Lithium in patients with amyotrophic lateral sclerosis (LiCALS): a phase 3 multicentre, randomised, double-blind, placebo-controlled trial

Summary Background Lithium has neuroprotective effects in cell and animal models of amyotrophic lateral sclerosis (ALS), and a small pilot study in patients with ALS showed a significant effect of lithium on survival. We aimed to assess whether lithium improves survival in patients with ALS. Methods The lithium carbonate in amyotrophic lateral sclerosis (LiCALS) trial is a randomised, double-blind, placebo-controlled trial of oral lithium taken daily for 18 months in patients with ALS. Patients aged at least 18 years who had ALS according to the revised El Escorial criteria, had disease duration between 6 and 36 months, and were taking riluzole were recruited from ten centres in the UK. Patients were randomly assigned (1:1) to receive either lithium or matched placebo tablets. Randomisation was via an online system done at the level of the individual by block randomisation with randomly varying block sizes, stratified by study centre and site of disease onset (limb or bulbar). All patients and assessing study personnel were masked to treatment assignment. The primary endpoint was the rate of survival at 18 months and was analysed by intention to treat. This study is registered with Eudract, number 2008-006891-31. Findings Between May 26, 2009, and Nov 10, 2011, 243 patients were screened, 214 of whom were randomly assigned to receive lithium (107 patients) or placebo (107 patients). Two patients discontinued treatment and one died before the target therapeutic lithium concentration could be achieved. 63 (59%) of 107 patients in the placebo group and 54 (50%) of 107 patients in the lithium group were alive at 18 months. The survival functions did not differ significantly between groups (Mantel-Cox log-rank χ2 on 1 df=1·64; p=0·20). After adjusting for study centre and site of onset using logistic regression, the relative odds of survival at 18 months (lithium vs placebo) was 0·71 (95% CI 0·40–1·24). 56 patients in the placebo group and 61 in the lithium group had at least one serious adverse event. Interpretation We found no evidence of benefit of lithium on survival in patients with ALS, but nor were there safety concerns, which had been identified in previous studies with less conventional designs. This finding emphasises the importance of pursuing adequately powered trials with clear endpoints when testing new treatments. Funding The Motor Neurone Disease Association of Great Britain and Northern Ireland.


Introduction
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease in which motor neurons in the brain and spinal cord degenerate, resulting in progressive paralysis ultimately leading to dependence on mechanical ventilatory support or death, usually within 3 years. Riluzole, a benzothiazole derivative, improves survival in patients with ALS; 1 however, the eff ect is moderate and there remains a pressing need for more eff ective disease-modifying treatments.
Lithium has neuroprotective eff ects in cell 2-6 and animal 7 models of neurodegeneration, including ALS, although not in all studies. 8,9 In a positive study, transgenic ALS mice treated with lithium showed improved survival compared with wild-type mice treated with saline, 7 and in a pilot study of lithium in patients with ALS there was a signifi cant eff ect on survival in the lithium plus riluzole group compared with the riluzole only group. 7 The design of that trial could be criticised because the method of randomisation was not stated, it was not placebo-controlled, and only 44 patients were included. 7,10 However, the reported diff erence in survival at 15 months (100% survival in the lithium plus riluzole group compared with 70% in the riluzole only group), 7 taken together with the neuroprotective and neuroregenerative eff ects of lithium in the cell and animal models, suggested that a defi nitive randomised placebo-controlled trial was warranted. Therefore, a UK group of Motor Neurone Disease Association-supported ALS centres within the Dementias and Neurodegenerative Diseases Research Network (DeNDRoN) designed and undertook the lithium carbonate in amyotrophic lateral sclerosis (LiCALS) trial to test the hypothesis that lithium improves survival in ALS. Although several other trials were either in progress at the time or were planned, we knew that none had the same survival-based design, and we argued that a defi nitive answer was needed to show whether lithium has a biologically and clinically signifi cant eff ect on survival and function and is safe and well tolerated in ALS.

Patients
LiCALS was a multicentre, double-blind, randomised, placebo-controlled trial undertaken at specialist ALS clinics at ten participating centres in the UK. The full protocol is described elsewhere. 11 Eligible participants were adults aged at least 18 years who met the following criteria: possible, laboratory supported probable, probable, or defi nite ALS according to the revised version of the El Escorial criteria (the Airlie House Statement); 12 an electromyogram compatible with ALS; disease duration of at least 6 months and no more than 36 months (inclusive); and receipt of riluzole treatment for at least 1 month before enrolment. Women of childbearing potential were excluded if pregnant, if breastfeeding, or if a urine pregnancy test before randomisation was not negative. Exclusion criteria included participation in another therapeutic study in the preceding 12 weeks, use of other investigational drugs, tracheostomy or other assisted ventilation in the preceding 3 months, an existing gastrostomy, a medical disorder that might interfere with diagnosis or functional assessment, hepatic or renal insuffi ciency, a major psychiatric disorder, clinically evident dementia, or allergy to lithium.
All participants gave written, informed consent to participate before screening. The study was ethically approved by the South East Research Ethics Committee, reference 09/H1102/15.

Randomisation and masking
Randomisation was done via an online system based at the King's Clinical Trials Unit (CTU) at the Institute of Psychiatry (London, UK). Patients were randomised (1:1) to lithium or placebo at the level of the individual by block randomisation with randomly varying block sizes, stratifi ed by study centre and site of disease onset (limb or bulbar). Active and placebo lithium tablets were identical in appearance, dimensions, mass, and disintegration time, and patients were prescribed up to three tablets daily, adjusted according to serum lithium concentrations. Active tablets contained 295 mg of lithium. The aim was to achieve therapeutic lithium concentrations, defi ned as 0·4-0·8 mmol/L. To maintain double-blind status while retaining the ability to monitor lithium concentrations, one physician from each site was unmasked but had no patient contact. The unmasked physician instructed the masked physicians and nurses to adjust the dose. Unmasked research nurses assisted unmasked physicians in the collection of laboratory results and in ensuring sign off by the physician administering lithium. Blood lithium concentrations were entered on a web-based electronic case report form system, which was accessible only to the study data manager and the unmasked physicians so that central monitoring of lithium concentrations was possible. The unmasked physician adjusted the lithium dose of patients in response to lithium concentrations and the King's CTU checked that this process was consistent and timely. For patients on placebo, sham dose adjustments were recommended to the unmasked physician by the CTU to avoid unmasking. Placebo dose adjustments were done by pairing patients in the lithium and placebo groups as they were randomised. Each time a patient in the lithium group was randomly assigned a group, the unmasked study data manager was alerted by the randomisation system when the next patient in any centre was randomly assigned to placebo. The data manager then monitored the dose adjustments of the patient in the lithium group who was randomised fi rst and instructed the study physician of the paired patient on placebo to adjust the dose of study drug at the same time after randomisation as for the active patient and to the same dose adjustment. This process was monitored to ensure masked physicians and nurses remained masked throughout the trial (appendix). Adverse events were managed by clinicians masked to treatment allocation and serum lithium concentrations.

Procedures
The primary outcome was death from any cause at 18 months, defi ned from date of randomisation and verifi ed with documented evidence of death or of survival beyond 18 months in all cases. Secondary outcome measures, which comprised functional health status measured with the ALS functional rating scale-revised (ALSFRS-R), mental health state measured with the hospital anxiety and depression scale, and quality of life measured with the EuroQol (the EuroQoL group 5-dimension self-report questionnaire health state tariff and health evaluation scale), were assessed at baseline and 3, 6, 9, 12, 15, and 18 months. Reports of adverse events, whether related to the study drug or not, were collected and recorded at each timepoint. The masked physician at each site was responsible for deciding whether an adverse event was serious according to the Good Clinical Practice guidelines. Completed serious adverse event forms were then sent to the coordinating centre for review by the Chief Investigator. Non-serious adverse events were recorded in the electronic case report form alone.

Statistical analysis
The sample size was based on detection of a diff erence in survival rates at 18 months using Fleiss's method for a proportion incorporating a continuity correction. Two groups of 110 patients would give 80% power to detect a diff erence of 17·5% in survival rates (65% vs 82·5%) assuming a two-sided type 1 error rate of 5%. The fi gure of 82·5% represents a treatment eff ect midway between a typical survival rate at 18 months of 65% and the 100% survival reported in the original positive lithium study. 7 The primary analysis was of survival rates at 18 months in patients randomised to lithium treatment versus patients randomised to placebo by intention to treat, compared by the Mantel-Cox log-rank χ² statistic. Originally, a test of survival proportions was planned, but See Online for appendix clinical feedback suggested that survival over the entire study was important, and a log-rank test would be more in keeping with other studies of other drugs for treatment of ALS. This decision was made in August, 2010, before the masking was broken. This endpoint was a protocol amendment that was approved by the trial steering committee before the protocol was published (September 2011). Two further pre-specifi ed analyses of the primary outcome measure were done: an intention-to-treat analysis of survival rates at 18 months using logistic regression with adjustment for randomisation strata, and a per-protocol analysis of time to death (censored at 18 months) using a Cox proportional hazards model. Patients were included in these analyses only if they had taken at least 75% of tablets prescribed in every quarter during participation.
Compliance was assessed using the number of tablets prescribed as recorded in the medication log compared with the number of tablets returned to the pharmacy at each study visit. The site nurses recorded the returns in the electronic case report form, site pharmacies recorded the returns independently in the pharmacy fi le, and the trial manager checked discrepancies between electronic case report form and pharmacy fi les during site visits, to ensure the correct returns were recorded on both, before drug returns were shipped back to the manufacturer for eventual destruction.
Rate of change in each of the secondary outcome measures was compared between groups using mixed models with variation between patients and variation between occasions nested within patients treated as random eff ects. Time in years was treated as a continuous variable. Results, adjusted for diff erences between randomisation strata, are given in the form of diff erences in rate of change with corresponding 95% CIs between patients assigned to lithium and those assigned to placebo. This approach was extended to assess the eff ect of missing data 13 by joint modelling of the survival data (with a Cox proportional hazards model) and the secondary outcomes (with mixed models similar to those described earlier). The Cox model was built by forward stepwise regression, but backward elimination gave the same model. The variables included were El Escorial category, age of onset, slow vital capacity, and sex. An eff ect of lithium was then added to the model. Variables excluded were handedness, site of symptom onset, time between onset and diagnosis, and pulse rate. Within both of these models, an additional latent variable was included that can be conceptualised as the propensity to experience poor outcomes. The inclusion of this latent variable is common to both models and allows us to adjust our estimates of the treatment eff ect to allow for the diff erent rates of dropout in each group. Both models were estimated simultaneously, thus maximising the joint likelihood over both the survival and repeated measures data. Results are given in the form of the diff erence in rate of change of each of the secondary outcomes with corresponding 95% CIs.
Time to fi rst serious adverse event was analysed with a Cox proportional hazards model. We calculated hazard ratios (risk of a serious adverse event) and associated 95% CIs for patients assigned to lithium compared with patients assigned to placebo for all serious adverse events and for all serious adverse events excluding death. For quality of life scores, data imputation was undertaken provided at least half of the items in any scale had been completed; the imputed missing value was defi ned as the mean value of the non-missing items.

Role of the funding source
The funding source had no role in study design, data collection, data analysis, data interpretation, writing of the report, or the decision to submit for publication. The corresponding author had full access to all data in the study, and the manuscript was written by the corresponding author with assistance from other members of the writing committee. The corresponding author had fi nal responsibility for the decision to submit for publication.

Results
The fi gure shows the trial profi le. Between May 26, 2009, and Nov 10, 2011, 243 patients were screened, 214 of whom were recruited from ten centres (appendix) and were randomly assigned to lithium (n=107) or to placebo (n=107). The groups were well balanced for demographic and clinical characteristics at baseline (table 1). Compliance was good, with 140 patients (65%) taking at least 75% of the prescribed drug in every quarter. Compliance was better in the placebo group (71%) than in the lithium group (60%). During the course of the study, 65 patients withdrew from the study drug (23 on placebo and 42 on lithium), 41 owing to adverse events (12 on placebo and 29 on lithium), seven because they were no longer able to travel to the centre (three on placebo and four on lithium), and 17 for other reasons (eight on placebo and nine on lithium). Of 107 patients randomly assigned to treatment with lithium carbonate, 104 (97%) had at least one blood lithium concentration measurement in the therapeutic range (0·4-0·8 mmol/L), with the mean number of such measurements being 6·6 (SD 2·9; appendix). Of the three patients who did not achieve the therapeutic range, one withdrew after becoming pregnant, one withdrew after an adverse event 1 week into the study, and one had a serious adverse event leading to death before the therapeutic range could be achieved.
63 (59%) patients in the placebo group (56 on study drug and seven off study drug) and 54 (50%) patients in the lithium group (41 on study drug and 13 off study drug) were alive at 18 months. The survival functions did not diff er signifi cantly between groups (Mantel-Cox log-rank χ² statistic on 1 df=1·64; p=0·20). In a post-hoc analysis, after adjusting for study centre and site of onset using a Cox proportional hazards model, the estimated hazard ratio (lithium vs placebo) was 1·35 (95% CI 0·90-2·02). The corresponding result in patients who complied with their treatment (64 in the lithium group and 76 in the placebo group) was 1·40 (95% CI 0·83-2·34). The relative odds of survival at 18 months (lithium vs placebo) adjusted for centre and site of onset was 0·71 (95% CI 0·40-1·24). Therefore, there was no evidence that treatment with lithium infl uenced survival in this patient population. Table 2 lists the mean scores for the secondary outcome measures. The analytical strategy was to compare the rate of change of health status in the two groups by fi tting statistical models that allowed for repeated measures within individuals, for diff erences between randomisation strata, and for loss to follow-up (appendix). There was a marked deterioration in functional health status in both treatment groups.
In the unadjusted analysis for the ALSFRS-R-a functional scale from 0 to 48, where 48 is maximal function-the annual rate of change was -9·31 (95% CI -10·5 to -8·58) in the placebo group and -9·50 (-10·31 to -8·70) in the lithium group (appendix). The diff erence between these rates was -0·19 (-1·28 to 0·90) and was not signifi cant. Based on joint modelling, the rate of change in ALSFRS-R adjusted for survival was -9·47 (95% CI -10·98 to -8·46) in the placebo group and -9·75 (-11·62 to -8·47) in the lithium group. These increases in the estimated magnitude of change are consistent with the assumption that those patients who were lost to follow-up were those with the poorest functional status. As in the unadjusted analysis, the diff erence between these rates of decline in this adjusted analysis was not statistically signifi cant (diff erence -0·28, 95% CI -2·40 to 1·67), and this remained non-signifi cant after adjustment for strata (appendix).
Anxiety scores increased over the period of the study but the increases were small and the diff erence between groups was not signifi cant (appendix). Patients in both groups became more depressed over time, but the diff erence in the rate of change was not signifi cant (estimated diff erence based on the joint model adjusting for randomisation strata was 0·29, 95% CI 0·33 to 1·02). Similarly, quality of life deteriorated over time in both groups but the diff erence in the rate of change was not signifi cant.
117 patients had at least one serious adverse event during the 18 month follow-up period (56 patients in the placebo group and 61 patients in the lithium group;

Discussion
In this double-blind, randomised controlled trial of lithium versus placebo in ALS using a survival design, we found that there was no evidence that treatment with lithium resulted in an increase in survival at 18 months (panel). There was a marked deterioration in functional health status and quality of life, with an associated increase in depression and anxiety over time in patients assigned to either treatment group. The fi rst reported study of lithium in patients with ALS 7 examined 44 patients, 16 given lithium and riluzole and 28 given riluzole alone. There were no deaths at 15 months in the riluzole and lithium group compared with eight (29%) in the riluzole only group. Furthermore, patients in the lithium group showed little progression on functional measures compared with typical progression in the control group. This study came after a study of mutant SOD1 transgenic mice that showed compelling evidence for a neuroprotective eff ect of lithium in ALS. 7 The study was not placebo controlled and was small, although adequately powered to detect an eff ect similar to that noted in the mice. 7 Our study was powered to detect an eff ect on survival but, additionally, neither standard analysis of function nor joint analysis of function and survival, which accounts for the loss of those dying with worse function, showed any benefi t of lithium.
Similar fi ndings have been noted in other studies (table 4), although none of these had the traditional design of the present study, in which survival was used as an endpoint in a fi xed-duration study with two treatment groups. Thus, despite several studies that failed to show evidence of benefi t for lithium therapy in ALS, none was a traditionally designed double-blind, placebo-controlled, randomised trial with a primary endpoint of survival, and arguably none of these trials could robustly resolve whether 0 months 3 months 6 months 9 months 12 months 15 months 18 months ALS functional rating scale-revised

EuroQoL health evaluation
Data are mean (SD). ALS=amyotrophic lateral sclerosis. HADS=hospital anxiety and depression scale. *For the two HADS scales, a higher score corresponds to poorer outcome; for all other outcomes a higher score corresponds to better outcome. Data are number of patients. One patient had a serious adverse event between two screening visits and before signing the consent form. Because this event occurred before randomisation it has been excluded from the analysis. The relative risk of a serious adverse event (including death) in the lithium group compared with the placebo group was 1·09 (95% CI 0·58-1·39).

Systematic review
We searched PubMed for reports published before Dec 1, 2012, using the following terms: "lithium" and "amyotrophic lateral sclerosis", "motor neuron disease", "motor neurone disease", "ALS", or "MND". We included randomised, placebo-controlled trials and trials of other designs in amyotrophic lateral sclerosis (ALS) or a related disorder that involved lithium. We identifi ed six previous trials, three of which were randomised, placebo-controlled trials, [14][15][16] one that was randomised but not placebo controlled, 7 one that used historical controls, 17 and one that used self-reporting by patients. 18 We noted that all six trials used diff erent, non-traditional methods for design or analysis or both. We did a randomised, double-blind, placebo-controlled trial with a primary endpoint of survival rates at 18 months in 214 patients with ALS, half assigned to lithium treatment and half assigned to placebo.

Interpretation
We noted no diff erence in survival between placebo and lithium groups. This study confi rms the absence of benefi t of lithium for treatment of ALS.
lithium could have a small but biologically signifi cant eff ect on ALS progression, as measured by the gold standard endpoint for phase 3 studies, survival. Although our study is underpowered for detection of a small change in survival (eg, 5-10%, as detected for riluzole 1 ), we accepted this limitation as a pragmatic compromise for an academic-led trial with limited fi nancial resources and without the need for regulatory approval had we detected a benefi cial eff ect. Adjusting for study centre and site of onset, the relative odds of survival at 18 months (lithium vs placebo) was 0·71 (95% CI 0·40-1·24). Our projected survival rates of 65% and 82·5% in an achieved sample of 214 patients would correspond to an odds ratio of 2·54signifi cantly higher than the upper limit of 1·24 in our estimate. We can therefore be confi dent that if there is any eff ect of treatment with lithium on survival, it is very much lower than that hypothesised. Studies of function in clinical trials of ALS are diffi cult because patients with the worst function are most likely to die and no longer contribute scores to the mean; any decline is masked and there is reduced power to detect an eff ect. We have approached this problem in two ways. First, we analysed the outcomes in patients who did not survive until the study end, confi rming outcomes were indeed poorer. Second, we adjusted for loss to follow-up, confi rming that this adjustment increases the estimated annual change in functional scores, mental health scores, and quality of life. By undertaking a joint analysis of survival and function, we were able to control for the loss to follow-up in the comparison between treatment groups and showed no benefi t of treatment. Further studies are needed to identify the optimum method for analysis of functional scores in trials of treatments in ALS.
In this phase 3, randomised, placebo-controlled, double-blind trial of lithium therapy in ALS, we did not fi nd evidence of benefi t, but nor were there safety concerns, which had been identifi ed in previous studies with less conventional designs. 14 This fi nding emphasises the importance of pursuing adequately powered trials with clear endpoints when testing new treatments, bearing in mind that a trial tests biologically important hypotheses as well as clinical effi cacy. Previous lithium trials could not adequately address this issue, but our results suggest that we can now be confi dent that lithium at these serum concentrations does not signifi cantly infl uence disease progression, as assessed either by a validated functional measure (ALFRS-R) or by survival.

Contributors
PNL liaised with colleagues to develop and write the protocol and all trial materials, was the original chief investigator until June 2010, advised on the undertaking of the trial, and participated in writing and revising the manuscript. AA-C was the chief investigator from June, 2010, to the trial end in 2012, undertook the literature search, participated in data collection and interpretation, and wrote and revised the manuscript. INS was the trial statistician, contributed to the design of the study, undertook the statistical analysis, and participated in writing the manuscript. JK was the senior data manager for the King's CTU; CLM was the manager of the King's CTU. MT was the overall clinical trial manager. TLW was a principal investigator and contributed to study design. CY was a principal investigator and contributed to the protocol, participated in data collection, and helped write the manuscript. PJS was a principal investigator, assisted in the inception and design of the study, advised on the conduct of the study, participated in data collection, and helped draft the manuscript. JE, COH, TM, KEM, RO, and KT were principal investigators. HHo was a research nurse and contributed to the study design. RA-J, JB, and MH were research nurses and participated in data collection, study visit management, and patient recruitment. SDh and RHor were research nurses and trial coordinators for their sites. EO, AS, CWi, LWy, WB, GM, and LPar were research nurses and contributed to data collection. JC, SEC, SDo, AD, RHi, CMe, KOH, LPat, PP, LS, DS, HV, TW, and CWh were research nurses, enrolled patients, and collected data. TKK was a trial physician, participated in data collection and entry, and did clinical assessments. DR and LWi were trial physicians and collected data. DJB, CME, HHa, CH, RHow, II, CMD, CLM, JP, MR, PR, RA, KS, MRT, JW, and IL were trial physicians. LC, FOK, JS, AT, and EW were trial administrators. CEz was a clinical research offi cer, recruited patients, and collected data. CP was a clinical study offi cer. LT was a clinical study offi cer and assessed patient eligibility and enrolment, supervised randomisation, and collected data for her site. CC was clinical trials unit manager for her site. RS was a trial physician. All authors revised the paper and approved the fi nal manuscript.

UKMND-LiCALS Study Group members
Data monitoring and ethics committee: C Allen, C Counsell, A Farrin. Trial steering committee: A Al-Chalabi, B Dickie (co-opted observer), J Kelly (co-opted observer), P N Leigh, C L Murphy (co-opted observer),