the elderly ≥80 years. However, the most frequently endorsed minimum level was 0.40 mmol/L for all age groups, whereas the maximum level varied with age; 1.00 mmol/L for adults, children and adolescents and between 0.70 and 1.00 mmol/L for the elderly.
Finally, for the prevention of specific episodes there was consensus for recommending similar levels for the prevention of mixed episodes vs manic episodes (endorsed by 87.9%) but not for recommending similar levels for the prevention of manic vs depressive episodes and for the prevention of hypomanic or subthreshold depressive episodes vs manic and depressive episodes, respectively.

| D ISCUSS I ON
The key conclusion from this review (and the major limitation) is that there is a lack of well-designed, double-blind randomized dose finding (or serum level finding) studies assessing optimal lithium serum level in the maintenance treatment of BP. This conclusion is truly remarkable given that lithium has been available for the treatment of BP since it was licensed in the early 1970s. This paucity of data makes it difficult if not impossible to formulate confident evidencebased recommendations for the dosing of lithium based on monitoring its serum level. Based on the initial inclusion criteria for our systematic review we were unable to identify any study that fulfilled all 6 original inclusion criteria for our review. Only after loosening these criteria, we identified 7 studies that addressed, at least to some extent, the questions we posed at the outset, and outlined in the introduction. TA B L E 5 Recommendations from the Delphi survey regarding lithium serum levels in the maintenance treatment of bipolar disorder endorsed by at least 80% (consensus) of the participating members of the task force, or by less than 80% (no consensus)

Domain Recommendation
Timing of blood sampling to measure lithium serum levels

| What is the optimal serum level for lithium in the maintenance treatment of bipolar disorder?
The main finding of our review is that 4 of the 7 included stud- There is also some additional information from some of the above studies. Hullin 40  however, there are no data supporting the notion that going beyond 1.00 mmol/L is associated with any further benefit, possibly with exception of its use in the treatment of acute manic episodes (not further being discussed in this paper). In addition, the upper limit of 0.80 or 1.00 mmol/L is of course also defined by the increased risk of side effects and toxicity at higher lithium serum levels. 41 Taking all these above data into account, they do support the outcome of the Delphi survey: that for adults (18-

| What is the best timing of blood sampling?
No study was found that addressed this question but most of the included studies followed the generally accepted recommendation to sample blood 12 ± 1 hours after the last (ie evening) intake. There was also consensus on this practice among the Task Force members for patients taking lithium twice daily or taking lithium once daily in the evening.
However, there was no consensus on whether to sample in the evening or in the next morning for patients taking lithium once daily in the morning, a dosing schedule probably hardly applied in clinical practice.

| Should the recommendations be different for dosing once daily vs twice daily and for immediate release vs extended release formulations?
There is pharmacogenetic evidence that a total daily dose of lithium given once daily results in a 10%-15% higher 12 ± 1 hours trough level than when given twice daily. 42

| Should recommendations take ethnicity or gender into consideration?
We did not find any study addressing these topics in relation to optimal lithium serum levels. Therefore, we also did not formulate recommendations on these issues.

| Practical considerations regarding lithium level monitoring
Due to lithium's low therapeutic index, its serum levels should always be monitored during treatment, at least when treatment is initiated, after each dose increase and every 3-6 months thereafter.
The blood samples for measuring the lithium level must be drawn at F I G U R E 2 Recommendation for optimal lithium serum levels in the maintenance treatment of bipolar disorder: endorsed in consensus by the Task Force members for adults (18-65 y) and by the majority of the members for children (<12 y) and adolescents (12)(13)(14)(15)(16)(17)  Change medication no steady state occurring after 5-6 times the half-life of lithium, which is approximately 24 hours in the nonelderly with normal renal function. Thus, the lithium level should be measured after 5-6 days at constant daily dosing. In the elderly or in patients with impaired renal function, the time to steady state increases, implying that the lithium level should also be measured after 10-12 days. In individual patients treated with a given schedule with a given lithium formulation, at steady state, the ratio of daily lithium dose, and the 12 hours lithium level is constant. An additional requirement for safe use of lithium is to monitor renal function at baseline and also every 3-6 months.
Finally, it should be taken into account that subjects with a relatively high renal function, in particular young men, may have an unexpectedly low 12-hours lithium level despite a relatively high daily lithium dose due to relative large variations in lithium levels over the day. If the dose is increased to obtain a lithium level within the recommended range, such patients may become intoxicated. 46

| CON CLUS ION
Based on our review and the responses obtained in our Delphi survey we recommend that the standard serum level in the maintenance treatment of adults with BD should be 0.60-0.80 mmol/L with the option to reduce the level to 0.40-0.60 mmol/L in case of sufficient response but poor tolerance, or to increase the level to 0.80-1.00 mmol/L in case of insufficient response and good tolerance. For children and adolescents the majority of the Task Force members endorsed the same recommendation, while for the elderly the majority endorsed a more conservative approach: usually 0.40-0.60 mmol/L, with the option to go to maximally 0.70 or 0.80 mmol/L at ages 65-79 years, and to maximally 0.70 mmol/L over age 80 years. Sampling should be in the morning 12 ± 1 hours after intake of the evening dose. Tentative recommendations for optimal serum lithium levels should be the same with once and twice daily dosing, and with immediate release formulations compared to extended release formulations.
While the above recommendations reflect the available evidence, more high-quality research is still needed to provide more refined advices.

ACK N OWLED G EM ENTS
The following authors reported acknowledgements: Dr. Baldessarini