| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Department of Psychiatry (P.S., V.M.), University of Ioannina, School of Medicine, Greece; and the Department of Psychiatry (P.S., G.L.), University of Bristol, UK.
Address correspondence and reprint requests to Dr. Petros Skapinakis, Department of Psychiatry, University of Ioannina School of Medicine, Ioannina 45110, Greece. E-mail: pskapin{at}cc.uoi.gr
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
METHODS: The authors examined this issue by using data from the World Health Organization collaborative study of psychological problems in general health care. Three thousand two hundred one subjects from 15 primary care centers in 14 countries were followed up for 12 months. The Composite International Diagnostic Interview was the main instrument used. Odds ratios and their 95% confidence intervals (CI) were calculated using logistic regression models adjusted for sociodemographic variables, physical morbidity and intercenter variability.
RESULTS: Cases of depression were found to have an increased risk of developing a new episode of unexplained fatigue at follow-up with an adjusted odds ratio of 4.15 (95% CI = 2.64–6.54). Similarly, cases of unexplained fatigue were found to have an increased risk of developing a new episode of depression at follow-up with an adjusted odds ratio of 2.76 (95% CI = 1.32–5.78). Further adjustment for subthreshold symptoms at baseline weakened the reported associations, especially between fatigue and development of a new episode of depression, but these remained significant.
CONCLUSIONS: The findings support the view that unexplained fatigue and depression might act as independent risk factors for each other.
Key Words: comorbidity, • depressive disorder, • fatigue syndrome, • chronic, • follow-up studies, • primary health care.
Abbreviations: CI = confidence interval;; CIDI = Composite International Diagnostic Interview;; GHQ-12 = 12-Item General Health Questionnaire;; GHQ-28 = 28-Item General Health Questionnaire;; GNI = gross national income;; ICD-10 = International Classification of Diseases, Tenth Revision;; WHO = World Health Organization.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Only a few longitudinal studies have been carried out to investigate this issue, and their results are contradictory. Van der Linden et al. (4), in a study in UK primary care, observed that patients with unexplained fatigue did not have an increased risk of developing a common mental disorder, nor did patients with common mental disorders have an increased risk of developing unexplained fatigue at 6 months. However, the focus of this study on subjects with fatigue without comorbid psychiatric disorder ("pure fatigue") excludes the majority of patients. Pure fatigue was also studied in an Australian primary care study, and the results were similar to the UK study (5). In contrast to these studies, a community study in the United States, using data from the Epidemiological Catchment Area Study, reported that both fatigue and depression predicted each other over 13 years’ follow-up (6). Some of these contradictory findings might be explained by the different setting (community vs. primary care) and the duration of follow-up (6 months vs. 13 years).
In order to investigate the temporal relation between depression and unexplained fatigue, we used data from a World Health Organization (WHO) longitudinal study in primary care (7). In a previous paper, we reported a strong cross-sectional association between unexplained fatigue and depression (8). In the current study, we aimed to clarify the nature of this association by investigating whether unexplained fatigue precedes the development of depression or vice versa at 12-month follow-up. A companion paper has reported the prognosis of fatigue syndromes at this longitudinal data set (9). Our study differs from the others in that we used an international sample from 14 countries, both developed and developing.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
The longitudinal part of the study included all subjects with a definite or a subthreshold psychiatric disorder and a 40% random sample of noncases. These became eligible and were asked to complete the follow-up assessment 12 months later. The follow-up assessment was identical to the baseline, but it excluded the physician assessment. Three thousand two hundred one subjects participated in the follow-up assessment. Taking into account that some centers selected to reinterview all subjects for their own reasons (these were Bangalore, Berlin, Nagasaki, Paris, and Seattle), the overall response rate for the 12-month follow-up was 68%, with a range from 29% to 88% (p < .001). Centers with response rates less than 60% were Nagasaki (53%), Santiago (47%), and Rio de Janeiro (29%). The likelihood of completing the follow-up interview was not related to age, gender, or the presence of unexplained fatigue at baseline, but it was lower for subjects scoring 20 or more in the GHQ-28 (p = .003).
The present paper used data from the 3201 subjects who completed both the baseline and follow-up assessment of the second phase. The advantage of the two-phase design is that it can provide estimates for the whole population screened at phase 1 of the study by applying appropriate sampling weights.
Instruments
The basic instrument of the study was the CIDI. The standard interview was slightly modified to be more compatible with use in primary care (this version is referred to as CIDI-PHC). Changes concerned the inclusion of a section on fatigue, and questions to assess both current and lifetime functional somatic symptoms (14). The instrument was translated into several languages, and there was extensive testing of its reliability. The intercenter reliability in the form of interviewer-observer reliability coefficient was found to be 0.92 overall, ranging between 0.81 and 1.00 for individual sections (14).
Assessment of Fatigue
Fatigue was assessed with a special section included in CIDI-PHC. It should be noted that these series of questions were different from the questions used to assess the symptom of fatigue as part of the syndrome of depression, and they were asked immediately after the somatization section of the CIDI-PHC and before the assessment of psychiatric conditions. Three screening questions were asked to all the subjects (Q1 "In the past month have you felt tired all the time?", Q2 "Do you get easily tired while performing everyday tasks?", and Q3 "Does even minimal physical effort cause exhaustion?"). Then the interviewer asked a specific sequence of questions to determine the clinical importance and possible cause of the symptom. Fatigue was considered clinically important if the subject had sought professional help, or took medication for it more than once or felt that it interfered with his/her life a lot. Fatigue was also considered "medically explained" if a doctor had been consulted and had given the patient a definite diagnosis or if there had been any abnormalities reported on examination or further investigation. It should be noted that unclear diagnoses were not considered as medical explanations, and this included the terms "myalgic encephalomyelitis" or "chronic fatigue syndrome." Subjects with clinically important unexplained fatigue were also asked a fourth question that assessed the severity of fatigue (Q4 "Is it difficult to recover from fatigue with rest?").
We used the following 2 criteria to classify subjects as cases of substantial unexplained fatigue: a) presence of clinically important unexplained fatigue and at least 1 positive answer to questions Q1-Q3 and b) a positive answer to the severity question Q4. This definition corresponds broadly to the first and third criteria of the "tenth-revision of the International Classification of Diseases" (ICD-10) definition of neurasthenia (15), but it differs in that a) it does not include the multiple somatic symptoms criterion (Criterion B in ICD-10), b) it refers to 1-month duration instead of 3 (Criterion D in ICD-10), and c) it does not exclude other comorbid psychiatric disorders (Criterion E in ICD-10). Therefore, it is a much broader definition compared with ICD-10 neurasthenia. This definition is also very similar to the concept of Idiopathic Chronic Fatigue as defined by an international group of experts in Atlanta (16) but with a duration criterion of at least 1 month instead of the 6 months required by the latter.
To estimate subthreshold fatigue symptoms, we calculated the number of fatigue symptoms reported by subjects. Subjects who had at least 1 fatigue symptom and who did not meet our criteria for substantial unexplained fatigue syndrome were considered as having subthreshold symptoms.
Assessment of Depression
Depression was assessed with the administration of CIDI-PHC. This includes a separate section that asks about symptoms of current (1-month) depression. The investigators used specific diagnostic algorithms to assign the diagnosis of depression to subjects fulfilling the ICD-10 criteria of depressive episode (F32) (15). These conditions are similar to the major depression category of the DSM-IV. According to the ICD-10 criteria, the depressive episode can be further subclassified to mild, moderate, or severe forms, but in the present paper, we considered all forms of severity as 1 category of depression. It should be noted that even the mild form fulfills the F32 criteria for depressive disorder and does not refer to subthreshold symptoms.
Because fatigue is 1 of the symptoms assessed in the depression module of CIDI-PHC, we excluded this symptom from the definition of depression to avoid artificial overlapping with the fatigue syndromes that we assessed separately. Therefore, the term "depression" in the context of the present study refers to ICD-10 depressive episode excluding the fatigue items. This definition therefore is a broader definition compared with the ICD-10 definition, but otherwise it is identical.
We also estimated subthreshold depressive symptoms by calculating the number of depressive symptoms assessed in the CIDI depression module (excluding the fatigue items). We defined as cases of subthreshold depression all those who did not fulfill criteria for depression and who were reporting more than 4 symptoms.
Assessment of Other Variables
Data on various sociodemographic variables were systematically collected by the interviewers: a) age was used as a continuous variable; b) marital status was grouped into 3 categories: married, unmarried, and a third category including all other (widowed, divorced, separated); c) number of children: this variable was grouped into 3 categories, none, 1, or 2 and more than 2; d) years of education: this variable refers to years spent in school or college. It was grouped into 0 to 6 years (elementary education), 7 to 12 (basic education), and more than 12 (higher education); e) employment status: this variable was coded into 3 groups: employed, not employed, and economically inactive (mainly housewives and students).
Physical morbidity was assessed by using both a checklist of 20 chronic medical conditions, which the patient completed ("chronic physical morbidity"), and a rating of severity on a 5-point scale by the treating physician ("current physical morbidity"). Both variables were coded in a binary form. Subjects with chronic physical morbidity were considered those with at least 1 self-reported chronic medical condition, and subjects with current physical morbidity were considered those who had at least a mild physical condition (as opposed to none or subclinical) according to the treating physician.
Analysis
Data analysis was carried out with the program STATA version 6.0 (17). We used the "svy" commands, which are appropriate for the 2-phase design of the study (11). In all the analyses, we used probability (sampling) weights to reflect the population that was originally screened at phase 1. We examined the hypotheses of the study with a series of weighted logistic regression models using the "svylogit" command. We first examined whether depression at baseline predicted a new onset of unexplained fatigue at follow-up. Only subjects free of fatigue at baseline were included in this analysis (N = 2714). The dependent variable was new onset of fatigue at follow-up, and the independent variable was depression at baseline. We adjusted for current and chronic physical morbidity at baseline, the various sociodemographic variables at baseline, and intercenter variability, using dummy variables for each center. A similar model was used to test whether substantial unexplained fatigue at baseline (the independent variable) predicted new onset of depression (the dependent variable) at follow-up, controlling for all other variables. Only subjects free of depression at baseline were included in this analysis (N = 2298). To examine whether any reported associations were due to subthreshold symptoms at baseline, we further adjusted for subthreshold fatigue or depression symptoms, respectively. The results are reported in the form of adjusted odds ratios with 95% confidence interval (CI).
We also examined whether there was any interaction between centers and the independent variables of interest (depression or fatigue) by calculating the likelihood ratio test statistic with the "lrtest" command in STATA (17). To reduce the number of variables in these models, we classified centers in 3 categories according to the level of economic development of the matching country. This classification was based on the gross national income (GNI) per head in 2000 as follows: high-income countries with more than $10,000 (Athens, Berlin, Groningen, Mainz, Manchester, Nagasaki, Paris, Seattle, Verona), middle-income countries with less than $10,000 but more than $1000 (Ankara, Rio de Janeiro, Santiago), and low-income countries with less than $1000 (Bangalore, Ibadan, Shanghai). Data for the GNI per head were derived from the World Bank databases available on the Internet (http//www.worldbank.org).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Development of Unexplained Fatigue and Depression at 12-Month Follow-up
There were 167 new onsets of substantial unexplained fatigue out of 2714 subjects free of fatigue at baseline, for a weighted average of 3.5% (95% CI = 2.7–4.4) and a range between 0.8 (95% CI = 0.3–2.1) in Paris and 10.5 (95 CI = 5.2–20.3) in Santiago (p = .006). Excluding these 2 outliers, differences between centers were not significant (p = .18).
Similarly, there were 161 new onsets of depression among 2298 subjects free of disorder at the baseline, for a weighted average of 5.1% (95% CI = 3.4–7.6) and a range between 0.6 (95% CI = 0.2–1.77) in Nagasaki and 45.9 (95% CI = 13.1–82.6) in Santiago (p < .001). Excluding these 2 outliers, differences between centers were nonsignificant (p = .27).
In Table 1 we present the cumulative incidence of both unexplained fatigue and depression across centers classified according to level of economic development. Middle income countries have a higher incidence of both depression and fatigue and this is due to Santiago, which also reported the highest prevalence of psychiatric morbidity in the original survey (10).
|
|
Regarding the other covariates in the model, only the presence of subthreshold fatigue symptoms were found to be significantly associated with development of fatigue at follow-up with an odds ratio of 1.49 (95% CI 1.20–1.86) for each additional fatigue symptom reported at baseline. Neither age nor sex was associated with development of fatigue at follow-up.
Substantial Unexplained Fatigue at Baseline and Risk of Developing Depression at Follow-up
As shown in Table 3, cases of fatigue at baseline were found to have an increased risk of developing a new onset of depression at follow-up with an odds ratio of 2.76 (95% CI = 1.32–5.78). Part of this association was due to the confounding effect of subthreshold baseline depressive symptoms. However, even after adjusting for subthreshold symptoms, the association was still significant (p = .04). We also tested for an interaction between baseline fatigue and centers, and the likelihood ratio test was 7.38 on 2 degrees of freedom and a p value of 0.03, which is significant. However, when we removed the observations from Santiago, the test was no longer significant (p = .09). It is worth noting that when we removed the observations from Santiago, the association between baseline fatigue and depression was strengthened (odds ratio after adjustment for subthreshold symptoms was 2.45, 95% CI 1.16–5.21, p = .02). In Santiago, there were 46 subjects free of depression at baseline, of whom 2 only had the exposure of interest (fatigue at baseline). At follow-up none of them reported depression, as compared with 11 of 44 with no fatigue, and therefore statistics cannot be calculated. With the exception of Santiago, the reported association between baseline fatigue and a new onset of depression can be considered to be similar across centers.
|
Regarding the effect of other covariates in the model, only sex, subthreshold depression, and fewer years of education were significantly associated with the development of depression at follow-up. Subthreshold depression had the strongest association, with an odds ratio of 3.96 (95% CI 2.08–7.54).
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Our findings should be interpreted in the context of the following limitations. First, the secondary nature of our study did not allow us to use the duration criteria of 3 or 6 months usually applied to syndromes of unexplained fatigue. However, in primary care, fatigue is usually of longer duration compared with community settings (18), and therefore fatigue of 1-month duration is not expected to be much different from fatigue of 3 months’ duration. Second, the exclusion of medical conditions was not based on detailed medical investigations but was elicited from the interviewers by asking subjects certain questions from their history. This might have resulted in misclassification, but because the original study did not aim to examine unexplained fatigue, we do not think that this could introduce a systematic bias into our study. Third, the study was not a real cohort, and the term "new onset" does not imply an incident case of fatigue or depression. Other patients could have reported fatigue or depression if they had been asked in another instance during the follow-up period. Fourth, the study was carried out in primary care rather than in the community, but the longitudinal design supports the idea that selection bias is less likely to have influenced the findings. And lastly, there was considerable variability in response rates between centers, both at the baseline and follow-up assessments, and one might question the validity of combining results across centers. We carried out likelihood ratio tests to examine the consistency of results across centers, and these tests showed that the results could be combined. Although the center of Santiago stood out as an outlier, we do not think that including the observations from this center threatens the validity of our results. In addition, the use of such a multicultural sample offers some advantages over studies that have used more homogeneous samples from developed countries, in that the possibility of a cultural bias in symptom reporting can be taken into account.
The association between unexplained fatigue and depression, found in numerous cross-sectional studies (2), is very strong, and is unlikely to be due to chance or systematic biases. There are at least 4 possible explanations for this association: a) that unexplained fatigue syndromes are "caused" by depression (19) (the "causal hypothesis"); b) that depression is secondary to a disabling illness of unknown origin (the "reverse causality" hypothesis, most relevant in the context of chronic fatigue syndrome); c) that depression and unexplained fatigue syndromes share common risk factors that confound the association between them (1) (the "common etiology" hypothesis); and d) that this association is caused by the overlapping criteria in their definition (the "overlapping hypothesis"). It is worth noting that these explanations are not mutually exclusive. Longitudinal studies can help distinguish between issues of causality or reverse causality. Our results do not support the view that fatigue is just a symptom of depression, because we also observed that subjects with fatigue (but no depression) had an increased risk of developing a new onset of depression at follow-up, even after adjusting for subthreshold symptoms. By removing the fatigue questions from the definition of depression, we also minimized the effect of overlapping criteria. Two previous studies in primary care provide data comparable to the present study, 1 from the UK (4) and the other from Australia (5). The UK study involved 1177 patients in primary care who completed the GHQ-12 and a fatigue questionnaire at 2 times 6 months apart. The response rate was 82.5% (791 patients). The researchers observed that cases of unexplained fatigue at baseline were not more likely to develop psychiatric disorder, noncomorbid with fatigue, at follow-up (odds ratio 1.0; 95% CI = 0.5–1.8). However, they did find an association between fatigue and development of psychiatric morbidity comorbid with fatigue (odds ratio 3.8; 95% CI = 2.3–6.1). Similarly, cases of psychiatric morbidity were not more likely to develop noncomorbid fatigue at follow-up (odds ratio 0.8; 95% CI = 0.4–1.5.), but were more likely to develop fatigue comorbid with psychiatric disorder (odds ratio 2.6; 95% CI = 1.5–4.2). The Australian study involved 1593 primary care attenders who completed the 30-item general health questionnaire and a fatigue questionnaire at 2 times 12 months apart. The response rate was 49.5% (789 patients), but the study used data from 652 patients with complete data at both assessments. The results were similar to the UK study, with the corresponding odds ratios being 0.95 (95% CI = 0.2–3.6) for development of noncomorbid fatigue and 1.4 (95% CI: = 0.6–3.4) for the development of noncomorbid psychiatric disorder. Comorbid forms of disorder were significantly associated with baseline fatigue or psychiatric disorder.
In our own analysis, we first examined whether fatigue or depression at baseline increased the risk of developing a new onset of depression or fatigue, respectively, without distinguishing between comorbid and noncomorbid forms of disorder at follow-up. Excluding comorbid forms is not completely justified given the large cross-sectional association. Our results are closer to the community longitudinal study that used data from the Epidemiologic Catchment Area study (6). This study found that both fatigue and depression predicted each other over 13 years of follow-up, and the odds ratios reported are similar to those in our study.
There is now enough evidence from the literature that fatigue and psychiatric disorder are not the same. Genetic studies have failed to find common genetic influences (20), and 1 study concluded that the association is most likely environmental in origin (21). Neurobiological studies have also shown differences between fatigue states and depression (22). Therefore, it is evident from the literature that fatigue and depression have distinct risk factors. On the other hand, given the magnitude of the association, there will also be common risk factors that confound the association between them. Another issue might be the overlapping criteria. However, this and the Addington et al. (6) study show that fatigue and depression could also be independent risk factors for each other in a manner that resembles an etiological vicious cycle. We present this vicious cycle model in Figure 1.
|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
The data reported in this article were collected as part of a World Health Organization’s Psychological Problems in General Health Care project. Participating investigators include: O. Ozturk and M. Rezaki, Ankara, Turkey; C. Stefanis and V. Mavreas, Athens, Greece; S. M. Channabasavana and T. G. Sriram, Bangalore, India; H. Helmchen and M. Linden, Berlin, Germany; W. van der Brink and B. Tiemens, Groningen the Netherlands; M. Olatawura and O. Gureye, Ibadan, Nigeria; O. Benkert and W. Maier, Mainz, Germany; R. Gater and S. Kisely, Manchester, UK; Y. Nakane and S. Michitsuji, Nagasaki, Japan; Y. Lecrubier and P. Boyer, Paris, France; J. Costa e Silva and L. Villano, Rio de Janeiro, Brazil; R. Florenzano and J. Acuna, Santiago, Chile; G.E. Simon and M. von Korff, Seattle, Washington; Y. He-Quin and X. Shi Fu, Shanghai, China; and M. Tansella and C. Bellantuono, Verona, Italy. The study advisory group includes J. Costa e Silva, D. P. Goldberg, Y. Lecrubier, M. von Korff, and H-U Wittchen. Coordinating staff at World Health Organization headquarters include N. Sartorius and T. B. Ustun.
Received for publication May 3, 2003.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. A. Rimes, R. Goodman, M. Hotopf, S. Wessely, H. Meltzer, and T. Chalder Incidence, Prognosis, and Risk Factors for Fatigue and Chronic Fatigue Syndrome in Adolescents: A Prospective Community Study Pediatrics, March 1, 2007; 119(3): e603 - e609. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. Ismail and G. Lewis Multi-symptom illnesses, unexplained illness and Gulf War Syndrome Phil Trans R Soc B, April 29, 2006; 361(1468): 543 - 551. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
