To evaluate how a prior affective disorder (major depression or generalized anxiety disorder) affects current fatigue among individuals with rheumatoid arthritis (RA). To determine whether that relationship is mediated by self-efficacy expectations.
Forty-eight RA patients with a prior affective disorder and 74 without a history of affective disorder completed a mailed questionnaire that included the Multidimensional Assessment of Fatigue and indicators of neuroticism and self efficacy.
RA patients with a history of affective disorder reported higher levels of fatigue than those with no previous affective disturbance. Controlling for neuroticism and self efficacy, affective disorder history continued to predict current fatigue. Mediational analyses revealed both direct and indirect effects (via self efficacy) of history of affective disorder on the experience of fatigue in RA.
History of affective disorder independently predicts higher levels of fatigue in RA patients, and self efficacy plays a mediating role in this relationship.
Although pain is the most common and frequently studied feature of rheumatoid arthritis (RA), fatigue is a significant source of distress and disability for many RA patients. The vast majority of RA patients complain of fatigue (1), and more than 40% of all RA patients experience clinically significant levels of fatigue on a daily basis (2, 3). The chronic fatigue associated with RA is not reliably reduced by sleep (4), and its vagaries often breed frustration and challenge RA patients' coping efforts. Although fatigue can be affected by disease factors, it is not generally considered a direct function of disease activity.
The demands of managing and coping with fatigue in RA are very different from those associated with RA pain. Managing pain generally consists of developing new ways to perform activities in a manner to reduce pain (4) and increase adaptation to illness and tolerance for activity demands. Efforts to relieve fatigue, such as sleeping or attempting to conserve energy, are often insufficient (5). These ineffective efforts can lead to an increased sense of helplessness (4), which is a frequent correlate of depression (6).
Several studies have examined the relationship between RA symptoms and depression. Depressive symptoms have been linked to more severe disease (6), higher levels of daily joint pain (7), and increased levels of fatigue (1). A related line of inquiry has begun to examine the ways in which a previous affective disorder may influence subsequent health outcomes, including RA symptoms (8). In this article, we describe a study that examined the extent to which a history of major depression or generalized anxiety disorder predicted subsequent levels of fatigue in RA. We also evaluated a potential mediator of this association.
Shea et al (9) noted that individuals with a history of affective disorder are characterized by high levels of neuroticism. Neuroticism (N) is the general tendency to experience a variety of negative emotions, including fear, anger, sadness, and guilt. High N individuals tend to be in a less pleasant mood, have lower self esteem, and are more impulsive, vulnerable, and helpless (10). Neuroticism appears to play an important role in the experience and manifestations of chronic illness, including RA. It has been linked with susceptibility to distress, difficulties with stress management (11), and less effective coping (12). Watson and Pennebaker (13) have suggested that the high N individual's hypervigilant, ruminative, and self-focused cognitive style may lead to the perceptual amplification of bodily sensations. Although N has yet to be studied specifically in relation to fatigue, we reasoned that by amplifying the perceived severity of symptoms and decreasing their effective management, N may influence the experience of fatigue in RA.
Prior affective disorder and current fatigue may also be linked through what may be a lingering consequence of the affective disturbance: an individual's diminished expectation that he or she can deal successfully with the demands of RA. According to Bandura (14), the self-efficacy (SE) construct captures this diminished expectation. Specifically, SE refers to a person's conviction that he or she can successfully execute the behavior required to produce a desired outcome. High SE in RA has been associated with increased pain tolerance (15), reduced depression, and lower levels of arthritis helplessness (6). Moreover, psychological interventions that enhance RA patients' SE also improve their health status (16).
Two opposing theories attempt to explain the relationship between personality factors (such as N) or self perceptions (such as SE) and history of affective disorder. The vulnerability model, or trait marker hypothesis, proposes that certain personality characteristics or self perceptions represent a risk factor for the onset of affective disorder. There has been some support for this model, particularly with regard to neuroticism (9). In contrast to the vulnerability model, Lewinsohn and colleagues (17) proposed that a first episode of depression creates a lasting change in personality and ways of interpreting everyday experience, which results in a residual deficit, or psychological scar. This scar, which is presumed to predispose the individual to future episodes of affective disturbance, may also influence the natural history or symptom trajectory of chronic illness by affecting how the patient responds to symptoms.
Fifield et al (18) considered this possibility in their longitudinal study of RA patients. They examined the experience of fatigue in RA patients with a history of major depression (MD), generalized anxiety disorder (GAD), or both compared with patients who had never experienced an affective disorder. Patients with a history of affective disorder reported 10% more fatigue at the beginning of the study and their fatigue level remained higher over the subsequent 8 years. Both recent and distant episodes of affective disturbance were related to greater RA fatigue.
Although Fifield et al (18) demonstrated that an episode of MD or GAD predicted levels of fatigue years later, their study had several limitations. First, their indicator of fatigue was limited to a single-item visual analog scale. Second, due to the prevalence rates of lifetime affective disorder, fewer than 20% of the participants in that study had a history of affective disorder. Finally, Fifield et al did not examine whether the link between a history of affective disorder and current fatigue was mediated by SE, a factor that appears closely associated with the vulnerability and scar hypotheses.
Our current study addressed each of these limitations. Specifically, we 1) examined several dimensions of fatigue in an effort to identify distinguishing characteristics of the fatigue experience of RA patients with and those without a history of affective disorder; 2) included a larger proportion of formerly depressed or anxious individuals; and 3) evaluated SE as a potential mediator and neuroticism as a potential moderator of the association between past affective disorder and current fatigue.
PATIENTS AND METHODS
RA patients were recruited from the National Rheumatoid Arthritis Study, a 10-year prospective panel study that took place from 1988 to 1998 (19). The original sample of 988 RA patients was recruited through private rheumatologists who were fellows of the American College of Rheumatology. At the conclusion of the study, 462 (47%) of the original participants remained. In 1999, 2 independent followup studies were conducted with those remaining eligible participants. The sample was split into 2 groups, resulting in 211 participants who were eligible for recruitment into this study. In dividing the sample of 462 into 2 groups, all of the participants who had endorsed prior affective disturbance (n = 91) were assigned to the current study.
Due to death, changed diagnosis, and returned mail marked undeliverable, 195 (92%) of the targeted 211 RA subjects were eligible for participation. Questionnaires were completed and returned by 122 of the 195 eligible participants (63%). In this sample, 48 (39.3%) participants had a history of affective disorder, i.e., MD or GAD, and 74 (60.7%) were without a history of affective disorder. These percentages are based on sample selection and recruitment and do not represent natural prevalence rates. Due to the small numbers of participants with a past diagnosis of GAD (n = 9 women) or comorbid MD and GAD (n = 6 women), these groups were collapsed with individuals with a history of MD (n = 33) into 1 group for analytic purposes. Table 1 presents participants' demographic characteristics.
Table 1. Study sample characteristics
Total (n = 122)
Mood history (n = 48)
No history (n = 74)
Age, mean ± SD years
59.9 ± 9.7
60.8 ± 10.2
58.5 ± 8.8
Education, mean ± SD years
13.3 ± 2.1
13.3 ± 2.0
13.3 ± 2.4
Illness duration, mean ± SD years
21.0 ± 8.3
22.0 ± 8.7
19.4 ± 7.4
We compared the sample used in this study with the remainder of the overall sample on demographic, illness-related, and mood variables as measured in year 8 (1996) of the original study. Year 8 data were selected on the basis of maintaining consistency with previously published work using this sample (18) and because data on history of affective disorder were specifically collected in that year's assessment.
Chi-square tests revealed no differences between groups on employment status, income, or marital status. Additionally, groups did not differ on RA disease stage, functional status of RA, or self-reported overall health in the last year. Finally, an independent samples t-test found no differences between groups on indicators of disability and current depression. Overall, there were no significant differences between the participants of this study and the complete RA sample.
The current sample had a mean ± SD age of 60 ± 9.67 years and included 9 men (7.4%) and 113 women (92.6%). The ethnic composition was predominantly white (94.3%). The mean ± SD level of education was 13.3 ± 2.14 years, and the majority were married. Duration of illness ranged from 11 to 47 years, with a mean ± SD of 20.96 ± 8.31 years. Demographic characteristics did not differ between affective history groups.
Affective disorder diagnosis, including Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) current and lifetime diagnoses of MD and GAD, was determined through a semistructured DSM-IV telephone interview in 1996–1997 using the Semi-Structured Assessment for the Genetics of Alcoholism (SSAGA) (20). The SSAGA provides a complete and detailed lifetime psychiatric history for adults aged 17 years and older. The SSAGA interview schedule covers the major Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised; DSM-IV; and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision defined axis I psychiatric disorders as well as antisocial personality disorder. The SSAGA has been shown to have good reliability for lifetime major depression (20). It is suitable for either telephone or face-to-face administration by lay interviewers (21).
Fatigue was measured using the Multidimensional Assessment of Fatigue (MAF) developed by Belza (22) to assess fatigue in RA samples. The MAF consists of 16 items that tap 4 dimensions of fatigue: severity, distress, degree of interference in activities of daily living, and timing. Responses are based on perceived fatigue patterns from the previous week. The 4 subscales are combined to create the global fatigue index (GFI), which ranges from 0 (no fatigue) to 50 (extreme fatigue). The MAF has demonstrated good convergent and discriminant validity with the Profile of Mood States fatigue and vigor subscales, respectively (1). Cronbach's alpha in our study sample for the overall scale was 0.92.
In addition to the MAF, the questionnaire contained several supplementary questions aimed at better understanding the nature and patterns of participants' fatigue. These questions related to qualitative aspects of fatigue, such as time of day fatigue is most bothersome, whether fatigue is continuous or intermittent, and whether fatigue precedes, follows, occurs during, or is unrelated to pain flares. Additionally, to assess of how fatigue compared with 2 other prominent symptoms of RA, pain and joint stiffness, subjects were asked to rate on a scale of 0–100 (from “not at all” to “extremely”) how bothersome each symptom had been for them over the past week.
Neuroticism was measured with the N scale of the NEO Five Factor Inventory (23). This 12-item self-report measure assesses the general tendency to experience a variety of negative emotions. Responses are made along a 5-point Likert scale ranging from “strongly agree” to “strongly disagree.” The NEO scales have been used extensively in arthritis research (12). Coefficient alpha in the current sample was 0.75.
Arthritis self-efficacy scales.
Lorig et al (24) constructed self-efficacy scales specifically designed for arthritis patients. These scales consist of 28 questions, beginning with the phrase, “How confident are you that you can… .” Responses are based on a 10-point Likert scale. Self efficacy is measured in relation to performing self-management behaviors (e.g., exercise regularly) to manage disease in general and to achieve outcomes (e.g., doing chores, participating in social and recreational activities, and managing symptoms). Lorig et al (24) reported test–retest reliability coefficients ranging from 0.82 to 0.89. Coefficient alphas for the SE scales in the current sample ranged from 0.84 to 0.96.
Participants were categorized into 2 groups based on the presence or absence of history of affective disorder according to results of the semistructured diagnostic telephone interviews conducted in 1996. Forty-eight participants with and 74 participants without prior affective disturbance responded to the questionnaire. Five weeks after the initial mailing from which there were 98 returns, a followup postcard reminder was sent to each person who had not returned the questionnaire. This followup produced 24 additional respondents. Individuals who completed and returned the questionnaire subsequently received compensation of $5.00.
The nature of fatigue in RA: descriptive findings.
Several questions included in the questionnaire were aimed at elucidating the features of fatigue in RA. Only 4% of the sample reported having no fatigue, whereas 61.2% described their fatigue as intermittent and 33.9% considered their fatigue to be continuous. All participants remained in the subsequent analyses. Although 27% of participants without a history of affective disorder described their fatigue as continuous, 45% of those with history of affective disorder described continuous fatigue (Pearson χ2 = 6.22, P < 0.05). More than three-fourths (76%) of the participants indicated that they had experienced an arthritis flare within the past month. They were asked about the pattern of their fatigue in relation to flares. More than half of the sample (51.7%) reported that their fatigue was unrelated to RA flares. A smaller proportion (28%) reported that their fatigue tended to occur during flares. Pearson chi-square tests revealed no differences between those with and without a history of affective disorder (Pearson χ2 = 6.22, P = 0.42).
The overall means ± SD for the single-item ratings of joint stiffness, pain, and fatigue were 48.12 ± 27.96 for joint stiffness, 51.55 ± 30.11 for pain, and 55.80 ± 29.41 for fatigue. Participants with a history of affective disorder rated each symptom as more bothersome than their counterparts without a prior affective disorder (P < 0.05). Table 2 presents the intercorrelations among affective history, RA symptoms, neuroticism, and self efficacy. Subsequent reports of fatigue scores are based on the multidimensional GFI rather than the single-item fatigue rating.
Table 2. Means, standard deviations, and intercorrelations among affective history, RA symptoms, neuroticism, and self efficacy*
Pain, fatigue, and joint stiffness are based on 0–100 numerical rating scale. GFI is based on the Multidimensional Assessment of Fatigue (range 0–50). Neuroticism is based on NEO Five Factor Inventory (range 12–44). SE total is based on the sum of 7 SE domain scores (range 10–70). N = neuroticism; SE = self efficacy; GFI = global fatigue index.
GFI scores for the affective history group (mean ± SD 34.31 ± 10.0) were higher than scores in the group without a prior affective disorder (28.77 ± 9.55, t  = 3.03, P < 0.01). These differences set the stage for closer inspection of factors that influence this difference in fatigue scores between groups.
Affective disorder history, neuroticism, and fatigue.
As anticipated, the affective history group (mean ± SD 21.77 ± 6.84) had higher levels of N than the group with no previous affective disorder (Table 3; mean ± SD 16.75 ± 6.40, t = 4.09, P < 0.001). Similarly, N and fatigue showed a significant association, r(120) = 0.44, P < 0.001.
Table 3. Comparison of fatigue, neuroticism, self efficacy (SE), and fatigue in individuals with and without a previous affective disorder
Affective disorder history, self efficacy, and fatigue.
To maintain the integrity of Bandura's original concept of self efficacy as a behavior-specific construct, 7 planned 2-tailed independent sample t-tests were performed between the history of affective disorder and no history groups. Four SE indicators yielded group differences (Table 3): SE to complete chores t(119) = −2.16, P < 0.05; SE to obtain help, t(119) −2.67, P < 0.05; SE to manage or control depression, t(120) = −2.60, P < 0.01; and SE to participate in social and recreational activities, t(118) = −2.72, P < 0.01. In each of these comparisons, the affective history group had lower levels of SE. In contrast, there were no group differences in SE to manage symptoms, SE to exercise regularly, or SE to manage disease in general (P > 0.05). Each of the 7 SE domains was negatively correlated with fatigue (GFI), with r values ranging from −0.23 to −0.49.
The 7 SE subscales were also combined to create a total SE score. Although it may be argued that a total SE score obscures the behavioral specificity of each indicator, preliminary analyses demonstrated that the domain-specific SE subscales that differed significantly between affective history groups performed similarly to the SE total variable. Thus, the total SE score provided a meaningful and parsimonious measure of self efficacy in the analyses to follow. Total SE differed between groups, t(119) −2.71, P < 0.01. Additionally, total SE showed an inverse association with GFI, r(116) = −0.47, P < 0.001.
Self efficacy and neuroticism.
As expected, each of the 7 SE indicators was negatively and moderately correlated with N scores. Correlations ranged from −0.31 to −0.53, P < 0.01. Total SE was also negatively correlated with N scores (r = −0.55, P < 0.001).
Contributions of N, SE, and affective history in predicting current fatigue.
Hierarchical regression analysis was used to examine whether N and SE made a unique contribution to the prediction of fatigue beyond the contribution of prior affective disorder. The first step in the regression included demographic variables (age, education, and illness duration). Affective history, entered next into the regression statement, was dummy-coded 0,1. The third step in the equation included the simultaneous entry of N and SE. The overall model accounted for 27% of the variance in fatigue scores. Presence or absence of affective history alone accounted for 8% of the variance F[4,114] = 3.13, P = 0.018. Self efficacy and neuroticism contributed an additional combined 17% of the variance F[6,114] = 6.77, P < 0.001. An examination of the individual beta coefficients revealed that SE accounted for a significant portion of the variance, whereas N did not contribute uniquely to the prediction of fatigue scores. Controlling for SE and N, affective history remained significant, indicating that affective history made a unique and independent contribution to the variance in fatigue scores.
Testing a mediational model.
The overall pattern of findings suggests that history of affective disorder has both direct and indirect mediated effects on current fatigue. Specifically, history of affective disorder predicted variability in fatigue scores after controlling for N and SE, indicating a direct effect. Additionally, although both N and SE were correlated with affective disorder history and fatigue, only SE uniquely predicted fatigue beyond the contribution of prior affective disorder.
To test whether SE mediated an indirect path from affective history to fatigue, we followed the approach suggested by Baron and Kenny (25). First, the potential mediator, SE, was regressed on affective history, F[1,119] = 7.36, P < 0.01. Next, fatigue was regressed on affective history, F[1,115] = 9.17, P < 0.01. Finally, affective history and SE were entered together, F[2,113] = 19.97, P < 0.001. A reduction in the standardized regression coefficient for affective history from β = 0.272 (P < 0.01) to β = 0.202 (P < 0.05) provided initial evidence for partial mediation (Figure 1). As suggested by Baron and Kenny (25), a test of indirect effects was conducted using Sobel's (26) formula to determine whether the indirect effect of affective history on fatigue via SE was significantly different from 0. The result (test statistic = 2.41, P = 0.016) confirmed that SE partially mediated the relationship between affective history and fatigue.
Next, a test of moderation was conducted to determine whether N moderated the link between affective history and current fatigue. The theoretical rationale for this test is that a history of affective disorder has varying degrees of “scarring,” and that at low levels of N, one would not expect to see a relationship between depression history and fatigue. Moderation was tested according to Baron and Kenny's (25) guidelines. Neuroticism was tested using a regression equation whereby controlling for the independent effects of affective history and N, the history × N interaction term was added to the equation. The nonsignificant interaction term (P = 0.96) indicated that the relation between affective history and fatigue did not change as a function of N.
Our findings confirm that fatigue is an important feature in the general symptom profile of RA. Not only was fatigue highly prevalent in this sample, it was also rated as the most bothersome symptom, followed by pain and joint stiffness. Despite its commanding presence in RA, we found that the characteristics of fatigue were diverse among RA patients. There was little consistency as to when fatigue was experienced or the pattern of fatigue in relation to arthritis flares. This pattern of findings highlights the individual variability in symptom manifestation, particularly in fatigue, which is not considered a sole product of disease activity. Numerous personal, psychological, social, and disease factors likely interact to determine individual patterns of symptoms over time.
Using a psychometrically sound indicator of fatigue, we obtained support for the finding of Fifield et al (18) that RA patients with a history of affective disorder experience higher levels of fatigue even years after the episode. We also found that compared with individuals without a prior affective disorder, individuals who have experienced affective disturbance in the past do not possess the same levels of perceived self efficacy necessary to meet successfully the adaptational challenges associated with their chronic illness. Furthermore, their higher levels of neuroticism are likely to increase somatic awareness and increase the tendency to focus on the sensations of pain, stiffness, and fatigue. Because individuals higher in neuroticism tend to feel more vulnerable and helpless (10), the perceived negative impact of their RA symptoms could be amplified (i.e., heightened sensitivity and reactivity to the sensations associated with fatigue). When efforts to combat a symptom such as fatigue are not rewarded, a deepened sense of helplessness may emerge. In turn, this could contribute to the lower levels of self efficacy and the higher reports of fatigue observed in this sample.
Tests of mediation demonstrated that affective history affects fatigue both directly and indirectly through self efficacy. In other words, the relationship between affective history and self efficacy exerted a combined influence on perceived levels of fatigue. Moreover, the influence of affective history on subsequent RA-related fatigue appears to occur through its effect on perceived self efficacy. This demonstrated relationship expands upon the existing understanding of the role of self perceptions of coping effectiveness in health outcomes by including the potential influence of premorbid affective disturbance.
Although promising, our replication of the association between affective disorder history and subsequent RA-related fatigue, and the support we found for mediation of that association through self-efficacy expectations, must be tempered by several caveats. The first caveat is our cross-sectional design. Although the lower levels of SE and higher levels of N we found among RA patients with a prior affective disorder fit the picture of poor adaptation to chronic illness, our study's cross-sectional design cannot help us elucidate the process by which these characteristics emerge. Similarly, the cross-sectional nature of this study does not allow us to distinguish the trait marker from the scar hypothesis in providing an explanation for the emergence of lower SE and higher N in the RA patients with a history of affective disorder. However, this study does provide initial evidence for the role of personality and self-perception factors in the relationship between affective history and fatigue.
Second, our study participants had been diagnosed with RA for at least 10 years and we do not know if similar patterns would emerge in a sample characterized by shorter illness duration; that is, the pattern of relationships observed in this sample may vary over time. Additionally, the mean ± SD age for this sample was 60 ± 9.67 years, which introduces a number of factors not accounted for in this study (e.g., life transitions, degrading health not associated with RA). The sample included only 9 men, which is consistent with RA prevalence rates. These proportions did not allow for tests of sex differences. Research in the area of fatigue and affective history in RA would benefit from newly recruited, younger samples and consideration of illness duration.
An additional caveat lies in the absence of measures of pain and disability in the tested model. The relationship between pain and fatigue in RA has been firmly established (2, 3, 8). Pain also has been shown to share an association with neuroticism (12) and prior affective history (8). Additionally, the association between perceived fatigue and functional impairment has been widely demonstrated (1). Inclusion of the contribution of pain to perceived fatigue in future studies and the relative contributions of affective history, pain, and fatigue on disability would further elucidate relationships among prior affective history and future health outcomes in RA patients.
Finally, we do not know how many of our participants met criteria for depression at the time they completed the study. Although this study presents evidence for a relationship between history of affective disorder and current fatigue, it is reasonable to assume that current affective disturbance would also contribute to fatigue levels. In a study of lupus patients, Tayer et al (27) found that depression was related to higher fatigue levels in cross-sectional analysis; however, the same was not found in a longitudinal analysis. This suggests that depression's relationship to fatigue may have more proximal effects at subclinical levels, whereas clinical levels of fatigue may be required to produce enduring effects on fatigue. This conceptualization is consistent with the scar hypothesis in explaining the relationship between affective disturbance and future health-related sequelae.
Based on prevalence rates and our previous investigations, we estimate that 6 (5%) of our participants were currently depressed. It is possible that some of the “never depressed” participants may have experienced a depressive episode during the 4 years between 1996 and the time they participated in this study; however, any such changes in group status would tend to make our analyses more conservative. Future studies of affective history and fatigue should attempt to evaluate current depression status, despite its relatively modest point prevalence. Obtaining information related to the onset, pattern, and severity of past and current affective disorders would provide the opportunity to explore numerous aspects of the relationship between fatigue and history of affective disorder.
Through the careful monitoring of fatigue in RA, strategies could be developed to assist individuals in managing their energy levels so that chores, activities, and health-promoting behaviors could once again be achieved with more confidence. Strategies aimed at exerting energy in such a way that selective tasks are given a greater likelihood to be accomplished could serve to enhance self efficacy and reduce feelings of helplessness. Moreover, cognitive–behavioral interventions that enhance perceived self efficacy by training formerly depressed RA patients in self-regulatory techniques (for example, reference 28) might interrupt the link between a history of affective disorder and subsequent RA-related fatigue. Whereas neuroticism is considered to be a stable trait, self efficacy is a domain-specific cognitive variable, thus allowing for a certain degree of malleability. Enhanced self efficacy has been shown to reduce depression (28) and reduce arthritis helplessness (6). The cumulative effect of these relationships could serve to reduce the severity of perceived fatigue in RA patients, especially in those with a history of affective disorder.