Psychopharmacology

, Volume 170, Issue 3, pp 225–234

The effects of lithium on cognition: an updated review

Authors

    • Brain Injury ProgramColumbia Health Centre
  • Amy M. Wisniewski
    • Pacific Graduate School of Psychology
Review

DOI: 10.1007/s00213-003-1592-x

Cite this article as:
Pachet, A.K. & Wisniewski, A.M. Psychopharmacology (2003) 170: 225. doi:10.1007/s00213-003-1592-x
  • 502 Views

Abstract

Rationale

Adverse cognitive effects associated with lithium are often implicated as contributing to vocational and social impairment, as well as medication noncompliance. As impaired cognitive functioning caused by lithium has clear clinical implications, it is important to determine whether evidence for or against impaired cognitive functioning exists in the literature.

Objectives

An attempt is made to synthesize findings from previous studies, which assess a variety of cognitive domains, to determine whether conclusions can be drawn regarding lithium-associated cognitive impairment. The "reversibility" of neuropsychological impairment following lithium discontinuation and whether lithium administration has negative cumulative effects on cognition were also reviewed.

Methods

Key word searches on "Medline" and "Psych Info" were completed for clinical articles that investigated the neuropsychological effects of lithium in clinical and normal populations between 1968 and 2000.

Results

Despite methodological flaws, poor replicability and the subtle cognitive effects of lithium, five consistent findings emerged from the review; impairment on tasks of psychomotor speed, impaired functioning in the majority of studies examining verbal memory, no impairment on tasks of visuo-spatial constructional ability or attention/ concentration, and no negative cumulative effect.

Conclusions

Many patients administered lithium carbonate complained of mental slowness. Lithium carbonate also appeared to have definite, yet subtle, negative effects on psychomotor speed. Studies reviewed also showed a trend toward impaired verbal memory. Recommendations with respect to future research, methodological and statistical problems, and additional clinical implications are presented.

Keywords

LithiumMemoryCognitionBipolar

Introduction

The discovery of the beneficial effects of lithium (Li) dates back to the early contributions of Hartigan (1963) and Baastrup (1964), who independently reported on its prophylactic effects in recurrent affective disorders. Since these early studies, Li has been accepted as an important mood stabilizing agent in the treatment of bipolar mood disorder (BMD; Joffe et al. 1988; Bowden et al. 1994; Emilien et al. 1996). However, many patients cannot tolerate Li's many side effects (Bone et al. 1980; Lyskowski et al. 1982). One of the often insidious adverse reactions to Li is the subjective complaint of "mental slowing" (Joffe et al. 1988; Lenzer et al. 1989). However, the relationship between Li treatment and its concomitant subjective neuropsychological side effects has been de-emphasized despite the frequency of patient complaints (Schou 1989).

In addition to the subjective experience of mental slowing, investigators have reported various cognitive deficits in Li-treated normal volunteers and BMD patients in areas such as attention, concentration, memory (Kocsis et al. 1993), creativity (Stoll et al. 1996), word fluency (Shaw et al. 1986; Kocsis et al. 1993) and psychomotor speed (Squire et al. 1980; Lenzer et al. 1989; Kocsis et al. 1993). Li has also been known to cause hypothyroidism at a rate between 2% and 15%; a disorder that is often associated with cognitive impairment (Maarbjerg et al. 1987; Salate and Klein 1987; Tremont and Stern 1997). These adverse drug effects may be an important cause of vocational, as well as lead to social impairment and medication noncompliance (Gitlin et al. 1989; Muller-Oerlinghausen 2000). Estimates of Li noncompliance range from 10% to 47% with cognitive side effects reported as one of the most important contributing factors (McCreadie et al. 1985; Gitlin, et al. 1989).

The suggestion that impaired neuropsychological functioning can be caused by Li administration has important clinical implications. Therefore, it is important to determine whether there is consistent evidence regarding impaired cognitive functioning in the literature and if any conclusions may be drawn.

This paper reviews studies between 1968 and 2000 that investigated the neuropsychological performance of psychiatric patients, as well as normal volunteers, treated with Li across the areas of verbal and visual memory, word fluency, psychomotor speed, attention/concentration, visuospatial constructional ability, processing speed, and reaction time. Attention is also given to two unique topics: (1) the "reversibility" of neuropsychological impairment following Li discontinuation; and (2) whether Li administration has negative cumulative effects on neuropsychological performance. If it was not possible to calculate the power of the study or if the calculated power was less than 0.20, the study was not included in the review.

This review is divided into three topic areas: (1) the neuropsychological performance of psychiatric patients on and off Li; (2) the neuropsychological effects of Li treatment of psychiatric patients compared with normals; and (3) the neuropsychological effects of Li in normals. The literature review concludes with a discussion of common methodological and statistical issues and a summary highlighting clinical implications of research findings and needed future research in this area.

Psychiatric patients on and off lithium

The first topic of review includes medication-free patients compared with patients receiving treatment or patients in a cross-over design where they act as their own controls. This body of literature has been divided into studies that found significant differences in neuropsychological functioning, and those studies that did not find significant differences. An overview of this section is presented in Table 1.
Table 1.

Psychiatric patients on and off lithium. √ impaired performance, X performance not impaired, n sample size, Dx patient diagnosis, IR immediate recall, DR delayed recall, PS processing speed, PM psychomotor speed, A/C attention/concentration, VSCA visual-spatial constructional ability, WF word fluency, CE cumulative effects

 

Study characteristics

Domain investigated

Author/s

n

Power

Dx

Duration Li treatment

Verbal memory

Visual memory

PS

PM

A/C

VSCA

WF

CE

     

IM

DR

IM

DR

      

Christodoulou et al. 1981

15

0.26

14 BMD 1 depression

Prophylactic with Li discontinuation

     

X

Kocsis et al. 1993

46

0.38

Mixed psychiatric

Prophylactic & placebo

   

  

 

Reus et al. 1979

24

0.22

BMD

Prophylactic

         

Sharma and Singh 1988

30

0.48

Primary affective disorder

Prophylactic with Li discontinuation

X

X

X

X

  

X

X

  

Shaw et al. 1986

22

0.41

21 BMD 1 depression

Prophylactic with placebo cross-over

        

 

Shaw et al. 1987

22

0.41

21 BMD 1 depression

Prophylactic with placebo cross-over

   

    

Squire et al. 1980

16

0.31

Mixed psychiatric

2 weeks

X

X

X

 

 

X

X

  

The effects of long-term Li treatment on immediate, short- and long-term memory was examined in a group of 24 BMD patients by Reus et al. (1979). This was the first report of comparison data for a control group of BMD patients not receiving medication. The patient pool was separated into a Li group consisting of 17 participants and a non-Li group consisting of seven participants and subsequently administered the Buschke Selective-Reminding Protocol in one testing session. Participants were not randomly assigned to treatment groups as patients in the non-Li group had discontinued Li treatment for medical or personal reasons. All volunteers were considered euthymic as determined by self-reports. The Li treatment group recalled significantly fewer words in each learning trial, and their word retrieval was less consistent following a long delay. The differences in performance between the two groups could have been explained on the basis of the non-Li group being healthier. However, the authors refuted this possible confound by stating that the groups were similar in past illness history and across demographic and descriptive variables.

The effect of Li on memory and other cognitive functions was explored by Squire et al. (1980) in 16 psychiatric patients (13 persons diagnosed with alcoholism, two diagnosed with BMD, and one diagnosed with schizoaffective disorder). The participants took part in this double-blind crossover study for 4 weeks beginning with either a 2-week trial of Li or a 2-week trial of a placebo. At the end of each treatment phase, participants were administered a battery of memory and other cognitive tests. The results indicated significant Li-related slowing of performance on the digit symbol subtest from the Wechsler adult intelligence scale and the Minnesota clerical test, but no evidence of Li-related deficits on the visual and verbal memory tests. The failure to find verbal performance deficits is in contrast to Reus et al. (1979), but may be the result of Squire et al. using different diagnostic groups, methodology, and tests.

Christodoulou et al. (1981) also examined Li discontinuation in their study of 14 BMD patients and one patient with recurrent depression. Patients were administered a battery of memory tests the day before Li withdrawal, and 16 days after Li withdrawal. Improved memory performance was noted after Li discontinuation on the immediate and delayed recall scores on the Rey 15-item test and the Benton visual reproduction test. Differences in degree of impairment were not related to the duration of exposure to Li. This suggests that Li does not have a cumulative deleterious effect on memory and, if there were memory impairments as a result of Li, they dissipate following Li discontinuation. These results are in contrast to the discontinuation study conducted by Squire et al. (1980) who did not find impaired memory performance. However, alternative forms for the administered tests were not used in the Christodoulou et al. (1981) study and this may account for the improved performance at the second testing session (Li discontinuation).

To determine the effect of Li on the word fluency, memory, and psychomotor speed, Shaw et al. (1986, 1987) assessed 20 patients diagnosed with BMD, primarily manic, one diagnosed with recurrent depressive disorder, and one diagnosed BMD mixed type, at weekly intervals for 5 weeks. During week 1, participants were maintained on their usual dose of Li and assessed at the conclusion of the week. During weeks 2 and 3, participants were switched to placebo administration and were assessed (once/week) at the conclusion of each week. For week 4 and week 5, participants were switched back to their original Li dosage and again tested (twice), at the conclusion of each week. Li plasma levels were monitored at weekly intervals. The authors administered a battery of tests assessing oral word fluency, attention, concentration, memory, and psychomotor speed. Several mood state questionnaires were administered to ensure patients were euthymic. The results indicated that Li-treated patients had a significant decrease in psychomotor speed, memory, and word fluency when compared with performance during the placebo condition. These findings were consistent with the results of Christodoulou et al. (1981), and suggest that performance on motor speed and memory tasks improved when Li was discontinued and declined when Li was reintroduced.

The effect of Li discontinuation and resumption on measures of cognition, creativity, and fine motor performance was examined by Kocsis et al. (1993) with 36 patients diagnosed as BMD, nine as major depression, recurrent, and one as bipolar mixed. Three assessments were completed: one at baseline while patients continued their normal Li dosage, the second at the end of the second week on the placebo, and the third after the patients resumed their normal dosage level of Li for 2 weeks. Several questionnaires assessing mood status were administered before the beginning of each testing session. Li serum levels were measured weekly and thyroid levels were determined at the beginning of the study. The patients were administered a battery of tests assessing psychomotor speed, memory and word association skills. Alternative forms of the administered tests were not used; however, Kocsis et al. stated that the administered tests have demonstrated robustness to guard against practice effects. IQ and education of the participants was not provided. The results indicated that immediate and delayed verbal memory, motor speed, and number of word associations significantly improved following Li discontinuation, and declined after the resumption of Li treatment.

Sharma and Singh (1988) examined the long-term effects of Li on cognitive functions, the discrepancy between objective and subjective evaluation of cognitive functions, and the effect of withdrawal of Li on cognitive functions in 30 affective disorder patients receiving prophylactic Li therapy. Patients were administered a battery of tests assessing memory, attention/concentration, and visual-spatial constructional skills while on their normal dose of Li. Thereafter, Li was withdrawn and the participants were kept on a placebo for 1 week then re-evaluated. No information was reported regarding when and how often Li serum levels were checked. Thirty controls matched with regard to age, education, occupation, domicile, and IQ were selected from the hospital staff and relatives. The controls were evaluated on the same tests and times as were the Li patients. No Li-associated decrease in performance was found for both the patient group in comparison with the control group. However, a 1-week period off lithium in prophylactically treated patients may fail to achieve a drug-free brain state. One study was not included in this section because the power analysis revealed an effect size below 0.20 (Smigan and Perris 1983) and two studies were not included because a power analysis could not be completed (Telford and Worrall 1978; Marusarz et al. 1981).

Psychiatric patients compared with normals

This section of the literature review critically examines studies that investigated the neuropsychological effects of Li treatment on patient volunteers in comparison with normal controls. An overview of this section is presented in Table 2.
Table 2.

Psychiatric patients relative to normals. impaired performance, X performance not impaired, N sample size, Dx patient diagnosis, IR immediate recall, DR delayed recall, PS processing speed, PM psychomotor speed, A/C attention/concentration, VSCA visual-spatial constructional ability, WF word fluency, RT reaction time, CE cumulative effects

 

Study characteristics

Domain investigated

Author/s

n

Power

Dx

Duration Li treatment

Verbal memory

Visual memory

PS

PM

A/C

VSCA

WF

RT

CE

     

IM

DR

IM

DR

       

Jauhar et al. 1993

43

0.36

20 BMD or Depression

Prophylactic

X

    

   

 

Kusumo and Vaughan 1977

26

0.23

11 BMD 2 Depression

Prophylactic

X

         

Loo et al. 1981

42

0.35

21 BMD

3 years

X

 

     

X

Lund et al. 1982

50

0.41

50 BMD

Prophylactic

    

X

   

X

Van Gorp et al. 1998

47

0.39

13 BMD & 12 BMD with ETOH abuse

Prophylactic

X

   

X

X

X

  

In an early study that explored the effect of Li on cognition in 13 BMD patients on the Halstead Impairment Index was completed by Friedman et al. (1977). The Halstead-Reitan neuropsychological battery was administered to all patients and the Halstead Impairment Index was calculated based on test scores. Friedman et al. found that five of the 13 BMD patients had elevated Impairment Indexes. Of the eight participants between the ages of 18 years and 53 years, only one had an elevated impairment index, but four of the five patients aged 59–68 years had elevated impairment indices. No significant differences between the two age groups appeared when daily Li doses, serum Li levels, duration of Li therapy, education level, or WAIS full-scale IQ were compared. However, the older patients had suffered from BMD for an average of 20 years longer then the younger patients. It is difficult to determine whether the older patients' elevated Halstead Indices were secondary to Li treatment or the effects of aging. Reitan and Wolfson (1993) have indicated that in normal subjects, at least, there is a progressive trend toward more impairment on the impairment index with advancing age.

Kusumo and Vaughan (1977) investigated the short- and long-term verbal memory performance of 13 affective disorder patients on prophylactic Li treatment in comparison with 13 medication-free normal controls. Evidence of impaired performance for the Li treatment group in comparison with the controls was found for the short-term memory task (immediate recall). However, when the performance of the Li treatment group was compared with the volunteer group on the delayed memory task, no impairment in performance was found.

In a 3-year longitudinal study examining the long-term effects of Li, Loo et al. (1981) investigated the cognitive functioning of 21 non-hospitalized BMD patients treated with Li. A control group of 21 participants matched for age, sex, and educational level was compared with the BMD group. The control group included outpatients with "relatively minor psychiatric difficulties." Loo et al. failed to clearly define the inclusion criteria for their control group. These criteria would have been useful in further elucidating the degree of psychopathology of subjects included in this group. The participants were administered a battery of cognitive tests after 1, 2, and 3 years of prophylactic Li treatment. In order to reduce the effect of learning on the results, Loo et al. used a parallel test battery for the second administration and the original test battery for the third administration. The Li patients performed significantly poorer than the control group on the processing speed and verbal memory. In contrast, no significant differences were found for the experimental group across testing intervals, suggesting that the performance of the Li-treated participants was not influenced by the time patients were on Li.

Lund et al. (1982) completed another study examining the long-term effects of Li prophylactic treatment in their investigation of 50 BMD patients. Two-thirds of the patients had ECT treatments within 2 years before beginning the study. The participants were administered five cognitive tests in one session. Test results were generally in the average range in comparison with normative data for the tests applied. When inter-test performance was analyzed, lower scores on memory and perceptual tests were indicated. There were also no significant differences between patients treated with prophylactic Li for 9–15 years and patients treated for 2–9 years. These findings should be interpreted cautiously because the authors only used simple descriptive statistics (SD, mean, and range) and qualitative comparisons when they interpreted the tests.

Jauhar et al. (1993) investigated the psychomotor performance of 20 patients with affective disorders who were on maintenance Li therapy in comparison with 23 normal controls. To confirm euthymic mood at testing, all participants were administered the Hamilton Rating Scale for Depression. To assess psychomotor performance, all participants were tested on the critical flicker fusion test, choice reaction time test, speed of recognition of a word list, and a driving simulator in one testing session. The Li group performed significantly slower on the choice reaction time test and their response time and the number of mistakes they made were significantly higher on the driving simulator test. The authors concluded that impairment in psychomotor functioning and primary reaction times in patients administered prophylactic Li should be of concern and not underestimated. These findings are also consistent with other studies that found impairment in psychomotor performance following Li treatment (Shaw et al. 1987; Kocsis et al. 1993).

A recent study by van Gorp et al. (1998) examined cognitive impairment in 13 BMD patients without alcohol dependence, 12 BMD patients with history of alcohol dependence, and 22 normal controls. Patients were administered several mood questionnaires to confirm they were euthymic at the time of testing. Participants were administered a battery of neuropsychological tests assessing the domains of verbal memory, nonverbal memory, executive functioning, visuospatial, and psychomotor skill. The test results revealed that the BMD groups, with and without alcohol dependence, performed significantly worse than controls in the verbal memory domain. In addition, the BMD patients with alcohol dependence performed significantly worse than the other two groups on tests in the executive functioning domain. However, due to the lack of randomization to treatment groups and the lack of a psychiatric group as a control, it is difficult to determine whether the identified memory impairments were due to bipolar disorder or to the effects of Li. Two studies were eliminated from this section due to their low power (Demers and Heninger 1971; Engelsmann et al. 1988) and one study was eliminated because a power analysis could not be completed (Rapp and Thomas 1979).

Li-treated normals

The systematic investigation of Li-treated healthy normal participants has attempted to isolate the neuropsychological effects of Li from those of bipolar disease process. An overview of this section is presented in Table 3.
Table 3.

Lithium-treated normals. impaired performance, X performance not impaired, N sample size, IR immediate recall, DR delayed recall, PS processing speed, PM psychomotor speed, A/C attention/concentration, VSCA visual-spatial constructional ability, WF word fluency, RT reaction time

 

Study characteristics

Domain investigated

Author/s

n

Power

Duration Li treatment

Verbal memory

PS

PM

A/C

VSCA

WF

RT

    

IM

DR

      

Calil et al. 1990

17

0.29

2 months

X

 

X

 

X

  

X

Judd 1979

42

0.52

14 days

  

 

X

X

X

Kropf and Muller-Oerlinghausen 1979

24

0.21

2 h & 14 days

      

Weingartner et al. 1985

10

0.20

14 days

       

In one of the most often cited articles in this body of literature, Judd (1979) summarized a series of controlled studies that investigated the effects of Li on mood and cognitive functioning in normal volunteers (Judd et. al. 1977a, 1977b, 1977c). A total of 42 normal male participants were included in the series of studies, in which a double-blind, randomized, split-half crossover design was employed. Li and an inactive Li placebo were administered to participants for 14 days each. There were three identical testing sessions: a baseline and two experimental sessions, each at the end of the 14-day placebo or Li maintenance period. The patients were administered a battery of tests that assessed affect/mood, cognition, sensory-motor functions, and creativity. Analyses of the cognitive tests found a significant learning effect on the digit symbol and block design subtests of the WAIS, and trails A and B; therefore, a learning effect correction factor was calculated and applied to comparisons between the placebo and Li conditions. Using this correction factor, there were Li-related performance deficits on several tasks assessing processing speed and visual-motor functions (Otis test, trails A, digit symbol, speed of closure, and Minnesota clerical test). Judd (1979) postulated that these results were likely related to a Li-associated slowing of performance.

In another double-blind study investigating the effects of Li 2-h after administration and after a 14-day trial, Kropf and Muller-Oerlinghausen (1979) assessed 24 paid healthy university students, across four specific variables: memory, learning, emotions, and mood. The volunteers were separated into a placebo group and a Li group; however, there was no mention of the volunteers being randomly assigned to the treatment groups. The participants were tested on the four variables at two intervals: 2-h after Li or placebo administration and 14 days after Li or placebo treatment. Participants were administered several learning and memory tasks, as well as a personality and mood questionnaire. No significant differences were found between the placebo group and the Li group after the initial assessment (single Li dose) on the learning and memory tasks. At the 2-week assessment, there was a significant difference in free recall memory, with the Li group remembering fewer words than the placebo group. As to learning, the Li group needed significantly more learning trials to retain the 16-word list (slower learning slope) than the placebo group. The results of the Kropf and Muller-Oerlinghausen (1979) study are consistent with one similar study that examined the cognitive effects of Li in healthy male volunteers (Weingartner et al. 1985). This study found a decrease in the accuracy and clarity of remembered words on a word presentation task (Weingartner et al. 1985).

Until 1990, the examination of the effects of Li in healthy volunteers beyond 14 days remained uninvestigated. In order to determine whether the effects of Li persist, decrease, or worsen subsequent to 2 weeks of treatment, a 2-month Li–placebo double-blind crossover study was carried out with 17 healthy university students by Calil et al. (1990). The participants were randomly assigned to begin in either a 1-month Li-treatment group or a 1-month placebo group and subsequently given the other treatment (either Li or placebo) for 1 month. At the beginning of the study and at the end of each treatment period, volunteers were administered a battery of tests assessing attention, verbal memory, processing speed, and reaction time. Participants also self-rated their mood daily using the visual analogue mood scale and three times using the profile of mood states (before and at the end of each treatment period). Blood serum levels were taken before administration of the tests to ensure both therapeutic levels of Li and normal thyroid levels.

When the performances of the participants on the administered tests were compared at baseline, at 1 month of Li treatment, and at 1 month of placebo, the participants' performances significantly improved on all measures except reaction time. These results are in sharp contrast to previously reviewed studies, but should be regarded with caution because the improvement in performance may indicate a learning effect on these tasks as no alternative forms of the administered measures were used and no correction for learning was employed. Three studies were eliminated from this section due to their low power (Karniol et al. 1978; Glue 1987; Kolk et al. 1993).

Methodological and statistical problems and suggestions

Methodological and statistical problems abound in the literature investigating the neuropsychological effects of Li. These problems may partially account for some of the disparate findings in the literature. Moreover, comparisons between studies are an onerous task because of the heterogeneous patient groups, diverse research designs, and varied diagnostic methods.

One issue that is inherent in the assessment of patients with psychiatric disorders is the need for reliable and accurate psychiatric diagnoses. However, the criteria that were used to classify patients into diagnostic groups was often not stated. Information regarding the time since the patient was first diagnosed with a psychiatric disorder was also commonly omitted. This additional information would provide data necessary to assess whether the disease process had a cumulative negative effect and would enable use of this information as a covariate. Several studies also did not assess the presence of psychopathology (e.g., depression, mania, etc.) at the time of testing. Moreover, several investigators solely used self-reports to determine current levels of psychopathology. As these measures are often only face valid, a multi-method approach incorporating a self-report and a psychiatric interview would seem advisable.

The exclusion of participants with other illnesses that can have neuropsychological consequences (e.g., hypothyroidism, diabetes, HIV infection, etc.) was commonly not addressed. Many studies also omitted valuable demographic and descriptive information of the participants they were testing (e.g., ethnicity, educational background, or IQ). These are important characteristics, not only to ensure equivalency between treatment groups, but also to enable replicability of the study. In contrast, the age of the participants was provided in most studies, but the effect of age on the neuropsychological tests often remained unexplored.

Several studies also did not report the Li serum levels of participants at the time they were tested to ensure treatment compliance and maintenance of a therapeutic dosage. Li often has a deleterious effect on the thyroid gland (Salate and Klein 1987); however, investigators rarely examined thyroid functioning. Assessing thyroid functioning in patient's receiving Li is crucial to identify participants who are hypothyroidic (Salate and Klein 1987). The addition of thyroid hormone has been reported to enhance cognitive functioning (Tremont and Stern 1997). In addition, the amount of time that participants had been treated with Li was also often omitted. Although the literature does not support the cumulative negative effect of Li in psychiatric patients (Christodoulou et al. 1981; Loo et al. 1981; Lund et al. 1982; Engelsmann et al. 1988), it would still be important to gather this data because of past methodological and statistical issues. It was also common that the patients' history of previous or recent ECT was omitted. While known to be beneficial in the treatment of major depression, it may not be without negative cognitive side effects (Chamberline and Tsai 1998; Sackeim et al. 2000).

Another problem was the lack of consistency in the neuropsychological tests administered. Some investigators administered standardized measures, while others developed their own measures. Reliability or validity coefficients of the chosen measure were rarely provided. The use of common standardized tests would facilitate comparisons between studies. It was also a common procedure to use a repeated-measures design; however, alternative or parallel forms of tests were commonly not used. In some cases, a learning coefficient was calculated to compensate for the lack of parallel forms. In others, the conclusions that could be drawn were suspect due to the possibility of practice effects. Also common in studies that used repeated-measures designs was the lack of baseline data, small sample size, and the associated loss of power and decrease in generalizability of findings.

Being able to determine the degree of impairment, Li patients experience with respect to standard deviations of impairment on standardized tests, as well as the effect size of each study would also have been useful in determining the relative magnitude of Li's impact on cognition. Although the effect sizes were not readily available or easily determined in the reviewed studies, the writers attempted to complete a power analysis on each study using the GPOWER statistical computer program (Erdfelder and Buchner 1996). Using the guidelines provided by Cohen (1992), an effect size was classified as small (0.20), medium (0.50), and large (0.80). Even assuming a liberal effect size (i.e., medium), scores from the power analyses were far below optimal (please see Table 1, Table 2, and Table 3). Power analyses were not completed for three studies because insufficient data was provided (i.e., experimental design, variable group sizes, etc.).

Discussion

The cognitive functioning of Li-treated psychiatric patients, as well as that of normal healthy volunteers was reviewed. Studies examining the subjective complaints associated with Li treatment have reported "mental slowing" (Joffe et al. 1988; Lenzer et al. 1989). Although the potential cognitive effects of Li are subtle, this subtlety and the lack of sensitivity of the administered cognitive tests may account, at least partially, for many of the equivocal findings in the areas of processing speed and word fluency. Despite the methodological flaws, difficulty with replication, and subtle effect of Li, five consistent findings emerged.

First, the studies of Kocsis et al. (1993) and Shaw et al. (1987) with psychiatric patients on and off Li found impairment on tasks assessing psychomotor speed. A third study by Jauhar et al. (1993) also found impaired psychomotor speed when psychiatric patients were compared with normals. However, the relationship between the effect of the psychiatric illness (e.g., disease process) versus lithium utilization was unclear. Second, the majority of studies assessing immediate and delayed verbal memory found impaired performance in Li-treated patients and normal controls. Ten of the 14 studies that investigated immediate verbal memory found impairment, and 7 of the 10 studies that evaluated delayed verbal memory found impairment. While there was an overall trend toward impaired verbal memory in the reviewed studies, the trend was relatively weak and additional investigation and replication is warranted to further elucidate this relationship. Third, psychiatric patients and normal volunteers treated with prophylactic Li were consistently not impaired on tasks of visuo-spatial constructional ability. Fourth, psychiatric patients and normal volunteers treated with prophylactic Li were consistently not impaired on tasks of attention/concentration. Lastly, studies examining the prophylactic effect of Li in psychiatric patients did not find a negative cumulative effect on neuropsychological functioning.

Clinical implications and recommendations

Li often remains one of the first choices for treating patients with BMD and it has been identified as an effective prophylactic mood stabilizing agent in the treatment of BMD and unipolar depression (Souza and Goodwin 1991; Bowden et al. 1994). Given that prophylactic Li patients spend an estimated 93% of their time in an inter-episodic period and that cognitive dysfunction is a significant risk factor for noncompliance, patients remain at significant risk for relapse (McCreadie et al. 1985; Gitlin et al. 1989; Muller-Oerlinghausen 2000). Subjective complaints of cognitive deficits in Li patients are common and this review supports the conclusion that they should not be ignored.

Anticonvulsant medications such as carbamezapine (CBZ) and valproate (VPA) may be considered possible alternatives to Li. A study that investigated switching BMD patients from Li to VPA found a positive subjective decrease in cognitive complaints (Stoll et al. 1996). Andrews et al. (1990) also investigated cognitive functioning in affective disorder patients administered either Li, CBZ, or a combination or CBZ and Li. They found that patients treated with CBZ performed significantly better than the other two groups on a cued recall memory task. To date, there remains a paucity of research comparing the cognitive performance of psychiatric patients taking Li versus alternative medications, such as CBZ and VPA. However, the risk/benefit ratio should also be considered when determining which medication to prescribe (e.g., the side effects of Li may be tolerable if treatment efficacy is superior). Although beyond the scope of this paper, the comparative efficacy of commonly used medications to treat BMD have been reviewed by Bowden et al. (1994) and Emilien et al. (1996). The negative cognitive side effects of anticonvulsant medications in epileptic patients has also been subject to several reviews (Vermeulen and Alderkamp 1995; Goldberg and Burdick 2001).

Although speculative, Honig et al. (1999) recommended the use of a long-acting preparation to minimize potential variations in peak serum levels. In theory, Li treatment attempts to achieve the same blood levels with slow-release and regular preparations. Moreover, a reduction in dose may be associated with an improvement in cognitive functioning (Squire et al. 1980; Kocsis et al. 1993). Polypharmacy is also fairly common in BMD. Li patients with complaints of impaired memory might also be referred for a neuropsychological evaluation to document cognitive status pre- and post- any change in medication regimen.

Honig et al. (1999) provided areas of focus and possible tests to administer when performing cognitive assessments/screens with patients administered Li. They also astutely pointed out the shortcomings of commonly administered brief cognitive screens, such as the mini-mental state examination in the detection of Li-associated subtle effects. Additional areas of focus should include reaction time tasks, visual memory tasks, and executive functioning. As psychomotor speed was consistently impaired in the studies reviewed, a referral for an on-road driving evaluation should also be considered (Honig et al. 1999). Although psychomotor slowing may be a small price relative to mood stability and beneficial prophylactic effects, informing patients of the possible deleterious effects Li on psychomotor speed and verbal memory will be important.

Given the consistencies and trends in the literature, cognitive impairments found following neuropsychological evaluation of BMD patients may be related to Li, and not to other pathological etiologies. Therefore, careful interpretation of neuropsychological test results is warranted. Patients and staff should be provided psychoeducation about the possibility of cognitive deficits associated with Li administration and their implications on day-to-day functioning.

Future research

Past research investigating Li in psychiatric patients and normals has suffered from many methodological and statistical flaws. In light of the aforementioned problems, future research in this domain is essential to aid in clarifying past research findings, as well as to expand and explore the issues and trends suggested in this review. Future research needs to use standardized diagnostic criteria, neuropsychological tests with demonstrated reliability and validity, and a multi-method approach to help rule-out acute psychopathology. Moreover, there are no studies investigating the consequences of Li-induced cognitive impairment on role performance and social functioning (e.g., employment status, activities of daily living).

There also remains a paucity of studies investigating alternative medications used to treat BMD patients. Studies by Stoll et al. (1996) and Andrews et al. (1990) provide some support for switching to valproate or carbamazepine from Li. However, this area of research remains in its infancy. Further comparisons between carbamezapine, divalproex, and Li in BMD patients or normal volunteers would be valuable to aid in differentiating the deleterious neuropsychological effects of these commonly used medications.

In contrast to the view that prophylactic Li treatment is associated with negative cognitive sequelae, there is a new and potentially revolutionary area of research investigating the neuroprotective effects of Li. The neuroprotective potential of chronic Li administration has been elucidated in several recent studies (Chen et al. 1999; Manji et al. 2000; Moore et al. 2000). Robust increases in the neuroprotective protein bcl-2 in the central nervous system was found in rodent and human neuronal cells in culture and in the hippocampus and striatum in vivo (Chen and Chuang 1999; Manji et al. 1999). Therapeutic doses of Li were also found to increase N-acetyl-aspatate (NAA; a putative marker of neuronal viability and function) concentration in humans (Moore et al. 2000). Both bcl-2 and NAA have been postulated to have beneficial effects mediated by neurotrophic/neuroprotective events. Moreover, chronic Li has been demonstrated to exert dramatic protective effects against middle cerebral artery occlusion, reducing infarct size, as well as neurological effects (Nonaka and Chuang 1998). For excellent reviews of this topic see Manji et al. 1999, 2000.

The continued investigation of the neuroprotective potential of Li treatment in mood disorder patients is of clear importance given that the long-term clinical course of many BMD patients is often associated with a progressive decline in overall functioning (Goldberg and Harrow 1999). Moreover, structural brain changes, such as reductions in regional central nervous system volume and cell numbers, a decrease in nonpyramidal neurons, and decreased cortical/laminar thickness have also been observed in BMD patients (Drevets et al. 1997; Rajkowska 1997; Benes et al. 1998). Some support for the neuroprotective effect of Li was found in this review; no cumulative negative effect or cognitive deterioration was found in mood disorder patients administered chronic Li treatment. As only three articles investigating the cumulative effects of Li treatment on cognitive functioning were identified, further study is warranted. Moreover, the relationship between Li as a neuroprotective agent and cognitive status remains to be investigated in a clinical sample (e.g., whether Li acts as a neuroprotective agent, why are some patients impaired on measures assessing psychomotor speed and verbal memory? Why do many patients subjectively report "mental slowing"?). Lastly, it will be important to further describe the relationships between chronic Li treatment, affective episode-induced cell loss or atrophy, cognitive functioning, and the clinical course of affective disorders.

Conclusion

Methodological problems in this area of research are numerous and likely account for some of the disparate findings in the literature and their lack of reliability. However, several consistencies emerged from this review. It would seem clear from the aggregate data reviewed that Li administration is associated with mild impairment in psychomotor speed. In addition, the majority of studies investigating verbal memory also found a mild decline in functioning associated with Li treatment. At the least, these data suggest that the subjective cognitive complaints should not be ignored, but should be investigated through a referral to a neuropsychologist to document pre-post medication changes or a possible switch to an alternative medication. From this review, it also appears that Li does not have a negative effect on visual-spatial constructional skills and attention/concentration, and does not have a negative cumulative effect on cognition.

The dissemination of information regarding Li's subjective and objective cognitive side effects to patients and relevant staff will be essential to provide adequate patient care. Neuropsychologists will need to be aware of the implications Li can have on their test data to avert misinterpreting findings and results. Further study in this area, especially between Li and other mood stabilizers, is needed to provide additional clarification in areas of consensus, as well as in areas of incongruity. Moreover, future research examining Li as a neuroprotective agent will be crucial in further describing the implications of long-term Li treatment with mood disorder patients and possibly even patients with neurodegenerative disorders.

Acknowledgements

I wish to thank Dr. Shawn Gray and Dr. Peter Wass for their attentive reading of an earlier version of this article.

Copyright information

© Springer-Verlag 2003