Assessing treatment response to prophylactic lithium use in patients with bipolar disorder

Objetive: To identify potential clinical and epidemiological predictors of long-term response to lithium treatment. Methods: A total of 40 adult outpatients followed in an university hospital, with confirmed diagnosis of bipolar disorder and with history of lithium use for at least a six months period, had their response to this medication assessed through the use of a standardized instrument. The ALDA scale is based on retrospective clinical data, in our study assessed through a thoroughly reviewed of the medical charts, and is used to evaluate the clinical improvement with the treatment (Criterion A), corrected by the acknowledgement of possible confounding factors, such as duration of the treatment, compliance and concomitant use of additional medications (Criterion B), in order to estimate the response that can be specifically attributable to lithium. Results: Our study found an inverse relation between the number of mood episodes with psychotic symptoms and lithium treatment outcome. Conclusion: The results reinforce the hypothesis that lithium seems to be less efficacious in patients with bipolar disorder who present psychotic symptoms.


INTRODUCTION
Bipolar disorder (BD) is a chronic and severe mental illness, with an estimated prevalence of 1% in the general population [1][2][3][4] .The first mood episodes usually appears in adolescence or early adulthood, between 20 and 30 years of age 5 , and the most prevalent clinical form of the disorder is the occurrence of mania or hypomania, interspersed with episodes of depression and periods of euthymia.Taking in account patients with subthreshold mood symptoms, the prevalence of bipolar spectrum disorders in the community may rise up to 6.4% 6 .The disorder is strongly associated with reduced quality of life, high degree of impairment, disability and higher health care costs [7][8][9] .Psychiatric comorbidity in BD is also significant, with a lifetime prevalence of 65% for other Axis-I (DSM) disorders 10 .
Since the famous paper from Cade (1949), in which he describes the use of lithium in patients with BD, many studies have explored the pharmacological treatment options for acute mania, depression, and BD maintenance.Still, considering the myriad of treatment alternatives with Food and Drug Administration (FDA) approval, lithium even now stands as a major option in bipolar maintenance 11 .Lithium received the FDA approval for the prophylactic treatment of BP in 1974, but only in 1994 the first study to assess its efficacy using a control group was published 12 .Later, in 2000, Bowden et al., published the first randomized double blind study with lithium against placebo 13 .
In this context, the attempts to identify possible predictors to lithium response date back to the first publications involving its prophylactic use in mood disorders.Since then, many studies explored the predictive power of a number of variables, with controversial results 14 .Nevertheless, available data in the literature suggest that lithium therapy seems to be more effective in patients with bipolar disorder presenting the so called "core phenotype", that is, with an episodic course, family history of BD, family history of lithium response, absence of rapid cycling and low comorbidity rate 15 .Furthermore, patients with atypical features, including mood incongruent psychosis, mixed states, rapid cycling, and the presence of psychiatric comorbidity, seem to have lower rates of response to lithium therapy [16][17][18][19] .
Concerning specifically the presence of psychotic symptoms in BD, more than half of the patients with BD develop psychotic symptoms at some point of their lives 18 , and the hypothesis that patients with psychotic features, like paranoid delusion and other formal thought disorder, have worst response to prophylactic lithium is old 17 .However, the attempts to confirm the predictive value of psychotic symptoms on the outcome of patients with BD and treated with lithium showed mixed and inconclusive results 14 .
And important aspect involving the predictive value of psychotic symptoms in BD concerns the definition of psycho-sis.In the most frequently adopted model, the psychotic symptoms are divided in mood congruent or mood incongruent.Some data in the literature point to worst treatment outcomes when mood incongruent psychosis is present 14 , what occurs in up to 55% of the patients with BD 20 .
Nevertheless, the literature shows no universally accepted definition of mood incongruent psychosis and even the identification of psychotic features in patients with BD may be challenging 21 .The most commonly adopted definition includes symptoms not directly related to mood polarity, like the presence of paranoid and reference delusions, insertion of thoughts, and control delusions 20,22,23 .Mood incongruent psychosis is associated with worst psychosocial outcomes, possibly because it is related to a higher overall severity of the disorder 20 , though diagnostic issues (such as differential diagnosis with schizophrenia spectrum disorders) may also play a role.Data in the literature also suggest that these patients have a poor response to prophylactic lithium 24 .
This study aims to evaluate, using a standardized method (the ALDA scale), the therapeutic response to lithium in the maintenance treatment of BD, as to assess possible clinical predictors of lithium response.It is important to notice that the present paper was conceived within the framework of a larger genetic study, the ConLiGen, a Genome-wide association study (GWAS) with the objective to determine the phenotype of lithium response, and of which we represent the Brazilian branch.The results presented here are, therefore, a preliminary analysis of the clinical and sociodemographical data of the studied population.The ALDA scale is the instrument adopted by the ConLiGen to determine lithium response.

METHODS
From August 2013 to August 2014, outpatients consecutively seen at the Unicamp General Hospital Psychiatric Clinic with confirmed diagnosis of bipolar disorder (according to DSM-IV) and history of current or previous lithium use were identified by one of the researchers (LFALS) during their routine psychiatric consultation and invited to participate in the study (which implied in contributing with a blood sample for genetic analysis).Only patients with age ≥ 18 years and who used lithium for at least a 6 months period were included.Patients with intellectual disability, neuropsychiatric symptoms possibly related to clinical illness, or inability to understand the study objectives and to freely consent to participate were excluded.
The 40 patients who consented to participate had their medical charts thoroughly reviewed in order to collect clinical and sociodemographic data.We used the collected information to evaluate lithium response in the maintenance treatment of BD, and to explore possible predictors of lithium response.

Response to lithium in the maintenance treatment of bipolar disorder
It can be challenging to evaluate improvement in patients treated in a clinical scenario and not following a strict research protocol.For this purpose, the ALDA scale is considered a valuable tool, and as far as our knowledge goes, is the only instrument that allows to retrospectively assess clinical improvement with lithium treatment in a standardized manner.
The ALDA scale is based on retrospective clinical data and is divided in two major criteria, namely A and B. Criterion A is used to determine the relation between clinical improvement and the treatment (illness activity while in lithium use), with scores ranging from 10 (complete response) to 0 (no change or worsening).
Criterion B concerns possible confounding factors, and is used to establish whether there is a causal relationship between clinical improvement and the treatment.Score 0, 1 or 2 points for each item: number of episodes before the treatment (B1), frequency of episode before the treatment (B2), duration of the treatment (B3), compliance during period(s) of stability (B4), and use of additional medications during the periods of stability (B5).
The ALDA scale total score (TA) is obtained by subtracting B from A. This way, it allows a categorical approach (above or below a cut-off score), or a dimensional approach in the assessment of clinical improvement with lithium treatment 25,26 .The Brazilian version of the ALDA scale adapted to Portuguese from the original English instrument within the context of this study is provided in the Appendix 1.
To use the ALDA scale, the first step is to determine the longest period in which lithium was used and well documented, with information regarding clinical improvement, frequency and duration of mood episodes, total duration of the treatment, compliance during treatment (including drug levels within therapeutic range), and description of additional medication use, including antidepressants and antipsychotics.
In our study, two different researchers (LFALS, JCL) were trained using the clinical vignettes from The International Consortium on Lithium Genetics (ConLiGen), in order to calibrate the scoring of the scale; after that they reviewed the clinical charts.This procedure generated a single data bank, including all the 40 patients, on which both the researchers based their independent score of the ALDA scale.Later they debated about their individual ratings aiming for a consensus score for each of the ALDA scale items.These consensus scores were used together with the other variables (described below) to explore possible predictors of lithium response.

Clinical and sociodemographical variables
Demographic and clinical variables were also obtained in this process such as gender, education, age of onset and frequency of mood episodes, presence and frequency of psychotic symptoms, suicide attempts, hospitalization, electroconvulsive therapy (ECT), use of tobacco, alcohol and other drugs, and family history of mental disorder.
The timeframe considered in this study was from the onset of the illness (that is, first episode of depression or mania) until the beginning of the interval chosen for the scoring of the ALDA scale.This approach was meant to test the predictive power of the studied variables in regard to the lithium response.
To determine the relation between these variables and the scores of the ALDA scale, all data was transformed in ranks and submitted to a logistic regression and linear regression (univariate and multivariate models).The significance level adopted is this study was of 5%.
This study was approved by the Unicamp Research Ethics Review Board and a signed consent form was obtained from all patients prior to entering the study.

Sample description
Of the 40 patients with BD assessed in this study, 15 were males and 25 were females, with a mean age of 40.03 (± 3.08) years.Forty percent never got married, and 57.5% never had children.Regarding education, most of our patients completed high school (28.21%) or had incomplete superior education (23.08%), with a mean of 9.72 (± 4.08) years of scholarship.Only 42.5% could be considered economically active.

Clinical variables
The mean age of onset of mood symptoms was 24.18 (± 9.18) years, while the length of the disorder until the chosen period for scoring the ALDA scale was of 12.85 (± 9.18) years.Manic symptoms manifested slightly earlier than depressive symptoms (26.55 vs. 28.14years), but mixed symptoms showed the smaller mean age of onset, of 24.55 years.A detailed description is shown on Table 1.
Concerning the manifestation of psychosis in BD, our study found that 67.5% of the patients (N = 27) presented with psychotic symptoms at some point of their clinical course, including delusions, hallucinations and severe disorganized behavior.The mean number of mood episodes with psychotic features in this group of patients was of 2.77 (± 2.62), with median of 2.00.
In the majority of the patients who presented psychotic symptoms at some point (62.96%),such symptoms occurred exclusively on manic episodes; only in 3.7% of the cases this symptoms occurred exclusively in depressive episodes.In 33.33% of the patients, the occurrence of psychotic features happened in both mania and depression.Most of the patients (75.86%) had at least one episode with mood incongruent psychotic symptoms, that is, paranoid and reference delusions, auditory hallucinations and severe disorganized behavior.Among the studied variables, there are those that point to a more severe clinical course, like history of suicide attempt.In our sample, 22.5% of the patients attempted suicide.The mean age of the first suicide attempt was 26.89 years (± 12.81).In 28.57% of the cases, the patient had clinical complications with the need to receive medical support.
Other important evidence of clinical severity is the need for hospitalization.In the vast majority of the cases (87.5%;N = 35), the patient was hospitalized at least once, with a mean of 3.4 (± 2.51) hospital admissions per patient, and median of 2.0.Finally, 25% had history of ECT treatment for mania, depression or mixed states.
Concerning the presence of comorbidity, at least one clinical disease was present in 40% of the cases, and at least one other psychiatric disorder was found 57.5% of the cases.Substance use disorders, including abusive use and dependence identified by the assistant doctor and reported in the medical chart, had a prevalence of 7.69% for alcohol related disorders and 10.0% for other drugs.History of current or previous tobacco use had a higher prevalence, of about one third of the patients (31.58%).
In relation to the instituted maintenance treatment for BD, the vast majority of the patients had history of current or previous use of more than one pharmacological agent beside lithium.The most frequent association was with antipsychotics (95.0%), followed by anticonvulsants (80.0%), benzodiazepines (60.0%) and antidepressants (42.5%).
Finally, concerning the presence of family history of mental disorder, it is interesting to notice that about 80% of the patients had family history of mood disorders, 20% for psychotic disorders and 42.5% for some other mental disorder, specified or not.

Lithium response
The consensus scores obtained for each of the ALDA scale items are presented in the Table 2.The mean duration of the selected intervals to score the scale was 37.35 months (± 39.74), with a median of 24.5 months.

Univariate and multivariate analysis
Despite the large number of variables we have analyzed, only some of them were chosen to integrate the multivariate model.This choice was based in their distribution in the studied population and data regarding possible predictors to lithium response in the literature.The selected variables are exposed in Tables 3 and 4.
Concerning the analysis involving the TOTAL ALDA scores (A minus B), we highlight that only four patients had TA ≥ 7, the most common cut-off score adopted in studies that use the scale.This distribution did not allow the creation of two distinct groups (responders vs. non-responders).Thus, we adopted a different approach to the data, and used the scores of Criterion A as a continuous variable with the exclusion of patients with TB scores > 4, a strategy also adopted by the ConLiGen 27 .As a result, we have obtained a subgroup of 26 patients, which was analyzed.The rationale for this approach is that patients with a high prevalence of confounders, that is, with high TOTAL B scores, can compromise the inference of causality between clinical improvement and lithium use.The multivariate analysis (N = 23), showed that the number of mood episodes with psychotic symptoms had and inverse relation with clinical improvement, accessed through criterion A of the ALDA scale (PE -1.24241).On the other hand, we have found a direct relation between clinical improvement and history of current or previous tobacco use ( PE 7.66608).

DISCUSSION
The available data in the literature suggests that the prevalence of BD is similar in males and females 28 , but the clinical presentation and the received treatment can differ significantly between sexes.Female patients are treated more often with antidepressants, benzodiazepines, neuroleptics, lamotrigine and psychotherapy than their male counterparts (52).In our sample, concerning lithium response, we have not found difference between male and female patients, which is in consonance with the data in the literature 29,30 .
As for the age of onset of mood symptoms, it is considered one of the variables with potential do discriminate patients with BD in clinical subgroups.Nevertheless, the validity of this age based groups is uncertain, and the criteria used to discriminate the patients are not consensual 5,31 .We found no relation between age of onset and lithium response.
Concerning the number and frequency of mood episodes, data in the literature suggest that a higher duration of the disorder or a higher frequency (number) of mood episodes is directly related with morbidity, including higher rates of mood cycling and length of the mood episodes 32 .However, other studies like the systematic review conducted by Kleindienst et al. found no relation between the number of mood episodes and lithium response in patients with BD 14 .Our results also found no relation between the number of episodes of mania, depression, hypomania or mixed states and the clinical response to prophylactic lithium.
Bipolar disorder is strongly related with premature mortality, with rates up to 2-3 times higher than in general population 33 .Suicide is the main cause of early mortality in this patients 34 .During the course of their lives, approximately 50% of the patients with bipolar disorder attempt suicide 35 , and between 11% and 19% commit suicide 36 .In our study, 22.5% of the patients attempted suicide, which is lower than the rates reported in the literature.We have not found positive or negative relation between history of suicide attempt and clinical improvement with prophylactic lithium.
We have also explored the relation of history of electroconvulsive therapy (ECT) and clinical improvement with lithium use.ECT can be a valuable therapeutic option in patients with BD, in mania or depressive episodes that are refractory to pharmacotherapy.In our sample, 25% of the patients had history of ECT treatment, prevalence similar with the one found in the literature.We did not found any relation between lithium treatment improvement and history of ECT.Finally, an interesting finding of our study was the positive relation between current or previous tobacco use and clinical improvement with lithium treatment.Prevalence of tobacco use among patients with BD varies between 31.2% 37 to 66% 38 .In our study the prevalence was of 31.58%.This result conflicts with literature data in which the use of tobacco is more frequently associated with worst treatment outcomes and higher recurrence and severity rates in patients with BD 37 .
It is interesting to notice that both tobacco alkaloids and lithium reduce the activity of glycogen-synthase-kinase 3 (GSK3), through the increase of its phosphorylation, thus modulating cellular inflammatory and apoptotic pathways 39,40 .Nonetheless, we did not perform a detailed characterization of tobacco use in the assessed patients and for this reason we consider this finding as a secondary outcome which would demand a different study design to be properly investigated.

Psychotic symptoms
Our study found, in the multivariate model, an inverse relation between the number of mood episode with psychotic symptoms and clinical improvement with lithium treatment.This finding reinforces the hypothesis that lithium treatment tends to be less efficacious in these patients [16][17][18][19] .It is important to notice that our variable does not discriminate between mood incongruent and congruent psychosis, but 76% of the assessed patients with psychotic features had, at some point of their clinical course, mood incongruent symptoms.
It is not clear if the presence of this symptoms indicate a particular form of BD or simply a non-specific manifestation that follows the subjacent mood symptoms 41 .It is a fact though, that psychotic features indicate a severe clinical presentation of the mood disorder, and is related with worst scores in the PANSS, YMRS and GAS scales 18 .Therefore, a possible explanation to this results concern the phenotype of the studied population, since what is labeled as bipolar disorder today is a significantly larger and more heterogeneous group than the manic-depressive disorder patients of the recent past 42 .
Our study has a number of limitations.First, as any other study in which patients are retrospectively assessed through the review of their medical charts, the obtained data is limited to the quality of these records, and the occurrence of missing data or poor descriptions of the patients' phenomenology during mood episodes are not rare.The sample size (n = 40) is also small, what can limit the predictive power of our conclusions.Hence, it is possible that the relation between psychosis and worse outcome with lithium treatment is incidental, and we could be dealing with a sample of patients affected with more severe forms of mental disorder.
Finally, we believe future studies exploring the relation between smoking habit and BD, with larger sample sizes and a more refined description of tobacco use, could lead to some interesting findings, possibly with implication in the clinical treatment of these patients.

CONCLUSION
We found that patients with higher number of mood episodes with psychotic symptoms had worse response to lithium maintenance treatment, what is in consonance with the available data in the literature.

Table 1 .
Clinical features of the mood episodes N < 40 are due to the nonoccurrence of a given phenomenon or, in some cases, to missing data. *

Table 2 .
Consensus Scores for the ALDA Scale