|Home | About | Journals | Submit | Contact Us | Français|
Psychotic symptoms are exacerbated by stressful life events in schizophrenia patients as a group. Some individuals appear to be more vulnerable than others in this regard. This study tested whether schizophrenia patients are highly emotionally reactive compared with controls and whether the level of trait emotional reactivity in patients influences the degree to which they respond to life stressors with exacerbations of psychosis. Schizophrenic outpatients and nonpsychiatric controls were assessed for levels of trait emotional reactivity, arousability, and trait anxiety. Severity of symptoms was also rated in the patients. Patients were then followed up 9 months later, assessed for independent stressful life events occurring during the month before the follow-up session, and reassessed for symptom levels. The patients scored higher than the control subjects on all 3 measures of reactivity at the initial assessment. At follow-up, the occurrence of potentially stressful life events predicted increases in psychotic symptoms in patients, and there was a significant interaction between level of initial trait reactivity and the occurrence of life events in the prediction of these increases. High-trait–reactive patients showed increases in psychotic symptoms in response to life stressors, whereas low-trait–reactive patients did not. These findings support the idea that patients as a group have higher than normal levels of trait reactivity and also that patients with very high levels of trait reactivity are at elevated risk of psychotic relapse under stress. Such patients might benefit particularly from interventions designed to assist them in coping with potentially stressful life events and circumstances.
Schizophrenia is a heterogeneous disorder in a number of ways, one of which is that the course of the illness is widely variable.1–3 Generally speaking, of the 3 major symptom dimensions that characterize schizophrenia, the negative and disorganization syndromes are relatively stable over time and the positive, psychotic symptoms are somewhat more variable in course.2,4,5 However, it is also true that the core positive symptoms, delusions and hallucinations, can range over time between very severe and fully remitted in some patients and be quite consistent and chronic in others.3–5 There is evidence that the occurrence of stressful life events3,6–8 or the presence of social relationship stressors such as high levels of familial “expressed emotion9–11” are associated with subsequent exacerbation of psychotic symptoms in patients as a group. These environmental stressors appear to exacerbate the psychotic symptoms of some patients much more than others. The question of what might differentiate patients who respond to such stressors with increased psychosis from those who do not, or who respond in this way only minimally, is an important one both scientifically and clinically. Differential stress responsiveness of symptoms could well reflect a difference among patients in underlying pathophysiological processes which, if identified, could inform ongoing efforts to subdivide the diagnostic category of schizophrenia into more process homogeneous and, therefore, more researchable, subtypes or dimensions. It could also suggest differential treatment approaches for different patients.
In the general population, normal life stressors are likely to lead to changes in mood and perhaps level of anxiety. People respond to stressors with depressed mood or increased anxiety to varying degrees, depending partly on their level of the trait-variable emotional reactivity.12 Emotional reactivity is defined simply as the degree to which an individual is emotionally responsive to social and sensory stimuli. People with schizophrenia, like the rest of the population, have been found to show increases in mood and anxiety symptoms in response to life stressors.13,14 In fact, some studies have found patients to demonstrate a heightened sensitivity in this regard.13,15 However, schizophrenia patients, who differ from the general population in that they are disposed to experience psychotic symptoms, may also suffer exacerbations in these symptoms under stress. It is arguable that psychotic symptoms, although very different from mood or anxiety symptoms, are also emotion driven.15–17 Delusions and hallucinations are generally classified as cognitive symptoms, yet they are typically quite emotional in nature. They may be emotionally negative or positive in content, but they are rarely neutral. Most often they are negative, which is consistent with findings of high negative affectivity in schizophrenia patients.18,19
Presumably, in the association between stressful life events and exacerbations of psychosis it is not the events per se, but the negative emotion aroused by them, that triggers the symptom increases. If this assumption is valid, then the degree to which schizophrenia patients’ psychotic symptoms are stress responsive might depend on individual difference variables similar to those that make other, nonschizophrenic people more or less prone to increases in depressed mood or anxiety in response to events. There is evidence that cumulative stressful events and “hassles” increase emotional reactivity in schizophrenic patients.14,20–22 However, differences among patients in trait emotional reactivity have not been examined in this context. The present study examined individual differences in trait reactivity as a potential predictor of psychotic exacerbation in patients exposed to stressful life events.
In the present study, we assessed levels of 3 related types of trait reactivity in a sample of schizophrenia patients and nonpsychiatric controls and severity of symptoms in the patients. Two of the reactivity measures targeted responsiveness to both positive and negative stimuli and the third targeted trait anxiety, which is a specific type of reactivity to specifically negative stimuli. We then followed up patients 9 months later, assessed for stressful life events during the month before the follow-up session, and reassessed trait reactivity and symptom levels. We hypothesized that (1) self-report measures of emotional reactivity, arousability, and trait anxiety all would be related to each other in patients and controls, (2) they would be elevated in schizophrenia patients compared with control subjects, and (3) they would show trait-like stability over time in patients. We also hypothesized that (4) higher baseline levels of reactivity in patients would interact with the subsequent occurrence of independent stressful life events to predict increases in psychotic symptoms over time; ie, when independent stressful events occurred in patients’ lives, high-trait–reactive patients would show greater increases in psychotic symptoms than low-trait–reactive patients.
Participants at the initial assessment included 49 individuals with schizophrenia (n=31) or schizoaffective disorder (n=18), ages 18–50, and 34 nonpsychiatric control subjects roughly comparable on age, race, and parents’ educational attainment. The patient participants were recruited from a large state-funded community mental health center where they were receiving outpatient services. All had been outpatients for at least 3 months preceding the initial assessment, were receiving antipsychotic medications, and were considered by their case managers to be clinically stable. Exclusion criteria for the patients included current (past year) substance abuse and history suggestive of possible organic damage (eg, head injury, solvent abuse, seizures). The control subjects were recruited from support staff at the university and from the community by means of advertisements. Exclusion criteria were the same as for the patients and also included any history of psychotic symptoms or psychiatric hospitalization, as determined by structured diagnostic interview. Neither patients nor controls were excluded for other Axis I disorders such as mood or anxiety disorders. Descriptive information on participants is presented in table 1.
Nine months later, all the patients who were still in treatment at the clinic, receiving antipsychotic medications, and not abusing substances were reassessed (n=29). Attrition was primarily due to the transience of the clinic population (ie, no longer in treatment at the clinic). However, those who participated in the follow-up did not differ in any significant way with respect to the demographic, symptom, or other variables assessed in this study. Control subjects were not included in the follow-up portion of the study.
Diagnoses were determined by a clinical psychologist with extensive research diagnostic experience, based on information gathered using a structured diagnostic interview, the Schedule for Affective Disorders and Schizophrenia,23 adapted slightly for use with Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, criteria.1 Symptoms were rated using the Positive and Negative Symptom Scale (PANSS),24 by 2 graduate research assistants trained in the measure. The positive and negative symptoms of schizophrenia correlate with each other to varying degrees depending on the way they are defined. When defined broadly they are negatively related to each other, whereas when defined narrowly they are nearly independent of each other. We wanted to measure the 2 syndromes as reflective of distinct processes as much as possible. Therefore, we defined them narrowly as consisting of delusions and hallucinations, and flat affect and poverty of speech, respectively, following Crow.25 The 2 syndromes are most clearly distinct from each other, ie, not highly correlated with each other in either direction, when defined in this way.25,26 The 2 raters independently rated the symptoms of a subset of patients and showed acceptable levels of interrater reliability for all the relevant symptoms: delusions, intraclass correlation (ICC) r=.87; hallucinations, ICC=.89; core negative symptoms (motor retardation and blunted affect), ICC=.91; conceptual disorganization, ICC=.85; and emotion symptoms (anxiety, depression, guilt, and hostility), ICC=.79.
Three measures of reactivity were used in this study. One was created for this study, and the other 2 were existing measures of constructs believed to be related to emotional reactivity: trait arousability and trait anxiety. The new measure was created because a search of the literature did not identify any scales designed to assess general emotional reactivity, worded in a way that would be useable with schizophrenia patients, some of whom have cognitive limitations. The new scale, the Emotional Reactivity Scale, is a 13-item scale that asks specifically about emotional reactivity to a range of social and sensory stimuli. The items of this scale are presented in table 2. Internal consistency in the patients and controls were computed separately and both were acceptable, alpha=.82 for patients and alpha=.83 for controls. Item-to-scale correlations were all positive and ranged from r=.24–.72 in the patient sample and r=.26–.75 in the control sample.
The second reactivity-related measure was the Arousal Predisposition Scale (APS),27 a 12-item scale that assesses general trait arousability. Subjects indicate on a 5-point scale the degree to which each item is true of them (eg, “I am a calm person”(reversed). “I get flustered if I have several things to do at once.” “I tend to remain excited or moved for a long period of time after seeing a good movie.”). This scale showed good internal consistency in the patients, alpha=.93, and controls, alpha=.89. The third reactivity-related measure was Part 2 of the State-Trait Anxiety Inventory (STAI-T),28 which assesses trait anxiety by means of 20 items rated on a 4-point scale about how the participant “generally” feels (eg, “I feel secure” (reversed); “I feel nervous and restless”). Internal consistency was good for this scale as well, with alpha=.91 for patients and alpha=.93 for controls. This measure targets a type of reactivity (anxiety) to specifically negative stimuli, in contrast to the other 2 scales, which assess reactivity to both positive and negative stimuli.
Life events were assessed using methods of Dohrenwend.29 A long checklist of potentially stressful life events was administered by interview for events occurring during the 1-month time period immediately preceding the follow-up session. After completion of the checklist, each event endorsed was then described by the patient in narrative form and recorded, with the interviewer asking questions as needed to obtain adequate information about the event, the surrounding circumstances, the impact on the person’s situation, and the possible causal contributors, for later assessment by an objective, blind rater. The rater used the definitions and anchor points of Dohrenwend29 in coding the events.
In assessing the impact of life events on symptoms, directionality of effect is an issue. Symptom increases can cause behavior that may then cause or contribute to the occurrence of some stressful events (eg, eviction, divorce). In the present study, it was important for the direction of causality to be as clearly as possible in the event-to-symptom-change direction. Therefore, we examined only “independent” events, ie, those that could not have been caused by the behavior or symptoms of the participant. An objective, blind judge rated each event as dependent, possibly dependent, or clearly independent; only clearly independent events were included in the analyses. A second rater corated a subset of events, for reliability purposes. There was complete agreement on the independence ratings of these events.
The same judge rated the valence of each independent event and its severity. We wanted a measure of event severity that was independent of individuals’ trait reactivity levels, so that we could then look at the interaction between event severity and reactivity in predicting symptom change. Thus, these ratings were based on the rater’s judgment of how distressing the event would be to the average person, given the circumstances detailed in the narrative. The judge rated severity without regard to subjective feelings stated by the participants. For example, the death of a partner who had been living with the participant for years would have been rated as negative and very severe, whether or not the participant reported much distress, whereas the death of a very recent acquaintance would have been considered less severe, even if the participant voiced deep love for the person and devastation over the loss. This approach was taken in an effort to keep the measures of trait reactivity and life events as independent from each other as possible. Subjective judgments by participants about the severity of events might have been influenced by, and therefore correlated with, the level of their trait reactivity. Sample events were used as anchor points for the coding of objective severity, and for most events, the determination was unambiguous. For events in which the level of objective severity was questionable, the rater consulted with the principal investigator, blind to participant identity, and consensus ratings were made. To summarize, events included in the analyses were those that occurred within the preceding month, were independent of participant behavior, were negative in valence, and were objectively rated as moderately to very severe.
The analysis was done in 4 parts. First, correlations were computed between the 3 measures of reactivity in each group. Second, patients were compared with controls on the measures. Third, correlations were computed between 2 of the reactivity measures at the initial assessment and at follow-up to test for stability of the measures over time in patients. Fourth, 2-way (life events × trait reactivity) analyses of variance (ANOVAs) were computed, with change in psychotic symptoms over time as the dependent variable, to test for main effects of stressful life events and trait reactivity on symptom changes, and for an interaction effect between life events and trait reactivity in the prediction of psychotic exacerbations. All analyses were 2 tailed.
Schizoaffective patients were compared with schizophrenia patients on the measures of trait reactivity, psychotic symptoms at each assessment, and symptom change. There were no significant differences on any of these variables, so the patients were treated as a single group in all analyses. The patient group differed significantly from controls in age and parents’ educational level and also differed somewhat in race and gender distributions (although not statistically significantly so). Associations of age, parents’ education, gender, and race with the reactivity measures in each group were small and nonsignificant (NS). Therefore, no adjustments were made for these group differences.
In the patients included in the follow-up portion of the study, summed PANSS ratings for delusions and hallucinations at the initial assessment ranged from 2 (not present) to 10 (moderately severe), mean=5.0, median=5.0, and the follow-up assessment, from 2 to 11, mean=4.3, median=4.0. In all, 24% showed no psychotic symptoms at the initial assessment and 35% at the follow-up.
Correlations were computed between the 3 measures of reactivity in each group. As expected, all measures were related to each other in each group. These correlations support the idea that the 3 scales measure similar and overlapping constructs. These results are presented in table 3.
Patients were compared with controls on reactivity scores at the initial assessment by means of t-tests. Patients scored higher than controls on all 3 measures (see table 4), indicating that as a group they were reporting higher-than-normal levels of emotional reactivity, arousability, and trait anxiety. Total symptom ratings on the PANSS were correlated with trait anxiety in the patients (r=.47, P<.01) but not with scores on either of the other reactivity measures. Severity of the core psychotic symptoms of delusions and hallucinations at the initial assessment was not related significantly to any of the reactivity measures. Correlations were as follows: for the Emotional Reactivity Scale, r=.09, P<.54; for the APS, r=.16, P<.27; and for the Trait Anxiety Scale, r=.27, P<.08.
The trait anxiety measure has been used extensively in previous studies and has shown stability over time. The other two reactivity measures used in the present study were readministered to patients at the 9-month follow-up session to test for stability over time. For the Emotional Reactivity Scale at initial and follow-up assessments, r=.71, P<.01 and for the APS r=.65, P<.01. These results suggest that the measures were reflecting trait-like variables as intended.
Three 2-way ANOVAs were computed to test the second hypothesis, that the occurrence of stressful events would predict increases in psychotic symptoms and that this would be true primarily in those patients with high levels of trait reactivity. Eight of the 29 patients had one or more independent, objectively stressful life events during the month immediately preceding the follow-up session. Patients were divided into high- and low-trait–reactive groups by means of a median split on each of the reactivity measures. For the dependent variable of change in psychotic symptoms, PANSS ratings for delusions and hallucinations at the initial assessment were subtracted from ratings at the follow-up assessment. This variable was also computed a second way, by residualizing the ratings at the first assessment out of the ratings at the second assessment.
With the Emotional Reactivity Scale, there was a main effect for life events (present vs absent), F1,28=10.41, P=.003; a main effect for trait reactivity level (high vs low), F1,28=4.74, P<.04; and a trait × event interaction effect, F1,28=4.63, P<.05. When residualized symptom ratings were used as the dependent variable in the ANOVA instead of difference scores, the results were very similar (interaction F=6.92, P<.02). The means for the difference scores are presented in table 5, and the individual scores for the patients in each group are presented in table 6.
These results indicate that patients who experienced potentially stressful life events during the month preceding follow-up showed increases in psychotic symptoms compared with those who did not experience any such events. They also indicate that patients high in trait reactivity showed greater symptom increases than those low in trait reactivity. Most importantly, they suggest that life events had a differential effect on the symptoms of those who were high vs low in trait reactivity (see figure 1).
A t-test looking only at those who had experienced independent, objectively stressful life events found a sizeable and significant difference in level of symptom change between those with high- vs low-trait reactivity. The high-trait–reactive patients showed substantial increases in psychotic symptoms, whereas the low-trait–reactive patients did not, mean (SD)=3.60 (1.95) and −0.67 (2.31), respectively, t(7)=2.81, P<.04. A similar t-test looking only at those who had not experienced any potentially stressful events found no difference in psychotic symptom change in the high- vs low-reactive patients, mean (SD)=−1.70 (1.70) and −1.73 (2.87), respectively, t(20)=0.07, NS. In fact, and notably, both of these groups showed a modest improvement in symptoms over time.
Results were different with the APS. There was still a main effect for life events, as would be expected, but the main effect for arousability as measured by this scale was small and NS, F1,28=0.30, NS, as was the interaction effect, F1,28=0.00.
With the STAI-T, results were very similar to those for the Emotional Reactivity Scale. There was the same main effect of life events; a main effect of trait anxiety, F1,28=5.91, P<.03; and an interaction effect, F1,28=6.29, P<.02. Once again, when residualized symptom ratings were used as the dependent variable in the ANOVA instead of difference scores, the results were very similar. The post hoc t-test results also were similar to those for the Emotional Reactivity Scale. Among patients who had stressful life events, those with high-trait anxiety showed significant increases in psychotic symptoms, whereas those with low-trait anxiety did not. Patients who had no stressful life events showed modest improvements in psychotic symptoms over time.
Effects of life events and reactivity on other symptoms were examined using the same analytic methods. Presence or absence of stressful life events, levels of reactivity, or interactions between the 2 did not predict changes in (1) the core negative symptoms of motor retardation (which includes poverty of speech) and blunted affect, (2) conceptual disorganization, or (3) the emotional symptoms of anxiety, depression, guilt, and hostility. All the effects in these analyses were small, and none approached statistical significance. Thus, the interaction effects of life events and reactivity in this sample were specifically on the core psychotic symptoms of delusions and hallucinations. When changes in these 2 symptoms were analyzed separately, the effects on delusions were somewhat stronger than on hallucinations.
To summarize the results, patients scored higher than controls on self-report measures of emotional reactivity, arousability, and trait anxiety. The reactivity measures were stable in patients over a 9-month interval. The occurrence of independent, objectively stressful life events during the month before follow-up predicted significant increases in psychotic symptoms in those patients who had scored high in emotional reactivity or trait anxiety at baseline but not in those who had scored low on these scales. These results support the idea that levels of trait emotional reactivity and trait anxiety influence the degree to which patients are vulnerable to exacerbations of psychosis in response to life stressors. Although the group of patients who experienced an independent life event was small, it is notable that while most patients (n=18) showed some improvement in symptoms over time, or no change (n=3), all 5 of those in the high-reactive-with-event group showed symptom increases.
The finding that psychotic symptoms were exacerbated in response to life stressors is not new. It is reasonable to assume that it is the emotional response to the life event (ie, the “stress” experienced in response to the stressor) that somehow precipitates the psychotic symptom exacerbation. Speculations may be made as to the mechanisms involved on several different levels. On a physiological level, there is evidence that delusions and hallucinations are associated with neurotransmitter dysregulation involving dopaminergic overactivation in limbic areas.30–32 Part of the normal response to stress involves increases in dopaminergic activity. Possibly, those patients with high levels of trait emotional reactivity are the ones in whom dopaminergic activity is most labile.33,34 Such patients might respond to the occurrence of events with relatively high levels of emotion and surges of dopaminergic activation, which could lead to exacerbations of psychotic symptoms. From another perspective, some patients have been found to demonstrate diminished psychophysiological inhibition.35 It is arguable that the emotional response to life stressors in such patients is less efficiently gated or habituated to, leading to a stimulus overload situation conducive to cognitive breakdown and psychosis. Arguing against this idea is the fact that despite many efforts, very little has been found to connect psychophysiological inhibitory failures in patients to severity of psychotic symptoms. Finally, on a psychological level, it is notable that schizophrenia patients as a group endorsed substantially higher levels of emotional reactivity than nonschizophrenic individuals, in this and other studies.36,37 Those patients who were in the “high reactivity” group in our study were reporting very high levels indeed. Patients who are this highly emotionally reactive, in the face of difficult events, may experience intolerable affect and may defend against it by retreating from reality.38
In contrast to the measures of emotional reactivity and trait anxiety, the APS was not a significant predictor of psychotic vulnerability to stressors. The reason for this is not clear. The scale scores correlated fairly highly with the Emotional Reactivity Scale scores, but one way in which it differs is that it puts a somewhat greater emphasis on reactivity to sensory stimuli and somewhat less on social stimuli. It also uses slightly more complex language, which could have resulted in more error, although the scale internal consistency argues against the latter interpretation. In any case, the results with this scale were null and the sample was not large, so interpretation is difficult.
The core negative, disorganization, and emotion symptoms were not significantly affected by independent life events or levels of emotional reactivity in this study. Psychotic symptoms are the ones that have been the most episodic and the most associated with relapse over time, and these were the only ones that showed significant reactivity in this sample. Again, however, null findings in a small sample do not necessarily indicate that there is no association. There have been laboratory research findings indicating stress responsiveness of disorganization symptoms39,40 and clinical research findings of reactivity of emotional symptoms13–15 reported in the literature. In a larger sample, these symptoms might well be significantly reactive too, although perhaps less so than the psychotic ones.
The construct of emotional reactivity examined here overlaps with the construct of negative affectivity. However, the Emotional Reactivity Scale in the present study differed from measures of negative affectivity in that it (a) focused on reactivity to events and (b) included positive as well as negative emotions and events. The positive and negative reactivity items cohered reasonably well, in that the scale's internal consistency was good and all items were correlated in the positive direction. However, it is not clear from the present study whether the characteristic most directly related to greater vulnerability to psychotic exacerbation was the broad trait reactivity to events measured by the Emotional Reactivity Scale, trait anxiety alone, or possibly general level of negative affectivity. Future research could test a wider range of emotion- and reactivity-related measures, including self-report measures of emotional reactivity and affectivity, measures of psychophysiological reactivity, and neuroimaging measures of brain activation during tasks that use emotional stimuli or that are performed under varying emotional conditions. Pinpointing the exact nature of a trait characteristic that interacts with life stressors to predict psychotic exacerbations could provide information relevant to understanding the psychological and pathophysiological processes that lead to relapse and, perhaps, even to the first onset of psychotic symptoms.
Future research also could be directed toward testing differential treatment approaches, depending on patients’ levels of emotional reactivity or trait anxiety. For example, psychosocial interventions such as stress management training,41–43 or family interventions that reduce family members’ stress-inducing critical or intrusive behaviors,44,45 may be particularly helpful to patients who are highly reactive. In the work rehabilitation arena, low-reactive individuals may be able to move directly into competitive employment without too great a risk of relapse, whereas highly emotionally reactive patients might find sheltered occupational programs a more efficacious first step toward employment. Pharmacologically, optimal maintenance medication may differ for patients depending on their levels of emotional reactivity. In addition, the occasional use of antianxiety medications during stressful times might help prevent relapse in highly reactive patients and not in others. These are all possibilities that could readily be tested in research.
The National Alliance for Research on Schizophrenia and Depression; the National Institute of Mental Health (R01-MH58783).
The authors have no conflicts, financial interests, or affiliations related to this research. The authors are grateful to the administration, staff, and clients of Community Support Services in Akron, Ohio, for their support and participation in this project. This article was presented at the International Congress on Schizophrenia Research, 2007.