Juntendo Medical Journal

Background: Influence of dopamine (DA) on working memory (WM) performance follows an inverted U-shaped function, with excessive levels of neurotransmitters impairing WM performance. Antipsychotics have an ameliorating effect on this impairment according to studies on monkeys and rats. DA release increases during illness exacerbation in schizophrenia. Therefore, WM performance may improve by increased DA release in patients with schizophrenia being treated with high-dose antipsychotics. If improvement in WM performance was an indicator for increased DA release, it could serve as a predictor for relapse. Method: We included 130 patients with schizophrenia taking≥1,000 mg chlorpromazine equivalent/day. The antipsychotic dose was reduced to ≤1,000 mg/day. The Digit Span, Vocabulary, and Block Design tests were used before the dose reduction and in comparing relapse and non-relapse cases. Results: 51 subjects relapsed. In comparison with the no relapse group, the relapse group scored significantly higher in the Digit Span test but significantly lower in the Vocabulary and Block Design tests. The logistic regression analysis indicated a correlation between relapse and cognitive function. Conclusion: Our study revealed a correlation between relapse and cognitive function. Our findings are consistent with the idea that relapse and changes in cognitive function are caused by increased DA release.


Introduction
The efficacy and safety of antipsychotics have been verified through numerous randomized controlled trials. These efficacy and safety outcomes are incorporated into all guidelines, and the current recommended dose of antipsychotics is approximately 200-800 mg/day with chlorpromazine equivalence. In a study that used raclopride, the risk of hyperprolactinemia was found to significantly increase when the D2 receptor occupancy in the striatum exceeded 72%, and extrapyramidal symptoms (EPS) were found to significantly increase when the occupancy exceeded 78% 1) . Furthermore, high-dose and combination antipsychotic therapy is associated with increased health costs, and thus these drugs should be used as monotherapy at the recommended dose by guidelines. Nonetheless, dose reduction may worsen symptoms or be a trigger of relapse in patients receiving high dose antipsychotic polypharmacy. High-dose antipsychotic therapy is associated with the severity of positive symptoms and multiple hospital admissions 2) . Worsening of symptoms as a result of antipsychotic dose reduction is accompanied by risks, such as an increased need for hospitalization and treatment. Slowing down the rate of dose reduction is the most effective means of lowering high-dose therapy 3) .
Acute stress studies in rats have shown that psychological or physiological stress increases DA release not only in the nucleus accumbens (NAc) but also in the prefrontal cortex (PFC) and even mild stress increases DA release in the PFC while having less effects on DA release in striatum 4)- 8) . A PET study reported that psychosocial stress also increased DA release in PFC of humans 9) .
Both too much and too little DA impairs performance in cognitive function. Hence, a stressinduced increase in DA release can result in excessive DA levels and cognitive function impairment. Studies on rhesus monkeys showed that the WM function (delayed response) was impaired by noise stress. However, when the monkeys were pretreated with antipsychotics, WM function could be spared from impairment or even improved 10) 11) . These findings suggest that WM impairment caused by increased DA release is prevented by the antagonism of D1 receptors using antipsychotics.
Cognitive function is also influenced by genetic and environmental factors, including sex, age, education, and ethnicity, as well as by stress and drugs 11)-14) . Performance in cognitive function requires not only a balanced stimulation of the DA receptor subtypes but also a balanced stimulation of the alpha-1 and alpha-2 receptors by norepinephrine (NE). Similar to DA, NE release is increased in PFC by acute stress 7) . Stimulation of the alpha-1 receptor by high levels of NE reduces PFC cell firing, resulting in impaired PFC function 15) . Similarly, excess or insufficient stimulation of the alpha-2A receptor reduces PFC cell firing 13) 16) . Reduced PFC function impairs cognitive functions such as WM, flexible thinking, and attention 13) . The clinical antipsychotic agents used for schizophrenia treatment have different affinities for the alpha-1 and alpha-2 receptors. It would be difficult to simultaneously optimize the stimulation of the alpha-1 and -2 receptors when prescribing high-dose and combination antipsychotics, and high levels of NE could cause some impairment of cognitive function.
If the DA release of schizophrenic patients undergoing high-dose antipsychotic therapy were increased, it is expected that WM performance would improve. Increased DA release causes worsening of symptoms and relapse. If increased DA release could be represented by WM performance, these could be used as predictive factors upon which decisions on whether or not to reduce high-dose therapy can be made. We therefore investigated the correlation between relapse and cognitive function in patients with schizophrenia undergoing high-dose antipsychotic therapy (≥ 1,000 mg CPZ equivalents (eq.)/day). We attempted to reduce the dose of antipsychotics in a group of patients with schizophrenia undergoing high-dose therapy (> 1,000 mg CPZ eq./day) to ≤1,000 mg CPZ eq./day. The dose was gradually reduced at a rate of ≤50 mg CPZ eq./week, and the reduction was discontinued if the subjects relapsed. The differences in baseline cognitive function were analyzed between the patients with no relapse and relapse groups.

Patients and Method
Participants were recruited from Okada Hospital. The protocol was accepted by the Juntendo University review board, and all patients provided a signed informed consent. This study was conducted at Okada Hospital, Japan, from 2012 to 2014 in accordance with the Declaration of Helsinki.
Eligibility criteria were 20 to 70 years of age, inpatients with a diagnosis of schizophrenia according to the fourth edition of the Diagnostic and Statistical Manual, and those in the chronic phase receiving mean daily antipsychotic doses exceeding 1,000 mg CPZ eq./day. Subjects were excluded if they were reluctant to reduce the dose. Subjects were also excluded because of mental retardation, any evidence of past or present substance abuse or dependence, a history of major head trauma, serious medical or neurological disorders, or depot antipsychotic injections within the previous 3 months and electroconvulsive therapy within the previous 6 months. In total, 139 patients (male: 84, female: 55) were eligible and participated in this study. Participants were all Japanese patients who had been taking antipsychotics for at least 2 years but had never taken clozapine. None of the patients had taken the weak D1 blockers aripiprazole, pimozide, sulpiride, or amisulpride (unavailable in Japan) and low potency-antipsychotic quetiapine, levomepromazine, or chlorpromazine as a main drug.

Baseline assessments
Cognitive function test must be standardized for comparison both across different test populations and among a test population. Because DA activity declines with age 18) 19) , we chose the Wechsler Adult Intelligence Scale-Revised (WAIS-R) 20) , which can be corrected for age, and the age correction and standardization were straightforward. Among subtests of WAIS-R, the Digit Span test was chosen to evaluate WM. Studies of semantic dementia indicated that the Vocabulary test performance is associated with the frontal cortex and temporal lobe 21) , which suggests the involvement of a top-down process. Studies have consistently shown that the Block Design test performance is associated with the right parietal lobe 22) 23) , which indicates a strong involvement of a bottom-up process from sensory input. Hence, the Vocabulary and the Block Design tests were selected as predictive factors.
Cortisol is a major stress hormone, and the cortisol increase during stress is greater in patients with schizophrenia than in healthy individuals. For cortisol measurements, blood was drawn at 10 am after abstaining from severe mental and physical exertion, eating, drinking, and smoking for at least 1 h.
Psychiatric and extrapyramidal symptoms were assessed using the Positive and Negative Syndrome Scale (PANSS) and Drug-induced Extrapyramidal Symptoms Scale (DIEPSS).
Remission was assessed according to the criteria defined by the Remission in Schizophrenia Working Group 24) .

Reduction of antipsychotic dose, discontinuation of reduction, and definition of relapse
The reduction of antipsychotic dose was initiated promptly after each subject completed the baseline clinical and cognitive tests. Antipsychotic dosage was reduced at a rate not exceeding 50 mg CPZ eq./ week, and anticholinergic dosage was reduced at the discretion of each investigator. We determined the discontinuation of antipsychotic withdrawal based on any one of following situations: (1) cancel-lation of informed consent (e.g., if the participant indicated his/her intention to discontinue antipsychotic withdrawal), (2) biochemical or physical adverse effect (e.g., malignant syndrome or a 50% or greater increase in total DIEPSS score), (3) an increase in the total PANSS score of 25% or more from baseline, (4) deliberate self-injury, (5) emergence of clinically significant suicidal ideation, and (6) violent behavior resulting in clinically significant injury to another person or property damage. Patients who discontinued antipsychotic withdrawal for reasons (3), (4), (5), or (6) were defined as relapsed 25) . The reducing period was from the start of reducing until point at which antipsychotics became 1,000 mg or less, and no follow up period was set after that.

Statistical analysis
Continuous variables were compared using t-tests and categorical variables using the chi-square test. All tests were two tailed. An adjusted alpha level of α < 0.004 was used for demographic variables because multiple comparisons were performed. WAIS-R scores are known to be affected by the age, sex, and years of education 26) 27) , the scores were corrected for age. To control for other confounding factors, the differences in WAIS-R scores between the patients with and without relapse were analyzed using one-way analysis of covariance (ANCOVA) with years of education or sex as covariates. Statistical significance between the groups was set at p < 0.05, and Bonferroni correction was set at α < 0.008. Intragroup differences in Digit Span vs. Vocabulary and Digit Span vs. Block Design performances were analyzed using ANCOVA. Comparison of Digit Span vs. Vocabulary performance was corrected for years of education, whereas that of Digit Span vs. Block Design performance was corrected for years of education and sex. Bonferroni correction was set at α < 0.01.
Binominal logistic regression was used to determine as to which baseline five characteristics of Digit Span, Vocabulary, Block design, dose of antipsychotics and dose of anticholinergics were associated with the relapse or no relapse group after adjusting the Digit Span and Vocabulary tests for education 26) 28) and Block Design for sex 27) . Antipsychotics influence cognitive function and psychotic symptoms. Anticholinergics also influence cognitive function, which includes reduction of Digit Span and Digit Symbol substitution test performances, and are associated with psychosis 29)-31) . The dosages of antipsychotics and anticholinergics, therefore, were selected as independent variables. Correlations between DA release and cortisol under stress are inconsistent 9) 32) . Furthermore, psychological stress suppresses the hypothalamic-pituitary-adrenal axis (HPA axis) 33) . For these reasons, cortisol was not used as an independent variable in the logistic regression analysis.
Finally, forward stepwise logistic regression (P enter < 0.05; P remove > 0.10) was used to determine which factors most strongly predicted relapse in a group versus no relapse in a group. Categorical variables were modeled using dummy variables. Assumptions of linearity were evaluated for all variables using the Spearmanʼ s rank correlation coefficient.
Analyses were conducted using the Statistical Package for Social Sciences (SPSS) version 21 (SPSS Japan, Tokyo, Japan).

Results
Informed consent was obtained from 139 subjects, of whom 6 declined the assessment and 3 withdrew the agreement after reduction of dosage was initiated. Of the 130 subjects, 79 tolerated the reduction of dosage without relapse and 51 showed relapse and the reduction was discontinued. In subjects who did not tolerate the reduction, symptoms were ameliorated by increasing dosage back to original levels.
When comparing groups before antipsychotics dose reduction, the WAIS-R Digit Span score was significantly higher [F(1, 127)= 12.70, p = 0.001] in the relapse group than that in the no relapse group (ANCOVA, α < 0.008) ( Baseline PANSS positive symptoms, negative symptoms, PANSS general psychopathology, and total scores were significantly higher in the relapse group (α < 0.004) ( Table-2). There were no significant differences in the mean antipsychotic dosage, anticholinergic dosage, cortisol values, and extrapyramidal symptoms (as measured by DIEPSS) between groups (   (Table-3). Although these intellectual function tests were significant predictors, the antipsychotic and anticholinergic dosages were not (Table-3).
There was no correlation indicative of multicollinearity (Table-4), and there were no outliers outside ± 2 standard deviation (SD). In a forward stepwise logistic regression analysis using the likelihood ratio, the predictive value of the 3 tests    The sensitivity and specificity for predicting relapse were calculated for subgroups with mean score differences ≥3 and ≥4 for Digit Span vs. Vocabulary and Digit Span vs. Block Design, respectively. For subjects with a mean score difference of ≥3, the sensitivity was 82% (95% CI: 0.74-0.87) and the specificity was 94% (0.89-0.97). For subjects with a mean score difference of ≥4, the sensitivity was 61% (0.53-0.64) and the specificity was 98% (0.93-0.99).

Discussion
Laruelle et al. 34) showed that increased DA release in the striatum is associated with the worsening of positive symptoms and relapse, and they characterized relapse as a hyperdopaminergic state, i.e., a situation where presynaptic dopamine function is augmented, indicating an increase in DA synthesis and DA stores. In human PET studies, stress-induced DA release in the striatum was higher in individuals with a risk of psychosis and in patients with schizophrenia compared with healthy controls 32) 35)-37) , indicating that elevated DA synthesis in the striatum correlates with the worsening of psychotic symptoms and impaired cognitive function 38) 39) .
In this study, we demonstrated that cognitive function can serve as a predictor for reduction of high-dose antipsychotic therapy. It is based on the theory that increased dopamine by relapse improves transmission of excessively blocked dopamine D1 receptors and improves WM function. Several hypotheses have contributed to the explanation of the mechanisms, and many issues need to be investigated in the future. In particular, the issues of WM improving because of increased DA release in patients with schizophrenia undergoing high-dose therapy and those of worsening symptoms caused by a reduction in high-dose antipsychotic therapy during times of increased DA release require clarification. We built this predictive model of relapse with WM as the predictor based on the fact that some patients have exacerbated symptoms, large antipsychotic doses, and high scores only in the digit span test of the WAIS (Japanese article), as well as the fact that antipsychotics improve the stress-induced impairment of WM. Although the decrease in dopamine D1 receptor transmission in models of attention-deficit hyperactivity disorder (ADHD) is the result of low DA levels, the decrease in D1 receptor transmission in high-dose antipsychotic therapy is the result of blockade of D1 receptors by antipsychotics. In either case, WM is improved by increased DA. If DA receptors were excessively blocked by antipsychotics, a change in the WM performance induced by increased DA release would be predictable. Despite the mechanisms being unclear, we investigated a predictive model of relapse with WM as a predictor because of the clinical demand for such an indicator 40) 41) .
Further examination is needed to determine whether dopamine supersensitivity, DA receptor upregulation, and other mechanisms are associated with the cause of relapse. Since increased DA release increases levels and proportions for the high-affinity state of the DA D2 receptor for dopamine (D2 High ), the worsening of symptoms could be caused by dopamine supersensitivity 42) . Nevertheless, further study on the possibility that dopamine sensitivity increased for reasons other than increased DA release (e.g., long-term antipsychotic use) is needed.
Differences arising from the type and combination of antipsychotics, in addition to the mechanisms, also need to be examined. Investigations also need to be conducted regarding differences of relapse prediction by cognitive function when a high-dose therapy with a low-potency antipsychotic as the main drug is used.
In the WAIS-R test, up to 15% of subjects showed score differences of at least 3 in 2 tests 43) 44) . Field (1960) calculated the difference that should be considered as indicative of abnormality (requiring additional tests) at a given significance level. Score differences fell between 3.5 and 4.3 at 5%, and any score difference ≥4 was considered as a discrepancy 44) . In our study, the mean score difference in the relapse group was 4.4 for Digit Span vs. Block Design and 3.7 for Digit Span vs. Vocabulary performances. These numbers indicate significant difference at a borderline level where an abnormality is suspected. In the chronic phase of schizophrenia, the cognitive function is impaired at approximately -1.2 < SD <-0.6 compared to normal control subjects as measured by WAIS-R Full scale IQ, Digit Span, Vocabulary, and Block Design tests 45) . In the present study, cognitive function was severely impaired in the patients with no relapse group (-1.8 < SD <-1.1) and further impaired in the relapse group (-2.7 < SD <-1.8). Digit Span performance, however, was -0.6 SD in the relapse group, which was roughly the same as the mean value reported in a meta-analysis of patients in the chronic phase of schizophrenia. The compromised intellectual function in the patients with no relapse group could be an adverse effect caused by highdose antipsychotic therapy.
In this study, we focused on changes in the cognitive function caused by stress and antipsychotics. Stress is generally defined as any stimulus that disturbs homeostasis and elicits a compensatory response to maintain homeostasis 46) . The main systems in the brain associated with the stress response are the HPA axis and locus coeruleus-norepinephrine system, and the involvement of DA has also been strongly suggested. The HPA axis is controlled by afferents from the PFC, NAc, amygdala, and hippocampus, and DA is involved in this modulation 47) 48) . Enhanced stress activates the hippocampal ventral subiculum, which in turn increases DA activity in the NAc via a polysynaptic pathway 47) .
Both DA and NE levels show an inverted U-shaped association with cognitive function, and cognitive function is impaired at excessively low and high levels of these transmitters in PFC. Increased DA release in the striatum, including the NAc, is associated with the worsening of symptoms and relapse of schizophrenia. In our study, baseline cognitive function was more severely impaired and the PANSS scores were significantly higher in the relapse group. Altered cognitive function, greater psychopathology, and relapse after reduction in antipsychotic dose can all be explained by increased DA release. The logistic regression analysis also indicated a strong correlation between cognitive function and relapse, consistent with the idea that DA release was elevated at baseline in the relapse group. Additional studies will be needed to assess the correlation between NE and relapse. A study on relapse of amphetamine psychosis using plasma monoamine values found a significant increase in DA levels in the relapse group, whereas the hyperactivity of NE was not as evident. Although DA and NE were both predictors of relapse, increased DA release appears to be more important 49) .
The anticholinergic effect of anticholinergics and antipsychotics are involved in both cognitive function and relapse. Although it is conceivable that a reduced dose of anticholinergics induced withdrawal symptoms, including physiological stress, it result in reduced DA release 50) , which hardly explains the high incidence of relapse following the dose reduction.
Although DA release is an adaptive response during cognitive function tests in unstressed healthy individuals, under stress it may be elevated to the point of excess, resulting in an impaired performance. Schizophrenic patients have trouble ignoring environmental stressors and are more susceptible to them 51) . In addition, DA release is increased in the hyperdopaminergic state, which tends to result in excessive DA levels. Excessive DA levels likely impair cognitive function, which would further compromise the ability to escape from the stressor or adapt to a stressful environment. Patients with schizophrenia are also reported to have abnormalities in the brain regions associated with stress response, including the PFC, striatum, and hippocampus. These abnormalities include reduced dendritic spine number on pyramidal neurons in the dorsolateral prefrontal cortex (Dlpfc) 52) , reduced hippocampal volume, pathological changes as revealed by functional and structural imaging 47) , and elevated capacity of DA synthesis in the striatum 53) . The pathological changes in the hippocampus enhanced hippocampal activity and increase the firing of DA neurons 54) , which is associated with DA dysregulation and abnormalities in the PFC, striatum, hippocampus are likely associated with vulnerability to stress.
In predicting relapse, detecting a change in the DA release is of foremost importance. The availability of nuclear medical tests such as PET scan is limited, and the high cost precludes their frequent use in routine examination. If a simple detection method was available, the risk of relapse associated with antipsychotic dose reduction could be assessed, and patients in whom dose may be safely reduced would benefit from the lessening of side effects and would decrease risk of admission to hospital as schizophreniaʼs main cost. We conclude that antipsychotic dosage should not be reduced in cases where increased DA release is suspected based on psychopathological observations. In such cases, increased DA release should be allowed to spontaneously decrease. Impaired cognitive function and improved WM performance shown in this study serve as indicators of increased DA release in patients with schizophrenia under high-dose antipsychotic therapy and are useful criteria for determining the relapse risk of dose reduction.

Limitations
This was an open-label study. Although DA and NE are significantly involved in cognitive function, many issues remain unclear, including the specific cognitive functions they control and to what extent. Furthermore, much of the previous data on the contributions of NE and DA to cognitive functioning were obtained in animal studies. In animal studies, WM is assessed by spatial and delayed response tests, whereas the WAIS-R WM test we used was the Digit Span test. This comprises separate forward and backward tasks, the digit forward test measure simple attention and the digit backward measures the verbal WM. Although spatial WM and verbal WM are associated with distinct brain regions, the Dlpfc is involved in both.