Introduction
The co-occurrence of psychopathological symptoms stemming from different functional domains is characteristic of psychiatric disorders. For instance, schizophrenia spectrum disorders (SSD) are characterized by psychopathological symptoms from various functional domains, including sensory–perceptual (e.g., hallucinations), cognitive (e.g., delusions, impaired working memory and executive functioning), affective (e.g., flat affect, anhedonia and depressive symptomatology [
1]), somatic (e.g., fatigue, muscle pain, gastrointestinal symptoms and tension [
2]), social (e.g., social isolation and autism), and sensorimotor (e.g., rigor, tremor, psychomotor slowing and akinesia [
3]), respectively. In mood disorders (MOD), such as major depressive disorder (MDD) and bipolar disorder (BD), patients not only experience low mood, sadness [
4], low self-esteem, lack of drive, and loss of interest or pleasure but also co-occurring cognitive changes (e.g., rumination [
5] and impaired working memory [
6]), somatic-neurovegetative (chest pain, gastrointestinal symptoms, irritable bowel, or other [
7]), interpersonal (e.g., social anxiety and withdrawal [
8]), sensory–perceptual (aberrant visual perception [
9]), and motor (psychomotor retardation or agitation [
10]) symptoms.
Furthermore, the majority of the psychopathological symptoms mentioned above are core diagnostic criteria for SSD and MOD by the most widely used diagnostic systems in clinical practice, such as the DSM-5 and ICD-11. The overlap and co-occurrence of psychopathological symptoms in SSD and MOD are specifically mentioned in both diagnostic systems as the foundation for differential diagnostic considerations. However, their intrinsic relationships—that is, in what ways do the various psychopathological symptoms interact and which symptoms are central or influential across SSD and MOD—remain understudied. In the last few years, novel statistical methods have been developed to explore the interplay between psychopathological symptoms and global functioning in psychiatric disorders [
11‐
13]. In particular, network analysis can help identifying which psychopathological symptoms are central or influential within a network of different clinical variables [
14,
15]. Central psychopathological symptoms are those that have strong connections to other symptoms or clinical variables and may play a crucial role in the development and maintenance of the disorder. Therefore, network analysis is a promising method to investigate the complexity of interactions between psychopathological symptoms stemming from different functional domains across different psychiatric disorders. Instead of seeing psychopathological symptoms as standalone entities, network analysis conceives of them as components of a complex network that interact with one another and frequently reinforce one another. In line with this, psychiatric disorders, such as SSD and MOD, might arise from the direct interactions of symptoms within a network architecture.
In the last two years, the network analysis method has contributed to a considerable increase in knowledge in the field of different psychopathological symptoms [
16‐
18]. However, the studies using network analysis method have investigated mainly the interaction between different categories of psychopathological symptoms using single items stemming from different rating scales such as the Brief Negative Symptom Scale (BNSS) or the Positive and Negative Syndrome Scale (PANSS) in uni-diagnostic psychiatric samples [
16‐
18].
Based on the recent evidence that distinct Neurological Soft Signs (NSS) have been shown to be differentially associated with psychopathology as well as cognition in SSD [
19‐
21] and BD [
22], the main goal of this study was to examine the associations among psychopathological symptoms, sensorimotor, cognitive, and global functioning in a transdiagnostic patient sample. We hypothesized that sensorimotor dysfunction, as examined with the Heidelberg NSS scale [
23] would be more closely connected with cognition (as assessed with the Brief Cognitive Assessment Tool for Schizophrenia [B-CATS] [
24]) than psychopathological symptoms examined with PANSS. Further, we hypothesized that sensorimotor dysfunction would be a central network component (as assessed with expected influence [EI] and strength). Unlike previous studies that conducted the network analyses at the single-item level, and because we had a strong prior theory on how variables are related, we decided to use subscale scores in order to examine associations between subdomains. If we confirmed our hypotheses, this study could point toward shared domain-based aspects of pathophysiology in SSD and MOD. This could point toward new treatment options for cognition, e.g., with stimulation techniques in specific brain regions associated with sensorimotor dysfunction [
25,
26] or virtual reality [
27].
Discussion
Using the network approach to psychopathology, we investigated—for the first time—the interrelationship between psychopathological symptoms, sensorimotor, cognitive, and global functioning in a transdiagnostic sample consisting of SSD, MDD, and BD patients. Two main findings emerged: First, NSS showed closer connections with TMT-B, CF and DSST than with GAF and PANSS. Second, DSST, PANSS general, and NSS MOCO scores showed the highest EI, while SI, DSST, and CF showed the highest strength. However, the EGA was unstable and could not be interpreted.
The first finding supports and extends our understanding of the relationship between sensorimotor and cognitive domains for several reasons: First, there has been prior research linking more severe hypokinetic sensorimotor abnormalities (e.g., NSS and parkinsonism) to executive functioning deficiencies in SSD [
52,
53]. Second, our findings are in line with Wolf et al. [
54] and confirm the precise link between the sensorimotor and cognitive domains. Our findings extend this previous work transdiagnostically. Third, the TMT-B and DSST assessments are well-known tools of cognitive performance, particularly in terms of processing speed, cognitive flexibility, and the phenomenon known as psychomotor slowing. This line of reasoning suggests that a more inclusive definition of this term may incorporate sensorimotor disorders such as NSS [
55]. Accordingly, Osborne et al. [
55] described that the cognitive (“psycho”) and sensorimotor sub-processes that lead to psychomotor slowing may overlap with the sub-processes of subtle sensorimotor abnormalities such as NSS. Finally, NSS and TMT-B are associated with overlapping circuitry, suggesting a biological link between sensorimotor and cognitive processes [
56]. On one side, changes in the inferior frontal gyrus, paracentral gyrus, inferior parietal lobe (IPL), bilateral putamen, cerebellum, and both the superior and middle temporal gyri (STG and MTG) are associated with NSS (for a summary, see also Hirjak et al. [
57,
58] and Zhao et al. [
59]). On the other side, the prefrontal cortex, the IPL, and the cerebellum play also a crucial role in the execution of TMT-B [
60]. This implies that a shared neurobiological process may lead to manifest sensorimotor and cognitive symptoms in SSD and MOD. Our results together with the above mentioned evidence extend our current understanding of interrelated behavior in the RDoC Matrix.
Last but not least, although there have been reports—also from our group [
19]—examining the bilateral link between NSS and psychopathology as well as NSS and cognition [
20,
21], the precise trilateral relationship between cognition, NSS, and psychopathology remains unclear. We interpret our results as suggesting that sensorimotor symptoms are closer to cognition than psychopathology, but this relationship does not exclude connections between NSS and psychopathology. Indeed, as shown in Table
2, PANSS positive and negative subscales showed (rather weak) partial correlations with NSS subscales. Psychopathology also showed few connections with cognitive symptoms, which is in line with assumptions that the psychopathology domain by itself (as outcome parameter) only incompletely reflects functional impairment, which is more closely connected with cognitive deficits [
61].
Second, the most central nodes in terms of EI were DSST, PANSS general, and NSS MOCO scores. All nodes showed a positive value in the centrality index. In terms of strength, SI, DSST and CF showed the highest centrality. These findings are crucial for a number of reasons: First, the DSST examines processing speed, working memory, visuospatial processing, and attention. Also, another recent network analysis in SSD showed a central place for processing speed evaluated with symbol coding [
42]. This said, DSST reflects the global cognitive functioning, which, when impaired, can lead to disturbances at the level of sensorimotor functions. In line with this, Morrens et al. [
62] used DSST (e.g., matching time and writing time) to examine sensorimotor and cognitive slowing in SSD. Although the authors concluded that both processes are unrelated, matching time was associated with neuropsychological test results. Second, this fits in well, because MOCO also plays a central role in our network. The MOCO subscale includes five items, such as Ozeretski’s test, diadochokinesia, pronation/supination, finger-to-thumb opposition, and speech articulation. Both DSST and MOCO are based on movement execution, predominantly at the level of sensorimotor and visuospatial control. Third, disturbances of sensorimotor and spatial–visual control can lead to various psychopathological symptoms, such as somatic concerns, anxiety, depression, motor retardation, disorientation, disturbance of volition, poor impulsive control, and preoccupation, respectively [
63,
64]. This could be a possible explanation for the centrality of the PANSS general scores. Fourth, as proposed by the European consensus on assessments and perspectives [
26], new treatment targets for stimulation techniques need to be identified. From a network point of view, variables are interrelated and targeting the most central nodes could affect other nodes as well. Consequently the connection between sensorimotor symptoms and cognition and the high centrality of sensorimotor symptoms suggest that sensorimotor symptoms and cognitive symptoms may share aspects of pathophysiology. Specifically, the high centrality of MOCO and SI together with the previously identified neurobiological correlates of NSS may suggest NSS MOCO and SI as possible treatment targets for stimulation techniques (such as transcranial direct current stimulation or repetitive transcranial magnetic stimulation) in order to improve cognition. In addition, one may wonder whether movement exercises that train the individual items of the two subscales could also lead to an improvement in cognition. This is in accordance with previous reports showing associations between NSS and cognition in longitudinal investigations [
20]. Previously reported structural correlates of NSS [
25] would be the primary neurobiological target regions for these stimulation techniques. This train of thought seems particularly relevant since cognitive deficits in SSD are associated with lower global and social functioning [
61,
65] and are difficult to treat so far. Yet, while structural correlates of NSS have been repeatedly demonstrated in SSD [
25,
66‐
69], we are not aware of such data in BD or MDD. When interpreting our results, it is important to bear in mind that there may not be a direct link between our non-direct, non-causal network analysis and neurobiological pathophysiology.
Taken together, complex phenomena (such as cognitive, psychopathological, and sensorimotor symptoms) may be best described at the systems level [
13,
70‐
72]. More precisely, a shift from focusing on individual components to studying the organization of the system’s components seems promising [
13,
70‐
72]. Network analysis provides tools to identify a system’s components and their relationships [
70]. In this context, the particular benefit of network analysis techniques may be provided by their focus on patterns of pairwise conditional relationships as well as by enabling powerful visualizations of those patterns [
70]. Furthermore, previous studies have established relationships between the sensorimotor and cognitive domain [
52,
53]. However, the methods applied in these studies were limited: they were not able to simultaneously take into account the dynamic effects of other relevant variables on the relationship of interest as network analysis can. Also, associations between the sensorimotor and psychopathological domains have been reported as well [
19]. Yet, how these domains interact simultaneously with each other has not been investigated so far. Here, our network analysis for the first time models the interactions between the sensorimotor and cognitive domains while accounting for psychopathological symptoms and global functioning. Our results suggest that there is a closer connection between the sensorimotor and cognitive domains than between the sensorimotor and psychopathological domains. Furthermore, the reported association between the sensorimotor and cognitive domains after accounting for psychopathological symptoms and global functioning may implicate several future investigative steps: First, the close connection between the sensorimotor and cognitive domains on the level of clinical tests suggest overlapping pathophysiology between both constructs [
25,
60,
73]. Longitudinal MRI studies with several network analyses sequentially excluding psychopathological symptoms and global functioning may shed more light on the precise relationship between sensorimotor and cognitive symptoms. Another road may be—in accordance with previous network analysis literature [
70,
71,
74]—to consider either sensorimotor symptoms or cognitive symptoms as treatment targets and investigate the effects on both domains after pharmacological and non-pharmacological (stimulation techniques) interventions.
Strengths and limitations
The study sample size, a solid theoretical framework, and the use of two network indices, such as EI and strength centrality, are the main strengths of this study. However, this study also has limitations: First, sample size of BD and MDD subgroups was limited, which could imply that our results may have been more influenced by our SSD subgroup. Due to small group sizes in our MOD participants, we were not able to perform the network analyses in this subgroup, thus uncertainty remains whether our findings truly extend to MOD. Yet, in separate analyses including only SSD participants (
n = 174), the results remained comparable. Still, the imbalance in the examined groups concerning the diagnosis could lead to sampling bias, making it difficult to generalize the findings to the broader population. Therefore, the question as to whether sensorimotor symptoms are really a transdiagnostic therapeutic target needs to be examined in future studies including larger and more balanced diagnostic groups. Second, by employing cross-sectional data, we cannot make inferences about long-term relationships between psychopathology, sensorimotor, and cognitive symptoms and we advocate longitudinal studies which could better test for causality. Third, we did not employ second-generation negative symptom scales such as the BNSS. However, our focus in this project was to analyze the entire psychopathological domain, for which PANSS remains the most widely used scale in SSD and other psychiatric disorders such as MOD (please refer to PANSS general symptoms). Fourth, we are aware of the previous network analyses [
14,
15,
75,
76] which examined data on several levels ranging from total scores to sub-scores to single items level. The authors stated that more knowledge could be accumulated through the multilevel analysis. The study by Fried et al. [
77] corroborate this view and performed network analyses on several levels. The authors [
77] reported that relationships between depression and inflammation were strongly attenuated after controlling for BMI. They also concluded that decomposing sum scores may lead to reduced reliability [
77]. Therefore, we assume that our network analysis could reveal more detailed insights if performed on several levels. However, in our study, we were interested in the interplay between psychopathology, sensorimotor, cognitive and global functioning more globally. Thus, we opted for subscores/subscales rather than single-item level scores. In addition, investigating many individual features (i.e., including more nodes in the network) makes network estimation more complex, requiring larger study groups in order to yield stable networks and consequently, we refrained from these analyses. Last but not least, our hypotheses were focused on investigations of the relationships between the different domains and not individual items.