Adjunctive sarcosine plus benzoate improved cognitive function in chronic schizophrenia patients with constant clinical symptoms: A randomised, double-blind, placebo-controlled trial.

Abstract Objectives Hypofunction of NMDA receptor is implicated in the pathophysiology, particularly cognitive impairment, of schizophrenia. Sarcosine, a glycine transporter I (GlyT-1) inhibitor, and sodium benzoate, a d-amino acid oxidase (DAAO) inhibitor, can both enhance NMDA receptor-mediated neurotransmission. We proposed simultaneously inhibiting DAAO and GlyT-1 may be more effective than inhibition of either in improving the cognitive and global functioning of schizophrenia patients. Methods This study compared add-on sarcosine (2 g/day) plus benzoate (1 g/day) vs. sarcosine (2 g/day) for the clinical symptoms, as well as the cognitive and global functioning, of chronic schizophrenia patients in a 12-week, double-blind, randomised, placebo-controlled trial. Participants were measured with the Positive and Negative Syndrome Scale and the Global Assessment of Functioning Scale every 3 weeks. Seven cognitive domains, recommended by the Measurement and Treatment Research to Improve Cognition in Schizophrenia Committee, were measured at weeks 0 and 12. Results Adjunctive sarcosine plus benzoate, but not sarcosine alone, improved the cognitive and global functioning of patients with schizophrenia, even when their clinical symptoms had not improved. Conclusions This finding suggests N-methyl-d-aspartate receptor-enhancement therapy can improve the cognitive function of patients with schizophrenia, further indicating this pro-cognitive effect can be primary without improvement in clinical symptoms.


Introduction
Schizophrenia is a severe mental disorder, influencing 1% of the global population (Schultz and Andreasen 1999). Its phenotypes include positive symptoms, negative symptoms and cognitive impairments. Cognitive dysfunction is thought to be the core manifestation of schizophrenia due to its detrimental impact throughout the lifelong illness (Kasper 2006;Carrió n et al. 2011;Green et al. 2011). There is controversy regarding the effect of current antipsychotic medications on cognitive dysfunction (Sergi et al. 2007;Meltzer 2013). Moreover, despite preclinical data, there has been limited positive clinical feedback on putative pro-cognitive drugs in humans (Millan et al. 2012;Preskorn et al. 2014).
Evidence of the role of N-methyl-D-aspartate receptor (NMDA) receptors in the cognition and pathophysiology of schizophrenia (Krystal et al. 1994;Sawa 2009;Javitt et al. 2012;Errico et al. 2013) suggests that enhancement of the NMDA receptor function may improve the cognitive impairment of schizophrenia (Krystal et al. 1994;Javitt et al. 2012;Hashimoto et al. 2013;Paoletti et al. 2013). Previous research showed that adjuvant NMDA-enhancing agents which directly or indirectly enhance the NMDA function via the NMDA-glycine site, including D-serine (Tsai et al. 1998;Heresco-Levy et al. 2005, glycine (Javitt et al. 2001;Heresco-Levy et al. 2004), glycine transporter I (GlyT-1) inhibitor (Tsai et al. 2004), and D-amino acid oxidase (DAAO) inhibitor , reveal beneficial efficacy on clinical symptoms. For example, sarcosine, increasing glycine supply to the synapse by blocking the reuptake of glycine through inhibiting GlyT-1 (McBain et al. 1989), could decrease positive and negative symptoms of some patients (Tsai et al. 2004;Lane et al. 2005). Whether sarcosine can improve cognitive function is not yet known. Roche also recently announced two Phase III studies of another GlyT-1 inhibitor, bitopertin, in adults with persistent, predominant negative symptoms, but failed to meet their primary goals (FirstWord Pharma 2014). In addition, the Cognitive and Negative Symptoms in Schizophrenia Trial (Buchanan et al. 2007) contained a relatively larger sample size along with a design relevant for interpreting negative symptoms and cognition effects of glycine and D-cycloserine. The results were negative. D-Serine may be more potent than glycine as the neurotransmitter for the coagonist site of the NMDA receptor (Bado et al. 2011;Balu and Coyle, 2014) and may produce pro-cognitive effects in rodents (Bado et al. 2011) and healthy subjects (Levin et al. 2015). However, Weiser et al. (2012) reported add-on D-serine showed no improvement in negative or cognitive symptoms of schizophrenia and the lower D-serine doses could be a possible factor. Although higher-dose D-serine may provide additional opportunities to test its pro-cognitive effects in schizophrenia, the nephrotoxicity is a concern. Another potential approach to simulate NMDA function is to prevent D-serine degradation. DAAO, a flavoenzyme of peroxisomes that exists in the central nervous system, is responsible for degrading D-serine and D-alanine (Verrall et al. 2010;Burnet et al. 2011). One of the potential agents is sodium benzoate (benzoate), an inhibitor of DAAO, which can slow the metabolism and elevate the synaptic concentration of D-amino acids such as D-serine and therefore enhance NMDA receptormediated neurotransmission (Van den Berghe-Snorek and Stankovich 1985). In an animal study, a single oral dose of sodium benzoate did not change D-serine levels in plasma or in the brain; however, sodium benzoate induced antipsychotic effects in the phencyclidine model of schizophrenia (Matsuura et al. 2015). Smallscale pilot trials showed that benzoate improves the cognitive function of patients with chronic schizophrenia ) and patients with early-phase Alzheimer's disease (Lin et al. 2014a), supporting the safety and pro-cognitive potential of this DAAO inhibitor.
While D-serine and glycine are endogenous coagonists of NMDA receptors, their availability are different (Papouin et al. 2012). The preferential affinity of D-serine is for synaptic NMDA receptors and that of glycine is for extrasynaptic NMDA receptors. Moreover, long-term potentiation relies on synaptic NMDA receptors, and long-term depression requires both the activations of synaptic and extrasynaptic receptors (Papouin et al. 2012). Whether modulating D-serine and glycine simultaneously can activate NMDA-related functions, such as cognitive function, more efficiently than one approach alone remains unclear. The feasibility of combining two NMDA-enhancement approaches by inhibiting both GlyT-1 and DAAO deserves further studies.
To date, few studies have applied comprehensive cognitive measures, such as the domains recommended by the Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Committee , to assess the procognitive effects of NMDA enhancing agents (Kantrowitz et al. 2010;D'Souza et al. 2013;Lane et al. 2013). This study compared the cognitive and clinical efficacy as well as safety of add-on sarcosine plus benzoate vs. sarcosine in patients with chronically stable schizophrenia in a 12-week, placebo-controlled trial.

Participants
Ethical approval was obtained from the Institutional Review Board. Patients with chronic schizophrenia were recruited from the inpatient units of the Department of Psychiatry, Changhua Hospital, Taiwan. All patients provided written informed consent following a complete description of the study. Patients were enrolled in this study if they: (1) were aged from 18 to 60 years, (2) were physically healthy and had laboratory assessments (including urine/blood routine, biochemical tests and electrocardiograph) within normal limits, (3) fulfilled the diagnosis of schizophrenia according to the criteria of the Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV) (American Psychiatric Association 1994), (4) remained symptomatic but without clinically significant fluctuation and the antipsychotic doses were unchanged for at least 2 months prior to this study, and (5) had a minimum baseline total score of 60 on the Positive and Negative Syndrome Scale (PANSS) (Kay et al. 1987). Exclusion criteria included DSM-IV (American Psychiatric Association 1994) diagnosis of mental retardation, substance/alcohol abuse or dependence, history of epilepsy, head trauma or central nervous system (CNS) diseases, pregnancy or breast-feeding, and an inability to follow the protocol.

Study design
All subjects had been receiving a balanced hospital diet and unchanged institutionalisation before and during the trial. After achieving optimal clinical treatment response, patients' antipsychotic doses remained constant for at least 2 months prior to enrolment into the study and during the study period. All patients were randomly assigned to receive the 12-week trial of stable antipsychotic regimens concomitant with placebo, sarcosine (2 g/day) or sarcosine (2 g/day) plus benzoate (1 g/day) in a 1:1:1 ratio (Figure 1). These doses were safely used in previous trials (Lane et al. 2005;Lane et al. 2013;Tsai et al. 2004). Study medications were given twice daily and were provided in coded containers with a supply of identical in appearance capsules of placebo or either of the active compounds. Patients, caregivers, and investigators (except for the investigational pharmacist) were all masked to the assignment. Patient's adherence and safety were closely monitored by the research psychiatrists and the nursing staff.

Measurement of clinical symptoms
Clinical assessments were measured with the PANSS and the Global Assessment of Functioning (GAF) scale Figure 1. Flow chart of the subjects throughout the study period. Seventy-five schizophrenia patients were screened. Sixty-three were eligible and randomised: 21 received placebo; 21, sarcosine; and 21, sarcosine plus benzoate. Participants were measured with PANSS and GAF every 3 weeks. Cognitive domains recommended by the MATRICS Committee, were measured at weeks 0 and 12. Side-effect assessments including the Simpson-Angus Rating Scale, the AIMS, the Barnes Akathisia Scale and UKU Side-effects Rating Scale, were conducted at weeks 0, 3, 6, 9 and 12. Routine laboratory tests were arranged at the baseline and endpoint. In total, 49 of the participants completed the trial: 16 received placebo; 16 received sarcosine; and 17 received sarcosine plus benzoate.
(American Psychiatric Association 1994), and the Clinical Global Impression-Severity scale (CGI-S; Guy 2000) at weeks 0, 3, 6, 9 and 12, and cognitive function was measured at the baseline and at the endpoint. The ratings were performed by the research psychiatrists who were experienced in the rating scales. Inter-rater reliability was analysed with the analysis of variance (ANOVA) test. Only raters reaching the intra-class correlation coefficients of 40.90 during pre-study training were allowed to rate the study patients.
Primary outcome measures were PANSS total, global composite and neurocognitive composite of seven cognitive domains (see later); secondary outcome measures were the three PANSS subscales, CGI, GAF and seven cognitive domains (see later).

Measurement of cognitive function
This study was started before a commercial Chinese version of Measurement and Treatment Research to Improve Cognition in Schizophrenia (MATRICS) Consensus Cognitive Battery (MCCB) was available.
Cognitive function in the current study was assessed using a battery of tests, which were the same tests or the analogues of tests from MCCB (Green et al. 2004), due to the lack of Chinese versions of some tests. The tests utilised in this study were reviewed by one of the developers of the MCCB, Dr Green, in our previous study ). There were seven domains: (1) speed of processing (consisting of three tests: Category Fluency, Trail Marking A and WAIS-III Digit Symbol-Coding); (2) sustained attention (Continuous Performance Test) (Sternberg 1966;Chen et al. 1998); (3) working memory, verbal (backward digit span, Silver 2003) and nonverbal (WMS-III, Spatial Span); (4) verbal learning and memory (WMS-III, word listing); (5) visual learning and memory (WMS-III, visual reproduction); (6) reasoning and problem solving (WISC-III, Maze); and (7) social cognition (the Mayer-Salovey-Caruso Emotional Intelligence Test [MSCEIT] Version 2) (Mayer et al. 2003;Green et al. 2005). The Chinese version of the MSCEIT tasks was translated and back translated from English to Mandarin Chinese with satisfactory reliability, validity (Ma et al. 2012), and applicability ). The first six domains were defined as neurocognition, the processes of linking and appraising information, while the seventh domain, social cognition, was defined as the mental operations underlying social interactions such as the perception, interpretation, and generation of responses to the intentions and behaviours of others (Mayer et al. 2003).
Cognitive performance data of 78 healthy comparison (HC) participants, matched with patients in age (P ¼ 0.11) and gender (P ¼ 0.71), were collected. Inclusion criteria for HC participants included no Axis I or Axis II psychiatric disorder, no neurological illness, no substance dependence or abuse, no family history of major psychiatric disorder, good general physical health, and age between 18 and 60 years. Participants provided written informed consent after the test procedures were fully explained. All cognitive raw scores of schizophrenia patients were standardised to t scores based on the data of the 78 HC participants. For the cognitive domain that included more than one test, the summary score for the domain was calculated by summing the t scores of the tests included in that domain and then standardising the sum to a t score (Kern et al. 2008). A global composite score (for all seven domains) and a neurocognitive composite score (for the six neurocognitive domains without social cognition) were also calculated.

Measurement of safety
Side-eGect assessments included the Simpson-Angus Rating Scale (Simpson and Angus 1970) for extrapyramidal side effects, the Abnormal Involuntary Movement Scale (AIMS) for dyskinesia (Guy 1976), the Barnes Akathisia Scale (Barnes 1989) and Udvalg for Kliniske Undersogelser (UKU) Side-effects Rating Scale (Lingjaerde et al. 1987) for systemic side effects. These assessments were conducted at weeks 0, 3, 6, 9 and 12. Routine laboratory tests, including complete blood count, biochemistry, urine routine and electrocardiogram were arranged at the baseline and endpoint.

Statistical analysis
At baseline, the demographic and clinical characteristics, including age, gender, years of education, duration of illness, classification of current antipsychotics, number of previous hospitalisations, the number of poor responses to antipsychotics, daily antipsychotic dose, severity of clinical symptoms and cognitive function were compared by Kruskal-Wallis tests for continuous variables and Pearson's 2 -tests or Fisher's exact tests for categorical variables.
The effects of the study drugs on the changes in PANSS, GAF, CGI-S and MATRIC T scores from the baseline to endpoint were assessed according to the linear mixed effects' model with treatment, visit and treatment-visit interaction as fixed effects; baseline value as the covariate (to adjust the effect of baseline severity); the intercept is the only random effect (to adjust the individual's effect). The autoregressive of order 1, named AR(1), was used as the covariance type to specify the within-patients' dependence due to repeated measurements from the same patient. In addition to the placebo group, the sarcosine group was also set as a reference group to compare the differences in clinical and cognitive change between the sarcosine group and the sarcosine plus benzoate group. All data were analysed by the SPSS 18.0 statistical package. All P values were based on two-tailed tests with a significance level of 0.05.

Patients' characteristics
To enhance drug adherence and minimise environmental factors, we enrolled inpatients in this study. A total of 75 patients were screened. Among them, 63 were eligible and randomised: 21 received placebo; 21 received sarcosine; and 21 received sarcosine plus benzoate ( Figure 1). The mean age of the 63 patients was 38.4 ± 9.3 (SD) years, the duration of illness was 13.7 ± 6.7 years, the education duration was 11.1 ± 2.7 years, the number of previous hospitalisations was 3.6 ± 5.6, the number of poor response to antipsychotics, defined as records of failure to achieve clinical improvement after 48 weeks treatment of an antipsychotic agent, was 4.1 ± 1.1, and antipsychotic dose was 13.7 ± 5.6 mg/day (olanzapine-equivalent) (Gardner et al. 2010). Patients' characteristics at the baseline were similar among the three treatment groups (Table 1).
In total, 49 of them completed the trial: 16 received placebo; 16 received sarcosine; and 17 received sarcosine plus benzoate (Figure 1). There was no significant baseline difference between the completers and the patients who dropped out (Supplementary Table 1 available online). Moreover, the characteristics of the three groups of completers were also similar (Supplementary Table 1 available online).

PANSS, GAF and CGI-S
Of all the 63 patients, their PANSS, GAF and CGI-S at the baseline were similar among the three treatment groups ( Table 2). The sarcosine plus benzoate group was better than the placebo group in improving the GAF (mean difference in score changing rate [±SE] ¼ 0.16 ± 0.06, P ¼ 0.005) after 12 weeks of treatment; however, there was no significant group difference in improvement in the PANSS and CGI-S scores ( Table 2).
Of the 49 completers, their PANSS, GAF and CGI-S at the baseline were also similar among the three treatment groups (Supplementary Table S2 available online). Likewise, the sarcosine plus benzoate group was also better than the placebo group in improving GAF after 12 weeks of treatment; however, there was no significant group difference in improvement in the PANSS and CGI-S scores (Supplementary Table S2 available online). Table 3 presents the cognitive functions among the three treatment groups over the 12-week treatment. The cognitive function of the patients who did not complete the 12-week study was not measured at the endpoint. Therefore, the improvement in cognitive function was compared among the 49 completers. In line with previous research (Corbera et al. 2013;McCleery et al. 2014), schizophrenia patients had lower baseline cognitive performance than HC participants (not shown). Similarly to previous reports on the psychometric   properties of multiple cognitive domains like the MATRICS Consensus Cognitive Battery , standardised scores from domains measured by more than one test tended to be lower than domains measured by a single test (Table 3). Moreover, the patients in this study exhibited severe levels of cognitive impairment in the neurocognitive composite and global composite scores (

Side effects
The three treatment groups had minimal extrapyramidal syndrome at the beginning of the study. The baseline score on the Simpson-Angus Rating Scale was 0.7 ± 1.8 in All cognitive raw scores of patients were standardised to t scores based the data of 78 healthy comparison participants. For cognitive domains that included more than one measure, the summary score for the domain was calculated by summing the t scores of the tests included in that domain and then standardising the sum to a t score (Kern et al. 2008). yFor assessing neurocognitive function, a composite t score including all six neurocognitive domains, excluding social cognition, was calculated by standardising the sum of the t scores. zFor assessing the global cognitive function, an overall composite t score including all seven domains was calculated by standardising the sum of the t scores. *Cognitive performance at baseline was similar among the three treatment groups by Kruskal-Wallis tests (Speed of processing P ¼ 0.17; Attention/Vigilance, P ¼ 0.69; Working memory, P ¼ 0.64; Verbal learning and memory, P ¼ 0.48; Visual learning and memory, P ¼ 0.50; Reasoning and problem solving, P ¼ 0.77; Social cognition, P ¼ 0.38; Neurocognitive composite, P ¼ 0.80; Global composite score, P ¼ 0.75. **Mixed-model repeated measure (MMRM) methods, adjusted with age, gender, education, treatment and visit times. P values were based on two-tailed tests. ***P50.05. the placebo group, 0.2 ± 0.7 in the sarcosine group and 1.0 ± 1.0 in the sarcosine plus benzoate group. The baseline AIMS was 0.2 ± 1.1 in the placebo group, 0.2 ± 0.7 in the sarcosine group and 0.1 ± 0.4 in the sarcosine plus benzoate group. The baseline Barnes Akathesia Scale was 0.5 ± 1.5 in the placebo group, 0.2 ± 0.6 in the sarcosine group and 0.1 ± 0.4 in the sarcosine plus benzoate group. There were no significant differences among the three groups in the Simpson-Angus Rating Scale (P ¼ 0.11), AIMS (P ¼ 0.79) and Barnes Akathesia score (P ¼ 0.56).
At the endpoint, the severity of extrapyramidal syndrome remained minimal and did not reveal significant differences among the three groups. The mean of the Simpson-Angus Rating Scale at the endpoint was 0.4 ± 1.3 in the placebo group, 0 ± 0 in the sarcosine group, and 0.4 ± 0.7 in the sarcosine plus benzoate group (P ¼ 0.14). The endpoint AIMS was 0 ± 0 in the placebo group, 0.1 ± 0.5 in the sarcosine group, and 0 ± 0 in the sarcosine plus benzoate group (P ¼ 0.36). The endpoint Barnes Akathesia score was 0.6 ± 1.8 in the placebo group, 0.1 ± 0.5 in the sarcosine group and 0 ± 0 in the sarcosine plus benzoate group (P ¼ 0.32).
The routine blood cell count, biochemistry, urine routine, and electrocardiogram after treatment remained unchanged (data not shown). No dropout occurred due to adverse effects.

Discussion
Cognitive impairment, a core feature throughout the course of schizophrenia, is related to functional outcome. However, it is poorly managed with currently available antipsychotics (Green et al. 2000;Mohamed et al. 2008;Allott et al. 2011). Moreover, there have been few positive responses from putative pro-cognitive drugs in humans (Millan et al. 2012;Preskorn et al. 2014). The present study suggests that adjunctive sarcosine plus benzoate, but not sarcosine alone, can improve the cognitive and global functioning of patients with chronic schizophrenia, even when their clinical symptoms cannot be improved. This finding lends support to the previous notion that NMDA-enhancement therapy can improve the neurocognition of patients with schizophrenia (Kantrowitz et al. 2010;Lane et al. 2013) and further indicates that this procognitive effect can be primary without improvement of the clinical outcome. Although it may not achieve a striking clinical difference, the effect of sarcosine plus benzoate on GAF was statistically significant after 12 weeks. It has been suggested that the global functioning of patients with schizophrenia may be related to cognitive function rather than to clinical symptoms (Carrió n et al. 2011;Green et al. 2011). This result indicates that pro-cognitive effects may reflect in the improvement of global functioning.
This clinical trial is the first to examine the combination effect of a DAAO inhibitor and a GlyT-1 inhibitor on symptomatic or cognitive domains of schizophrenia. The finding on its safety is akin to those of previous studies (Tsai et al. 2004;Lane et al. 2005Lane et al. , 2006Lane et al. , 2010Lane et al. , 2013, indicating add-on sarcosine and sarcosine plus benzoate are well tolerated. The mild side effects were likely coincidental observations because there was no significant group difference. Use of direct NMDA receptor agonists has been limited by the high doses that must be given or the relatively poor penetration of the CNS (Javitt 2008). Therefore, combining two kinds of NMDA receptor-enhancing agents is of a great interest in research (Labrie and Roder 2010). For example, D-serine in combination with a DAAO antagonist produced greater ameliorative effects than either compound applied alone in animals (Hashimoto et al. 2009). Previous pre-clinical and clinical studies supported compounds providing selective modulation of the NMDA receptor D-serine/glycine site such as diminishing D-serine catabolism by inhibition of DAAO or stimulating the glycine modulatory sites by blocking GlyT-1 could demonstrate beneficial effects on schizophrenia (Tsai and Lin 2009). Our study supports that coadministration of GlyT-1 and DAAO inhibitors could provide a potential therapeutic approach. However, there has not yet been a preclinical study examining the combination effect of a DAAO inhibitor, such as benzoate, and a GlyT-1 inhibitor, such as sarcosine in the NMDA receptor models of cognitive impairment associated with schizophrenia. Further studies are warranted.
Although there was no significant improvement in global composite score from the sarcosine add-on treatment, the sarcosine group also displayed improved reasoning and problem solving (P value of 0.04) when compared to the placebo group. Therefore, it remains possible whether a dose higher than the dose in the current study, to reach a higher level of NMDA activation, would generate a better response (Lane et al. 2008;Kantrowitz et al. 2010).
The current result shows that adjunctive sarcosine plus benzoate improved the cognitive function of patients but not their negative symptoms. The relationships between positive, negative and cognitive symptoms have been equivocal in both cross-sectional and longitudinal studies. While some studies suggest that cognitive deficits are modestly associated with negative symptoms (Berman et al. 1997;Savilla et al. 2008), no significant associations between changes in negative symptoms and cognitive impairment have been raised (Bell and Mishara 2006;Umbricht et al. 2014). Our current result supports the notion that negative symptoms and cognition should be viewed as independent targets for intervention (Nasrallah et al. 2014). As in benzoate alone treatment , the study drugs in the current study did not affect social cognition. This may reflect the multi-determined nature of the social cognitive domain (Couture et al. 2006) and may be related to the inadequate study duration. The current results encourage further study of longer treatment duration and NMDA-enhancement therapy plus social rehabilitation in patients with chronic schizophrenia.
Different from most sarcosine or benzoate trials, the current study showed that all three groups have modest improvement in clinical manifestation; however, the improvements by sarcosine or sarcosine plus benzoate treatment were similar to that of the placebo group. In comparison with two previous sarcosine trials which showed effects on clinical symptoms, including positive and negative symptoms, in chronic schizophrenia, the current subjects tended to be older [38.4 (mean value) vs. 31.8 (Tsai et al. 2004) and 30.9 years old (Lane et al. 2010)], have longer illness durations [14.7 vs. 9.6 (Tsai et al. 2004) and 9.5 years (Lane et al. 2010)], and have more hospitalisations [(3.6 vs. 2.8 (Lane et al. 2010)]. The age and illness duration of the current study subjects were closer to those in the trial of sarcosine added to clozapine, which failed to reduce positive and negative symptoms of schizophrenia [38.4 vs. 36.1 years old and 14.7 vs. 14.9 years (Lane et al. 2006)], suggesting that adjunctive 2 g/day sarcosine reveals limited efficacy in older patients and those with longer illness duration.
A recent study ) demonstrated that benzoate was beneficial for patients with schizophrenia. The age, illness duration, and previous number of hospitalisations (38.4 years, 13.7 years and 3.2, respectively) of the current study subjects were similar to those (37.3 years, 14.7 years and 3.6, respectively) in the previous trial of benzoate . However, the current study showed that sarcosine plus benzoate had limited effects in patients who had failed to achieve clinical improvement with multiple trials of antipsychotic agents. Without a group of only add-on benzoate, it remains unclear whether sarcosine plus benzoate may have generated overactive neurotransmission; thereby somewhat hampering the beneficial effect of benzoate per se. Studies using only add-on benzoate is necessary to clarify this issue.
Although the precise mechanisms are unknown, the involvement of neuroimmune dysregulation and the glutamatergic system in the pathophysiology of schizophrenia is intriguing (Steiner et al. 2012;Heresco-Levy et al. 2015). Agents with both anti-inflammatory and glutamatergic transmission modification properties are promising in the treatment of schizophrenia (Hashimoto 2014). It was reported that sodium benzoate can suppress the mevalonate pathway and reduce microglial and astroglial inflammatory responses (Brahmachari et al. 2009). Whether patients can benefit from benzoate treatment through both the anti-inflammatory and NMDA receptor-mediated pathways deserves further investigation.
This study has several limitations. First, there were only three study arms: (1) placebo, (2) sarcosine, and (3) sarcosine plus benzoate. Therefore, comparisons of benzoate vs. the current three arms remain unclear. For example, whether sarcosine plus benzoate is superior or equal to benzoate alone is unknown. That is, whether there is a synergistic or additive effect from sarcosine plus benzoate remains to be determined. Second, the sample size was small and the study subjects were chronic patients in a single psychiatric hospital; therefore the generalizability of the findings would be limited. Studies focussing on subjects who are younger, at an earlier stage of illness or have less severe psychotic symptoms are necessary. Third, the changes in blood levels of amino acids such as glycine and D-serine after treatment were not measured. Further studies should be conducted. Fourth, the treatment duration of 12 weeks may have been insufficient for assessment of the efficacy and safety for patients with chronic schizophrenia. Fifth, the sarcosine plus benzoate group tended to have more females and a shorter duration of illness, albeit insignificantly, which might have influenced the practise effect on cognitive testing. Sixth, we did not follow cognitive function at a time point post the trial to examine the sustained influence of NMDA-enhancement treatment after the subjects discontinued the study drugs. Finally, the study was started before a commercial Chinese version of MCCB was available; therefore the t scores in the current study could not be compared with other studies directly. Further trials which refer to the normative data for Chinese version of MCCB should be encouraged.
In conclusion, a combination of NMDA-enhancing agents (sarcosine and benzoate), but not sarcosine alone, can improve cognitive function in patients with chronic schizophrenia, even when their clinical symptoms cannot be improved. These findings support the NMDA theory of cognitive impairment in schizophrenia (Krystal et al. 1994;Javitt et al. 2012;Hashimoto et al. 2013;Paoletti et al. 2013;Lin et al. 2014b). Future larger-sized studies in other racial populations are warranted.