Relationships between self-reflectiveness and clinical symptoms in individuals during pre-morbid and early clinical stages of psychosis

Procedure

This study was conducted at the Shanghai Mental Health Center and its affiliate, the Shanghai Psychotherapy and Psychological Counselling Center, with the recruitment of subjects from February 2019 to May 2021. The inclusion criteria for participants were individuals with APS or patients with FEP aged between 11 and 45 years who had completed at least 6 years of primary education. The exclusion criteria included a history of psychotic disorder, severe somatic diseases, mental retardation (intelligence quotient (IQ)<70), dementia, ingestion of antipsychotic medication for more than 2 weeks, and the exhibition of positive symptoms induced by other mental disorders or psychoactive substances. The study was carried out according to the tenets of the Declaration of Helsinki. All the participants, as well as the next of kin or legal guardian for those younger than 18 years, signed written informed consent forms.

Participants

The study enrolled 185 subjects who had been referred by psychiatrists to the Shanghai At Risk for Psychosis-extended (SHARP-extended) programme. They were preliminarily screened and interviewed by trained raters using the general interview section of the Structured Interview for Psychosis-risk Syndromes (SIPS). They were also assessed with the five positive symptom items of the Scale of Psychotic-risk Syndromes (SOPS) and the Global Assessment of Functioning (GAF) Scale. Four enrollees were excluded, three because of taking antipsychotic medications for longer than two weeks, and one for bipolar disorder symptoms. One hundred and twenty-two subjects met the criteria of APS with or without the genetic risk and deterioration syndrome, and 59 subjects met the criteria for FEP with the presence of psychotic symptoms. The APS and FEP groups were combined for the analysis. All participants were classified into three groups according to the quartile of self-reflectiveness: low level (≤11 scores of self-reflectiveness, LSR, n=59); medium level (12–16 scores of self-reflectiveness, MSR, n=67) and high level (≥17 scores of self-reflectiveness, HSR, n=55). See the flowchart of subject recruitment in figure 1.

Request permission

Measures

Statistical analysis

SPSS V26.0 was used to conduct the analysis. A one-way analysis of variance (ANOVA) was conducted to compare demographic continuum variables among the LSR, MSR, and HSR groups. A χ² test was conducted to compare the differences in the proportions of categorical variables among different groups. Adjusted standardised residuals (ASR) were calculated for contingency tables to analyse the contribution of different cells to the significance of the χ² test. For ASR>1.96, statistical significance was set at p<0.05 and, for ASR>2.58, at p<0.01. Multivariate analysis of covariance (MANCOVA) was used to compare clinical variables controlling for covariates. Multiple comparisons were adjusted using Bonferroni’s method. Binary logistic regression analysis was conducted to differentiate among the LSR, MSR, and HSR groups. Receiver operating characteristic (ROC) analysis was conducted, and the area under the curve (AUC) was calculated. Linear relationships in each group and non-linear relationships in total sample of self-reflectiveness with clinical insight and clinical symptoms were explored using linear and non-linear regression analysis. The level of significance was set at 0.05 (two-tailed).

Demographics and clinical characteristics

The one-way ANOVA revealed significant differences in the father's education among subjects with different levels of self-reflectiveness (F=4.662, p=0.011). The post hoc analysis (Bonferroni correction) showed that the father's education in the LSR group was significantly lower than in the MSR group (p=0.008). APS individuals and patients with FEP were distributed differently in low and medium levels of self-reflectiveness (χ²=7.904, p=0.019). A larger proportion of APS individuals were distributed in the MSR group (ASR=2.2, p<0.05), while a larger proportion of patients with FEP were distributed in the LSR group (ASR=2.6, p<0.01) (see table 1).

The categorical variable of diagnoses (APS and FEP) was transformed into a dummy variable (APS=0, FEP=1) and then added into the covariates, along with age, education and gender, when MANCOVA was conducted. The results showed a significant main effect of groups on each dimension of clinical insight (p<0.05). Pairwise comparisons adjusted using the Bonferroni method revealed that compared to the MSR and HSR groups, the LSR group had poorer awareness of illness (p=0.002, p=0.021, respectively) and poorer total clinical insight (p<0.001, p=0.026, respectively). The MSR and HSR groups demonstrated no significant differences in clinical insight. In terms of relabelling of specific symptoms and treatment compliance, the LSR group scored lower than the MSR group (p=0.008, p=0.011, respectively).

Comparison of the five PANSS factors among the groups

Controlling for age, education, gender and diagnoses, MANCOVA revealed that there were significant main effects (p<0.05) of groups on the factors of disorganisation, affect, and resistance, and the items of N1 (blunted affect), N5 (difficulty in abstract thinking), N7 (stereotyped thinking), G13 (disturbance of volition), G2 (anxiety), G3 (guilt feelings), and G14 (poor impulse control) (see table 2). Multiple comparisons adjusted using the Bonferroni method revealed that compared to the MSR group, the LSR group got higher scores on stereotyped thinking (p=0.043), poor impulse control (p=0.020), and the resistance factor (p=0.019); however, compared to the HSR group, the LSR group got lower scores on blunted affect (p=0.037) and disturbance of volition (p=0.020). Compared to the MSR and HSR groups, the LSR group also had lower scores on anxiety (p=0.017, p=0.006, respectively), guilt feelings (p=0.028, p=0.003, respectively), and the affect factor (p=0.019, p=0.006, respectively). The MSR and HSR groups had no significant differences. In addition, the MSR group showed a lower score on stereotyped thinking than the HSR group (p=0.039), demonstrating no significant difference with the LSR group.

Binary logistic regression analysis for differentiating different groups

LSR vs MSR

Binary logistic regression analysis, controlling for age, education, gender and diagnoses revealed that awareness of illness, stereotyped thinking, anxiety, and poor impulse control entered the model for differentiating between the LSR and MSR groups (omnibus tests of model coefficients: χ²=36.727, p<0.001) (see figure 2A). Higher levels of stereotyped thinking and poor impulse control were associated with lower levels of self-reflectiveness. However, individuals with lower levels of self-reflectiveness demonstrated lower levels of anxiety. As expected, self-reflectiveness was positively associated with awareness of illness. ROC analysis showed that the model differentiated between the LSR and MSR groups by 79.9% AUC (95% CI: 0.722 to 0.876).

Request permission

Figure 2

Forest plot and receiver operating characteristic curve of binary logistic regression differentiating individuals with different levels of self-reflectiveness. LSR, low scores on self-reflectiveness; MSR, medium scores on self-reflectiveness; HSR, high scores on self-reflectiveness; N1, blunted affect; N7, stereotyped thinking; G2, anxiety; G3, guilt feelings; G14, poor impulse control; SAI, Schedule of Assessment of Insight.

Non-linear relationship between self-reflectiveness and symptoms

Linear and non-linear regression analysis were conducted to explore the linear relationships in each group and non-linear relationships in the total sample of self-reflectiveness with awareness of illness, blunted affect, stereotyped thinking, anxiety, guilt feelings, and poor impulse control (see figure 3). From the low to medium level of self-reflectiveness, the trend line for each clinical variable was relatively consistent, although it did not reach the significance level for some variables at a low or medium level. However, at a high level, all variables reached a relative plateau, except for stereotyped thinking (which was opposite to the previous direction and elevated (F=3.748, p=0.058)) and blunted affect (which was slightly raised (F=1.248, p=0.269)). Although a significant difference in blunted affect existed between the LSR and HSR groups, the linear regression in each group and the non-linear regression in the total sample did not reach the level of significance (p>0.05). In the non-linear regression with statistical significance, the relationships of self-reflectiveness with anxiety and guilt feeling tended to be straight; the relationships of self-reflectiveness with stereotyped thinking and poor impulse control tended to be U-shaped; and the relationship of self-reflectiveness with awareness of illness tended to be inverse S-shaped.

Request permission

Main findings

Based on the self-reflectiveness paradox and the inconsistent results among different groups reported in previous studies, we examined whether self-reflectiveness presents non-linear relationships with clinical symptoms. The results showed that awareness of illness improved with the increase of self-reflectiveness from the low to medium level but reached a plateau when self-reflectiveness reached a high level. Anxiety and guilt feelings generally decreased with the increase of self-reflectiveness, although they demonstrated a plateau or decreased slightly at a high level of self-reflectiveness. Stereotyped thinking and poor impulse control decreased with the increase of self-reflectiveness from a low to medium level but demonstrated a reverse increase at a high level of self-reflectiveness, especially stereotyped thinking which showed a marked rise.

Regarding the differences in self-reflectiveness between APS individuals and patients with FEP, this study found the two categories of subjects distributed differently among the low, medium, and high levels of self-reflectiveness, demonstrating that APS individuals accounted for more subjects in the MSR group, while patients with FEP accounted for more subjects in the LSR group. There was no difference in the proportion of subjects in the HSR group between the two categories. This result indicates that a medium level of self-reflectiveness may be desirable, while low and high levels of self-reflectiveness may have different negative effects. Compared with patients with FEP, APS individuals not only had milder symptoms but also exhibited less functional impairment,24 and even remained relatively intact in some aspects of their functionality.25

Previous studies consistently reported that self-reflectiveness was closely related to clinical insight26 27 and emotional symptoms, such as depression and anxiety,3 6 which was further reinforced by this study. However, although there were significant differences in each dimension of clinical insight among the three levels of self-reflectiveness, logistic regression analysis revealed that only awareness of illness entered the regression model for distinguishing the LSR from the MSR and HSR groups. The results indicate that self-reflectiveness may play an indirect role in relabeling specific symptoms and treatment compliance through awareness of illness. Lui et al 26 proposed that cognitive insight complements clinical insight in affecting medication adherence in patients with schizophrenia. In addition, we found that anxiety, guilt feelings, and clinical insight, as well as blunted affect, entered the regression model for differentiating the LSR from the MSR or HSR groups. Our previous study also found that self-reflectiveness positively correlated with negative symptoms.28 However, this study further selected self-reflectiveness as the independent variable and blunted affect as the dependent variable to conduct regression analysis and found that it did not reach significance. Blunted affect refers to diminished emotional responsiveness, characterised by a reduction in facial expression, modulation of feelings, and communicative gestures. Jeganathan and Breakspear29 proposed a computational framework for emotional expression based on active inference and stated that blunted affect derives from uncertainty about whether one’s emotional expression is effective in responding to others’ affective responses. It is possible that blunted affect influenced self-reflectiveness. When a person has weak inferences about the validity of emotional responses, their self-reflectiveness might be reduced.

Stereotyped thinking and poor impulse control presented a U-shaped curve with self-reflectiveness. Poor impulse control refers to disordered regulation and control of action on inner urges, resulting in sudden, unmodulated, arbitrary, or misdirected discharge of tension and emotions without concern for consequences. Individuals with a low level of self-reflectiveness lack introspection and openness to external feedback,13 indicating that they may not be able to observe their own internal emotional impulses, while conversely lacking tolerance for external stimuli. They may, therefore, be more impulsive. However, it is difficult to understand that the HSR group also scored higher on poor impulse control, although there was no significant difference between the HSR and MSR groups. In addition, stereotyped thinking refers to decreased fluidity, spontaneity, and flexibility of thinking. The subjects may refuse to accept another idea or have difficulty in switching from one idea to another, which is understandable in the LSR group, but contrary to the concept of self-reflectiveness in the HSR group. García-Mieres et al 3 reported that although self-reflectiveness was positively correlated with depression, it also reduced the degree of psychological distress associated with depressive mood. It remains to be investigated whether higher self-reflectiveness correlated with higher poor impulse control and whether stereotyped thinking has a special role to play as García-Mieres et al have reported.

Limitations

This study has several limitations. First, this study assumes a continuous spectrum from the varying severity of CHR to the first episode of psychotic symptoms, so we included both CHR and FEP to analyse the three groups at different levels of self-reflectiveness. Although the categorical variable of diagnoses was selected as a covariate in the statistical analysis, there may be qualitative differences between the CHR individuals and patients with FEP, therefore, the differences in different groups at different levels of self-reflectiveness can be further explored after the subsequent expansion of the sample size. Second, this study did not investigate psychological stress or distress associated with symptoms and so failed to further explore the regulatory role of self-reflectiveness demonstrating positive or U-shaped correlations with symptoms, as previous researchers have done. Third, the transition rate of APS individuals to psychosis at 24 months was reported to be 19%.30 Those who did not convert to psychosis may ultimately have a different diagnosis,31 such as mood disorders, and may score relatively high on self-reflectiveness.1 Stereotyped thinking and poor impulse control have been reported to be common and perhaps even more pronounced in mood disorders.32 Therefore, whether higher scores on these items in the HSR group are related to the final diagnosis needs to be clarified by further follow-up studies. The follow-up results of some subjects could not explain this issue because the information from the subjects in this study was not completely collected until May 2022.

Implications

In summary, self-reflectiveness is beneficial to the maintenance of clinical insight but may also have a positive linear correlation with anxiety and guilt feelings, which supports the notion of the self-reflectiveness paradox. Additionally, self-reflectiveness may present U-shaped relationships with stereotyped thinking and poor impulse control, meaning that moderate levels of self-reflectiveness may be more beneficial. Therefore, the U-shaped relationship between self-reflectiveness and some variables should be considered when the causal associations between self-reflectiveness and clinical symptoms are explored. Conversely, changes in some relevant clinical symptoms should be noted during interventions to improve self-reflectiveness.