Abstract
There is evidence that environmental stressors can have an impact on the course of schizophrenia [1, 2]. A careful review of this literature suggests that the role of stress is best understood in the context of a stress–vulnerability interaction, such that increasing levels of stress are most apt to result in increased symptoms in those individuals who have underlying vulnerability to development of the disorder [2–4].
Norman and Malla [5] reported the results of one of the very few longitudinal studies of the relationship between stressors and symptoms in a group of outpatients with a DSM-III-R diagnosis of schizophrenia. This study was unique in several ways, the two most noteworthy for present purposes were that: (i) daily hassles or minor stressors rather than major life events were used to predict subsequent symptoms; and (ii) correlations between level of stress and subsequent symptoms were calculated for each individual across time, using monthly data. This latter procedure allowed examination of the extent to which there was variation between individuals in their symptoms’ apparent reactivity to stress. The results showed that while a greater number of individuals than would be expected by chance showed changes in their symptoms (particularly positive symptoms such as reality distortion) that were related to stress in the previous month, the majority of patients did not show such a relation to a statistically significant extent. None of the variables of gender, age, number of previous psychiatric admissions and level of medications were significantly related to likelihood of a patient showing a change in symptoms following changes in stress level.
Docherty [6] has argued that reactivity of symptoms to stress or negative affect may be a major aspect of individual differences in the processes which lead to schizophrenia. Furthermore, there is evidence that negative affect (manipulated by asking subjects to talk about bad or stressful vs pleasant memories) is more likely to lead to an increase in ratings of thought disorder on a speech sample in patients with a family history of schizophrenia than in those without such a family history [7]. Docherty et al. [7] suggest that these findings are consistent with those of an earlier study [8] showing that patients with a positive family history of schizophrenia demonstrate an increasing evoked response to repeated stimuli in contrast to a diminishing response for patients without a positive family history.
The above findings suggest that positive family history may be a moderator of the extent to which the symptoms of patients suffering from schizophrenia vary as a function of changing level of stressors in their natural environment. In this paper we provide additional findings related to this possibility using data from our earlier longitudinal study [5].
Method
Subjects
All subjects in the original study were recruited from outpatient clinics at the Victoria Campus of London Health Sciences Centre and London Psychiatric Hospital in London, Ontario, as well as North Bay Psychiatric Hospital in North Bay, Ontario. The inclusion criteria were having a DSM-III-R diagnosis of schizophrenia as confirmed by a structured clinical interview [9] conducted by an experienced psychiatrist; being between the ages of 17 and 60; having had at least one previous admission to hospital for treatment of schizophrenia; and having had a stable level of medication for at least 3 months prior to entry into the study. Exclusion criteria included organic brain disease or history of head injury and significant drug or alcohol dependence. Data from 55 patients who participated in an earlier longitudinal analysis have been presented elsewhere [5].
Procedure
At the time of each patient's entry into the study, information was collected concerning demographic characteristics and current and past psychiatric treatment. In order to collect longitudinal data, patients were interviewed on a monthly basis. The mean number of consecutive monthly assessments carried out per patient was 19, with a range of 12–29. During the monthly interview, information concerning daily stressors was collected using the Hassles Scale [10]. This self-report instrument has been widely used to measure stressors in the form of common every day difficulties and challenges. There are a total of 118 items in the scale which assesses reported pressures, problems or difficulties associated with such circumstances as ‘not enough money for clothing’, ‘family responsibilities’, ‘transportation problems’ etc. Respondents are asked to indicate which of the difficulties have occurred to them over the previous month, and for each that had occurred, they are asked to rate its severity on a three-point scale. In scoring the Hassles Scale we followed the most widely used procedure of summing the severity ratings for all items endorsed (e.g. [11]). This scale has been found to have adequate levels of test–retest reliability, and investigations in other populations have found it to be predictive of later psychological and physical symptoms [10,12–14].
A trained interviewer also completed a monthly symptom assessment of the patient using the Scale for the Assessment of Positive Symptoms (SAPS) [15] and the Scale for the Assessment of Negative Symptoms (SANS) [16]. These symptom ratings were completed without knowledge of the subjects’ reported levels of stress. The interviewer had been given extensive training in the use of the SANS and SAPS by the principal investigators. All ratings were carried out on sixpoint scales and composite scores were calculated by total of individual items, not including global ratings. Interrater reliabilities of the SANS and SAPS were established by calculating intraclass correlation coefficients (ICCs) on 15 patients who were assessed by two raters. All ICCs for items and summary scores were significant at a minimum of the p = 0.005 level and varied between 0.61 and 0.98. Interviewers were not informed of the family history of patients, nor was the hypothesis of family history as a moderating variable salient when data were being collected, so systematic bias is unlikely to have been present.
In addition to calculating total score on the SAPS and SANS, scores were calculated for each of the three syndromes of reality distortion, disorganization and psychomotor poverty [17, 18]. Consistent with several previous studies (e.g. [19, 20]), psychomotor was scored using items assessing poverty of speech, decreased spontaneous movement and the average of four items reflecting aspects of blunting of affect (affective non-responsivity, unchanging facial expression, paucity of expressive gestures and lack of vocal inflections). The disorganization syndrome was the sum of scores for items measuring inappropriate affect, poverty of content of speech and the global rating of positive formal thought disorder. The reality distortion syndrome was the sum of items for auditory hallucinations commenting on the patient's behaviour, persecutory delusions and delusions of reference.
As part of each patient's assessment prior to entry into the respective treatment programs, information was collected regarding family psychiatric history. Sources for this information include the patient, information from first-degree relatives of the patient plus any available case records. All subjects are well known in their respective clinical services and information was collected carefully by treating psychiatrists and case managers. It should also be noted that recording of family history was always cross-checked with other informants in the family rather than being based solely on patients’ reports. Family history of schizophrenia was recorded in a reasonably conservative manner (i.e. only when a specific relative was identified as having received a formal diagnosis of schizophrenia), and this information was collected prior to subjects’ entry into the research protocol without knowledge of the outcome of their longitudinal data concerning stressors and symptoms.
We recognize that our method of identifying family history of schizophrenia cannot be considered perfectly reliable. Roy et al. [21] have reported on the relationship of family psychiatric histories established through family informant reports to those established through personal interviews with all available relatives of probands. Although the level of agreement between the two methods was generally disappointingly low, the best level of agreement tended to occur when more than one family informant was used, when the target diagnosis was schizophrenia and when the proband's and relatives’ diagnoses were the same, all of which characterized the data being used in this study. Furthermore, all family history data collected in this study were always with reference to relatives of individuals with a diagnosis of schizophrenia thereby minimizing the systematic bias introduced by comparing reports from relatives of individuals with severe mental illness with those of relatives of normal controls or individuals suffering from less severe illness [21]. There appears to be little reason for assuming that identification of a positive family history would be systematically biased as a function of the relationship between stress and symptoms.
A diagnosis of schizophrenia in a blood relative was reported for 12 patients who had been included in our longitudinal study [5]. For four of these patients the relative was a cousin or grandparent; for three it was an uncle or nephew; for two it was a sibling; for one it was a parent; for one patient it was parent, grandparent and uncle; and for another it was a parent and uncle.
Twelve subjects were selected from patients in the earlier study with no known family history of psychiatric illness solely on the basis of matching to those patients with a reported positive family history of schizophrenia on gender, age and number of previous psychiatric admissions. For purposes of the issues being addressed in this paper, focus is on the comparison of the matched samples of the 12 patients with a reported positive family history of schizophrenia with 12 matched patients with no known family history of psychiatric illness.
Results
Comparison of groups on demographic and clinical characteristics
As can be seen in Table 1 subjects were predominantly male and unmarried. The baseline SAPS and SANS total scores in Table 1 are based on the sum of individual items, excluding global items for each subject. Comparisons between groups using either paired sample t-tests or a comparison of proportions for paired samples, as appropriate, revealed no statistically significant differences between patients with a family history of schizophrenia and those with no family psychiatric history on the characteristics in Table 1.
Demographic and clinical characteristics for subjects with family history of schizophrenia and no family history of psychiatric illness
Although the differences on the baseline symptoms, medication levels, and Hassles index suggest higher levels in the group with a positive family history, they did not reach significance (t = 0.86, p < 0.40 for chlorpromazine (CPZ) equivalent; t = 1.23, p < 0.20 for SAPS; t = 1.25, p < 0.20 for SANS and t = 1.73, p < 0.10 for Hassles score). The failure to reach statistical significance is likely a result of the small sample size.
Relationship of stress and symptoms as a function of family history
For each subject, Pearson correlation coefficients were calculated over time between stress level as assessed by the total Hassles score and symptom indices for the following month. In the case of patients for whom there were significant serial correlations in symptom data, the correction to the degrees of freedom suggested by Holtzman [22] was adopted for assessing the significance of the correlations between Hassles score and subsequent symptoms. In order to assess whether patients with a positive family history showed evidence of stronger symptom reactivity to stress we compared the number of patients showing significant stress–symptom correlations in the two groups.
Table 2 shows the number of longitudinal correlations reaching statistical significance in each group. Differences between the family history groups were examined using tests of the significance of differences in proportions for paired samples [23]. The proportions of patients showing significant correlations with Hassles scores differed between family history groups on both total SAPS scores (z = 2.24, p < 0.05) and reality distortion (z = 2, p < 0.05), but not for the other symptom measures.
Number of subjects showing a statistically significant correlation over time between stress and symptoms
As has been argued elsewhere [3, 5], it could be that daily stressors are themselves a reflection of level of symptoms. For instance, patients may become less careful in managing their finances, or more argumentative as a result of initial increases in symptoms and this could result in an increase in apparent stressors such as lack of funds or interpersonal difficulties. In order to provide a more conservative estimate of the possible causal role of stress in bringing about changes in symptoms, a stepwise, multiple-regression analysis was carried out on each patient's data, first of all entering symptoms from the previous month to predict subsequent month's level of symptoms and then assessing whether Hassles score added significantly to the prediction of variance in symptoms for the following month. For instance, one month lagged total SAPS score was used to predict SAPS scores across months, then it was tested whether variation in Hassles score over time was an additional significant predictor of variation in subsequent months’ SAPS score. When this procedure was followed, the proportions of patients in each family history group showing a significant relationship of Hassles score as a predictor of subsequent total SAPS and reality distortion scores remained exactly the same as that presented in Table 2.
It is also possible to treat the stress symptom correlations as continua rather than simply counting the number of statistically significant correlations in each family history group. In doing so, we first contrasted the magnitude of the correlations between the Hassles score and each of the symptom indices for the two family history groups using paired t-tests. Parallel to the findings in Table 2, the difference in correlations of Hassles score with total SAPS score differed significantly between the two groups (mean r for family history of schizophrenia group = 0.33; for no family history, mean r = 0.10, t = 2.19, p = 0.05). The difference in mean correlations on the reality distortion index approached twotailed significance (r = 0.32 and 0.10 for positive and negative family history groups, respectively, p = 0.08), and none of the differences for the other symptom indices approached significance.
Given the substantial, albeit not statistically significant, differences in some baseline symptoms and Hassles scores, we carried out multiple regression analyses using baseline Hassles score, family history, and baseline levels of each symptom to predict the respective magnitudes of the correlations between Hassles score and each symptom complex. As an example, a stepwise regression procedure using baseline Hassles score, baseline total SAPS score and family history condition resulted in only family history being entered as a significant predictor of the magnitude of the correlations between Hassles and total SAPS score (F = 5.24, df = 1,22, p < 0.05). Similarly, family psychiatric history was the only variable to enter the regression equation as a significant predictor of reality distortion (F = 4.7, df = 1,22, p < 0.05). Baseline levels of Hassles score or symptoms did not enter into either of these regression equations (in all cases F < 1). The only other significant predictor of the magnitude of Hassles symptom relationship was baseline Hassles score as a predictor of the magnitude of Hassles–psychomotor poverty correlations (F = 4.9, df = 1,22, p < 0.05).
It is important to consider the possibility that the differences found between groups are the result of a statistical artefact. Two possible data parameters which might increase the likelihood of finding a statistically significant correlation are a larger number of data points (months) in the collection of data for the group with a family history which would increase degrees of freedom for testing significance and/or a greater range of scores over time on either the Hassles index or symptom index for the group with a positive family history. We therefore compared both the average number of time points and the average range of scores over time for individuals in each group. There were no significant differences.
For 18 patients (eight with family history of schizophrenia, 10 without) ratings were available from the treating psychiatrist of the extent to which the patient was compliant in taking prescribed medication. These ratings were made on a three-point scale (1 = no problem with compliance, 2 = minor problems with compliance, 3 = often problems with compliance). There was no significant difference between family history groups in mean compliance ratings (1.1 vs 1.2, t = 0.402, NS).
Discussion
Our findings suggest stronger relationships between stress and subsequent symptoms of psychosis in patients with a positive family history for schizophrenia than for those with no known family history of psychiatric illness and, thereby, parallel past findings by suggesting that negative affect has a greater effect on the cognitive functioning of patients suffering from schizophrenia who have a positive family history [6, 8]. There are, however, several provisos which should be considered in interpreting the results of the current study.
The sample size is small and so all findings should be considered preliminary and tentative. The small sample size also means that there is not sufficient power for meaningfully carrying out more formal statistical tests examining interactions between family history and stress as predictors of symptomatology. Nevertheless, the unique nature of the database on which these preliminary findings are based encourages us to provide this report.
It should also be acknowledged that although the relationship of family history of schizophrenia and the stress– symptom relationship was not explained by the demographic and clinical characteristics that were assessed, other confounds may have been present. For example, although drug or alcohol dependence was an exclusion criterion, there is still the potential for differences between the groups in mild to moderate use of substances, concerning which relevant measures were not available.
The stressors assessed in the current study take the form of daily hassles, typically stressors that would be considered to be of comparatively minor or moderate magnitude rather than major life events or traumas. A genetic vulnerability to schizophrenia may leave an individual more susceptible to the effects of stressors, parallel to evidence of an interaction between genetic vulnerability and life events influencing the onset of depression [24], and other evidence of genetic influences on responsivity to stress (e.g. [25]). Such a suggestion is consistent with a stress–vulnerability model which postulates that as levels of biological vulnerability increase, individuals become more susceptible to the development of psychosis in response to stress and that such changes may be preceded and/or mediated by changes in more general psychopathology such as anxiety or depression [26, 27]. At the extremes of low or high vulnerability there might be less clear evidence of a correlation between naturally occurring stressors and symptoms, because so many or so few stressors would be capable of eliciting the symptoms of interest [2, 3, 28]. If we assume that a family history, by itself, denotes a moderate level of risk then we would expect our current findings with reference to the apparent effects of daily stressors. Such findings, however, would not necessarily imply that greater stressors are of no significance for other patients.
While Docherty et al. [7] focused on the effect of negative affect on thought disorder, we did not find positive family history patients and negative family history patients to show differential response to stress on the syndrome of thought disorganization. The reason for this discrepancy in findings is difficult to determine. It may reflect the difference in the nature of the source of negative affect (naturally occurring daily stressors vs experimental induction of negative affect through being asked to talk about negative events) or the subtlety of the measures of thought disorder (overall clinical symptom ratings vs highly structured ratings of a recorded speech sample). Certainly, findings that disorganization and reality distortion syndromes may have somewhat differing neurophysiological bases [20, 29, 30] suggest that this is an issue worthy of further investigation. It may well be that what is inherited is the vulnerability to a particular dimension of psychopathology and that heritability of some dimensions may have a stronger interactive relationship with environmental stress than others. Given the association of the temporal lobes to the limbic system, past findings implicating temporal lobes or frontal– temporal relationships in the reality distortion syndrome [20, 29] suggest that this syndrome may be particularly responsive to changes in stress or affective tone.
Differences in the reliability of assessment of each of the three syndromes could also be of potential relevance to understanding our results. In a previous paper [31] we have examined levels of interrater reliability of items on the SAPS and SANS. Although our previous findings (and those of other studies) suggest that items reflecting psychomotor poverty may have slightly lower levels of reliability than those for reality distortion and disorganization, the latter two syndromes do not appear to differ with respect to the reliability of the assessment of their constituents. Differences in reliability do not therefore appear to explain our finding of greater correlations of stress with reality distortion than with disorganization.
The current study contrasted only patients with family history of schizophrenia and those with no known family psychiatric history. As noted earlier, we acknowledge that our method of assessing family psychiatric history is unlikely to be perfectly accurate, but there is no credible reason to suppose that likelihood of false negatives would be systematically biased with reference to stress– symptom relationships. The operationalization of positive family history used in this study may have identified only the most severe cases of schizophrenia among relatives as we relied on reporting of cases in which diagnosis had been confirmed in a relative. The role of family history of other schizophrenia spectrum disorders in bestowing a vulnerability to psychosis in response to stress remains to be investigated. There is evidence of increased likelihood of other types of psychiatric illness in the family history of individuals with schizophrenia [32, 33]. Only one of the patients included in the positive family history of schizophrenia group in the current study also had a known family member with another psychiatric diagnoses (a sibling with major depressive illness). In future, it will be important to examine whether a family history of only non-schizophrenia spectrum disorders is associated with increased responsivity to stressors in those suffering from schizophrenia. Among the initial sample not included in this study were a substantial proportion of individuals with only a reported family history of other psychiatric illnesses or unidentified psychiatric illness, but it was not possible to positively identify a sufficient number with the same illness to justify including them as a comparison group.
This study was designed to examine correlations of changes in level of stress with variation in levels of symptoms of schizophrenia. The issue of how many of the changes in symptoms would meet some a priori criterion of being ‘clinically significant’ is not directly addressed. The average range of scores across the months per person on the SAPS was 15.08 and for the SANS it was 25.12. Such ranges do suggest that the variation that was occurring over time was clinically noteworthy. We admit, however, that the inclusion criterion of having been on stable level of medication for at least 3 months prior to entering the study may well have implications for generalization of our findings to more unstable patients.
Finally we would note that Walker and Diforio [1] have recently postulated a neural mechanism for explaining the effects of stress on symptoms in schizophrenia. Their model suggests that activation of the hypothalamicpituitary- adrenal (HPA) axis in response to stressors may interact with the neural basis for susceptibility to schizophrenia (most likely involving abnormalities in neurotransmission in the thalamo-limbic-cortical system) to leave some patients hypersensitive to stress. Walker and Diforio suggest that this may be the mechanism for development of symptoms in some but not all patients with schizophrenia. Much of the human evidence used in support of their model involves high-risk individuals defined on the basis of family psychiatric history. This raises the possibility that family history may well moderate the neurotransmitter mechanisms underlying responses to stress in schizophrenic patients. Another possibility is that a family history of schizophrenia may have resulted in disruptions of early family life for patients which could leave them more susceptible to the effects of subsequent stressors. These are issues worthy of further investigation.
Footnotes
Acknowledgements
This research was supported by a grant from the Upjohn London Neurosciences Program.