OUP user menu

Effects of add-on mirtazapine on neurocognition in schizophrenia: a double-blind, randomized, placebo-controlled study

Jan-Henry Stenberg , Viatcheslav Terevnikov , Marina Joffe , Jari Tiihonen , Evgueni Tchoukhine , Mark Burkin , Grigori Joffe
DOI: http://dx.doi.org/10.1017/S1461145709990897 433-441 First published online: 1 May 2010


Mirtazapine added to antipsychotics appears to improve the clinical picture of schizophrenia, including both negative and positive symptoms. This study explored the effect of adjunctive mirtazapine on neurocognition in patients with schizophrenia who had shown an insufficient response to first-generation antipsychotics (FGAs). Thirty-seven schizophrenia patients, who were at least moderately ill despite their FGA treatment, received add-on mirtazapine (n=19) or placebo (n=18) in a 6-wk double-blind, randomized trial. Widely used neuropsychological tests were performed to explore visual-spatial functions, verbal and visual memory, executive functions, verbal fluency and general mental and psychomotor speed. The data were analysed on the modified intent-to-treat basis with last observation carried forward. False discovery rate was applied to correct for multiple testing. Mirtazapine outperformed placebo in the domains of visual-spatial ability and general mental speed/attentional control as assessed by, correspondingly, Block Design and Stroop dots. The difference in the degree of change (i.e. change while on mirtazapine minus that on placebo) was 18.6% (p=0.044) and 11.1% (p=0.044), respectively. Adjunctive mirtazapine might offer a safe, effective and cost-saving option as a neurocognitive enhancer for FGA-treated schizophrenia patients. Mirtazapine+FGA combinations may become especially useful in light of the currently increasing attention towards FGAs. Larger and longer studies that incorporate functional outcomes, as well as comparisons with second-generation antipsychotics are, however, still needed for more definite conclusions.

Key words
  • Mirtazapine
  • neurocognition
  • RCT
  • schizophrenia


Schizophrenia is a neuropsychiatric disorder with neurocognitive deficits as a major component (Keefe & Fenton, 2007). These neurocognitive deficits are independent of both clinical state and medication, and present from the first episode of disease (Albus et al. 2006; Heinrichs & Zakzanis, 1998; Saykin et al. 1994). While no particular type of neurocognitive deficit is unique for schizophrenia, tasks that require active retrieval of verbal material from long-term memory, visual-motor processing, attention, vigilance or integrity of executive functions seem to be difficult for most patients with schizophrenia (Blanchard & Neale, 1994; Heinrichs & Zakzanis, 1998; Nopoulos et al. 1994; Raine et al. 1992; Saykin et al. 1994; Sullivan et al. 1994).

In schizophrenia, neurocognitive deficits are important predictors and can account for up to 20–60% of the functional outcome (Green et al. 2000; Keefe & Fenton, 2007; Keefe et al. 2007; Weinberger & Gallhofer, 1997). This association is even more pronounced in chronic, highly symptomatic patients (Verdoux et al. 2002). Despite the use of different pharmacological or psychosocial strategies, most patients with schizophrenia still suffer from neurocognitive impairments that severely limit their social and vocational functioning. Therefore, development of therapeutic strategies to improve neurocognitive performance remains one of the primary goals in the treatment of schizophrenia (Hyman & Fenton, 2003). None of these strategies explored until now have resulted into approved treatment guidelines due to insufficient or inconsistent data.

The neurocognitive effects of add-on antidepressants in schizophrenia have been explored in only a few randomized controlled trials (RCTs), and with conflicting results. For example, the serotonin reuptake inhibitor citalopram did not yield any benefits (Friedman et al. 2005), while add-on mianserine, a serotonin receptor-blocking antidepressant with a mechanism of action somewhat close to mirtazapine, did improve neurocognition in first-generation antipsychotic (FGA)-treated schizophrenia patients (Poyurovsky et al. 2003). One of the explanations for this effect of mianserin was its serotonin 5-HT2 receptor inhibition – a feature that FGAs [in contrast to second-generation antipsychotics (SGAs)] are lacking.

Mirtazapine is an antidepressant with a unique mechanism of action that includes antagonism of presynaptic noradrenaline α2, serotonin 5-HT2 and 5-HT3 receptors, and indirect agonism in the 5-HT1A receptor (Anttila & Leinonen, 2001), which have all been shown to enhance neurocognition (Akhondzadeh et al. 2009; Galletly, 2009; Sumiyoshi et al. 2007).

In schizophrenia, adjunctive mirtazapine has been repeatedly shown to improve negative symptoms (Berk et al. 2001; Joffe et al. 2009; Zoccali et al. 2001). In our recent study, we found mirtazapine to also improve positive symptoms in FGA-treated schizophrenia patients (Joffe et al. 2009). Neurocognitive functioning improved when mirtazapine was added to clozapine in one open-label schizophrenia study (Delle Chiae et al. 2007). However, to the best of our knowledge, no RCT has reported a neurocognitive enhancing effect for mirtazapine in schizophrenia.

This study aimed to explore, with an appropriate scientific methodology, the effects of adjunctive mirtazapine on neurocognition in schizophrenia with a negligible or suboptimal response to different FGAs in stable dosages.

Patients and methods

This study was a neurocognitive arm of our earlier trial performed in the Psychiatric Hospital, Matrosy, Priaža District, and Day Treatment Unit, Psychoneurological Dispensary, Petrozavodsk, Republic of Karelia, Russia. The rationale, design, patients and procedures have been described elsewhere (Joffe et al. 2009). The study protocol and its amendments were approved by the Ethics Committee of the Karelian Republic, Russian Federation and conducted in compliance with the Declaration of Helsinki, Guideline for Good Clinical Practice (ICH GCP), and current national regulations.

In brief, the study was a double-blind, add-on, randomized placebo-controlled trial of 30 mg/d mirtazapine added to ongoing FGAs in stable dosages (⩾200 mg chlorpromazine equivalents for ⩾6 wk prior to and throughout the trial). The participants were adults suffering from schizophrenia or schizoaffective disorder depressive type according to DSM-IV criteria (DSM-IV-TR; APA, 2000) who demonstrated suboptimal response to FGAs defined as persistent positive or negative symptoms or both [Clinical Global Impression Scale (CGI), severity item ⩾4 (Guy, 1976)], with the clinical condition remaining unchanged during ⩾6 wk. A predefined randomization schedule assigned patients using a block size of 4 and a randomization ratio of 2:2 to ensure equal numbers of patients in each group.

Neuropsychological assessments

Russian translations (latest available versions) of internationally widely used, well established neuropsychological tests with a broad coverage of neurocognitive domains (Szoke et al. 2008) were employed to measure neurocognitive functions. To ensure the same level of basic intelligence functions in the groups, participants were first tested with four subtests of the Wechsler Adult Intelligence Scale – Revised (WAIS-R; Wechsler, 1981). When evaluating general intelligence domains, verbal intellectual function was measured with the Information and Similarities subtests, and non-verbal intelligence with the Block Design and Digit Symbol subtests of the WAIS-R.

Memory was assessed with the three subscales of the Wechsler Memory Scale (WMS; Wechsler, 1945) as follows: verbal memory with the Logical Memory and Verbal Paired Associations subscales, and visual memory with the Visual Reproduction subscale. Executive functions were evaluated with the Trail Making Test (TMT), part B (Lezak, 1995; Reitan & Wolfson, 1985), Stroop coloured colour-names part of the Stroop test (Stroop words; Jensen & Rohwer, 1966; Lezak, 1995) and the Digit Symbol subtest of the WAIS-R (Lezak, 1995). To explore initiation and conduct strategic mnemonic processing, verbal fluency was examined with a semantic category (animals) and letter word-list generation (words beginning with K) within 60 s (Lezak, 1995). Visual-spatial ability and fluency was tested with the Block Design of WAIS-R. General mental speed/attention control was assessed with the TMT-A (Reitan & Wolfson, 1985) and with the naming of coloured dots in the Stroop test (Stroop dots; Jensen & Rohwer, 1966; Lezak, 1995).

The neuropsychological examinations were conducted at baseline (week 0) and endpoint (week 6) by trained psychologists who were blinded to the identity of the treatment groups. Patients were tested individually. All neuropsychological tests used were considered efficacy variables.


Cross-sectional statistical differences between the study groups were tested using χ2 for categorical variables and Mann–Whitney U test for others. A non-parametric approach was used, due to the skewed distributions of the continuous response variables. Possible within-group changes in the efficacy variables over time were tested using the Wilcoxon test. Mann–Whitney U test was used to examine and test the differences between treatments over time. Spearman's correlation coefficient was used post hoc to explore associations between longitudinal changes in neurocognitive measures (i.e. those that differed with statistical significance from placebo) and both baseline clinical and demographic data. p Values <0.05 were considered statistically significant. False discovery rate (FDR) was applied to correct for multiple testing. FDR controls the expected proportion of incorrectly rejected null hypotheses (type I errors). When FDR is used, the inference is based on ‘so-called’ q values which are defined as being the FDR analogues of the p values. In practical terms, the q values measure the minimum FDR that is incurred when designating a test result significant. The q value is useful for assigning a measure of significance to each of many tests performed simultaneously whereas the p value is commonly used for performing a single significance test. In our study, the FDR procedure was carried out using q value software (http://genomics.princeton.edu/storeylab/qvalue/manual.pdf). The detailed estimation methodology and computational procedure implemented into the software is described in Storey (2002). Analyses were performed using SPSS for Windows 14.0 software (SPSS Inc., USA).


As shown in Fig. 1, of the 46 subjects screened, 41 were randomized. One patient on placebo was withdrawn during the first week. No withdrawals occurred in either group thereafter; however, one more placebo patient was excluded from the analyses due to a protocol violation. The remaining 39 patients were screened for the neurocognitive arm of the study. One patient in the mirtazapine group and one patient in the placebo group were excluded at baseline due to profound intellectual decline and therefore inability to perform in neurocognitive tasks. Thirty-seven patients (19 on mirtazapine, 18 on placebo) were included in the modified intent-to-treat population.

One patient (in the placebo group) was diagnosed with schizoaffective disorder, depressive type, the remaining subjects had chronic (mean duration of illness 20 yr), difficult to treat schizophrenia. Patients received haloperidol, fluphenazine, zuclopentixol, chlorprotixene, levomepromazine, trifluoperazine, periciazine, haloperidol decanoate, fluophenazine decanoate, zuclopentixol decanoate, or a combination of ⩾2 FGAs.

Baseline demographics, clinical characteristics, medication history, current FGA dose and measured neuropsychological performance did not differ statistically significantly between the mirtazapine and placebo groups (Table 1).

View this table:

After 6 wk of treatment with mirtazapine, baseline vs. retests within groups showed 5/21 measured parameters improved with statistical significance for mirtazapine vs. only one parameter with placebo at the FDR-corrected significance level (Table 2). No parameters worsened during either treatment. The between-group comparison showed that mirtazapine outperformed placebo with statistical significance (again, at the FDR-corrected significance level) on two parameters – Block Design and Stroop dots, while no single measure favoured placebo. The improvement in Block Design and Stroop dots did not show any associations with age, gender, out- or in-patient status, duration of illness, doses of antipsychotics or number of previously received antipsychotics, nor with improvements in either positive, negative or general symptoms. Depression scores [item 20 of the Positive and Negative Syndrome Scale (PANSS)] did not differ between the groups at baseline (t=0.434, p=0.667), nor did baseline depression scores correlate with the changes in Block Design (rs=−0.08, p=0.649) or Stroop dots (rs=−0.085, p=0.629).

View this table:


To the best of our knowledge, this is the first report of a RCT on neurocognition for add-on mirtazapine in the treatment of schizophrenia. We found a statistically favourable effect for mirtazapine added to FGAs compared to placebo in domains of visual-spatial ability and general mental speed/attention control, which were correspondingly assessed with Block Design and Stroop dots. The difference in the degree of change (i.e. change on mirtazapine minus that on placebo) was 18.6% (p=0.044) in Block Design and 11.1% (p=0.044) in Stroop dots. These improvements were not related to the changes in psychopathology. In within-group analyses, 5/21 measured parameters improved with (FDR-corrected) statistical significance with mirtazapine vs. only one parameter with placebo.

In the only previously published study (8 wk open-label, outpatients, n=15) with add-on mirtazapine in schizophrenia (Delle Chiae et al. 2007), mirtazapine added to clozapine was reported to improve neurocognition. In contrast to our findings, Delle Chiae and co-authors reported improvement in immediate and delayed memory, but attention, visual-spatial abilities and psychopathology parameters remained unchanged. This difference might be due to a difference in study population (i.e. stable outpatients with a duration of illness of 10 yr vs. highly symptomatic in-patients with a duration of illness of 20 yr in our study), different receptor affinity spectrums in mirtazapine+FGAs combinations (i.e. inhibition of a wide range of various receptors in addition to D2) vs. mirtazapine+clozapine combinations (i.e. an additional inhibition in only a limited number of receptors, e.g. α2), or methodological issues, e.g. different neuropsychological measures, a smaller sample size or open design in the earlier study. Both Block Design and Stroop dots are somehow related to general executive functions which in our study were measured also with TMT-B. Interestingly, TMT-B also tended to favour mirtazapine (improvement by 13% with mirtazapine vs. worsening by 4.8% with placebo), although difference after the FDR correction did not reach statistical significance.

The actual mechanism of a potential neurocognitive enhancing effect of mirtazapine in schizophrenia remains unknown, but it may be explained by its receptor-binding profile. First, like SGAs, mirtazapine could increase prefrontal catecholamine activity (and thereby improve neurocognitive performance) via 5-HT2A or 5-HT2C receptor blockade (Liegeois et al. 2002; Meneses, 2007; Zhang et al. 2000). Second, the 5-HT3 receptor modulation by mirtazapine could also improve neurocognition (Akhondzadeh et al. 2009), presumably through increased release of acetylcholine (Ramírez et al. 1996). Third, neurocognition could improve as a consequence of the indirect agonism of 5-HT1A receptors by mirtazapine (Sumiyoshi et al. 2007).

The α2 receptors continue to be a focus for neurocognitive research and they can improve neurocognition through noradrenaline-mediated modulation of response to environmental stimuli (Friedman et al. 2004). Moreover; α2 receptor antagonism appears to enhance hippocampal neurogenesis (Rizk et al. 2006). Indeed, mirtazapine may boost the levels of brain-derived neurotrophic factor (BDNF; Rogoz et al. 2005), a major mediator of neurogenesis and neuroplasticity. BDNF is often abnormal in schizophrenia (Rizos et al. 2008). Mirtazapine is a more potent α2 receptor antagonist than clozapine, which may explain its additional neurocognition-enhancing effect, even if (as in the trial by Delle Chiaie et al. 2007) added to clozapine.

As a potential neurocognitive enhancer in schizophrenia, mirtazapine added to FGAs might become useful, since following recent large independent studies (Jones et al. 2006; Lieberman et al. 2005; Tiihonen et al. 2006), FGAs may in future increasingly re-enter the everyday psychopharmacological armamentarium. Both FGAs and mirtazapine are out of patent time and thus cheap – an issue of an increasing importance. The neurocognitive effects of mirtazapine in our study did not depend on baseline demographic or clinical parameters. The statistically significant differences fell mostly into the positive symptoms domain, which appears to be independent of neurocognition (Heydebrand et al. 2004).

The sedative effect of mirtazapine did not seem to prevent its neurocognitive effects, at least by endpoint (week 6) when the soporific effect of mirtazapine might have already been attenuated through tolerance (Ramaekers et al. 1998). Moreover, the antidepressive effect of mirtazapine is an unlikely explanation for our findings, as additional transporter-blocking antidepressants which are devoid of noticeable receptor affinity do not seem to improve neurocognition in schizophrenia (Friedman et al. 2005). Further, in our study, baseline depression scores neither differed between the groups nor correlated with changes in Block Design and Stroop dots scores.

The short duration was a considerable limitation in our study. The desirable effects of neurocognitive enhancers in schizophrenia tend to emerge several weeks after the initiation of treatment, with further improvements over time. For example, during treatment with ziprasidone, neurocognition improved modestly after 6 months and moderately after 12 months (effect size, respectively, 0.24 and 0.61; Gibel & Ritsner, 2008). Although our sample size for add-on mirtazapine is the largest so far, it is still relatively small. However, these disadvantages were balanced by a low discontinuation rate, controlled design, and the beneficial effect observed in a chronic schizophrenia population.

There is need of larger and longer studies that comprise different types of schizophrenia patients, and also include functional outcome measures. In addition, widely used comprehensive test batteries; e.g. recently introduced MATRICS (http://www.matrics.ucla.edu; Harvey & Cornblatt, 2008; Nuechterlein et al. 2008) should be considered. The research should be expanded to compare between add-on mirtazapine and SGAs, especially clozapine. Since sedation may counteract the probable neurocognitive-enhancement properties of mirtazapine, higher doses are worth additional investigation, as they are paradoxically less sedative (Fawcett & Barkin, 1998).


Mirtazapine might improve neurocognitive functioning in patients treated with FGAs. Unlike some other neurocognitive enhancers, mirtazapine seems to increase the antipsychotic effect of FGAs, rather than provoke psychosis. Mirtazapine+FGA combinations may be a cost-effective treatment. As the use of FGAs may in future re-emerge in clinical practice, adjunctive mirtazapine as a neurocognitive enhancer may become a good choice for patients with schizophrenia. Larger and longer studies are, however, needed to verify these findings.


This trial (01T-67) was supported by a grant from the Stanley Medical Research Institute (SMRI), Bethesda, MD, USA. The SMRI had no role in study design, the gathering or analysis of data, the writing or publication of this report. [The study was registered by Current Controlled Trials (http://www.controlled-trials.com, trial registration number ISRCTN00721331).]

Statement of Interest



View Abstract