Relations between changes in safety behavior, paranoid ideations, cognitive biases, and clinical characteristics of patients with a psychotic disorder over time

Background: Safety behaviors, both positive (maladaptive coping behavior) and negative (avoidance behavior), are used by people with paranoid delusions to avoid perceived threats. Safety behaviors contribute to the persistence of paranoid delusions by preventing disconfirmation of threat beliefs and may influence other psychiatric symptoms. This study investigated how changes in safety behaviors are related to changes in paranoid ideation, social anxiety, depression, cognitive biases and self-esteem over time. Methods: This study included 116 patients diagnosed with a psychotic disorder (DSM-IV) and at least moderate levels of paranoid ideations (GTPS > 40).The data were collected as part of a multi-center randomized controlled trial where patients were randomized to VR-CBT ( n = 58) or treatment as usual (TAU; n = 58). Assessments were completed at baseline (T0), after three months (T3) and after six months (T6). For all variables, change scores between T0 and T3 and T3 and T6 were calculated and Pearson correlations between change scores were computed. Results: A decrease in total safety behavior was related to diminished paranoid ideation, social anxiety, and depression. No significant temporal associations were found between changes in safety behavior and changes in cognitive biases and self-esteem. Similar but less robust results were found for respectively negative safety behavior and positive safety behavior. Conclusion: Dropping safety behavior can be specifically targeted in behavioral interventions. Whereas there appears to be a relation with reduction in anxiety, paranoia, and depressive symptoms, changes in safety behavior do not seem to align with changes in cognitive


Introduction
Individuals who feel threatened carry out actions to prevent feared catastrophes from occurring, so called 'safety behaviors' (Salkovskis, 1991).Different types of safety behaviors have been identified, such as avoidance, in-situation, escape, compliance, help-seeking, and aggression, which can be divided into two global categories (Freeman et al., 2007).The most common category of safety behavior is avoidance behavior (i.e.negative safety behavior (Freeman et al., 2007)) to prevent a perceived potentially threatening situation from occurring.On the other hand, once individuals are in a perceived potentially threatening situation they may display maladaptive coping behavior as an active effort to reduce the threat (i.e.positive safety behavior (Freeman et al., 2007)).The concept of safety behaviors was originally developed in cognitive accounts of anxiety disorders (Salkovskis, 1991), and it has been applied to paranoia in psychotic disorders as well.Nearly all (96 %) people with paranoid delusions use safety behaviors (Freeman, 2016), of which avoidance is the most common type (Freeman et al., 2007;Freeman et al., 2001a).
According to the cognitive model of paranoid ideation (Fig. 1) (Freeman, 2016;Freeman et al., 2002), paranoia is actually maintained by safety behaviors.Positive safety behavior brings short-term relief, since it provides the feeling of having escaped the feared catastrophe, but the cognitions on the event do not change.Complete avoidance (i.e.negative safety behavior) of the threatening situation, prevents the opportunity to perceive and process evidence that contradicts the paranoid ideation (Freeman et al., 2007).
In the treatment of paranoia, dropping safety behavior is a treatment target in order to reduce negative appraisal and fear by experiencing no bad consequences in perceived threatening situations.Most Cognitive Behavioral Therapy (CBT) protocols for paranoid ideations, start with cognitive interventions in preparation of behavioral interventions.Recently, Virtual Reality-Cognitive Behavioral Therapy (VR-CBT) has been introduced as an effective method to immediately target safety behavior in patients with paranoid ideations, lowering the threshold to engage in behavioral experiments and exposure exercises (Pot-Kolder et al., 2018).Since safety behavior is highly prevalent in patients with paranoid delusions, maintains paranoia and is an important treatment target, insight into interrelated factors and different types of safety behavior is needed to increase understanding of psychosis and improve treatment strategies.
Several cross-sectional studies as well as the cognitive model of paranoia have suggested associations between safety behaviors and other factors related to paranoia in psychosis.Safety behavior has been related to higher levels of anxiety and depression, lower self-esteem and internalized stigma (Freeman et al., 2007;Upthegrove et al., 2014;Cooke et al., 2007;Yanos et al., 2008).Cognitive biases such as belief inflexibility (i.e.holding onto beliefs despite disconfirming evidence), and jumping to conclusions (i.e.making assumptions without sufficient evidence), and social cognition problems (i.e.impairments in understanding, interpreting, and responding to social cues and information) appear linked to the use of safety behavior.Cognitive biases and social cognition problems may contribute to the formation of paranoid beliefs, and the use of safety behaviors as a strategy to cope with anxiety, and may prevent challenging or disconfirming these beliefs.Indeed, crosssectional research has shown positive associations between belief conviction, cognitive biases and safety behavior (Moritz and Van Quaquebeke, 2014;Freeman et al., 2014;Birchwood et al., 2005).A cognitive model of persecutory delusions (Freeman et al., 2002) suggests a relationship between cognitive biases and safety behavior as well.
In addition to general safety behavior, a few cross-sectional studies have differentiated between positive and negative safety behavior in relation to factors associated with paranoid delusions in psychosis.In particular, the negative safety behavior of avoidance rather than other types of safety behavior has been related to emotional distress and higher levels of anxiety (Freeman et al., 2007;Cooke et al., 2007;Birchwood et al., 2005;Freeman et al., 2001b), internalized stigma (Yanos et al., 2008) and poor self-esteem (Freeman et al., 2001a;Cooke et al., 2007;Yanos et al., 2008).Poor self-esteem has been related to greater use of the safety behavior compliance (Freeman et al., 2001a) (one form of positive safety behavior), but not with other forms of safety behavior.Moreover, longitudinal research found that reductions of avoidance behavior specifically were associated with enhanced selfefficacy (Strous et al., 2005).
While previous studies have investigated relations between safety behavior and paranoid ideations, depression, anxiety, cognitive biases and self-esteem by means of cross-sectional designs, it remains unclear whether and how changes in safety behavior over time relate to changes in the other factors.Therefore, this longitudinal study of patients with a psychotic disorder who received treatment for paranoid ideations aimed to investigate the temporal relationships between changes in safety behaviors and changes in paranoia, social interaction anxiety, depression, cognitive biases, and self-esteem.Based on cross-sectional studies, we hypothesized that a reduction of safety behavior is related to a decrease of paranoid ideation, social anxiety, depression, and an increase in cognitive biases and self-esteem.

Participants and procedure
This study is a secondary analysis of data from the randomized controlled trial (RCT) 'Virtual Reality exposure therapy for psychosis'(VRET.P), which was designed to compare VR-CBT plus treatment as usual (TAU) to TAU (Moritz and Van Quaquebeke, 2014).Assessments took place at baseline (T0), three months later (T3) and six months later (T6).A total of 116 participants were included in the RCT.The VR-CBT intervention consisted of 16 individual sessions of which 40 min comprised exercises in VR.An individualized case formulation guided the exposures in VR to elicit anxiety and paranoid ideations.During VR sessions participants were guided to explore and challenge paranoid thoughts, drop safety behavior and test harm expectancies.
Inclusion criteria were a DSM-IV diagnosis of a psychotic disorder; avoidance of either shops, streets, public transport, or bars or restaurants; paranoid ideation in the past month (defined as a score >40 on the Green et al.Paranoid Thoughts Scale (Weisman and Rodebaugh, 2017)); and age 18-65 years.Exclusion criteria were an estimated IQ of 70 or lower; insufficient mastery of the Dutch language; and a history of epilepsy.The protocol was approved by the medical ethical committee of VU University Medical Center Amsterdam (METC number NL37356.058.12).From all participants written informed consent was obtained.

Safety behaviors
Safety behaviors were assessed by The Safety Behaviors Questionnaire persecutory delusions (SBQ) (Ross et al., 2015).Total SBQ scores were reported (which are the sums of the frequencies of each of the safety behaviors).Seven subscale scores of safety behaviors are distinguished: avoidance, in-situation, escape, compliance, help-seeking behavior, aggression and delusional.Consistent with Freeman et al. (2007), safety scores are divided into negative safety behavior, and positive safety behavior.Negative safety behavior was based on that avoidance subscale.Positive safety behavior was calculated by the total score of the other subscales, which represent active efforts to reduce the threat.The SBQ has a high inter-rater reliability, adequate test-retest reliability, and adequate validity (Freeman et al., 2002).

Paranoid ideation
Paranoia symptoms were assessed with the Green Paranoid Thoughts Scale (GPTS) (Weisman and Rodebaugh, 2017).The GPTS is a self-report questionnaire that consists of two subscales each including sixteen items: part A measures paranoid delusions of social reference and part B measures social persecution in the past month on a five-point Likert scale.Both scales have good internal consistency and validity (Weisman and Rodebaugh, 2017).

Anxiety
Interaction anxiety symptoms were assessed with the Social Interaction Anxiety Scale (SIAS) (Berkhof et al., 2021).The SIAS consists of 19 items that assess the tendency to fear and avoid social situations.High levels of internal consistency and test-retest reliability have been established for both scales (Berkhof et al., 2021;Mattick and Clarke, 1998).

Depression
Depression was assessed by the Beck Depression Inventory -II (BDI-II) (Roelofs et al., 2012) which is a self-report measure of 21 items assessing symptoms and level of depression over the past two weeks.Psychometric properties of IDS-SR are satisfactory and the instrument has been recommended in research (Roelofs et al., 2012).

Beliefs
The Davos Assessment of Cognitive Biases Scale (DACOBS) (Van der Gaag et al., 2013) is a self-report measure to assess cognitive biases in psychosis on seven independent subscales (jumping to conclusions, belief inflexibility bias, attention for threat bias, external attribution bias, social cognition problems, subjective cognitive problems and safety behavior) in the past two weeks on a seven-point Likert scale.In this study, we only used the subscales jumping to conclusions, belief inflexibility, and social cognition problems.These three subscales have shown to have good reliability and jumping to conclusions and inflexibility bias have acceptable.The social cognition problem subscale has shown mixed concurrent and divergent validation and should be interpreted with more caution (Pugliese et al., 2022).

Self-esteem
Self-esteem was assessed with the Brief Core Schema Scales (BCSS) (Fowler et al., 2006) which consists of 24 items measuring beliefs about the self and others on a 5-point rating scale.As a measure of self-esteem, the positive and negative core schemas about the self were used in this study.The BCSS has good psychometric properties including construct validity (Fowler et al., 2006).

Statistical analyses
The data analysis was performed according to the intention-to-treat principle.Descriptive statistics were calculated for each of the assessment points (T0, T3, T6).To provide detailed information on changes in safety-seeking behavior, (sub)scores were described and change scores for each included individual were plotted in a graph.Change scores were calculated for all variables (T0-T3 and T3-T6).Pearson correlations between change in safety behavior (total scores and the subscales) and all other outcome variables were computed.If assumptions were violated, non-parametric Spearman correlation analyses were performed.A two-sided p-value <0.05 was considered statistically significant.

Participants
In total 116 participants completed the baseline assessment, 103 completed the post-treatment assessment, and 99 also completed the follow-up assessment.The sample was predominantly male (71 %) with a diagnosis of schizophrenia (82 %), see Table 1.The means and standard deviations of the clinical characteristics at the three different assessment points (T0, T3, T6) are shown in Table 2.

Total safety behavior
Across the observed timeframe, a slight decrease was found in both positive and negative safety behavior.Changes in total safety behavior and paranoid ideation were positively correlated over time (Table 3).A decrease in the use of safety behavior was related moderately to a decrease in paranoid ideation.Reductions of safety behavior were also significantly associated with a decrease in social anxiety over time.A similar pattern was found for depression; a decrease in safety behavior was related to a decrease in depression, however, this finding was only found for one of the two 3-month periods.No significant associations were found between changes in safety behavior and changes in belief inflexibility, jumping to conclusions, social cognition problems, positive self, and negative self.

Positive safety behavior
A decrease in positive safety behavior (i.e.coping) was significantly related to a decrease in paranoid ideation (Table 4).However this relation was only significant for the first period between 0 and 3 months.A significant positive correlation was found for positive safety behavior and social anxiety over time.A significant positive correlation was also found for positive safety behavior and depression for the first period.Only small and non-significant associations were found between changes in positive safety behavior and changes in belief inflexibility, jumping to conclusions, social cognition problems, positive self, and negative self.

Negative safety behavior
Changes in negative safety behavior and paranoid ideation were positively correlated over time (Table 5); a decrease in avoidance was related to a decrease in paranoid ideation.A decrease in negative safety behavior was also significantly related to a decrease in social anxiety over 3-6 months, but not over 0-3 months.Significant weak positive correlations were also found for negative safety behavior and depression over 0-3 months, but not over 3-6 months.No significant correlations were found between changes in negative safety behavior e and changes in belief inflexibility, jumping to conclusions, social cognition problems, positive self and negative self.

Discussion
The current study investigated the temporal relationships between changes in safety seeking behaviors and changes in paranoid ideation, social anxiety, depression, cognitive biases and self-esteem, in patients treated for a psychotic disorder.A decrease in total safety behavior was related to diminished paranoid ideation, and depression after three months, and paranoid ideation and social anxiety at three and six months.No significant temporal associations were found between changes in safety behavior and changes in cognitive biases and selfesteem.Similar but slightly less robust results were found for respectively negative safety behavior and positive safety behavior.
In line with the cognitive model of psychosis (Freeman, 2016;Freeman et al., 2002), the results support the idea that the vicious circle of paranoia and social anxiety can be broken by dropping safety behavior.Safety behavior prevents the learning of new meanings of the feared stimulus.When safety behavior is dropped, adequate exposure to a perceived potential threatening event occurs and patients experience high levels of anxiety.However, subsequently patients may learn that they are safe and that they can endure distress, i.e. threat beliefs are falsified and maladaptive fear associations may be replaced with new, more adaptive ones.As a result, anxiety and depression may decrease as well (Ten Broeke and Rijkeboer, 2017).While, previous cross-sectional research have shown that safety behavior is associated with paranoid ideation (Freeman et al., 2002), anxiety and depression (Freeman et al., 2007;Upthegrove et al., 2014;Cooke et al., 2007;Yanos et al., 2008), the current study replicated this with longitudinal data.In all, findings further underline the importance of safety behavior as a treatment target.
In this study, no significant temporal associations were found between changes in the use of safety behavior and changes in cognitive biases.While CBT in general and CBT for paranoid delusions in psychosis (CBTp) in particular start with interventions aiming to affect    dysfunctional cognitions, these findings suggest that behavioral interventions aimed at dropping safety behaviors may be a good alternative starting point for reducing paranoia.Although in CBT it is assumed that cognitive processes determine emotions and behavior, it might be possible that meta-cognitive processes such as cognitive biases follow behavioral change with a delay.Previous research (Freeman et al., 2002;Gaynor et al., 2013) showed that holding unusual beliefs was not related to safety behavior, suggesting that it is not the cognitions themselves that are being responded to but rather anxiety.Thus, our findings could suggest that changing safety behavior affects the body (leading to reduced anxiety by fostering a sense of safety) more than the cognitive processes.Moreover, metacognitive processes, including jumping to conclusions, belief inflexibility and social cognition problems, may be resistant to change (Van Oosterhout et al., 2016).Studies on metacognitive training aiming to affect cognitive biases, with jumping to conclusion as outcome, had a small, non-significant effect (Van Oosterhout et al., 2016).Also, cognitive biases as jumping to conclusions may play a role in delusion formation, but may not be important in the maintenance.It has been shown that jumping to conclusions is associated with delusions, but the effect-size is very small in the general population and absent in the population with current delusions (Ross et al., 2015).Furthermore, no significant association was found between changes in the use of safety behavior and self-esteem over time.Since self-esteem is a complex and multifaceted phenomenon, it likely requires a more comprehensive and targeted approach to change (Freeman, 2016;Freeman et al., 2002).Moreover, the data showed little change on selfesteem over time, which makes it difficult to determine if a change in self-esteem is associated with a change in safety behavior.Also, selfesteem was assessed with the BCSS which measures core beliefs persons hold about themselves.Core beliefs are robust to change, because they are long established and deeply ingrained, and people tend to actively seek confirmatory evidence to reinforce them (Fowler et al., 2006).
When we differentiated between positive and negative safety behaviors, a less consistent pattern was apparent.A reduction of both types of safety behavior separately were related with decreased paranoid ideation, depression, and social anxiety in one or both of the threemonth periods.While most previous studies and treatment protocols tend to focus on negative safety behaviors (i.e.avoidance), these results emphasize the importance of positive safety behaviors, as these seem to contribute to maintenance of paranoid ideations and negative emotions as well (Tully et al., 2017).To increase the effect of exposure, it is necessary to focus on eliminating positive safety behavior in addition to negative safety behavior (Blakey and Abramowitz, 2016;Weisman and Rodebaugh, 2017).
The current study had several strengths.To our knowledge, this is the first longitudinal study of safety behavior in relation to other components of the cognitive model of paranoia, allowing to investigate temporal relationships.A relatively large sample of 116 patients with clinical levels of paranoia and varying degrees of safety behavior was included.The consistency of the observed correlations over time, combined with existing literature and theoretical support, suggests reliability and generalizability of the findings.Further replication studies can help confirm the reliability and generalizability of these findings.The study also had several limitations.First, data were derived from a RCT investigating VR-CBT and TAU at baseline, posttreatment and at follow-up.Although this design provided assessments at three different time points enabling to analyze two periods of three months, a longer follow-up period would have been more powerful to provide further evidence regarding the consistency of the relationships over an extended duration.Second, more fine-grained measures, i.e. repeated measurements within an individual at different times of the day, would be more suitable to assess affective components of the cognitive model, e.g.cognitive biases and self-esteem, as it captures dynamic fluctuations in emotions.Third, all measures including safety behaviors were assessed retrospectively by means of self-report, which may have led to underestimation or overestimation of actual safety behavior.Fourth, due to the design and analyses no claims on causation can be made.Finally, it is important to note that changes in safety behavior and the factors associated with paranoia, as well as the correlations observed over time, can be influenced by potential confounding factors.
To conclude, the vicious cycle of paranoid ideations in patients with a psychotic disorder can be broken by dropping safety behaviors, in which behavioral change by exposure and behavioral experiments can lead to a reduction in paranoia without a change in cognitions.While CBTp typically starts with cognitive interventions, findings of the current study emphasize the key role of behavioral interventions targeted at dropping safety behavior for treatment of paranoid ideations.However, it can be challenging for highly distressed and anxious patients to be willing to engage in exposure and behavioral experiments directly at the start of therapy.Moreover, behavioral interventions are often not applied sufficiently in clinical practice because they are time consuming for therapists.A way to overcome this difficulty is by using VR, in which treatment protocols include behavioral interventions from the start, that can be done in the therapist's office and have a lower threshold for patients (Pot-Kolder et al., 2018;Freeman et al., 2016) A currently ongoing RCT compares VR-CBT with standard CBTp, investigating the added value of VR in the treatment of paranoid delusions, in particular time to clinical change, engagement in behavioral interventions and treatment response (Berkhof et al., 2021).Future research should prioritize further exploration of the relationship between safety behaviors and symptom change in the treatment of paranoia, by utilizing a research design that targets one group specifically on safety behavior while comparing it to a control group.This will provide guidance for the improvement of effective treatment strategies for paranoid delusions.

Declaration of competing interest
The authors declare that there are no conflicts of interest in relation to the subject of the study.
This work was supported by Fonds NutsOhra; Stichting tot Steun VCVGZ; the European Research Council (ERC) under the European Union's Horizon 2020 research and innovative programme ERC-CoG-2015 -681466.Funders had no involvement in the study design; in the collection, analysis and interpretation of data; in the writing of the report; nor in the decision to submit the article for publication.CRediT authorship contribution statement M. Berkhof: Writingoriginal draft, Writingreview & editing, Conceptualization, Formal analysis.E.C.D. van der Stouwe: Writingoriginal draft, Writingreview & editing, Conceptualization.C.N.W. Geraets: Formal analysis, Writingreview & editing, Conceptualization.R.M.C.A. Pot-Kolder: Conceptualization, Investigation, Writingreview & editing, Project administration.M. van der Gaag: Conceptualization, Funding acquisition, Supervision, Writingreview & editing.W. Veling: Conceptualization, Supervision, Writingreview & editing.

Table 2
Clinical characteristics over time.

Table 3
Pearson correlations between change in safety behavior and the corresponding change in other variables related to paranoia.

Table 4
Pearson correlations between change in positive safety behavior and the corresponding change in other variables related to paranoia.

Table 5
Pearson correlations between change in negative safety behavior and the corresponding change in other variables related to paranoia.