| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
University of Texas at El Paso (K.B.S., J.O.B.), El Paso, TX; and University of Washington (J.I.F., D.S.B., W.J.K.), Seattle, WA.
Address reprint requests to: Karen B. Schmaling, PhD, Office of the Provost, University of Texas at El Paso, 500 W. University Ave., El Paso, TX 79968. E-mail: schmaling{at}utep.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
METHODS: A cohort of 100 patients with unexplained chronic fatigue (UCF), which encompasses both chronic fatigue syndrome (CFS) and idiopathic chronic fatigue (ICF), completed questionnaire measures and medical and psychiatric evaluations on four occasions, each six months apart.
RESULTS: Approximately 21% of the sample did not meet criteria for either CFS or ICF at their last research appointment 1.5 years after their index visit. Vitality increased over time, and physical functioning tended to improve, but UCF symptoms did not decrease significantly. Less education, being unemployed, worse mental health, more use of sedating and antidepressant medications, and more somatic attributions for their symptoms were associated with worsening symptom severity over time. Older age, current depression, and more somatic attributions predicted worsening physical functioning. Better mental health, less use of sedating medications, and fewer somatic attributions for illness were significant predictors of increases in vitality.
CONCLUSIONS: Demographic and clinical variables predict outcomes over time among a cohort of patients with unexplained chronic fatigue.
Key Words: chronic fatigue, • prognosis, • prospective, • prediction, • attribution, • depression.
Abbreviations: CFC = Chronic Fatigue Clinic;; CFS = chronic fatigue syndrome;; DIS = Diagnostic Interview Schedule;; ICF = idiopathic chronic fatigue;; M = mean;; N = sample size;; SD = standard deviation;; UCF = unexplained chronic fatigue.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Unexplained chronic fatigue is associated with a poor prognosis: patients tend to improve over time but few improve to the extent that they may be considered remitted. For example, in one study that followed patients over an average of 3.5 years, only 3% of patients remitted (no longer met CFS criteria) during that period although 57% were rated as improved (5). In another study that observed patients over an average of 2.5 years, 44% remitted (6). The illness appears to persist in the majority of patients over time. In addition to monitoring CFS case criteria and CFS-related symptoms over time, global measures of functional status also have been used to reflect illness outcomes. The Medical Outcomes Study Short Form-36 (SF-36) is perhaps the most well established survey of functional status across multiple domains (7). CFS patients score below the normal range across virtually all domains (3, 8). Several years of repeated evaluations have revealed nonsignificant trends for physical functioning to decline and vitality to improve (8).
Age and marital status are demographic characteristics that have been linked to UCF outcomes in previous studies. Older age has predicted both better (9, 10) and worse outcomes (5, 6, 11) . Being married has been associated with greater improvement in fatigue over 2 years in one study (10).
Psychiatric disorders generally have been linked to poorer outcomes. For example, dysthymia has been associated with lack of improvement over an average 1.5-year follow-up (9) and to illness persistence over an average 2.5-year follow-up (12). Any Axis I anxiety or depressive disorder was linked to illness persistence and disability over a 2.5-year follow-up (6). More numerous medically unexplained somatic symptoms predicted less improvement over a 2-year follow-up (10) and over a 2.5-year follow-up (12). In contrast to these results, Tiersky et al. (5) found that psychiatric diagnoses at base-line were associated with more improvement and a higher likelihood of employment an average of 3.5 years later. The authors suggested that patients with psychiatric illness in addition to UCF may have a different etiology associated with a more advantageous course than patients without psychiatric illness. These results warrant further investigations of the role of psychiatric illness in UCF.
Patient attributions refer to patients’ beliefs in the causes of their illness. Patients with CFS are generally more likely to make somatic attributions (eg, attribute their illness to a physical cause, such as a virus) than psychological or emotional attributions (eg, attribute their illness to stress or depression) for their illness (13–16). Somatic attributions have been associated with poorer functioning in cross-sectional studies (13, 15–19), and poorer outcomes in longitudinal studies (10, 18, 20). However, not all studies have supported these general trends. For example, Deale et al. (21) found that more patients with CFS attributed their illness to both physical and stress/lifestyle factors (47%) than physical causes alone (30%), although these differences did not achieve statistical significance. In a community-based sample, Taylor et al. (22) found that more patients attributed their illness to both physical and psychological causes than solely physical causes and that patients who improved over an approximate 3-year period were more likely to endorse both physical and psychological causes than patients whose fatigue persisted, but these differences did not achieve statistical significance. Edwards et al. (14) did not find an association between attributions and fatigue scores. That the preponderance of studies has found associations between attributions and outcome suggests that attributions warrant further examination.
In studies of the natural course of UCF, the treatments received by the participants should be assessed and examined for their contribution to patient outcomes. Although reviews of the efficacy of medication trials for CFS (eg, of immunologic, antidepressant, corticosteroid therapies) have generally concluded that the evidence is insufficient to support beneficial effects (eg, 23, 24), Bombardier and Buchwald (9) found that antidepressant use was related to improvement over an average 1.5-year follow-up. Specific forms of cognitive and behavioral therapies have demonstrated beneficial effects (see references 23 and 24 for reviews).
The purpose of the study was to determine changes in clinical status related to UCF over an 18-month period, and to identify demographic and clinical variables that predicted changes in UCF clinical status, using variables recognized previously in the literature as predictors of course and outcome.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
The records of patients scheduled for CFC appointments were reviewed before their appointment for meeting CFS/ICF inclusion and exclusion criteria. Within 6 months of the CFC appointment at which the patient was recruited for participation, medical and psychiatric evaluations were conducted to exclude other causes of chronic fatigue. These consisted of: (a) a physical examination; (b) a set of laboratory tests that included a complete blood count with differential, erythrocyte sedimentation rate, 12-factor automated chemical analysis, liver function tests, thyroid stimulating hormone, and any other tests deemed medically necessary to fully evaluate other potential causes of fatigue; (c) a review of the patient’s current and former medical records; (d) a self-report questionnaire that probed for the presence and absence of all the major and minor CFS case criteria; and (e) a computer-assisted structured psychiatric diagnostic interview, the Diagnostic Interview Schedule (DIS) Version III-A (25). (In otherwise eligible patients for whom the CFS case criteria questionnaire or psychiatric diagnostic interview occurred more than 6 months before the CFC appointment, these components of the evaluation were repeated and updated if the patient was interested in participating.)
The initial review of medical records revealed that 228 patients were potentially eligible for recruitment. Of these, 60 were ineligible, because they lived outside of the Seattle area (the majority of cases) or because of other conditions not exclusionary for UCF that would preclude participation (eg, blindness). After the screening process, the study was explained to 168 eligible subjects at their clinic appointment and they were invited to participate. Of these patients, 104 agreed to participate and 64 declined to participate. The study was reviewed and approved by the Institutional Review Board at the University of Washington; all participants indicated informed consent by signing a consent form. The longitudinal study involved four research appointments at 6-month intervals, for a total of 1.5 years: the index visit, and visits 6, 12, and 18 months subsequent to the index visit. An appointment for the index visit was scheduled and participants were given questionnaires to complete and return at the time of their appointment. Of the 104 participants who initially agreed to participate, 4 did not complete an index visit, resulting in a final sample of 100 participants who completed one or more research visits. At each appointment, a physical exam and structured interviews were conducted, and questionnaires and a blood sample were obtained. Subjects were paid $35 for completing each of four research appointments.
When participants were contacted for research appointments subsequent to the index visit and were unable or unwilling to come to research appointments, they were asked if they would complete the questionnaires by mail and interviews by phone. If participants were willing to provide some but not all measures by mail or phone, an attempt was made to obtain the measures in prioritized order, beginning with the sentinel measures of outcome (see below). Participants who were unable to be contacted or refused further participation were considered lost to follow-up (14 participants did not complete all four research appointments, see below).
Measures
The outcome and predictor measures described in the following sections were gathered through questionnaires and interviews.
Clinical Status Measures
CFS/ICF Criteria
For study enrollment, subjects underwent a comprehensive evaluation including a self-report questionnaire that probed for the presence of the CFS case criteria symptoms. At subsequent visits, and in reference to the 6-month period before each research appointment, subjects were queried about decreased functioning as a result of fatigue and the persistent or recurrent presence of the eight CFS/ICF symptoms: memory or concentration problems, sore throat, painful lymph nodes, muscle pain, joint pain, headaches, unrefreshing sleep, and feeling poorly after exertion. Subjects with functional impairment as a result of fatigue and four or more of the eight other symptoms occurring simultaneously were coded as CFS; those with impairment and less than four symptoms as ICF; and those with less than significant functional impairment as not UCF (remitted). It should be noted that the question regarding functional impairment for the initial clinic evaluation for enrollment ("Has your fatigue resulted in a substantial reduction of occupational, educational, social or personal activities as compared with before you became ill?") differed from the question asked at subsequent research appointments ("Have you been so fatigued in the past six months that you have had to reduce your average activity level below half of your normal level before you became ill?"), which may not yield comparable results.
Unexplained Chronic Fatigue Symptom Severity
We designed a questionnaire to assess the severity of 9 CFS/ICF symptoms; fatigue, memory or concentration problems, sore throat, painful lymph nodes, muscle pain, joint pain, headaches, unrefreshing sleep, and feeling poorly after exertion. Participants rated how often they experienced each symptom on a five-point scale (0 = not at all to 4 = constantly), yielding a possible range of scores of 0 to 36. For the index visit, the alpha for the symptom severity items was 0.65, and patients classified as ICF had lower scores (M = 18.0, SD = 4.4) than did patients classified as CFS (M = 22.5, SD = 4.4) (t (97) = 2.6, p< .05). Although the value of the alpha coefficient at the index visit approached but did not achieve the standard criterion of 0.70 for acceptable internal consistency, the coefficients for all subsequent visits exceeded this criterion (
= 0.73, 0.79, 0.82, at the second, third, and fourth visits, respectively).
Illness Outcome
The SF-36 (7) was used to evaluate functional status. This 36-item questionnaire may be scored for eight scales that reflect physical functioning, role-physical functioning, role-emotional functioning, social functioning, mental health, bodily pain, vitality, and health perceptions. Higher scores indicate better functioning, each score ranges from 0 to 100, and the normal range of scores is considered to be 80 and above. We retained two scores, physical functioning and vitality, for analysis on the basis of these scores’ demonstrated tendencies to be more sensitive to change over time (8), and the particular relevance of their content for persons with fatiguing illness.
Predictors of Clinical Status
Demographic Variables
Age, gender, years of education, ethnicity (recoded as either Caucasian or non-Caucasian), relationship status (recoded as either married or living with an intimate partner versus single, widowed, or divorced), and employment status (recoded as either working part- or full-time versus not working) were retained for analysis.
Psychiatric Status
The DIS is a highly structured clinical interview for psychiatric disorders (25) that is particularly indicated for the evaluation of psychiatric disorders in nonpsychiatric patients (26). The version of the DIS that applied the Diagnostic and Statistical Manual of Mental Disorders III-Revised (DSMIII-R) (27) criteria was used (the most recent version of the DIS for DSM-IV was not available at the beginning of the study). For each symptom queried in the interview, the interviewer follows a set of standard questions to arrive at a final classification of the symptom as absent; present but not to a clinically significant degree; or present and attributable to physical illness; to the use of medications, drugs, or alcohol; or to emotional or psychiatric causes. The appropriate attributions of potential psychiatric symptoms in the context of a medically unexplained illness that has symptom overlap with psychiatric disorders, such as CFS, is controversial. It has been demonstrated that changing coding assumptions results in variability in cases of psychiatric disorder (eg, 28). For the purposes of this study, possible psychiatric symptoms that overlapped with the current CFS case definition (1) and were attributed to CFS during the interview were coded as attributable to physical causes and not to psychiatric disorders. For example, among the symptoms for major depression, sleep disturbance, fatigue, and cognitive problems could be coded as attributable to physical causes if the above criteria were met, whereas dysphoria, anhedonia, weight change, psychomotor change, suicidal ideation, and guilt were not coded as due to physical causes if the patient stated they were attributable to CFS since these latter symptoms are not part of the current CFS case definition. This approach to symptom classification is conservative, resulting in fewer symptoms being coded as psychiatric than if other classification strategies had been used.
Three variables from the DIS were retained for analysis: current depressive disorder (dysthymia or major depression), any current Axis I disorder, and the number of medically unexplained somatic symptoms coded as clinically significant but not attributable to medical illness or drug/alcohol use was retained, of 44 possible such symptoms, at each appointment. In addition, the SF-36 (see above) mental health scale was retained as a continuous measure of emotional and psychological functional status.
Treatment
The treatments the subjects engaged in were assessed at each research appointment; no treatments were prescribed or proscribed as part of the study. The total number of outpatient visits for mental health and nonmental health reasons were tallied, and the number of medications taken by class of medication was tallied. The variables retained for analysis were the total number of outpatient mental health and nonmental health clinic visits and the number of antidepressants (selective serotonin reuptake inhibitors, tricyclics, or atypical antidepressants) and sedating medications (benzodiazepines, muscle relaxants, narcotic analgesics) used at each appointment. Medications that may have fatiguing effects have not been examined for their potential deleterious consequences for illness outcomes.
Illness Attribution
Wessely and Powell (29) described a five-item attribution scale; participants were asked to endorse one statement. We modified the scale for use in the present study: subjects rated two of Wessely and Powell’s items ("My illness is a physical one," "My illness is psychological in nature") on a 1- ("not at all") to 5-point ("extremely so") scale, the latter item was reverse keyed, and the responses were added. For the resulting score (range = 2–10), higher responses indicated more endorsement of somatic attributions and less endorsement of psychological attributions.
Statistical Analysis
Because of the repeated measures involved in the data, general linear mixed model regression analysis was used to examine the main study questions of the ability of variables identified previously in the literature to predict continuous outcome measures (UCF symptom severity, SF-36 vitality, and physical functioning scores) among patients with UCF. The number of days elapsed between research appointments was used as a time-varying covariate. To conform with normality of residuals, Box-Cox transformations were performed for physical functioning and vitality (30). UCF symptom severity was used with no transformation.
In addition to the mixed model regression analyses, three logistic regression analyses were used to predict improved or unimproved status: time 2 variables were used to predict improved or unimproved status between time 1 and time 2; time 3 variables were used to predict change between time 2 and time 3; and time 4 variables were used to predict change between time 3 and time 4. For example, subjects classified as improved at time 2 were coded as ICF at time 1 and remitted at time 2, or CFS at time 1 and ICF or remitted at time 2; subjects not meeting these criteria were classified as unimproved. For comparisons of time 2 and 3, and time 3 and 4, subjects who were classified as remitted at both time points were deleted from analysis. Although we had originally intended to control for previous visit values when subsequent visit values were used as predictors (eg, control for time 1 values when time 2 variables were used to predict improvement from time 1 to time 2), due to the high correlation (multicollinearity) between time points, previous visit values were excluded.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
Table 2 lists the percentages of subjects that met criteria for CFS, ICF, or neither condition, at each research appointment. Approximately 21% of persons did not meet criteria for either CFS or ICF at their last research appointment 1.5 years after their index visit. It should be noted that no exclusionary medical or psychiatric condition was revealed during any of the appointments, so exclusionary conditions do not account for any of the noncases of UCF. Despite changes in diagnostic status, UCF symptom severity decreased less than two points (of 36 possible points) on average over the 1.5-year period. Nearly one third of participants’ (N= 29) symptom severity scores increased when index visit values were compared with visit 4 values (range = 1–9 points; M = 2.9, SD = 2.0), potentially counterbalancing the effects of those who remitted on overall trends in symptom severity over time.
|
The overall test of the prediction of symptom severity was significant (F (16, 281)= 4.88, p< .001). Table 3 shows the results of each predictor of symptom severity. Less education, being unemployed, worse mental health, more use of sedating and antidepressant medications, and more somatic attributions for their illness were associated with worse symptom severity over time. For example, for each additional sedating medication used, there was a corresponding increase of nearly a point (0.72 points) in symptom severity.
|
|
|
2 (10) = 12.91, p= .23).
At time 3, 9 subjects were improved and 67 subjects were classified as unimproved compared with time 2 (8 subjects were deleted from this analysis as they maintained their remitted status at both time points). The overall model was significant (2 (10)= 28.83, p= .001). An inspection of individual variables revealed that age (p= .008), years of education (p= .027), and antidepressant use (p= .049) were significant predictors with estimated odds ratios of 0.77, 1.90, and 0.10 respectively (95% confidence intervals = 0.63–0.93, 1.08–3.37, and 0.01–0.99, respectively). A 1-year increase in age or the use of one additional antidepressant was associated with significant decreases in the likelihood of improvement, while a 1-year increase in education was associated with a near two-fold increase in the odds of improvement from time 2 to 3.
At time 4, 8 subjects improved, 63 subjects were unimproved, and 11 subjects were classified as remitted at both time 3 and time 4. The overall model was not significant (2 (10) = 11.18, p= .34) and none of the individual variables were significant predictors of improvement.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
The main purpose of the study was to test variables identified previously in the literature as predictors of course and outcome in a cohort of patients with UCF, and to add measures of medical utilization (outpatient mental health and nonmental health visits; use of sedating and antidepressant medications) as potential predictors. We identified significant demographic and clinical predictors for each of three outcome measures: UCF symptom severity, physical functioning, and vitality. Several variables were significant predictors for multiple outcome variables. A stronger tendency to attribute one’s illness to physical rather than emotional causes predicted worse outcomes across all three outcome variables. These results align well with the results of most previous cross-sectional and longitudinal studies of attributions in UCF that suggest that patients’ beliefs in physical causes are associated with poorer functioning and worse outcomes over time (10, 13, 15, 17–20) . It follows from these data that modifying dysfunctional beliefs (ie, beliefs associated with poorer functioning and outcomes) may be crucial for improvement. Cognitive and behavioral therapies that reactivate the patient and also help patients identify, test, and modify their beliefs about the illness have been shown to be effective (23, 24).
Poorer mental health functioning was associated with worse outcomes, specifically greater UCF symptom severity and less vitality, over time. The presence of a current depressive disorder (major depression or dysthymia) also was associated with worse physical functioning over time. Dysthymia has been associated with poorer outcomes in previous longitudinal studies of UCF (9, 12). Although unlike a past study (6), we did not find that a categorical variable reflecting the presence versus absence of any Axis I psychiatric disorder predicted outcomes, our positive finding for mental health functioning may reflect a similar result. Poorer mental health functioning and the specific presence of a depressive disorder bode ill for improvement among persons with UCF.
We also examined the ability of several treatment and medical utilization parameters to predict outcomes. Unlike a previous longitudinal study (9), we found that more use of antidepressant medications predicted worse outcomes (increased in UCF symptom severity and less likelihood of improvement from time 2 to time 3). Our data are perhaps more in line with meta-analyses and other reviews of clinical trials of antidepressants for UCF that have not found sufficient evidence to support their efficacy (23, 24). More use of sedating medications such as benzodiazepines, muscle relaxants, or narcotics was associated with worse outcomes, specifically greater UCF symptom severity and less vitality, over time.
To test the possibility that the use of antidepressants and sedating medications predicted worse outcomes because they were used by the most depressed and ill participants, respectively, the mixed model analyses were redone adding post-hoc tests of the interaction terms of antidepressant use and current depression (present vs. absent), and of sedating medication use and clinical status (CFS vs. ICF vs. remitted). The interactions were not significant predictors of outcome, suggesting that depression, use of antidepressants, and sedating medications function as independent predictors of poorer outcomes.
Although UCF includes several pain symptoms (sore throat, tender lymph nodes, myalgia, arthralgia, headaches) that may prompt muscle relaxant or narcotic prescriptions, and also disturbed sleep that may lead to benzodiazepine use, these results are the first to our knowledge that demonstrate that these medications are associated with poorer outcomes. Although untested by these data, a possible pathway through which these medications have deleterious effects on outcome is through the impairment of patients’ abilities to remain active or engage in physical conditioning programs because of their sedating effects. Limiting exercise has been associated with more functional impairment in one previous longitudinal study (18).
Less education and being unemployed were associated with increased symptom severity over time. The associations between unemployment and worsening physical functioning and vitality approached significance (see Tables 4 and 5
). The loss of the structure and activity associated with employment may have negative effects on outcomes in UCF. More years of formal education was associated with significantly greater odds of improved clinical status at the third visit compared with the second visit. More education may be associated with relatively higher incomes and more enjoyable work environments, which may motivate recovery and return to work. These variables have not been found to predict outcome in previous studies. Older age was associated with decreases in physical functioning over time. Being older was associated with significantly lower odds of improved clinical status at the third visit compared with the second visit. As reviewed previously, past studies have found mixed results of the association between age and outcome. Some studies have found a relationship in the negative direction between age and outcomes as did the data in the present study (5, 6, 11), but others have found an association in the positive direction (9, 10). Unlike past studies of predictors of the course of UCF over time, we did not find that the number of medically unexplained symptoms predicted outcome. Although being partnered tended to predict increased vitality over time, this result did not attain significance, unlike in previous studies.
Although the repeated measurements and prospective design were strengths of the study, several limitations were noted that may have affected the results and would be issues to address in future research. First, although the demographic characteristics of the participants in this sample were similar to those reported previously for the patient population in this clinic (3), patients in a tertiary care clinic may not be representative of the entire population of persons with UCF. Second, slightly different questions were used in the process to initially determine if potential participants met case criteria for CFS or ICF than were used to track their case status over time. Use of the same questionnaire that could flexibly address initial and subsequent case status would be preferable to avert the possibility that changes in the instrument, rather than the construct being measured, determined the resulting data. Third, the coding scheme used to classify some symptoms in the DIS such as fatigue as psychiatric or medical in nature likely resulted in fewer cases of psychiatric illness, with unknown effects on the subsequent results. In future studies, it would be useful to identify the set of coding assumptions most sensitive and predictive of changes in UCF clinical status.
In conclusion, we identified predictors of clinical status in a cohort of patients with UCF followed prospectively over an 18-month period. Several significant predictors of outcome identified in this study (eg, use of sedating medications) and found in this and other studies (eg, illness attributions) are potentially modifiable. Targeting these potentially modifiable variables for change in future invention studies may result in a more encouraging course of illness for patients with UCF.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Received for publication September 8, 2002.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
