Disentangling the Relationships Between the Clinical Symptoms of Schizophrenia Spectrum Disorders and Theory of Mind: A Meta-analysis

Abstract

Background and Hypothesis

Previous studies have suggested links between clinical symptoms and theory of mind (ToM) impairments in schizophrenia spectrum disorders (SSD), but it remains unclear whether some symptoms are more strongly linked to ToM than others.

Study Design

A meta-analysis (Prospero; CRD42021259723) was conducted to quantify and compare the strength of the associations between ToM and the clinical symptoms of SSD (Positive, Negative, Cognitive/Disorganization, Depression/Anxiety, Excitability/Hostility). Studies (N = 130, 137 samples) including people with SSD and reporting a correlation between clinical symptoms and ToM were retrieved from Pubmed, PsycNet, Embase, Cochrane Library, Science Direct, Proquest, WorldCat, and Open Gray. Correlations for each dimension and each symptom were entered into a random-effect model using a Fisher’s r-to-z transformation and were compared using focused-tests. Publication bias was assessed with the Rosenthal failsafe and by inspecting the funnel plot and the standardized residual histogram.

Study Results

The Cognitive/Disorganization (Zr = 0.28) and Negative (Zr = 0.24) dimensions revealed a small to moderate association with ToM, which was significantly stronger than the other dimensions. Within the Cognitive/Disorganization dimension, Difficulty in abstract thinking (Zr = 0.36) and Conceptual disorganization (Zr = 0.39) showed the strongest associations with ToM. The association with the Positive dimension (Zr = 0.16) was small and significantly stronger than the relationship with Depression/Anxiety (Zr = 0.09). Stronger associations were observed between ToM and clinical symptoms in younger patients, those with an earlier age at onset of illness and for tasks assessing a combination of different mental states.

Conclusions

The relationships between Cognitive/Disorganization, Negative symptoms, and ToM should be considered in treating individuals with SSD.

mentalizing, social cognition, negative symptoms, Cognitive/Disorganization symptoms, positive symptoms

Introduction

People with schizophrenia spectrum disorders (SSD) present with impairments in their everyday functioning. Multiple factors account for these impairments including clinical symptoms of SSD (eg, negative symptoms such as social withdrawal, positive symptoms such as delusions) and cognitive impairments.¹ Social cognition, defined as the mental operations that underlie social interactions, including perceiving, interpreting, and generating responses to the intentions, dispositions, and behaviors of others² has shown an overall medium association with functioning in people with SSD.^1,3 One domain of social cognition that has been most strongly and consistently linked with functioning in SSD is theory of mind (ToM).^1,4–6 ToM is defined as the ability to represent other people’s mental states and/or infer their mental states (ie, emotion, intention, belief/knowledge)⁷ and is impaired in individuals with SSD throughout all stages of the illness.^8–10 ToM is a complex cognitive domain that involves integrating information perceived in the environment as well as information retrieved from memory to form hypotheses regarding the mental states of others.¹¹ The integrity of ToM performance can thus be influenced by several factors such as neurocognition (eg, attention),¹² cognitive bias (eg, jumping to conclusions) as well as the clinical symptoms of SSD (eg, delusions, blunted affect).

Several previous studies have indeed suggested a link between ToM and the clinical symptoms of SSD. For instance, some researchers have put forth the idea that individuals presenting with more severe negative symptoms often underinterpret the mental states of others.^13,14 There has been growing evidence of a strong relationship between negative symptoms and ToM impairment, notably in studies using mediation models and demonstrating a link with poor social functioning.^15–19 Pelletier-Baldelli and Holt²⁰ have even hypothesized that negative symptoms may represent a “real-life” manifestation of ToM deficits.²⁰ By contrast, people with more severe positive symptoms such as delusions may be overinterpreting the mental states of others.^13,14 Many authors have demonstrated that ToM abilities could predict the presence and severity of positive symptoms, as well as their negative impact. For instance, in accordance with Frith²¹ theory, Mehl et al²² have found that one’s ability to infer intentions in a ToM task explained a significant proportion of variance in delusions. Overall, it seems as though the relationship between ToM and the clinical symptoms could be bidirectional, such that ToM impairments may contribute to more severe clinical symptoms, while clinical symptoms may also influence the perception, interpretation, and behaviors in social situations. Symptoms of SSD may indeed be contributing to how information is perceived, processed, and interpreted, which, in turn, could influence how the mental states of others are being interpreted. However, it remains unclear whether certain symptom dimensions or individual symptoms are more strongly associated with ToM than others and the results of the associations observed in the literature are heterogeneous. While some studies have found significant associations between ToM and several dimensions of symptoms,^23,24 others have only observed significant associations with some clinical symptoms (eg, negative symptoms) and not others (eg, positive symptoms).^25,26

This heterogeneity could potentially be explained by several factors. Firstly, the measures used to assess ToM present with significant heterogeneity, as well as psychometric limitations. For instance, some clinical symptoms may be more closely linked to certain types of mental states (eg, intention, emotion) than others; in some studies, the ability to infer intentions predict delusions^22,27 while in other studies, the ability to infer emotions is closely linked to the negative symptomatology.²⁸ Even tasks assessing the same type of mental state use different stimuli (eg, picture, text), methods (eg, irony, metaphor), complexity level (eg, first- vs second-order attribution), and level of ecological validity, all of which could influence the relationship between ToM and clinical symptoms of SSD. Secondly, the conceptualization of clinical symptoms is heterogeneous, such that there are several different scales to measure symptoms as well as methods of classification. For example, the Positive And Negative Syndrome Scale (PANSS) was initially developed as a 3-dimension conceptualization (Positive, Negative, General), but more recent factorial analyses have suggested a refined approach using 5 dimensions including Positive, Negative, Cognitive/Disorganization, Depression/Anxiety, and Excitability/Hostility dimensions.²⁹ The traditional approach to classify clinical symptoms includes a wide range of clinical manifestations within the same dimension which could result in a loss of information regarding potential relationships with certain clinical symptoms that are better represented by a 5-factor classification. Thus, the 5-factor classification offers a more nuanced conceptualization of the clinical profile that could help in identifying important relationships between clinical symptoms and ToM.

Previous reviews have provided insightful information regarding the relationship between ToM and other impairments recognized in people with SSD such as neurocognition¹² and/or functioning.^1,4,6 One meta-analysis revealed that negative symptoms moderated the relationship between ToM and functioning⁶ while a second meta-analysis found that total symptoms moderated the relationship between ToM and neurocognition.¹² The meta-analysis by Sprong et al³⁰ assessed how the strength of ToM impairments in SSD may be related to different groups of clinical symptoms and revealed that patients with symptoms of disorganization performed more poorly on ToM tasks than healthy controls and individuals with other predominant clinical symptoms. Another literature review³¹ has investigated the link between ToM and symptoms of SSD by comparing groups of patients based on their ToM performance and their profile of clinical symptoms. The authors highlighted that categorizing patients into groups based on their primary clinical symptoms may lead to an overlap across dimensions resulting in indistinct categories. This increases the risk of losing valuable information concerning the relationship or interaction between different individual symptoms since people with SSD seldom manifest just 1 symptom dimension. Further, the authors suggest that a symptom-based approach could provide informative results regarding the link between ToM and clinical symptoms of SSD.

The overarching goal of the present study was to better understand the relationships between ToM and the clinical symptoms of SSD. More specifically, we aim to conduct a meta-analysis to quantify and compare the strength of the associations between ToM and each symptom dimension of SSD, as well as each individual symptom within these dimensions. An exploratory objective was to assess the potential moderating effects of various sample characteristics (eg, duration of illness) and study variables (eg, quality of the study) on the association between ToM and clinical symptoms. We first hypothesized that there would be a stronger association between ToM and the Negative^16,20,28 and Cognitive/Disorganization¹² dimensions. Secondly, we hypothesized that 1 specific symptom from the Positive dimension, namely delusions, would be associated with ToM.^22,27

Methods

Data Sources and Literature Search

Articles were retrieved on June 16, 2021, from Pubmed, PsycNet, Embase, Cochrane Library, and Science Direct. Gray literature was explored with Proquest, WorldCat, and Open Gray. Keywords for clinical symptoms, ToM, and diagnosis are presented in table 1 and an example of a search strategy is presented in supplementary table S1. This study was registered in Prospero prior to the beginning of the searches in the databases (PROSPERO 2021 CRD42021259723) and all PRISMA guidelines were followed³² (see PRISMA checklist in supplementary table S2).

Table 1.

Search Terms

Keywords
Construct	Search Terms
Clinical symptoms	(“Positive Symptoms” OR “Negative Symptoms” OR Disorgani* OR “General Symptoms” OR Hallucination OR Hallucinations OR “Perceptual Abnormality” OR delusions OR delusion OR Paranoia OR Suspiciousness OR Persecution OR “Unusual thought content” OR “Thought disorder” OR Avolition OR Anhedonia OR Alogia OR Apathy OR Anhedonia OR Asociality OR Amotivation OR “Flat Affect” OR “Blunted Affect” OR “Emotional Blunting” OR “Emotional Discomfort” OR “Emotional Withdrawal” OR “Social withdrawal” OR “Negative Affect” OR “Bizarre Behavior” OR Excitement OR “Elevated Affect” OR Mania OR Agitation OR Elation OR Grandiosity OR Passivity OR “Social Withdrawal” OR “Social Dysfunction” OR “Scale for the Assessment of Positive Symptoms” OR SAPS OR “Scale for the Assessment of Negative Symptoms” OR SANS OR “Positive and Negative Syndrome Scale” OR PANSS OR “Brief Psychiatric Rating Scale” OR BPRS OR “Scale for the Assessment of Thought Language and Communication” OR TLC OR “Negative Symptom Assessment” OR “NSA-16” OR “The Clinical Global Impression Schizophrenia Scale” OR “CGI-SCH” OR “Clinical Assessment Interview for Negative Symptoms” OR CAINS OR “Brief Negative Symptom Scale” OR BNSS)
Theory of mind	(“Social cogniti*” OR “theory of mind” OR mentalizing OR mentalizing OR mentalization OR mentalization OR “social perception” OR “social adjustment” OR “mind reading” OR “social judgment” OR “social cue” OR perspective-taking OR “cognitive empathy”)
Diagnosis	(schizophrenia OR “schizoaffective disorder” OR “Schizophrenic Disorder” OR “Schizoaffective Disorders” OR Psychotic Disorder OR Psychosis OR Psychoses OR “Schizophreniform Disorder” OR “first-episode psychosis”)

Inclusion Criteria

Studies were included if they were written in English, French, or Spanish, they involved participants with SSD (at least 85% of the sample) according to the DSM-III to 5 or to the ICD-9 or 10 and reported at least 1 correlation between ToM and a clinical symptom of SSD. Studies including clinical high-risk samples were included only if they presented data specific to the group of participants who transitioned to SSD and only data following the transition to SSD were used. The inclusion of ToM tasks was based on an operational definition used in previous meta-analyses^6,12,33,34 presented in supplementary table S3, along with the list of ToM tasks.

As presented in supplementary table S4, the classification of clinical symptoms was based on the 5 dimensions of the PANSS including Positive, Negative, Cognitive/Disorganization, Excitement/Hostility, and Depression/Anxiety.^29,35,36 This classification was chosen because it includes a wide range of clinical symptoms of SSD measured in several clinical scales and it offers a more refined distinction of symptoms that may get lost in larger symptom dimensions. To reduce the large number of individual negative symptoms retrieved in the studies, they were further subdivided into the domains proposed by Kirkpatrick et al,³⁷ namely anhedonia, avolition, asociality, alogia, and blunted affect. According to the most recent guidelines presented in Galderisi et al,³⁸ these symptoms represent 2 distinct factors: experiential negative symptoms (anhedonia, avolition, and asociality) and expressive negative symptoms (blunted affect and alogia).

Procedure

Two evaluators (E.T. and J.R.) conducted an initial screening of the studies based on the title and abstract in Covidence and a second screening based on the entire article. In case of discrepancies, the 2 evaluators met to discuss whether to include the studies and a third evaluator (D.R.-C.) was consulted if needed. The included studies were then extracted by 4 people (E.T., J.R., D.R.-C., and A.B.) in an Excel template, in which 15% of the papers extracted from each person was double checked by E.T. The details regarding the data extracted are presented in supplementary table S5. In case of missing data, authors were contacted. The same 4 people also assessed the quality of the studies using the Quality appraisal checklist quantitative studies reporting correlations and associations of the National Institute for Health and Care Excellence (NICE).³⁹ This checklist is rated on the internal, external, and total quality of the studies. The internal and external quality scales were rated on a Likert scale of 1 to 3 and the total quality was rated on a Likert scale ranging from 2 to 6.

Statistical Analyses

The meta-essentials workbooks for meta-analysis (Meta-Essentials for Correlational data 1.5)⁴⁰ were used to conduct the analyses. Each correlation between ToM and a clinical symptom was extracted and categorized based on the symptom dimension it belonged to as well as the specific symptoms, if applicable. If a study reported more than 1 correlation for the same symptom dimension, a mean score was calculated to avoid overlap. Correlations were then entered into a random-effect model using a Fisher’s r-to-z transformation and can be interpreted as follows: Zr ~ 0.10 = small; Zr ~ 0.30 = moderate; Zr ~ 0.50 = strong.⁴¹ First, an association between ToM and the overall clinical symptoms was calculated, irrespective of symptom dimensions. Then, an effect size was calculated for each symptom dimension, respectively (Positive, Negative, Cognitive/Disorganization, Excitement/Hostility, and Depression/Anxiety). Also, the association between ToM and specific symptoms within each dimension (eg, hallucination in the Positive dimension) was calculated. To determine if some symptom dimensions were more strongly associated with ToM than others, a focused-test approach⁴² was used to assess the significance between contrast of effect sizes. The same approach was used to determine whether specific symptoms within each dimension were more strongly associated with ToM than others.

Moderation analyses were conducted to assess the potential effects of demographic and clinical characteristics on the relationship between ToM and symptom dimensions. These included sex, age, age at illness onset, duration of illness, stage of illness (ie, first-episode psychosis, multiple-episode psychosis), and antipsychotic dosage. The potential effects of study quality and the type of mental states assessed in the ToM tasks were also explored. For continuous variables, Pearson correlations were used and for categorical variables, a focused-test approach was used.⁴² Publication bias was assessed using the Rosenthal failsafe-N and by examining the funnel plot as well as the standardized residual histogram.

Results

19 308 studies were retrieved from the databases and 3019 duplicates were removed. As illustrated in figure 1, 16 288 studies were assessed for the initial screening based on the title and the abstract. After this initial screening, 15 549 studies were excluded and 739 remained to be assessed for the second screening based on the entire article. Following the second screening, 609 studies were excluded (see flowchart in figure 1 for detailed reasons of exclusion) and 130 studies (137 samples) were included as presented in table 2.

Table 2.

Studies Included in the Quantitative Meta-analysis and Their Characteristics

Sample	Study	N	% Men	Age	Age at Onset	Duration of Illness	Clinical Dimension	ToM Task
1	Abdel-Hamid et al43	50	48	37.1	28.4	9.0	Pos, Dep/Anx, Cog/Dis, Neg, Exc/Host	ToM Picture Sequencing task
2	Abramowitz et al44	32	68.8	35.8	N/R	16.2	Pos, Neg, Cog/Dis	Empathic accuracy (Derntl)
3	Abu-Akel and Abushua’leh45	24	100	36.2	23.3	12.6	Neg, Pos, Total	False-belief task
4	Akgül et al46	48	52.1	39.9	23.2	17.1	Pos, Neg	RMET
5	Andrade-González et al47	53	67.9	37.4	N/R	15.7	Pos, Neg, General	MASC
6	Andrzejewska et al48	20	50	36	N/R	13	Pos, Exc/Host, Neg, Cog/Dis	RMET
7	Ayesa-Arriola et al49	160	46.3	32.2	N/A	N/R	Pos, Neg, General	RMET
8	Balogh et al50	43	55.8	40.4	27.4	13.6	Neg	RMET
9	Bambini et al51	47	61.7	39.7	24.5	15.5	Total	ToM Picture Sequencing task
10	Bardamascou52	34	97.1	36.6	N/R	N/R	Neg	RMET, Hinting
11	Bazin et al53	15	80	35.4	20.2	N/R	Total	V-SIR
12	Bell et ala,54	66	60.6	42.7	23.5	N/R	Pos, Neg, Cog/Dis, Dep/Anx	SAT-MC
13	Bell et ala,55	77	55.8	43.4	22.7	N/R	Neg	SAT-MC, Hinting
14	Bengochea-Seco et al56	43	55.8	42.3	25.6	N/R	Cog/Dis, Pos, Exc/Host, Neg, Dep/Anx	Hinting
15	Bliksted et alb,57	36	52.8	22.7	N/R	N/R	Pos, Neg	Hinting
16	Bliksted et alb,58	59	61	22.9	N/R	N/R	Neg, Pos	Animated triangles task
17	Bonfils et al59	39	61.5	44.1	N/R	N/R	Pos	False-belief task
18	Bora et al60	50	66	30.6	21.4	9.1	Pos, Neg	RMET, Hinting
19	Bourgou et al61	12	58.3	14.8	12.3	2.5	Total	Moving shape paradigm
20	Bratton et al62	27	100	37.6	N/R	10.8	Pos	TASIT-III
21	Brown et al63	45	51.1	36.2	N/R	12.2	Pos, Neg, General	Hinting, RMET
22	Brüne64	23	78.3	38.8	26.5	12.3	Pos, Neg, General, Total	Cartoon Picture Stories
23	Brüne et alc,25	38	47.4	35.9	27.4	8.8	Total, Pos, Neg, General	Cartoon Picture Stories
24	Brüne et al65	50	44	39.2	29.3	9.9	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	False-belief picture sequencing task
25	Brüne et alc,66	69	65.2	36.3	25.9	10.6	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	ToM Picture Sequencing task, MSAT
26a	Canty et al (early Sz)23	26	50	23.2	22.3	0.6	Pos, Neg, Cog/Dis	VAMA
26b	Canty et al (chronic SZ)23	33	66.7	31.6	20.9	10.9	Pos, Neg, Cog/Dis	VAMA
27	Charernboon and Patumanond67	36	38.9	37.1	N/R	8.2	Pos, Neg	RMET
28	Champagne-Lavau et al68	32	65.6	41.6	N/R	15.8	Pos, Neg, General	Irony understanding
29	Corcoran et al69	115	55.7	38.3	24.5	N/R	Pos	False-belief task
30	Couture et al70	178	63.5	45.9	N/R	N/R	Neg	Hinting
31	Craig et al71	16	68.8	31.7	N/R	N/R	Pos	Hinting, RMET
32	Csukly et al72	58	37.9	37.8	28.3	10.8	Neg, Total, Pos, General	RMET
33	Davidson et al73	48	58.3	51	21.2	N/R	Pos, Neg, Total, Cog/Dis, Dep/Anx, Exc/Host	Hinting
34	de Sousa et al74	68	52.9	38.4	N/R	N/R	Pos, Neg	RMET
35	Dorn et al75	64	54.7	37.5	24.6	14.1	Pos, Neg, Cog/Dis	Animated Triangles task
36	Dwyer76	25	64	45.3	N/R	N/R	Neg	Hinting
37	Eşsizoğlu et al77	47	48.9	39.3	N/R	N/R	Total, Pos, Neg, General,	RMET, Hinting
38	Fernández-Modamio et al78	284	60.2	45.9	23.3	22.9	Total	Hinting
39	Fiszdon et al79	119	64.7	44.9	22.7	N/R	Total, Pos, Neg, Cog/Dis, Ec/Host, Dep/Anx	Hinting
40	Fretland et al80	52	63.5	28.8	N/R	5.7	Pos, Neg, Cog/Dis	MASC
41	Frøyhaug et al81	30	60	N/R	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	Hinting
42a	García-Fernández et al (FEP)82	43	69.8	27.3	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	Hinting, RMET
42b	García-Fernández et al (chronic schizophrenia)82	44	63.6	40.2	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	Hinting, RMET
43	Giannakou et al83	40	80	37.1	26.5	10.6	Pos, Neg, General	Faux-pas test
44	Giusti et al84	20	70	36.2	N/R	11.6	Pos, Neg, General, Total	ToM stories, RMET
45	Greig et al85	128	79.7	N/R	21.9	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	Hinting
46a	Green et al (FEP)86	81	75.3	22.0	N/R	N/R	Neg, Pos	TASIT-III
46b	Green et al (chronic sz)86	53	66	34.8	N/R	N/R	Neg, Pos	TASIT-III
47	Griffiths87	98	78.6	23.6	N/R	N/R	Neg	False-belief task
48	Harenski et al88	46	100	37.6	N/R	14.1	Pos, Neg, Cog/Dis	Empathic accuracy (Derntl)
49	Herold et al89	18	61.1	28.7	25.2	3.4	Pos, Neg, General, Dep/Anx, Total	Faux-pas test
50	Ho et al90	41	56.1	27.7	N/R	1.7	Neg	Yoni task
51	Horan et al91	55	76.4	40.1	22.1	17.4	Neg	Animated triangles task
52	Johansson et al92	49	77.6	36.9	N/R	10.1	Pos, Neg	MASC
53	Jung et al93	80	55	33.4	N/R	N/R	Pos, Neg, General	MASC
54	Kalin et al94	179	65.4	42.1	N/R	N/R	Neg	RMET, TASIT-III, Hinting
55	Keats95	38	63.2	43.2	N/R	N/R	Pos, Neg, Cog/Dis, Total	Frith-Happé animations
56	Kelemen et al96	52	67.3	37.3	N/R	N/R	Neg	RMET
57	Kern et al97	50	64	34.5	N/R	N/R	Pos, Neg	TASIT-III
58	Kincaid et al98	66	78.8	45	N/R	N/R	Pos	Gardner Hinting
59	Kim et al99	25	40	22.9	N/R	N/R	Pos, Neg, General	RMET
60	Koelkebeck et al100	23	47.8	24.5	N/R	3.0	Pos	Animated triangle task
61	Konstantakopoulos et al101	58	58.6	42.3	N/R	17.2	Pos, Neg	False-belief task, Hinting, Faux-pas task
62	Lahera et al102	49	57.1	40.4	N/R	15.7	Pos, Neg, General, Total	Hinting
63	Lam et al103	58	50	40	25.9	13.4	Neg, Pos, General	Faux-pas task, RMET
64	Langdon et al104	20	45	33.4	23	10.3	Neg, Pos,	ToM Picture Sequencing task
65	Langdon et al105	25	N/R	23	N/R	9	Pos, Neg, Cog/Dis	ToM Picture Sequencing task, Strange stories
66	Langdon et al106	25	N/R	33.5	23.1	10.9	Pos, Neg	ToM Picture Sequencing task, ToM Stories
67	Langdon et al107	35	65.7	35.9	22.8	12.6	Pos	Story comprehension, Joke comprehension, false-belief task
68	Langdon et al108	23	95.7	20.9	19.9	0.9	Pos, Neg	Picture sequencing task, Joke appreciation task; Story comprehension task
69	Langdon et al109	50	58	42.4	N/R	17.5	Pos, Neg	False belief task, Happé Story comprehension, ToM Picture Sequencing task, Cartoon jokes
70	Langdon et al110	23	69.6	45.8	23.2	23.1	Pos	Animated triangle task
71	Lee and Kim111	55	49.1	41.4	28.5	12.9	Pos, Neg, General	ToM Picture Sequencing task
72	Lee et al112	60	50	33.8	22.1	12.7	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	SAT-MC
73	Li et al113	35	57.1	16.5	N/R	1.3	Total, Pos, Neg, General	Yoni task, Faux-pas test
74	Li et al114	18	44.4	28	N/R	N/R	Neg	RMET
75	Lincoln et al115	75	62.7	33.9	N/R	N/R	Neg	Movie Task of Social Situations
76	Lindgren et al116	54	55.6	26.7	N/R	N/R	Pos, Neg, Dep/Anx	Hinting
77	Lo and Siu117	62	41.9	37.9	25.9	12.0	Pos, Total	SCSQ
78a	Majorek et al (forensic)118	33	96.9	31.8	21.8	9.7	Exc/Host, Cog/Dis, Neg	Cartoon Picture Stories
78b	Majorek et al (non-forensic)118	38	47.4	35.4	27.4	8.8	Exc/Host, Cog/Dis, Neg	Cartoon Picture Stories
79	Mansueto et al119	757	75	27.66	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	Hinting
80	Martinez et ald,120	36	83.3	23.4	N/R	N/R	Cog/Dis	MASC
81	Martinez et ald,121	51	82.4	23.35	19.1	N/R	Cog/Dis	Animated triangles task
82	Martínez-Domínguez et al122	21	71.4	39.2	N/R	13.3	Pos, Neg, General	RMET, Hinting
83	Marjoram et al123	20	60	39.8	28.4	10.9	Pos, Neg, Dep/Anx, Cog/Dis	Cartoon jokes
84	Martino et al124	21	57.1	32.7	N/R	N/R	Pos, Neg, Cog/Dis	Faux-pas test
85	Mazza et al125	20	85.7	42.7	N/R	12.3	Neg	ToM stories
86a	Mazza et al (FEP)126	49	67.4	26.8	N/R	1.01	Neg	ToM stories
86b	Mazza et al (chronic SZ)126	178	69.7	34.5	N/R	10.9	Neg	ToM stories
87	Mazza et al127	35	85.7	33.9	N/R	7.1	Neg, Cog/Dis, Pos	ToM stories
88	Mehl et al128	185	42.2	38.6	N/R	15.4	Pos	Cartoon Picture Stories
89	Mike129	88	70.5	24.9	N/R	N/R	Neg	Hinting, TASIT-III
90	Mizrahi et al130	71	83.1	33.0	N/R	N/R	Pos, General, Neg, Total	Hinting
91	Montag et al131	80	58.8	39.1	29.2	9.8	Neg, Pos	MASC
92	Morozova et al132	20	65	26.1	N/R	N/R	Pos, Neg, Dep/Anx, Total	RMET
93	Morrison133	26	88.5	39.7	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host	RMET, Hinting
94	Mothersill et al134	27	74.1	41.1	N/R	N/R	Total	RMET
95	Ng et al135	193	64.3	46.2	N/R	N/R	Pos, Nef, Cog/Dis, Exc/Host, Dep/Anx	Hinting
96a	Ntouros et al (SZ + TOC)136	38	73.7	26.4	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	PESIT
96b	Ntouros et al (SZ-TOC)136	27	88.9	26.3	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	PESIT
97	Ozguven et al137	20	100	27	18.8	6.8	Pos, Neg	False-belief task
98	Park138	51	64.7	37.5	24.5	12.1	Pos	Hinting, False-belief task
99	Pentaraki et al139	21	100	24.4	19.4	N/R	Pos, Cog/Dis, Exc/Host	False-belief task
100	Pinkham et al140	92	70.7	27.8	N/R	N/R	Pos, Neg, General	Cartoon Picture Stories
101	Piovan et al141	30	63.3	45.4	29.4	16.0	Pos, Cog/Dis, Neg	ToM Picture Sequencing task
102	Popolo et al142	37	94.6	27.2	N/R	N/R	Pos, Neg, Cog/Dis, Exc/Host, Dep/Anx	Hinting, ToM Picture Sequencing task
103	Portillo143	61	78.7	47.9	21.2	N/R	Neg, Cog/Dis	TASIT-III
104	Riccardi et al144	30	70	37.8	N/R	13.9	Total, Pos, Neg, Cog/Dis	Visual jokes (Marjoram)
105	Rominger et al145	20	80	37.7	N/R	12.5	Pos, Neg	RMET
106	Roncone et al146	44	77.3	33.4	N/R	10.7	Pos, Neg, Cog/Dis	False-belief task
107	Russell et al147	61	N/R	33.9	23.5	10.5	General	Animated Triangles task
108	Sampedro et al148	101	81.2	41.6	23.3	N/R	Pos, Neg, General, Total	Strange stories
109	Savina149	84	70.2	43.8	25.2	18.5	Total	ToM Picture Sequencing task, Faux-pas test, ToM stories
110	Sayar-Akaslan et al150	26	15.4	43.7	N/R	N/R	Neg	DEToM, RMET
111	Schenkel et al151	42	59.5	41.7	N/R	N/R	Pos, Neg, Exc/Host, Dep/Anx	Hinting
112	Schiffer et al152	47	100	36.9	24.2	12.7	Exc/Host, Dep/Anx	RMET
113	Sen et al153	30	70	27.4	N/R	0.75	Pos	False-belief task
114	Shamay-Tsoory et al154	22	59.1	32.6	N/R	N/R	Neg, Cog/Dis, Pos, General,	Yoni task
115	Stratta et al155	20	85	38.5	22.7	15.2	Pos, Neg, Cog/Dis	Cartoon jokes
116	Tso et al156	33	66.7	38.5	N/R	17.9	Total, Pos, Neg	RMET
117	Uhlhaas et al157	48	70.8	38.4	20.5	N/R	Total, Cog/Dis, Dep/Anx, Exc/Host, Pos, Neg	False-belief task, Hinting, RMET
118	Urbach et al158	206	72.3	42.7	N/A	N/R	Neg, Cog/Dis, Pos, Dep/Anx, Exc/Host	V-SIR
119	Valaparla et al159	51	43.1	32.4	26.7	5.5	Total, General, Pos	False-belief stories, metaphor irony detection, faux-pas recognition
120	Varela et al160	51	82.4	35	N/R	11.4	Total, Pos, Neg, General	RMET
121	Varga et al161	21	42.9	37.9	26.8	11.9	Pos, Neg, General, Dep/Anx, Total	Experimental irony task
122	Ventura et al162	77	77.9	21.5	N/R	0.6	Pos, Neg, Cog/Dis	SAT-MC
123	Vidarsdottir et al163	70	87.1	24.1	22.2	2.13	Pos, Neg, General	Hinting
124	Vohs et al164	26	80.8	23.8	N/R	N/R	Pos, Neg, Cog/Dis	Hinting, RMET
125	Wang et al165	48	43.8	33.4	N/R	9.9	Pos, Neg, General	False-belief task
126	Whitton and Henry166	34	41.2	43.6	20.5	23.1	Pos, Neg	RMET
127a	Wójciak et al (men)167	40	100	35	24	N/R	Neg, Dep/Anx,	RMET, Strange stories, Faux-pas test
127b	Wójciak et al (women)167	40	0	41	27	N/R	Neg, Dep/Anx,	RMET, Strange stories, Faux-pas test
128	Zhang et al24	56	51.8	27.7	N/R	5.9	Pos, Neg, General, Total	Yoni task
129	Zhu et al168	40	45	30.2	N/R	9.2	Pos, Neg, General, Total	Faux-pas test
130	Ziv et al26	30	70	37.8	23.5	13.2	Pos, Neg, General, Total	False-belief task
Total		7863	66.0	35.1	23.8	13.2

Note: Cog/Dis, Cognitive/Disorganization symptoms; Dep/Anx, Depression/Anxiety symptoms; DEToM, Dokuz Eylul Theory of Mind Index; Exc/Host, Excitability/Hostility symptoms; MASC, Movie of Assessment of Social Cognition; MSAT, Mental State Attribution Task; Neg, Negative symptoms; PESIT, Perception of Social Inference Test; Pos, Positive symptoms; RMET, Reading the Mind in the Eyes Test; SAT-MC, Social Attribution Task—Multiple Choices; SCSQ, Social Cognition and Screening Questionnaire; TASIT-III, The Awareness of Social Inference Test; ToM, theory of mind; VAMA, Virtual Assessment of Mentalising Ability; V-SIR, Versailles—Situational intention reading.

^aA part of the sample of Bell et al⁵⁵ include participants from Bell et al.⁵⁴ Only the correlations with negative symptoms were reported in Bell et al.⁵⁵ We included Bell et al⁵⁴ for all other correlations with other dimensions, as well as the correlation with SANS total that was not reported in Bell et al.⁵⁵ To avoid overlap of the samples, only data from Bell et al⁵⁵ were used to calculate sociodemographic information.

^bA part of the sample of in Bliskted et al⁵⁸ include participants from Bliksted et al.⁵⁷ Both studies were, however, included as separate sample since the studies reported correlations between clinical symptoms and different ToM tasks. To avoid overlap of the samples, only data from Bliksted et al⁵⁸ were used to calculate sociodemographic information.

^cA part of the sample of Brüne et al⁶⁶ include participants from Brüne et al.²⁵ Both studies were, however, included as separate sample since the studies reported correlations between clinical symptoms and different ToM tasks. To avoid overlap of the samples, only data from Brüne et al⁶⁶ were used to calculate sociodemographic information.

^dA part of the sample of Martinez et al¹²¹ include participants from Martinez et al.¹²⁰ Both studies were, however, included as separate sample since the studies reported correlations between clinical symptoms and different ToM tasks. To avoid overlap of the samples, only data from Martinez et al¹²¹ were used to calculate sociodemographic information.

Fig. 1.

PRISMA flowchart.

Associations Between ToM and Symptom Dimensions

The effect sizes, confidence intervals, and statistics of heterogeneity regarding the associations between ToM and symptoms are presented in table 3 and are illustrated in figure 2. The overall size of the association observed between ToM and total symptoms irrespective of the specific dimension was small to moderate (Zr = 0.22) with no evidence of heterogeneity. The moderate association of the Cognitive/Disorganization dimension showed the strongest link to ToM (Zr = 0.28) compared to the Positive (X²(1) = 37.21, P < .001), Negative (X²(1) = 5.03, P = .025), Excitability/Hostility (X²(1) = 38.49, P < .001), and Depression/Anxiety (X²(1) = 54.67, P < .001) dimensions. A similar pattern was observed for the Negative dimension (Zr = 0.24), revealing a stronger association with ToM than the Positive (X²(1) = 23.03, P < .001), Excitement/Hostility (X²(1) = 25.21, P < .001), and Depression/Anxiety (X²(1) = 40.51, P < .001) dimensions. The association with the Positive dimension was small (Zr = 0.16), but showed a stronger association with ToM compared to the Depression/Anxiety dimension (X²(1) = 7.33, P = .007). The associations between ToM and the Excitability/Hostility (Zr = 0.12) and Depression/Anxiety (Zr = 0.09) dimensions were both small.

Table 3.

Effect Sizes of the Associations Between ToM and the Symptom Dimensions and the Specific Symptoms Within Each Dimension

Symptom Dimension	Effect Size (Zr)	95% Confidence Interval	95% Prediction Interval	Total Sample Size (N)	Number of Samples (k)	Heterogeneity
						I2	Q	P
Total	0.22	0.20–0.24	0.20–0.24	8069	137	0.00	115.74	.895
Positive	0.16	0.13–0.18	0.13–0.18	5894	103	0.00	85.89	.874
Delusions	0.17	0.11–0.22		1590	26	13.18	28.80	.273
Hallucinations	0.17	0.05–0.28		732	15	33.37	21.01	.101
Bizarre behavior	0.16	−0.96 to 0.98		82	2	49.66	1.99	.159
Lack of judgment	0.14	−0.11 to 0.37		299	3	0.00	1.96	.376
Unusual thought content	0.26	0.04–0.46		413	7	57.99	14.28	.027*
Negative	0.24	0.22–0.26	0.22–0.26	6628	112	0.00	108.62	.546
Experiential negative symptoms	0.23	0.17–0.28		1125	23	0.00	17.13	.756
Anhedonia	0.24	0.16–0.31		398	9	0.00	3.44	.904
Avolition	0.19	0.13–0.26		960	18	0.62	17.11	.447
Asociality	0.23	0.15–0.32		595	12	0.00	10.06	.525
Expressive negative symptoms	0.23	0.18–0.31		1098	24	11.74	26.06	.298
Blunted affect	0.23	0.15–0.31		934	19	19.41	22.34	.217
Alogia	0.25	0.15–0.35		426	12	0.00	10.09	.522
Cognitive/disorganization	0.28	0.24–0.33	0.13 to –0.43	3075	47	25.98	62.14	.056
Difficulty in abstract thinking	0.36	0.27–0.44		442	8	0.00	4.72	.695
Attention	0.25	0.13–0.37		471	9	12.02	9.09	.335
Disorganized behavior	0.20	0.00–0.38		173	4	0.00	1.92	.590
Disorientation	0.16	0.00–0.32		371	5	10.83	4.49	.344
Mannerism	0.15	−0.05 to 0.33		406	5	41.23	6.81	.147
Stereotyped thinking	0.38	0.01–0.66		345	5	81.87	22.06	<.001*
Preoccupation	0.20	−0.26 to 0.58		325	4	71.55	10.54	.014*
Conceptual disorganization	0.39	0.27–0.49		554	10	33.45	13.52	.140
Excitement/hostility	0.12	0.08–0.16	0.08–0.16	2245	25	0.00	16.48	.870
Excitement	0.09	−0.05 to 0.23		408	7	8.76	6.58	.362
Hostility	0.07	−0.05 to 0.19		485	7	4.32	6.27	.394
Poor impulse control	0.05	−0.09 to 0.19		299	3	0.00	0.66	.720
Uncooperativeness	0.07	0.03–0.12		299	3	0.00	0.07	.967
Depression/anxiety	0.09	0.05–0.13	0.05–0.13	2378	29	0.00	22.89	.739
Anxiety	0.10	−0.11 to 0.29		299	3	0.00	1.29	.524
Depression	0.18	−0.02 to 0.37		339	6	45.14	9.11	.105
Distress	0.20	−0.14 to 0.51		106	5	56.67	9.23	.056
Guilt	0.19	−0.38 to 0.65		299	3	72.99	7.40	.025*
Somatic concern	0.04	0.00–0.08		299	3	0.00	0.05	.978
Tension	0.07	−0.19 to 0.32		299	3	4.62	2.10	.350

Note: ToM, theory of mind.

*indicates a significant heterogeneity < .05.

Fig. 2.

Illustration of the associations between ToM and symptom dimensions as well as the specific symptoms within each dimension. Note: ToM, theory of mind.

Associations Between ToM and Specific Symptoms Within Each Dimension

The associations between ToM and specific symptoms within each dimension are presented in table 3 and illustrated in figure 2.

Cognitive/Disorganization Dimension

Difficulty in Abstract thinking showed a stronger relationship with ToM than Disorientation (X²(1) = 9.61, P = .002), Mannerism (X²(1) = 13.91, P < .001), Stereotyped thinking (X²(1) = 6.45, P = .011), and Preoccupation (X²(1) = 13.83, P < .001). Conceptual disorganization showed a stronger relationship than Attention (X²(1) = 4.20, P = .040), Disorientation (X²(1) = 11.12, P < .001), Mannerism (X²(1) = 16.06, P < .001), Stereotyped thinking (X²(1) = 7.49, P = .006), and Preoccupation (X²(1) = 15.73, P < .001). A significantly stronger association was observed between ToM and Attention than Mannerism (X²(1) = 3.82, P = .050) and Preoccupation (X²(1) = 4.19, P = .041). The symptom of Stereotyped thinking was significantly heterogenous which was explained by the presence of an outlier.¹⁵⁸ Once removed, there was no longer a significant difference between Stereotyped thinking and Difficulty in abstract thinking and with Conceptual disorganization. Further, after removing the outlier, there were significantly stronger associations between Stereotyped thinking and ToM than Attention (X²(1) = 5.69, P = .017), Disorganized behavior (X²(1) = 5.67, P = .017), Disorientation (X²(1) = 10.62, P = .001), Mannerism (X²(1) = 13.52, P < .001), and Preoccupation (X²(1) = 13.82, P < .001). The symptom Preoccupation was also significantly heterogenous, due to the presence of an outlier.¹³³ Once this outlier was removed, a stronger association between ToM and Stereotyped thinking compared to Preoccupation was observed (X²(1) = 17.00, P < .001).

Negative Dimension

There was no significant difference between the symptoms of this dimension when the symptoms were compared between the 5 symptoms (ie, anhedonia, avolition, asociality, blunted affect, and alogia) as well as between the 2 factors (ie, experiential and expressive). No significant heterogeneity was observed for the symptoms of the Negative dimension.

Positive Dimension

There was no significant difference between the symptoms of the Positive dimension. A significant heterogeneity was observed for the symptom Unusual thought content, which was explained by the presence of an outlier.¹³² Once removed, no significant difference emerged between the symptoms of the dimension.

Excitement/Hostility

No significant difference between the symptoms nor heterogeneity was observed for this dimension.

Depression/Anxiety

There was no significant difference between the symptoms of this dimension. A significant heterogeneity was observed for the symptom Guilt, which was explained by the presence of an outlier.⁴³ Once removed, there was still no significant difference between symptoms.

Moderation Analyses

Variables Linked to the Participants

A significant moderating effect was observed, such that there was a stronger association between ToM and overall clinical symptoms in younger participants (r = −0.230, P = .008) as well as in those with earlier age at onset of illness (r = −0.402, P = .002). No significant moderating effect was observed for the other participant variables, nor for the specific symptom dimensions.

Variables Linked to the Tasks and the Studies

Regarding the type of mental state assessed in the ToM tasks, those assessing a combination of mental states were more strongly associated with overall symptoms compared to tasks assessing only belief/knowledge (X²(1) = 4.65, P = .031). A significant effect was also observed for the Positive dimension, such that there was a stronger association between ToM and positive symptoms when the tasks assessed a combination of mental states compared to intentions (X²(1) = 8.59, P = .003) and emotions (X²(1) = 4.29, P = .038) alone. Within the Positive dimension, there was a significant effect for the symptom Delusions, such that it was more strongly associated with ToM when the task assessed emotions compared to intentions (X²(1) = 4.65, P = .031). Within the Cognitive/Disorganization dimension, a significant effect was observed for Stereotyped thinking, such that there was a stronger association with ToM when the tasks assessed emotions compared to intentions (X²(1) = 6.84, P = .009). However, this effect was no longer significant when the significant outlier was removed. Within the Negative dimension, there was only 1 significant difference for Avolition, such that it was more strongly associated with ToM when the task assessed emotions compared to intentions (X²(1) = 3.84, P = .050). No significant effect was observed within the dimensions of Excitability/Hostility and Depression/Anxiety.

Finally, study quality showed a moderating effect in the Positive dimension for the external (r = −0.221, P = .023) and total (r = −0.240, P = .013) quality. Significant moderating effects were also observed in the Depression/Anxiety dimension for the internal (r = −0.430, P = .020) and total (r = −0.495, P = .006) quality. This suggests a stronger relationship between ToM and clinical symptoms in studies of a lower quality.

Publication Bias

Regarding the association between ToM and the overall clinical symptoms, the examination of the funnel plot (see supplementary figure S1), the standardized residual histogram and the Rosenthal failsafe-N (N = 19 174) suggested no publication bias. There was no evidence for publication bias regarding the association between ToM and the Cognitive/Disorganization (supplementary figure S4; Rosenthal failsafe-N = 3449), Negative (supplementary figure S2; Rosenthal failsafe-N = 14 862) and Positive (supplementary figure S3; Rosenthal failsafe-N = 6055) dimensions. As far as the Excitability/Hostility (supplementary figure S6; Rosenthal failsafe-N = 273) and the Depression/Anxiety (supplementary figure S5; Rosenthal failsafe-N = 258) dimensions, both revealed a potential publication bias.

Discussion

The goal of this study was to better understand the relationships between ToM and the clinical symptoms of SSD. Specifically, we aimed to quantify and compare the strength of the associations between ToM and each symptom dimension. While 2 previous studies^30,31 have conducted reviews comparing groups of patients based on their ToM performance and profile of clinical symptoms, there is, to the best of our knowledge, no prior meta-analysis that has attempted to quantify the relationships between ToM and each symptoms dimension of SSD. This large meta-analysis comprising 130 studies (137 samples), provides relevant and novel insights regarding the relationships between ToM and the clinical symptoms of SSD. A small to moderate overall association was observed between ToM and total symptoms as well as for the Cognitive/Disorganization and the Negative dimensions. Further, in line with our initial hypothesis, the symptom dimension most strongly related with ToM was the Cognitive/Disorganization dimension in comparison to all other dimensions, followed by the Negative dimension, in comparison to all the other domains (with the exception Cognitive/Disorganization). A small association was also observed between ToM and the Positive, Excitability/Hostility and Depression/Anxiety dimensions, with the Positive dimension showing a stronger association with ToM compared to the Depression/Anxiety dimension.

Examining the associations between ToM and the 5 dimensions of clinical symptoms provides new insights regarding how clinical symptoms are linked with ToM. These results also highlight the benefits of using a more nuanced approach to delineate the links between clinical symptoms and ToM. Further, this thorough examination provides a better understanding of how specific symptoms within these dimensions correlate with ToM. For instance, several previous studies have demonstrated a link between ToM and Positive symptoms and put forth theories suggesting that ToM impairments contribute to the development of specific symptoms such as delusions.^22,27 Contrary to our initial hypothesis, the results of the current meta-analysis did not reveal a stronger association between ToM and delusions in the Positive dimension. While a small to moderate association was observed, it is possible that the strength of the relationship with the Positive dimension was not as significant as expected due to the classification of symptoms used in this study. For instance, the original 3-factor PANSS includes Conceptual disorganization in the positive symptoms, while it was reclassified as a Cognitive/Disorganization symptom in the most recent factorial analyses suggesting a 5-factor solution.

The results of this meta-analysis support the hypothesis of a stronger association between the Cognitive/Disorganization and the Negative dimensions and ToM compared to the other dimensions of clinical symptoms. It has been recognized in the literature that ToM, cognitive impairments and negative symptoms are important predictors of the functioning difficulties in SSD.^1,35,48 Thus, these stronger relationships have clinical implications such that these dimensions of symptoms may be targeted in treating ToM impairments. On the other hand, ToM and clinical symptoms could be improved through psychotherapeutic approaches such as the mentalization-based treatment for psychotic disorder (MBTp)¹⁶⁹ or cognitive remediation therapy¹⁷⁰ that have been shown to be efficacious in improving ToM, clinical symptoms and functioning in SSD.

ToM and the Cognitive/Disorganization Dimension

The strongest association between clinical symptoms and ToM was with the Cognitive/Disorganization dimension. Additionally, significant differences emerged regarding the strength of the association between ToM and the individual symptoms of this dimension. There was a stronger association with Stereotyped thinking than several other Cognitive/Disorganization symptoms, but this was only true when the outlier of this symptom was removed. The stronger association between ToM and the Cognitive/Disorganization domain seemed particularly driven by the symptoms Difficulty in abstract thinking and Conceptual disorganization.

In a previous meta-analysis investigating the associations between ToM and neurocognition, abstraction was the most strongly associated aspect of executive functioning to ToM in SSD.¹² Abstraction can be defined as the process of formulating general concepts by abstracting common properties of instances.¹⁷¹ This stronger relationship with abstraction in the current meta-analysis supports the hypothesis that while a mental state can never be directly observed, it can be derived from an abstract emulation of what is observed and what is known.¹⁷² Thus, ToM could act as a social form of abstraction, allowing the integration of perceived and retrieved information to interpret and predict other people’s mental states. The results of the current meta-analysis suggest that difficulties in abstract thinking could be an important treatment target to improve ToM deficits in SSD (eg, using concrete material in cognitive remediation therapy for ToM impairments to diminish the impact of abstraction difficulties).

Our meta-analysis also revealed a stronger association between ToM and Conceptual disorganization compared to other Cognitive/Disorganization symptoms. This symptom refers to the difficulty for individuals with SSD to communicate their thoughts adequately, which can lead to manifestations such as derailment, incoherence, or tangentiality.¹⁰⁵ As suggested by Langdon et al,¹⁰⁵ ToM impairments could diminish patients’ ability to use referents adequately in their speech due to the difficulty to monitor and adjust their speech. Thus, ToM may not only impact the understanding of other people’s mental states, but also influence the ability to adjust social interactions accordingly.

ToM and the Negative Dimension

The current meta-analysis revealed a significantly stronger association between ToM and the Negative dimension compared to the other symptom dimensions, with the exception of Cognitive/Disorganization.

Several previous studies have reported a link between negative symptoms and ToM in SSD.^{48,55,173,174} As suggested by Pelletier-Baldelli and Holt,²⁰ negative symptoms might emerge specifically, or at least become more prominent in social settings, and social cognitive impairments, including ToM deficits, might manifest behaviorally as negative symptoms in these settings. While the link between negative symptoms and ToM has been established in previous studies and in the current meta-analysis, the mechanisms behind this association remain poorly understood. Pelletier-Baldelli and Holt²⁰ proposed that changes in sensorimotor processing in SSD could affect social cognition broadly and in turn support the development of negative symptoms. On the other hand, they also proposed that specific domains of social cognition might be linked to specific negative symptoms. For example, in Andrzejewska et al,⁴⁸ ToM performance was more strongly related to blunted affect and alogia. It is possible that the expressive negative symptoms may reflect an underlying disorder of communication caused by a failure to understand that intonation, gestures, and expression confer mental states.^21,175 Interestingly, the small to moderate association between the Negative dimension and ToM observed in the current meta-analysis was not significantly driven by any specific negative symptoms. Further, no significant difference was observed in the associations between ToM and the 2 factors of negative symptom (ie, experiential and expressive). This suggests that when considered as a unique domain, ToM deficit may be equally associated to all negative symptoms, however this may not be the case when considering specific mental states (eg, intention vs belief or emotions). It is important to recognize, however, that the measures used to assess negative symptoms are heterogenous, such that there are a wide range of different scales all assessing different aspects of negative symptoms. The use of momentary ecological assessment for negative symptoms and ToM tasks with good psychometric properties (eg, ecological validity) could help better delineate the mechanisms underlying the links between these 2 variables.

Moderator Analyses

The current meta-analysis explored the effects of different variables on the associations between ToM and clinical symptoms. Regarding the participant variables, a stronger relationship between ToM and overall clinical symptoms was observed in younger participants, as well as in those with an earlier age of onset of the illness. ToM capacities have been demonstrated to continue to develop from adolescence to young adulthood^176,177 but has been found to be blunted in youth who exhibit early expressions of clinical symptoms. For instance, Davidson et al¹⁷⁸ found that adolescents who were at clinical high risk for psychosis showed a poorer trajectory in ToM development, compared to healthy controls. Similarly, Piskulic et al¹⁷⁹ showed that individuals at clinical high risk for psychosis who did not transition to SSD exhibited significant improvement in ToM compared to those who transitioned. Interestingly, other authors have demonstrated that ToM abilities may protect against psychotic symptom formation and severity.^128,180 Further, previous studies have demonstrated better social cognitive performance in patients with schizophrenia with a later age at onset of illness.¹⁸¹ Thus, ToM abilities might be less efficient in younger patients and those with an earlier age at illness onset. Given the negative impact of ToM impairments on functioning and recovery in SSD,^1,6 these results highlight the importance of early interventions focusing on ToM such as cognitive remediation therapy, along with usual treatment for clinical symptoms.

Previous studies have suggested that the inference of certain types of mental states could influence the relationship between ToM and the clinical symptoms of SSD, such as a stronger link between delusions and ToM when the task assesses intentions compared to other mental states such as emotions.²² These previous studies have investigated this link by assessing specific types of delusions such as persecutory delusion, which could explain this particular finding. The results of the current meta-analysis, however, suggest an opposite effect, such that a stronger relationship was found between ToM and delusions when the tasks assessed emotions compared to intentions. Since all types of delusions were combined into a single category in the current study, this might have prevented us from observing these specific effects reported in prior studies. Further, in the current meta-analysis, the results indicate a stronger relationship between ToM and overall symptoms when using tasks assessing a combination of different mental states compared to the inference of beliefs only. Similarly, a stronger association with ToM was observed for the Positive dimension when the tasks assessed a combination of mental states compared to intentions or emotions alone. Altogether, these results suggest that the choice of ToM task and the type of mental states assessed has the potential to influence the observed relationships between ToM and the clinical symptoms of SSD.

Finally, the quality of studies was found to significantly moderate various associations in this meta-analysis. Regarding the Positive dimension, a stronger relationship with ToM was observed for studies with poorer external and total quality. Regarding the Depression/Anxiety dimension, a stronger association with ToM was observed for studies with a poorer internal and total quality. These results suggest that studies of a poorer quality might overestimate the strength of the relationships between ToM and clinical symptoms and thus require careful interpretation. This is particularly the case for the Depression/Anxiety dimension in which a potential publication bias was identified.

Limitations

Firstly, the Cognitive/Disorganization domain nearly reached significant heterogeneity despite the absence of an outlier, which suggests that different symptoms may be unequally contributing to this domain. Further, in some dimensions, several individual symptoms (ie, unusual thought content in the Positive dimension, stereotyped thinking and preoccupation in the Cognitive/Disorganization dimension and guilt in the Depression/Anxiety dimension) reached significant heterogeneity, which requires careful interpretation. This could be explained by the limited number of studies that included these particular individual symptoms. In line with this, it was not possible to conduct moderation analyses for all individual symptoms due to insufficient studies including some specific symptoms. Another limitation is the potential publication bias observed for the domains of Depression/Anxiety and Excitability/Hostility. There were fewer studies reporting correlations within these symptom domains which might reflect the relative recency of the 5-factor analysis. Further, it is worth noting that we received very few answers from authors contacted for missing information, which could add potential bias to the results of the meta-analysis. In addition, some studies only reported correlations with clusters of symptoms and not individual symptoms. For instance, some studies reported correlations between ToM and the 3-factor classification of the PANSS. In these cases, it was not possible to divide the individual symptoms into the appropriate 5-factor domains. Another potential limitation is the large heterogeneity of ToM tasks included in this meta-analysis that could influence the relationship between ToM and the clinical symptoms. Finally, it was not possible to assess the qualitative relationship between certain profiles of ToM (eg, hypermentalizing) and the clinical symptoms given the very few studies reporting this information.

Conclusion

This meta-analysis provided a novel attempt at quantifying and comparing the relationships between ToM and the clinical symptoms of SSD. This study highlights ToM’s stronger relationship with the Negative and the Cognitive/Disorganization dimensions. ToM impairments, negative and cognitive symptoms have been consistently associated with functioning difficulties, lower quality of life and difficulty in recovery in SSD. A better understanding of the mechanisms underlying the links between ToM and these 2 dimensions is warranted in future research so that more effective treatments may be developed, targeting these mechanisms in SSD.

Funding

E.T. was supported by a postdoctoral fellowship from the Canadian Institutes of Health Research (171198). M.L. is supported by a James McGill Professorship from McGill University.

Acknowledgments

M.L. reports grants from Roche, Otsuka Lundbeck Alliance, personal fees from Otsuka Canada, personal fees from Lundbeck Canada, grants and personal fees from Janssen, and personal fees from MedAvante-Prophase, outside the submitted work. Template data collection can be shared upon request. The authors report no conflict of interest for this study.