Journal of Affective Disorders 84 (2005) 53 – 62 www.elsevier.com/locate/jad Research report Stability and change of emotional functioning in late life: modelling of vulnerability profiles Edwin de Beursa,*, Hannie Comijsb, Jos W.R. Twiskc,d, Caroline Sonnenbergb, Aartjan T.F. Beekmanb, Dorly Deegb a Department of Psychiatry, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands b Department of Psychiatry, Vrije Universiteit, Amsterdam, The Netherlands c EMGO Institute, Vrije Universiteit, Amsterdam, The Netherlands d Department of Clinical Epidemiology and Biostatistics, Vrije Universiteit, Amsterdam, The Netherlands Received 23 July 2004; received in revised form 21 September 2004; accepted 27 September 2004 Abstract Background and aims: The present study investigated stability and change in emotional well-being in a prospective study of a large sample of community-dwelling older adults (z55 years). Emotional functioning was conceptualized according to the tripartite model distinguishing three aspects: general negative affect (NA), depression, and anxiety. The study tested models for the decline of mental health in late life based on the diathesis–stress model. In previous studies, support has been found for the diathesis–stress model (for an overview, see [Goldberg, D.P., Huxley, P., 1992. Common mental disorders: a biosocial model. Routledge, London; Zuckerman, M., 1999. Vulnerability to psychopathology. American Psychological Association, Washington, DC.]). The predictive ability of vulnerability factors (the personality characteristics mastery and neuroticism) and stressful life events and their interaction was tested for an increase in general negative affect, decreased positive affect (PA), and increased anxiety. More specifically, we tested the hypothesis that loss leads to decreased positive affect in subjects with low mastery, whereas threat leads to anxiety in subjects with high neuroticism. Methods: Data from the Longitudinal Aging Study Amsterdam (LASA) were used. LASA is a longitudinal study in a large representative sample of adults aged 55 to 85 (N=1837). Self-report data on depression, anxiety, and negative affect were collected from adults over a 6-year period in three waves. The data were analyzed using multilevel analysis. Results: The findings revealed an association between low mastery, high neuroticism, and an increase in negative affect, lack of positive affect, and anxiety. Furthermore, high mastery protected against the negative impact of loss events, but neuroticism did not augment the negative impact of threat events on emotional health. Conclusion: Partial support was found for a diathesis–stress model of change in emotional functioning in late life. Furthermore, support was found for distinguishing between symptoms of negative affect, depression, and anxiety. D 2004 Published by Elsevier B.V. Keywords: Anxiety; Depression; Risk factors; Vulnerability; Life events; Multilevel analysis * Corresponding author. Tel.: +31 71 526 3416; fax: +31 71 524 8156. E-mail address: e.de_beurs@lumc.nl (E. de Beurs). 0165-0327/$ - see front matter D 2004 Published by Elsevier B.V. doi:10.1016/j.jad.2004.09.006 54 E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 1. Introduction Anxiety and depression are the most prevalent psychological problems in late life. Traditionally, research into emotional aspects of the aging process focused on depression. However, recently, studies have been undertaken investigating anxiety in late life, and knowledge regarding the prevalence and comorbidity patterns of anxiety in older persons is growing (Flint, 1994; Fuentes and Cox, 1997). These studies show that anxiety disorders are quite prevalent among older persons. In an overview, the 6 months prevalence of anxiety in older persons was estimated at 10% (Flint, 1994). Although prevalence rates for major depression among older persons are generally lower (ranging between 0.5 and 3%), an estimated 10% to 15% suffers from sufficient depressive symptoms to have a significant diminished quality of life and increased medical consumption (Beekman et al., 1999). Similar findings regarding the impact of anxiety disorders in late life have been reported (de Beurs et al., 1999; Flint, 1994). Commonly, a strong association between depression and anxiety is reported; correlation coefficients between self-report measures for depression and anxiety usually range from r=0.50 to r=0.70. Although this may in part be explained by lack of specificity of these measures for depression and anxiety (Clark and Watson, 1991), there is also considerable true overlap in symptoms between both disorders. Furthermore, there is a high comorbidity of anxiety and depression at the disorder or caseness level (Barbee, 1998), especially in late life (Kirby et al., 1997). Among elderly, about 50% of the clinically depressed suffers from comorbid anxiety disorders, and 25% of patients with anxiety suffers from major depression (Beekman et al., 2000). Finally, depression and anxiety respond equally well to similar therapeutic interventions, such as Selective Serotonin Reuptake Inhibitors, further blurring a distinction between both disorders. Current knowledge on the interplay of depression and anxiety in late life is limited. The scarce available data suggest that distinctive features of depression and anxiety become even less pronounced with rising age (MacKinnon et al., 1994). Findings such as these raise doubt on the usefulness of distinguishing between both conditions (Tyrer, 1989) and made some researchers suggest that a dimensional approach to affective psychopathology would do more justice to the state of affairs (Andrews, 1996; Ormel et al., 1995). To explain the high comorbidity and high concurrence of symptoms in anxiety and depression, Clark and Watson (1991) proposed a tripartite model, postulating three dimensions for symptoms commonly found in patients suffering from depression and/or anxiety disorders. The model states that anxiety and depression share a common component, negative affectivity (NA) with presumably a genetic/temperamental background (Clark and Watson, 1991; Watson and Clark, 1984). This general dimension can be considered as a core component of anxiety disorders and depressive disorders (Brown et al., 1998) and consists of symptoms such as feeling depressed, hopeless, sad, afraid, and nervous. The second dimension comprises symptoms specific to depression, which they named positive affect (PA). Relatively unique to depression is a lack of emotional states such as feeling good about oneself and feeling optimistic or successful; the third dimension consists of symptoms specific to anxiety, predominantly somatic symptoms of anxiety and therefore named physiological hyperarousal or Somatic Anxiety (SA) (Keogh and Reidy, 2000). An important aspect on which these dimensions of anxious and depressive symptomatology may be distinguished are the risk profiles they are associated with. Some vulnerability factors or life events may be uniquely associated with negative affect, while others rather predict lack of positive affect or somatic anxiety symptoms. The predominant model for the etiology of psychopathology is the diathesis–stress model (Zuckerman, 1999). Initially formulated to describe risk factors of schizophrenia, the model has been applied with success to study etiological factors in depression and anxiety, where it became known as the vulnerability–stress model (Brown and Harris, 1978; Goldberg and Huxley, 1992). Basically, the model states that destabilization (getting symptoms) is the result of long-lasting vulnerability factors, acting in concert with exposure to environmental stressors, usually one or more highly stressful events. Regarding the question why some persons develop depression whereas others develop anxiety, Finlay-Jones and E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 Brown (Finlay-Jones and Brown, 1981) suggested that the type of event may be a decisive factor: stressful life events involving loss (e.g., death of a loved one, retirement) are more likely to lead to depression, whereas stressful events involving threat (e.g., being victimized by crime) are more likely to lead to anxiety. Furthermore, interactions of these life events and specific vulnerability factors are suggested. Events involving loss would be especially troublesome for subjects with a low sense of mastery, whereas threat events would have a stronger impact on subjects high in neuroticism. Put differently, the model proposes a moderator effect of mastery and neuroticism on the destabilizing effect of stressful life events. The notion that low mastery or an external locus of control is important for the development of depression is founded in the work of Rotter (1975) and can also be related to Seligman’s concept of learned helplessness in animal models of depression (Seligman, 1978). Also from other theoretical perspectives, a relation is proposed between depression on the one hand and low mastery, low perceived control, or dependency on the other (Beck, 1996; Blatt and Zuroff, 1992). Experiencing adverse events increases the risk for depression (Paykel, 2003), especially in those persons who feel limited control over their life’s course (Brown and Siegel, 1988). High neuroticism is a notorious risk factor for both depression and anxiety (Clark et al., 1994; Eysenck and Eysenck, 1991; Zuckerman, 1999). The importance of threat operationalized as the perception of looming vulnerability for the development of anxiety has been underlined by Riskind (1997). In the present study, the prognostic value of the risk factors is assessed for change in negative affect, positive affect, and somatic anxiety symptoms. Risk profiles for these three groups of symptoms are compared, evaluating main effects of risk factors and their interactions. The main research questions are the following: which factors predict a significant increase of emotional problems in late life and what decides whether the outcome of destabilization is predominantly negative affect, lack of positive affect, or somatic anxiety? These research questions are broken down into the following: (1) Are different long-standing vulnerability factors involved for the three dimensions? (2) (3) 55 Are certain types of stressful life events (loss vs. threat events) specific for developing negative affect, decreased positive affect, or somatic anxiety symptoms? Do stressful events and vulnerability factors (mastery*loss and neuroticism*threat) interact in the predicted manner, or do they rather have an additive effect on destabilization? Data from the Longitudinal Aging Study Amsterdam (LASA; Deeg et al., 1993) were used to investigate risk factors for changes in negative affect, positive affect, and somatic anxiety in the elderly. LASA is a longitudinal community-based study in a large representative sample of adults aged 55 to 85. In a previous cross-sectional study (Beekman et al., 2000), clear differences were found between subjects with a diagnosis of major depression and subjects with anxiety disorders. Major depression was associated only with age and an external locus of control, whereas having an anxiety disorder was associated with a wide range of vulnerability factors and stressors. Subjects with a comorbid condition had again a quite distinct risk profile from those with pure depression or anxiety (Beekman et al., 2000). We also investigated the development of depression and anxiety in longitudinal data from two waves of the LASA study. We used cutoff scores on symptom measures to delineate adults as either anxious, depressed, or both. Results revealed that depression and anxiety have many risk factors in common, but specific risk factors were also found, especially in subjects with both depression and anxiety (de Beurs et al., 2001). The present study builds on these findings by using longitudinal data from three waves of LASA. A different data analytic strategy is used, no longer dichotomizing patients as disordered (yes–no) but using the full range of scores on assessment instruments. This approach is more suitable to test the dimensional approach to psychopathology. Moreover, this way of analyzing the data yields more statistical power by preserving the continuous character of the symptom scores. Emotional problems were conceptualized according to the tripartite model for negative emotional states, distinguishing negative affect, lack of positive affect, and somatic anxiety. 56 E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 2. Method Table 1 Descriptives of the sample at baseline and 6 years later 2.1. Sample and procedure Data for LASA were collected by interviewing participants in their homes by trained and intensively supervised interviewers. All interviews were audiotaped to allow random quality checks. In 1992, a random sample of older (55–85 years) men and women stratified for age and sex was drawn from the population registers of 11 municipalities in three regions of The Netherlands. Older men were initially oversampled to ensure sufficient participants in these strata for a later phase of the study. In the first cycle of LASA (T 1), 3107 participants were interviewed. This sample has been described extensively in previous publications of LASA (Beekman et al., 1995; de Beurs et al., 1999; Deeg and Westendorp-de Seriere, 1994). Three years later (T 2), the participants were contacted again, and 2302 (74%) were willing and still able to be interviewed again. When we contacted the participants again in 1998–1999, 2076 (66.8%) of the initial sample could be interviewed again. Of the 1031 nonrespondents, 761 (24.5% of the T 1 sample) had died, 81 (2.6%) were too ill or cognitively impaired to be interviewed, 160 (5.2%) indicated that they were no longer interested in participating in the study, and 29 (0.9%) could not be contacted. Due to item nonresponse on measures for depression and anxiety at T 1, T 2, or T 3, a further 239 subjects were lost, leaving 1837 subject for whom scores on emotional functioning were available at all time points (60.1% of the T 1 participants). The present analysis was performed on this sample. The 1270 nonrespondents were compared with the 1837 participants on key variables to check for selective attrition. No significant difference was found between participants and nonrespondents regarding anxiety symptoms, but lower positive affect [t(3105)= 7.43; pb0.001] and higher negative affect [t(3105)= 6.59; pb0.001] at T 1 did make attrition more likely. For other variables, we also found an association with attrition: nonrespondents were more likely to be male [v 2(1)=22.7; pb0.001], unmarried [v 2(1)=32.3; pb 0.001], older [t(3105)=21.8; pb0.001], had lower socioeconomic status (SES; t=9.05; pb0.001), less education [v 2(2)=60.7; pb0.001], and a lower score on the Mini Mental State Exam (MMSE) [t(3105)=23.5; N Negative affect, mean (S.D.) (Lack of) Positive affect, mean (S.D.) Somatic anxiety, mean (S.D.) Sex Male Female Marital status Married Unmarried Age, mean (S.D.) SES, mean (S.D.) Education Low Middle High MMSE, mean (S.D.) T1 T3 3107 1.72 (2.87) 3.28 (3.04) 1837 1.80 (2.74) 3.53 (3.02) 2.53 (3.30) 3.12 (3.58) 1506 (48.5%) 1601 (51.5%) 816 (44.4%) 1021 (55.6%) 1942 1165 70.8 33.7 (62.7%) (37.3%) (8.8) (19.3) 1236 601 71.1 29.9 (67.3%) (32.7%) (8.1) (19.0) 1376 1370 353 26.6 (44.3%) (44.1%) (11.4%) (3.8) 713 908 216 27.4 (38.8%) (49.4%) (11.8%) (2.9) pb0.001]. Thus, the subjects who were lost from T 1 to T 3 comprised the less healthy and worse functioning part of the initial participants at T 1. Baseline characteristics of the baseline sample and the final study sample are presented in Table 1. 3. Measures 3.1. Negative affect, lack of positive affect, and somatic anxiety symptoms In LASA, depressive symptoms were measured with the Center for Epidemiological Studies Depression scale (CES-D; Radloff, 1977). This 20-item scale comprises Likert-type items which describe feelings such as depression, hopelessness, and the blues. The respondent is asked to indicate whether these feelings were experienced in the past week on a four-point scale, ranging from 0=brarely or neverQ to 3=balways or almost always.Q Two subscales of the CES-D were used. To assess negative affect, we used a subscale of seven items: having the blues, feeling depressed, life is a failure, feeling fearful. Lack of positive affect was assessed with four items: feeling as good as others, hopeful about the future, being happy, enjoying life. Thus, we limited use of the available CES-D data to two of the four subscales of this instrument. These two subscales mea- E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 sure the closest intended concepts of negative affect and lack of positive affect. Anxiety was measured with the anxiety subscale of the Hospital Anxiety and Depression Scale (HADS-A; Zigmond and Snaith, 1983). The HADS-A comprises seven Likert-type items in which the respondent is asked to indicate whether he/she has experienced feelings such as restlessness, tenseness, or panic on a scale scoring from 0=bseldom or neverQ to 3=balways or almost always.Q The anxiety subscale score has a theoretical range from 0 to 21. 57 of the events. Weights for various life events were derived from Tennant and Andrews (Tennant and Andrews, 1976). As both composites had a skewed and peaked distribution, they were log-transformed to approach a normal distribution. All assessment instruments used in the study had been previously validated in The Netherlands or their psychometric properties had been evaluated for their use in the older population in LASA pilot studies (Deeg et al., 1993). 3.4. Statistical analysis 3.2. Risk factors Two stable psychological characteristics of the participants were assessed: mastery and neuroticism. Mastery was assessed in the interview with a fiveitem scale adapted from Pearlin and Scooler (1978). A higher score means a more external locus of control or less mastery. Neuroticism (15 items) was measured through the abbreviated subscale of the Dutch Personality Inventory (Luteijn et al., 1985). This self-report scale was completed after the interview and mailed in by the respondent. Not all participants complied: 138 of the 1837 participants (8%) failed to return fully completed questionnaires. Nonresponse on the self-report data was not related to sex of the participant [v 2(1)=0.92, p=0.34] but was related to higher age [more nonresponse in older participants, t(3105)=7.69; p=0.002]. 3.3. Life events In the T 2 and T 3 interviews, it was assessed retrospectively whether stressful life events had occurred in the 3-year time interval prior to the interview. The following stressful events were assessed: illness of one’s partner, death of one’s partner, illness of a relative, death of a relative, a major conflict with others, income loss (of at least $50 a month), victimized by crime, and relocation. To investigate the impact of these life events, we constructed two composites: one for events associated with loss (e.g., death of a partner or relative) and one for events associated with threat (e.g., illness of a partner or relative, crime). Composites scores were made by differentially weighting life events and combining them in one score, representing the impact To examine the association between neuroticism, mastery, and the outcome variables, multivariate multilevel analysis were carried out. The general idea behind using multilevel analysis for multivariate problems is that the measurements of the different outcome variables can be seen as dclusteredT within the subject (Goldstein, 1995). As this was a longitudinal study, three levels of analysis can be distinguished: (1) the outcome variables negative affect, lack of positive affect, and anxiety, (2) the observations at different time points, and (3) the participants. This multivariate multilevel analysis is basically a multivariate linear regression analysis. There is, however, an additional level because we have repeated observations within each participant. In addition to the main effects for neuroticism and mastery, the interaction between neuroticism*threat and mastery*loss were added to the models. In a second analysis, the association between the predictor variables and the three outcome variables were examined in separate multilevel analyses. Figs. 1 and 2 present the associations of main interest for a simple model with one outcome variable and for the Fig. 1. Model for the association between predictors and change in emotional functioning. 58 E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 Fig. 2. Hypothesized model for the associations among predictors and three dependent variables. complete model. All analyses were adjusted for gender, age, socioeconomic status, and health (number of somatic diseases) and were performed with MLwiN (version 1.10.0007; Centre for Multilevel Modeling, Institute of Education, London, UK). For socioeconomic status, we used a weighted score composed of level of education, occupation, and income (range 0–100; van Tilburg et al., 1995). 4. Results First, the complete model was analyzed. Results are presented in Table 2. Overall, adjustment for control variables (age, sex, and SES) did not change the results of the analyses in a meaningful way. The results of analysis of the entire model revealed main Table 2 Results of multivariate multilevel analyses (main effects) Adjusteda Unadjusted 95% CI B Mastery 0.15 Neuroticism 0.10 p 0.13 to b0.001 0.17 0.09 to b0.001 0.11 B 95% CI 0.13 0.09 p 0.11 to b0.001 0.15 0.08 to b0.001 0.11 a Adjusted for sex, age, socioeconomic status, and health of the respondent. effects of mastery ( pb0.001) and neuroticism ( pb0.001) on deterioration of emotional functioning (the overall outcome, not distinguishing negative affect, lack of positive affect and anxiety). Less mastery and higher neuroticism were significantly associated with decreased emotional well-being (see Table 2). Furthermore, a significant interaction effect of mastery and loss events (B=0.002; CI=0.001– 0.003; p=0.002) was found. Apparently, loss events have a greater impact for participants who report less mastery. This finding is in accordance with our prediction. However, the predicted interaction between neuroticism and threat events is not supported by the data since this interaction failed to reach significance (B=0.001, p=0.44). Finding significant effects in the overall model justifies a closer look into the associations by analyzing three models separately, one for each dependent variable. Table 3 presents the results for main effects in the separate models for general negative affect, lack of positive affect, and somatic anxiety. Regarding main effects, the results reveal that general negative affect has a positive association with neuroticism and an equally strong but negative association with mastery (B=0.09 and B= 0.14, respectively). Lack of positive affect (specific to depression) was associated with low mastery. Neuroticism was also 59 E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 Table 3 Results of multilevel analyses for three outcomes (main effects and interaction effects) Adjusteda Unadjusted B 95% CI B p Negative affect Mastery 0.14 0.11 to 0.16 b0.001 Neuroticism 0.09 0.07 to 0.10 b0.001 Interaction effects Significant interaction between loss and mastery ( pb0.011) No significant interaction between threat and neuroticism ( p=0.90) Lack of positive affect Mastery 0.17 0.14 to 0.19 b0.001 Neuroticism 0.09 0.08 to 0.11 b0.001 Interaction affects No significant interaction between loss and mastery ( p=0.34) No significant interaction between threat and neuroticism ( p=0.25) Somatic anxiety Mastery 0.14 0.11 to 0.17 b0.001 Neuroticism 0.13 0.11 to 0.15 b0.001 Interaction affects Significant interaction between loss and mastery ( pb0.001) No significant interaction between threat and neuroticism ( p=0.98) a 95% CI p 0.12 0.09 0.10 to 0.15 0.07 to 0.10 b0.001 b0.001 0.15 0.09 0.12 to 0.17 0.07 to 0.10 b0.001 b0.001 0.13 0.13 0.10 to 0.16 0.11 to 0.15 b0.001 b0.001 Adjusted for sex, age, socioeconomic status, and health of the respondent. associated with lack of positive affect but according to the regression coefficients (B) to a somewhat lesser extent than mastery (B=0.09 vs. B= 0.17). In contrast, high somatic anxiety was equally strongly associated with high neuroticism and low mastery (B=0.13 and B= 0.14). Controlling for sex, age, education, SES, and health did not alter these results in a meaningful way. Regarding interaction affects of personality characteristics and both type of events, the results reveal some significant associations. No modifying effect of mastery or neuroticism on events leading to decreased positive affect is apparent in the data. Mastery moderates the impact of loss on somatic anxiety ( pb 0.001), and the same interaction effect is found with negative affect as outcome variable ( p=0.011). The predicted interaction effect of threat events and neuroticism on somatic anxiety was not found in the data. In general, the findings of the overall model and the three separate analyses converge: of the two predictors, mastery has the strongest association with the outcome variables. However, when comparing the strength of the associations of mastery and neuroticism for the three dimensions of psychopathology according to the tripartite model, neuroticism has the strongest association with somatic anxiety, a finding in accordance with the hypothesized distinctness of these three dimensions. The interaction of neuroticism and threat events is not found in the overall model and neither in the analyses of dependent variables separately. Instead, mastery modifies the effects of loss events on somatic anxiety ( pb0.001) and, to a lesser extent, on negative affect ( p=0.011). The direction of the interaction effect of mastery and loss events on anxiety is as one would expect: high mastery protects for the effect of loss events. 5. Discussion Overall, support has been found for the proposed model predicting different pathways for symptoms of depression, anxiety, and general negative affect. Increases in anxiety and in negative affect were both associated with an interaction effect of life events involving loss and low mastery. Contrary to our hypothesis, no significant interaction effect was found between neuroticism and negative life events involving 60 E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 threat. What do the present finding reveal about the validity of the tripartite model for anxiety and depression in late life? The findings suggest clearly explanatory power for the present model. By demonstrating different effects for risk factors on depression, anxiety, and general negative affect, the findings underline the validity of separating out these distinct groups of symptoms according to the tripartite model. Furthermore, distinguishing vulnerability in lack of mastery and high neuroticism proved fruitful given the distinct outcomes for these variables: mastery appeared a better predictor of decreased negative affect than neuroticism. The same holds for distinguishing between loss and threat events. In spite of the general support for the proposed model, differences among predicting variables for distinct outcomes were modest (e.g., B= 0.09 vs. B=0.14). In interpreting this outcome, one should also bear in mind that the present model for developing emotional problems or destabilization in late life is far from complete. The decisive factor for whether the outcome of life events and vulnerability is depression or anxiety may very well be of a biological (e.g., genetic (Kendler et al., 2001)) or social nature. Regarding the latter, the buffering effect of an extensive social network on the impact of negative life event is well documented, in part by studies on data from LASA (Penninx et al., 1998). Regarding the former, we have previously reported on the profound impact of physical health on anxiety (de Beurs et al., 2000). Other biological factors, such as HPA-axis functioning in depression, are currently under study. These factors were not included in the present analysis. More conclusive evidence regarding the utility of the tripartite model for a dimensional conceptualization of psychopathology will come from studies linking the dimensions to biological factors or to response to interventions in clinical populations. The present findings may have been influenced by a potential lack of variance in our composite scores for events. Particularly, the composite for threat events yielded low scores. For instance, very few subjects reported being victimized by crime. Furthermore, by their very nature, loss events may have a stronger impact than stress events. For instance, mourning a close relative has a greater impact on emotional functioning as compared to the same relative falling ill (a threat event mirroring losing a relative). Furthermore, life events were assessed by putting information from different sections of the interview together. Potentially, a more thorough assessment of events would have yielded different results. (e.g., the life history calendar approach to assessing life events (Caspi et al., 1996). Furthermore, the lack of support for some of our predictions, especially those regarding distinct predictions for somatic anxiety and lack of positive affect may be due to the fact that the present study was undertaken in a nonclinical elderly population. The present sample was population-based. Moreover, as time progressed, it comprised predominantly of those with the best chances of survival. The hypotheses may get more support when studied in a patient group, where symptom profiles will be more clearly distinct and where potentially more aggravating life events will be reported. Another limitation to the present study was the use of subscales of the CES-D and the HADS-A to measure symptoms according to the tripartite model. Although both instruments are well researched and valid, they are not the best choice for an optimum distinction between the symptoms of anxiety and depression. The subscales of the MASQ (Watson et al., 1995) or the Beck Depression Inventory and the Beck Anxiety Inventory (Beck et al., 1961; Beck and Steer, 1990) show less concordance between anxiety and depression and may have yielded more pronounced differences between correlates of depression and anxiety than the subscales of the CES-D and HADS-A. Future research with a dependent variable that is better suited for distinguishing the three factors of the tripartite model may yield more clear-cut results. In addition, the predictive power of the model might be improved by incorporating additional predictors. Promising predictors could be found in the biological realm (Charney, 2004) or in other psychological traits, such as attributional style or cognitive biases, that produce vulnerability for the development of psychopathology. Currently, in The Netherlands, three universities have started a collaborative effort to study (determinants of) the course of depression and anxiety in primary and secondary care patients in a longitudinal design (NESDA). A large number of patients will be repeatedly assessed over a 10-year time span. Data will be gathered on diagnostic status, symptoms of anxiety and depression, biological parameters, such as E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 cortisol, and psychological parameters (personality traits and other vulnerability factors such as anxiety sensitivity). Availability of these data will allow testing a more comprehensive model for different age groups. 6. Conclusion The present findings lend partial support to the diathesis–stress model. Low mastery and neuroticism are both associated with a decrease in emotional wellbeing over time, mastery most strongly with depression, and neuroticism with anxiety. The predicted modifying effect of neuroticism on the association between threat events and anxiety was not corroborated by the data. However, the significant interaction between mastery and loss events delineates those specifically at risk: individuals who experience negative life events and tend not to feel in control of their lives. A possible clinical implication of this finding is that efforts to increase perceived control, e.g., through cognitive behavior therapy, may enhance the outcome of pharmacological treatment for depression in late life and may prevent relapse after successful therapy. Finally, the study findings support distinguishing three types of symptoms: negative affect, lack of positive affect, and anxiety. References Andrews, G., 1996. Comorbidity and the general neurotic syndrome. Br. J. Psychiatry 168, 76 – 84. Barbee, J.G., 1998. Mixed symptoms and syndromes of anxiety and depression: diagnostic, prognostic, and etiologic issues. Ann. Clin. Psychiatry 10, 15 – 29. Beck, A.T., 1996. Beyond belief: a theory of modes, personality and psychopathology. In: Salkovscis, P.M. (Ed.), Frontiers of Cognitive Therapy. Guilford Press, New York, pp. 265 – 290. Beck, A.T., Steer, R.A., 1990. Manual for the Beck Anxiety Inventory. Psychological, San Antonio, TX. Beck, A.T., Ward, C.H., Mendelson, M., Mock, J.E., Erbaugh, J., 1961. An inventory for measuring depression. Arch. Gen. Psychiatry 4, 561 – 571. Beekman, A.T.F., Deeg, D.J.H., van Tilburg, T., Smith, J.H., Hooijer, C., van Tilburg, W., 1995. Major and minor depression in later life: a study of prevalence and risk factors. J. Affect. Disord. 34, 41 – 49. Beekman, A.T., Copeland, J.R., Prince, M.J., 1999. Review of community prevalence of depression in later life. Br. J. Psychiatry 174, 307 – 311. 61 Beekman, A.T.F., de Beurs, E., van Balkom, A.J.L.M., Deeg, D.J.H., Van Dyck, R., van Tilburg, W., 2000. Anxiety and depression in later life: co-occurrence and communality of risk factors. Am. J. Psychiatry 157, 89 – 95. Blatt, S.J., Zuroff, D.C., 1992. Interpersonal relatedness and selfdefinition: two prototypes for depression. Clin. Psychol. Rev. 12, 527 – 562. Brown, G.W., Harris, T.O., 1978. Social Origins of Depression. Tavistock, London. Brown, J.D., Siegel, J.M., 1988. Attributions for negative life events and depression: the role of perceived control. J. Pers. Soc. Psychol. 54, 316 – 322. Brown, T.A., Chorpita, B.F., Barlow, D.H., 1998. Structural relationships among dimensions of the DSM-IV anxiety and mood disorders. J. Abnorm. Psychology 107, 179 – 192. Caspi, A., Moffitt, T.E., Thornton, A., Freedman, D., Amell, J.W., Harrington, H., Smeijers, J., Silva, P.A., 1996. The life history calendar: a research and clinical assessment method for collecting retrospective event-history data. Int. J. Methods Psychiatr. Res. 6, 101 – 114. Charney, D.S., 2004. Psychobiological mechanisms of resilience and vulnerability: implications for successful adaptation to extreme stress. Am. J. Psychiatry 161, 195 – 216. Clark, L.A., Watson, D., 1991. A tripartite model of anxiety and depression: psychometric evidence and taxonomic implications. J. Abnorm. Psychology 100, 316 – 336. Clark, L.A., Watson, D., Mineka, S., 1994. Temperament, personality, and the mood and anxiety disorders. J. Abnorm. Psychology 103, 103 – 116. de Beurs, E., Beekman, A.T., van Balkom, A.J., Deeg, D.J., Van Dyck, R., van Tilburg, W., 1999. Consequences of anxiety in older persons: its effect on disability, well-being and use of health services. Psychol. Med. 29, 583 – 593. de Beurs, E., Beekman, A.T., Deeg, D.J., Van Dyck, R., van Tilburg, W., 2000. Predictors of change in anxiety symptoms of older persons: results from the Longitudinal Aging Study Amsterdam. Psychol. Med. 30, 515 – 527. de Beurs, E., Beekman, A., Geerlings, S., Deeg, D., Van Dyck, R., van Tilburg, W., 2001. On becoming depressed or anxious in late life: similar vulnerability factors but different effects of stressful life events. Br. J. Psychiatry 179, 426 – 431. Deeg, D.J.H., Westendorp-de Seriere, M., 1994. Autonomy and Well-Being in the Aging Population: I. Report from the Longitudinal Aging Study Amsterdam 1992–1993. VU University Press, Amsterdam. Deeg, D.J.H., Knipscheer, C.P.M., van Tilburg, W., 1993. Autonomy and well-being in the aging population: concepts and design of the Longitudinal Aging Study Amsterdam. NIGOverzichtstudies, 7 edn. Nederlands Instituut voor de Gerontologie, Bunnik. Eysenck, H.J., Eysenck, S.B.G., 1991. Eysenck Personality Scales. Hodder and Staughton, London. Finlay-Jones, R., Brown, G.W., 1981. Types of stressful life event and the onset of anxiety and depressive disorders. Psychol. Med. 11, 803 – 815. Flint, A.J., 1994. Epidemiology and comorbidity of anxiety disorders in the elderly. Am. J. Psychiatry 151, 640 – 649. 62 E. de Beurs et al. / Journal of Affective Disorders 84 (2005) 53–62 Fuentes, K., Cox, B.J., 1997. Prevalence of anxiety disorders in elderly adults: a critical analysis. J. Behav. Ther. Exp. Psychiatry 28, 269 – 279. Goldberg, D.P., Huxley, P., 1992. Common Mental Disorders: A Bio-Social Model. Routledge, London. Goldstein, H., 1995. Multilevel Statistical Models. Wiley, New York. Kendler, K.S., Thornton, L.M., Prescott, C.A., 2001. Gender differences in the rates of exposure to stressful life events and sensitivity to their depressogenic effects. Am. J. Psychiatry 158, 587 – 593. Keogh, E., Reidy, J., 2000. Exploring the factor structure of the mood and anxiety symptom questionnaire (MASQ). J. Pers. Assess. 74, 106 – 125. Kirby, M., Bruce, I., Radic, A., Coakley, D., Lawlor, B.A., 1997. Mental disorders among the community-dwelling elderly in Dublin. Br. J. Psychiatry 171, 369 – 372. Luteijn, F., Starren, J., van Dijk, H., 1985. Manual for the Dutch Personality Quaestionnaire [Handleiding bij de NPV]. Swets and Zeitlinger, Lisse, The Netherlands. MacKinnon, A.J., Christensen, H., Jorm, A.F., Henderson, A.S., Scott, R., Korten, A.E., 1994. A latent trait analysis of an inventory designed to detect symptoms of anxiety and depression using an elderly community sample. Psychol. Med. 24, 977 – 986. Ormel, J., Oldehinkel, A.J., Goldberg, D.P., Hodiamont, P.P., Wilmink, F.W., Bridges, K., 1995. The structure of common psychiatric symptoms: how many dimensions of neurosis? Psychol. Med. 25, 521 – 530. Paykel, E.S., 2003. Life events and affective disorders. Acta Psychiatr. Scand., Suppl., 61 – 66. Pearlin, L.J., Scooler, C., 1978. The structure of coping. J. Health Soc. Behav. 19, 2 – 21. Penninx, B.W., Tilburg, T., Boeke, A.J.P., Deeg, D.J.H., Kriegsman, D.M.W., Eijk, J.T.M., 1998. Effects of social support and personal coping resources on depressive symptoms: different for various chronic diseases? Health Psychol. 17, 551 – 558. Radloff, L.S., 1977. The CES-D scale: a self-report depression scale for research in the general population. Appl. Psychol. Meas. 1, 385 – 401. Riskind, J.H., 1997. Looming vulnerability to threat: a cognitive paradigm for anxiety. Behav. Res. Ther. 35, 685 – 702. Rotter, J.B., 1975. Some problems and misconceptions related to the construct of internal versus external control of reinforcement. J. Consult. Clin. Psychol. 43, 56 – 67. Seligman, M.E., 1978. Learned helplessness as a model of depression. Comment and integration. J. Abnorm. Psychology 87, 165 – 179. Tennant, C., Andrews, G., 1976. A scale to measure the stress of life events. Aust. N. Z. J. Psychiatry 10, 27 – 32. Tyrer, P., 1989. General neurotic syndrome and mixed anxietydepressive disorders. In: Tyrer, P. (Ed.), Classification of Neurosis. Wiley, New York, NY, pp. 132 – 164. van Tilburg, T., Dykstra, P., Liefbroer, A.C., Broese van Groenou, M., 1995. Sourcebook of Living Arrangements and Social Networks of Older Adults in The Netherlands. VU University Press, Amsterdam. Watson, D., Clark, L.A., 1984. Negative affectivity: the disposition to experience aversive emotional states. Psychol. Bull. 96, 465 – 490. Watson, D., Weber, K., Assenheimer, J.S., Clark, L.A., Strauss, M.E., McCormick, R.A., 1995. Testing a tripartite model: I. Evaluating the convergent and discriminant validity of anxiety and depression symptom scales. J. Abnorm. Psychology 104, 3 – 14. Zigmond, A.S., Snaith, R.P., 1983. The hospital anxiety and depression scale. Acta Psychiatr. Scand. 67, 361 – 370. Zuckerman, M., 1999. Vulnerability to Psychopathology. American Psychological Association, Washington, DC.