A systematic review of the effects of psychiatric medications on social cognition

Social cognition is defined as the mental processes which underlie the ability to understand and act on the thoughts, intentions, and behaviours of others [1]. Deficits in social cognition have been found in psychiatric disorders including depression, schizophrenia and bipolar disorder [2‐4] and can lead to the faulty interpretation of others’ intentions and thinking, as well as inaccuracies in identifying others’ emotions [5].

Social cognition can be separated into individual testable domains. However, many of these domains overlap, and there is no consistent agreement between cognitive scientists as to which are the most important. In psychiatry research the domains most frequently studied tend to reflect those identified by the National Institute of Mental Health (NIMH) at their meeting to define social cognition in schizophrenia in 2006 [6, 7]. These domains can be seen in Fig. 1, and include: Theory of mind (ToM) - the ability to ‘infer intentions, dispositions, and beliefs of others’ [8]; emotion processing - the ability to perceive emotions and interpret them appropriately [9]; social perception - the ability to process social cues and context to decipher social situations [10]; attributional bias - how people interpret the causes of events, which may be positive or negative in nature [11]; and social knowledge – how mental schemas of social situations guide behaviour [12]. Additional domains of social cognition tested in research include emotional intelligence, prejudice and stereotyping, and empathy [6].

Social functioning deficits are a core feature of most psychiatric diagnoses and considered an integral treatment target for many conditions, in order to promote recovery [13, 14]. Social cognition deficits have been associated with poor social functioning in several psychiatric disorders including schizophrenia, bipolar disorder, anorexia, Alzheimer’s disorder, and depression [2, 3, 15‐17]. Additionally, in schizophrenia better social cognitive ability has been linked to better social functioning outcomes [18]. This makes social cognition a potential target for treatment interventions across psychiatric care.

The existing studies showing social cognition deficits in psychiatric disorders often involve people who are currently taking psychiatric medication [19]. However, it is unclear how these medications might affect social cognition. Many psychiatric medications have sedative effects, including benzodiazepines, antipsychotics (to varying degrees) and drugs used as mood stabilisers (22), and evidence suggests these drugs impair neurocognitive functioning in volunteers [20‐23]. Antipsychotics also impair cognitive functioning in people with Alzheimer’s disease [24], but evidence on the effects of antipsychotics in people diagnosed with schizophrenia is inconclusive. Some evidence suggests that antipsychotics improve neurocognitive functioning [25, 26] and some that they impair it [27, 28].

Along with sedative effects, psychiatric medications affect emotion and motivation. Antipsychotics, for example, reduce motivation and suppress emotions in volunteers [29], effects which are also reported by patients [30], and these effects may impact on social cognition. On the other hand, psychiatric drugs may improve social cognition by alleviating symptoms that impair social interaction, such as psychotic symptoms and anxiety. Moreover, different agents within the same class may have different effects on social cognition, depending on their sedative profile and other effects [31].

Therefore, there is good reason to believe that psychiatric medications may influence social cognition, especially those with sedative actions that are known to impair neurocognitive functioning in volunteers. Clarifying these effects is important in order to understand the nature of social cognitive deficits in psychiatric disorders, and to evaluate the effects of treatment on social cognition and associated outcomes, such as social functioning. A previous review highlighted the paucity of evidence on the effects of antipsychotic treatment, but it did not explore the use of other medications or effects in volunteers [31]. Volunteer studies help to distinguish those effects that occur in the absence of symptoms of psychiatric disorders from those that are related to the disorder itself, or to the interaction of the treatment with the disorder. They can help with the interpretation of studies with patients who are taking medication, and ultimately improve our understanding of this complex area.

The realisation that social disability may be linked to cognitive dysfunction has led to the employment of neuroimaging techniques to study this phenomenon in psychiatric populations, including the use of electroencephalography (EEG) and functional magnetic resonance imaging (fMRI). EEG can identify temporal changes in brain activity in response to specific tasks via event-related potentials (ERP), and fMRI is used to detect the location of changes in brain activity via variations in blood-oxygen-level-dependence (BOLD) [32, 33]. ERPs typically associated with social cognitive emotional stimuli are the P300-P400 potentials, where the brain shows activations 300-400 ms post-stimulus [34]. In fMRI, a social cognitive brain network has been identified and includes consistent activation of regions, including the temporo-parietal junction (TPJ), anterior cingulate cortex (ACC), superior temporal sulcus (STS), ventral and dorsal medial prefrontal cortex (VMPFC and DMPFC), precuneus, and inferior frontal gyrus (IFG) [35]. Neuroimaging studies investigating the effects of sedative medications on social cognition will help to identify any temporal or spatial neural changes in social cognitive brain regions, as a result of medication effects. This research is important in allowing researchers to assess the biological impact of psychiatric pharmaceutical treatments. In studies where patients have been using psychiatric medications with sedative effects long-term, permanent changes to structural and functional brain systems may inhibit the identification of medication effects on social cognition, Therefore, healthy volunteer and drug-naïve patient studies will be integral to our understanding of medication effects on social cognition in neuroimaging studies.

Despite evidence of effects on neurocognitive functioning, there has been little consideration of how psychiatric medications affect social cognition. We hypothesised that psychiatric medications that produce sedative effects might affect social cognition, and we conducted a systematic review of the literature in this area. We included research on healthy volunteers as well as research conducted with patients with diagnosed psychiatric disorders.

An additional aim of this review was to explore any temporal or spatial brain differences between healthy volunteers and clinical populations with psychiatric diagnoses conducting social cognition tasks after administration of psychiatric medication using neuroimaging technology. Notable differences in brain activity may reflect the effects of medication on social cognitive processing.

The review will help to clarify the nature of any underlying deficits in social cognition in people diagnosed with psychiatric disorders, and this will help in the development of targeted treatments for social cognition, which may also improve social functioning and general outcomes [28].

This review follows the PRISMA guidelines for reporting systematic reviews [36]. The review protocol is available on the PROSPERO registry, ID: CRD42018092883.

The scope of our narrative synthesis was to examine the effects of sedative medications on social cognition. Following guidance from Popay et al. [37] we used four iterative elements shown in Fig. 2. As the first point suggests, we conducted an initial scoping of the literature to summarise the current research in the field, and in order to construct our search strategy. To address point two, we reported our findings in the results section and summarised relevant data from the included papers in a table (Table 1). In our discussion we critically explored relationships between the reported studies and went on to discuss the strengths and limitations of the current review, to address points three and four.

We searched the following major databases: MEDLINE (OViD), Embase, Psychinfo, Web of Science, Lilacs, and Scopus as well as grey literature through greylit.​org and opengrey.eu. Database-specific search terms included the keywords ‘social cognition’, ‘mental disorder’, ‘neuroleptic agents’, ‘sedatives’, and ‘tranquilisers’ with intervention-specific terms (including names of drug classes, and individual agents in classes that were not included as a whole, e.g., some sedative antidepressants), diagnosis-specific terms, outcome-specific subtypes and synonyms (see Additional file 1: Appendix A for full list of search terms and search strategy). An attempt to find additional studies was made through a backward reference search and contacting experts in the field.

Citations were imported to Mendeley and all duplicates were removed. ZH independently screened all citation titles for their applicability [77]. Titles that did not meet eligibility criteria were removed. Full-text papers were then screened and any uncertainties about inclusion were discussed with a second reviewer (AJW).

Study quality was evaluated using the Downs and Black checklist [78] as it allows for assessment of both randomised and non-randomised studies. The checklist evaluates papers on reporting, external validity, and internal validity (bias and confounding). It consists of 27 items scored with 0 points for ‘no/unable to determine’, or 1 point for a ‘yes’ response. Item 5 is scored differently with 0 points for a ‘no’ response, 1 point for a ‘partially’ response, or 2 points for a ‘yes’ response. The last item on the checklist regarding power was altered in concordance with a previous review conducted by Trac et al. [79] to rate whether a power analysis was calculated (1 point), or not (0 points). The maximum score for the checklist was 28, with the scoring ranges being (< 14) poor quality, (15–19) fair quality, (20–25) good quality, and (26–28) excellent quality.

The search identified 2931 papers fitting the eligibility criteria, with 2681 remaining after de-duplication. The abstracts and titles of those records were then screened and 2511 were excluded due to not meeting the inclusion criteria. This resulted in 170 papers for full-paper screening. A further 130 papers were excluded during this stage, for reasons shown in Fig. 3. The remaining 40 full-text papers were used in the narrative synthesis. Data from these papers including study design, sample, medication (name, dosage), pre-intervention treatment, placebo (yes/no), social cognitive domains tested, social cognition measures, follow-up timepoints, key findings, and study limitations, were extracted and can be viewed in Table 1. Notably there were no studies of mood stabilisers, barbiturates, pregabalin or any sedative antidepressants.

Data quality was rated by ZH on all 40 full-text papers and AJW on 20% randomly selected papers. An interrater reliability score Cohens Kappa Coefficient of 0.85 was calculated [80], indicating good agreement between authors. Of the 40 included papers, 11 were rated poor and 20 were rated fair. Only 9 total papers were rated good, and none were rated excellent. Scores for each paper are reported in Table 1.

Four benzodiazepine studies conducted in healthy volunteers showed significant impairments in emotion recognition social cognition tasks following diazepam administration [39, 41, 44, 81], suggesting that sedative medication at a therapeutic dose impairs emotion processing. One of these papers also incidentally investigated the effects of Metropolol, a beta-blocker with mild sedative effects, and found no significant effect of the drug on emotion recognition [41]. A further study [42] showed a selective effect of diazepam on recognition of angry expressions only. This result may be due to using a small dose in comparison to other studies. One benzodiazepine study using oxazepam showed no effect on a measure of empathy [38].

One study was conducted in patients with schizophrenia, which looked at the effects of benzodiazepine withdrawal. Patients who were in the process of withdrawing from benzodiazepines were significantly impaired in recognising negative emotions compared to healthy volunteers, in contrast to patients who had already withdrawn, who were not. However, all patients were likely to have been on other medications [81].

All neuroimaging studies compared social cognition before and immediately after administration of the experimental drug. Del-Ben et al. [43] showed that a single dose of diazepam in healthy volunteers resulted in attenuated activation of the right amygdala when responding to fearful faces, although no evidence was found for this interaction when participants viewed angry faces. In another healthy volunteer study, Paulus et al. [82] showed that lorazepam attenuated activation in the amygdala and insula, and that the activation was significantly lower after 1 mg compared to 0.25 mg, suggesting a dose-dependent reaction in emotional processing regions. However, a study by Olofsson et al. [45] found no interaction between benzodiazepines and EEG activity during response to an affective processing task.

A study investigating benzodiazepine effects on patients with ‘catatonic’ schizophrenia and patients with bipolar disorder found that lorazepam induced BOLD signal decreases in the occipital cortex and medial prefrontal cortex (MPFC) in patients with schizophrenia when undertaking a negative emotion recognition task. This resulted in BOLD patterns resembling those of healthy volunteers taking a placebo drug during the same emotion recognition tasks [46]. However, at the time of the fMRI task all patients were taking either antipsychotic or antidepressant medications in addition to the administered lorazepam.

One of the most important strengths of this review was establishing the current literature on the effects of sedative psychiatric medications on social cognition using a rigorous search strategy of published and unpublished work.

We included all psychiatric populations, and healthy volunteer studies in our search. However, in our review we found the research was largely limited to studies of benzodiazepine effects in healthy volunteers, and studies of antipsychotics in patients with schizophrenia with one study of patients with the neuropsychiatric disorder, Huntingdon’s disease [86]. Research on neurocognitive function suggests that antipsychotics, in particular, may have specifically detrimental effects in people with psychiatric disorders, such as Alzheimer’s, and further research on their effects on social cognition in people with these disorders would be valuable.

We made efforts to also include all prescribed psychiatric medications with sedative effects in our search, but we may have omitted some medications that are not commonly used. We also excluded drugs that are prescribed for mental disorders but are predominantly used for non-psychiatric indications, such as gabapentin or beta-blockers. We also did not include drugs with sedative effects that are routinely used for physical disorders, such as opiate anaesthetics, for example, and we also did not include recreational sedatives such as alcohol or heroin in our search. The review focused on prescribed psychiatric medication in order to clarify the effects of these medications in people with diagnoses of mental disorders, but recreational drugs are commonly used amongst patients with a mental health diagnosis [87], and their sedative effects may also influence social cognitive ability. Therefore, this should be an important consideration for future research in the area.

An integral strength of our search for this review was the inclusion of all known terminology for social cognitive domains and measures. However, this was difficult due to the use of interchangeable terms for similar items, exposing a feature of the social cognition field that needs to be addressed.

A limitation of the current review was the poor quality of available studies. Our data quality analysis tool allowed us to identify several deficiencies with current papers available in the field, such as small sample sizes, non-randomised designs, and few adherence to medication measures. Only four of the studies found conducted power analysis to qualify their sample size. This resulted in many of the included studies having small numbers of recruits, undermining the internal and external validity of the research findings. During quality analysis, researchers also found only three of the longitudinal studies included were recording medication adherence. In addition to this, very few studies considered the influence of practice effects, which have an important influence on the results of longitudinal studies of cognitive performance, and there were few randomised placebo-controlled studies that would allow conclusions about whether changes in patients taking antipsychotics were attributable to specific medication effects, or whether they were the result of unrelated symptom improvement or of practice effects. Additionally, only one study of antipsychotics, and no benzodiazepine studies, looked at dose-dependent effects. This variability in studies also restricted analysis of the papers included, making a meaningful meta-analysis impossible to conduct.

One other major limitation of this review was that 80% of the included studies explored emotion processing tasks, leaving the other domains of social cognition largely ignored in the literature. Resultingly, our review is more of an insight into the effects of sedative medications on emotion processing, rather than the broader area of social cognition as a whole.

Finally, benzodiazepines have reasonably consistent effects, but antipsychotics vary widely in their receptor targets, pharmacological actions and sedative profiles [47, 48]. The studies examined did not enable a comprehensive comparison of the effects of different agents within any class of drugs. In addition, no studies were found that assessed effects of other prescribed psychiatric drugs with sedative properties, such as mood stabilisers, sedative antidepressants or pregabalin.

We suggest that further research of higher quality is needed to clarify the effects of sedative medications on social cognition in healthy volunteers and patients with psychiatric diagnoses.

Further studies conducted with neuroimaging techniques will allow better insight into structural or functional brain changes resulting from administration of psychiatric medication with sedative effects. Conducting these studies with a behavioural performance element will also allow researchers to identify if brain changes are consistent with changes in social cognitive ability.

Further studies also need to control for practice effects, and studies involving patients should include placebo or no treatment control groups in order to distinguish the effects of medication from the natural evolution of psychiatric symptoms. Studies should be conducted across a range of social cognition domains, to ensure we are getting an accurate picture of complete social cognitive ability. Additionally, studies should be conducted across a range of psychiatric medications with sedative properties, to ensure we are able to identify any significant differences between drugs, and in different psychiatric diagnoses to clarify the effects of medication across conditions.

Notably, recent research in the field of social cognition and psychiatry has focused on the potential benefits of non- pharmacological interventions, such as social cognition and interaction training (SCIT), metacognitive reflection and insight therapy (MRIT), and metacognitive training. A review in 2009 by Choi and colleagues [88] found that five intervention studies showed promising results for social cognitive improvements in patients with schizophrenia, and a comprehensive review by Kurtz et al. [89] showed large effect sizes for training on facial affect recognition, moderate effect sizes on theory of mind, and small effect sizes on attribution bias for patients with schizophrenia. Although some studies included in these reviews used control groups, the majority of studies failed to control for potential medication effects in participants with schizophrenia. Psychiatric medication use alongside a social cognitive training intervention may cause improvements or deficits in participant outcomes, which we consider an important clinical implication in treatment implementation. Therefore, we suggest future research in this area accounts for psychiatric medication use in the analysis of intervention effectiveness.