Transdiagnostic comparison of visual working memory capacity in bipolar disorder and schizophrenia

Abstract

Background

Impaired working memory is a core cognitive deficit in both bipolar disorder and schizophrenia. Its study might yield crucial insights into the underpinnings of both disorders on the cognitive and neurophysiological level. Visual working memory capacity is a particularly promising construct for such translational studies. However, it has not yet been investigated across the full spectrum of both disorders. The aim of our study was to compare the degree of reductions of visual working memory capacity in patients with bipolar disorder (PBD) and patients with schizophrenia (PSZ) using a paradigm well established in cognitive neuroscience.

Methods

62 PBD, 64 PSZ, and 70 healthy controls (HC) completed a canonical visual change detection task. Participants had to encode the color of four circles and indicate after a short delay whether the color of one of the circles had changed or not. We estimated working memory capacity using Pashler’s K.

Results

Working memory capacity was significantly reduced in both PBD and PSZ compared to HC. We observed a small effect size (r = .202) for the difference between HC and PBD and a medium effect size (r = .370) for the difference between HC and PSZ. Working memory capacity in PSZ was also significantly reduced compared to PBD with a small effect size (r = .201). Thus, PBD showed an intermediate level of impairment.

Conclusions

These findings provide evidence for a gradient of reduced working memory capacity in bipolar disorder and schizophrenia, with PSZ showing the strongest degree of impairment. This underscores the importance of disturbed information processing for both bipolar disorder and schizophrenia. Our results are compatible with the cognitive manifestation of a neurodevelopmental gradient affecting bipolar disorder to a lesser degree than schizophrenia. They also highlight the relevance of visual working memory capacity for the development of both behavior- and brain-based transdiagnostic biomarkers.

Background

Cognitive impairment across a wide range of domains is a central common characteristic of both bipolar disorder and schizophrenia (Martínez-Arán et al. 2004; Kahn and Keefe 2013; Vöhringer et al. 2013; Bora and Pantelis 2015; Miskowiak et al. 2018). Consequently, both have been conceptualized as information processing disorders (Kahn and Keefe 2013; Bortolato et al. 2015). This paradigm supports the notion that transdiagnostic comparisons of crucial cognitive constructs are a central element of translational strategies to establish a psychiatric nosology based on the assessment of cognitive dimensions and the brain networks which give rise to them (Cuthbert 2014; Insel 2014). Ultimately, this should lead to the identification of neurobiologically distinct biotypes across diagnostic boundaries (Clementz et al. 2016) and the development of behavioral and brain-based biomarkers (Oertel-Knöchel et al. 2011). Furthermore, it might also facilitate a better understanding of the neurophysiological disturbances underlying impaired information processing and the development of more effective pro-cognitive interventions.

The need for transdiagnostic studies is underscored by the substantial phenomenological and pathophysiological overlap of bipolar disorder and schizophrenia (Ivleva et al. 2010; Pearlson 2015). They have the highest amount of shared heritability among neuropsychiatric disorders (Anttila et al. 2018; Lee et al. 2019). Both are also regarded to different degrees as neurodevelopmental disorders (Bortolato et al. 2015; Pearlson 2015), possibly forming a neurodevelopmental continuum (Owen and O’Donovan 2017). This implies that risk factors disturbing brain development and cognition play a larger role in schizophrenia than in bipolar disorder. Interestingly, most studies have reported a gradient of cognitive impairment with patients with schizophrenia generally more affected than patients with bipolar disorder (Goldberg 1999; Schretlen et al. 2007; Ivleva et al. 2010; Lewandowski et al. 2011; Hill et al. 2013; Reilly and Sweeney 2014).

Working memory is universally regarded as a central cognitive domain for transdiagnostic studies of impaired information processing (Insel et al. 2010). It is a crucial determinant of essential cognitive functions such as language comprehension and reasoning (Baddeley 1992), as well as an important mediator of cognitive development and learning (Baddeley and Hitch 1974; Cowan 2014). Working memory dysfunction is a central cognitive deficit in both bipolar disorder and schizophrenia (Glahn et al. 2006; Barch and Smith 2008). It has been reported in a large number of behavioral studies in schizophrenia across all stages of illness (Lee and Park 2005; Barch and Smith 2008; Luck and Gold 2008; Fuller et al. 2009; Hahn et al. 2010; Anticevic et al. 2011b; Leonard et al. 2017; Mayer et al. 2018). Working memory impairment has also been demonstrated in bipolar disorder (Adler et al. 2004; Glahn et al. 2006; Thompson et al. 2007; Mayer and Park 2012; Jensen et al. 2016). While working memory deficits appear to be particularly pronounced in manic or depressive phases (Townsend et al. 2010), they persist during euthymic phases of the illness (Xu et al. 2012), at least in a sizable number of patients (Volkert et al. 2015). Direct comparisons between patients with bipolar I (BP-I) and bipolar II (BP-II) disorder indicate overall a similar degree of working memory impairment (Bora et al. 2011; Bora 2018). Additionally, there is evidence for a modestly greater degree of impairment in bipolar patients with a history of psychosis, compared to bipolar patients without a history of psychosis (Bora 2018).

One particularly relevant aspect of working memory is its limited capacity (Cowan 2001), which appears to constitute a core cognitive trait with high intra-individual stability over time (Kane and Engle 2002). Working memory capacity differs considerably between individuals and has strong links to high-level cognitive measures including global fluid intelligence, abstract reasoning, language abilities, mathematics, and overall scholastic performance (Daneman and Carpenter 1980; Cowan et al. 2005; Fukuda et al. 2010; Johnson, McMahon et al. 2013; Cowan 2014; Unsworth et al. 2014). Finding pro-cognitive interventions which increase patients’ working memory capacity should therefore also be a promising way to improve their general level of cognitive functioning (Johnson et al. 2013). Quantifying the degree to which working memory capacity is constrained across the schizo-bipolar spectrum is an important step toward this goal.

Based on the extensive body of work in the field of cognitive neuroscience (Luck and Vogel 2013), visual working memory capacity has been proposed as an especially suitable construct for this purpose (Barch et al. 2012). This is supported by its good construct validity and a number of specific properties. Visual working memory capacity correlates closely with measures of verbal working memory capacity but is less prone to chunking or rehearsal mechanisms (Luck and Vogel 1997; Cowan 2001), which could confound the estimation of pure working memory capacity. It has also been studied extensively using functional neuroimaging (Linden et al. 2003; Todd and Marois 2004; Vogel and Machizawa 2004). Conversely, spatial span paradigms are generally regarded as poorly suited for functional neuroimaging studies (Barch and Smith 2008). Additionally, paradigms assessing visual working memory capacity have good test–retest reliability (Xu et al. 2018; Dai et al. 2019) and have been employed successfully in animal studies (Wright et al. 2010).

Visual working memory capacity has been studied most commonly using change detection paradigms. Here, subjects have to remember one or more features such as color, location or orientation of an array of simple visual items. Subsequently, after a short delay interval they are shown a test array and have to make a judgment, whether the test array is identical or if a single item had changed. Healthy individuals are able to store information of about four objects at one time as integrated features (Luck and Vogel 1997; Wheeler and Treisman 2002). They are able to remember three to four items when required to encode a single feature such as color, or even two features of each item such as color and location. Variations of the ‘canonical’ change detection paradigm have also been implemented (Feuerstahler et al. 2019). In change localization paradigms, subjects need to specify which item has changed. In partial-report change detection paradigms, the change decision during the test array is limited to a single item. In multiple change detection paradigms, more than one item might change during the test array.

Reduced visual working memory capacity has been observed in schizophrenia (Gold et al. 2010; Mayer et al. 2012; Hahn et al. 2018) and in bipolar disorder I with a history of psychosis (Gold et al. 2018). However, to our knowledge, no study has compared visual working memory capacity in cohorts of patients with schizophrenia and schizoaffective disorder (PSZ) and patients with bipolar disorder (PBD) representing the full spectrum of both disorders. The main goal of our study was to assess working memory capacity in PBD of all illness subtypes, as well as PSZ using a canonical change detection paradigm. We expected to observe a gradient of reduced working memory capacity with greater impairment in PSZ than in PBD.

Methods

Participants

We recruited 62 PBD (42 female, mean age 42.05, range: 20—61), and 64 PSZ (26 female, mean age 38.56, range: 20–57, n = 41 with schizophrenia and n = 23 with schizoaffective disorder) from psychiatric outpatient clinics in and around Frankfurt am Main, Germany. We established diagnoses of all patients according to DSM-5 criteria based on a clinical interview and careful chart review at a consensus diagnosis meeting chaired by one of the authors (R.A.B.). We pooled both patients diagnosed with schizophrenia and schizoaffective disorder because long-term diagnostic stability and inter-rater reliability of schizoaffective disorder is relatively poor (Maj et al. 2000).

The Positive and Negative Syndrome Scale (PANSS) was used to assess current psychopathology in PSZ (Kay et al. 1987). In order to establish euthymic mood state in PBD, participants were evaluated with the Young Mania Rating Scale (YMRS) and Montgomery-Åsberg Depression Rating Scale (MADRS) (Young et al. 1978; Montgomery and Åsberg 1979). Participants with YMRS values of ≥ 11 or MADRS values of ≥ 11 were excluded from our analysis.

70 matched healthy control subjects (HC), (44 female, mean age 38.61, range: 21–61) also participated. HC had no reported history of psychiatric illness, as well as no history in first-degree family members. They were recruited from the Frankfurt University campus and surrounding areas, as well as by online and printed advertisements. Current and past symptoms of psychiatric illness were ruled out using the German version of the Structural Clinical Interview SCID-I, from the Diagnostic and Statistical Manual, Version IV (Saß et al. 2003).

All participants reported no history of neurological illness and no drug use (excluding nicotine) within the past six months. All participants ranged in age from 20–61 years old. We matched subjects at the group level by conducting Kruskal–Wallis tests based on age (H(2) = 3.902, p = 0.142), and participants’ years of education (H(2) = 1.254, p = 0.534), as well as parental years of education (H(2) = 0.834, p = 0.659).

We assessed handedness as a continuous variable using the Edinburgh Handedness Inventory (Oldfield 1971). We compared handedness scores between groups using a Kruskal–Wallis test and did not find a significant difference (H(2) = 0.962, p = 0.618).

The German Mehrfachwahl-Wortschatz-Intelligenz Test (MWT-B) (Lehrl et al. 2005) was administered to assess premorbid verbal intelligence.

Further socio-demographic information for all cohorts can be found in Table 1. Prior to signing the informed consent form, participants were informed of its contents by the investigator and what to do in the case of experiencing distress, and how to end participation in the study. The ethics committee of the University Hospital Frankfurt approved all study procedures.Table 1 Demographic and clinical characteristics of all groupsFull size table

Change detection task

We implemented a ‘canonical’ color change detection task (Fig. 1) on a personal computer using Presentation software in Version14.9 (www.neurobs.com). Stimuli were presented on a grey background (RGB values: 191, 191, 191) in a dimly lit room with a viewing distance of approximately 60 cm. Throughout the experiment, a black fixation cross was displayed at the center of the screen. Each trial began with the alert phase, during which the fixation cross turned to red for 500 ms. This was followed by a preparation phase of 500 ms. During the encoding phase a sample array of four colored circles was presented for 200 ms. Each circle had a visual angle of approximately 0.95°. These circles were spaced equally apart on an imaginary circle with 12 possible locations around the black fixation cross covering a visual angle of approximately 5.25°, and the minimum distance between two circles was 0.29°. Each circle had one of seven easily discriminable possible colors with the following RGB values: black (0, 0, 0), red (255, 0, 0), white (255, 255, 255), blue (0, 0, 255), green (0, 255, 0), yellow (255, 255, 0), and magenta (255, 0, 255), with no repetitions of colors within a trial. During the delay phase, the black fixation cross remained on the screen for 1800 ms. A whole-display recognition test array followed, in which participants had a maximum duration of 3000 ms to decide if the test array was identical to the sample array presented in the encoding phase, or if one of the circles had changed color. Half of the trials were change trials (right mouse button), the other half no-change trials (left mouse button). In change trials, a randomly chosen circle changed its color. The total duration of each trial was 6000 ms followed by an inter-trial interval of 3000 ms. All participants received the same instructions prior to the beginning the task, and were asked to perform as accurately as possible, and to keep their eyes fixated constantly on the center of the screen. A total of 60 trials were tested in each participant, which required approximately nine minutes of testing time.

This entry was posted in News & updates. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s