Source: Psychological Medicine Preprint Date: 25 september 2007 URL: http://journals.cambridge.org/action/displayJournal?jid=PSM Risk factors for chronic fatigue syndrome/myalgic encephalomyelitis: a systematic scoping review of multiple predictor studies S. Hempel(*), D. Chambers, A.-M. Bagnall and C. Forbes Centre for Reviews and Dissemination, University of York, York, UK * Address for correspondence: Dr S. Hempel, Centre for Reviews and Dissemination, University of York, York YO10 5DD, UK. (Email: susanne.hempel@googlemail.com) Received 9 October 2006; Revised 11 July 2007; Accepted 18 July 2007 Background. The aetiology of chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) is still unknown. The identification of risk factors for CFS/ME is of great importance to practitioners. Method. A systematic scoping review was conducted to locate studies that analysed risk factors for CFS/ME using multiple predictors. We searched for published and unpublished literature in 11 electronic databases, reference lists of retrieved articles and guideline stakeholder submissions in conjunction with the development of a forthcoming national UK guideline. Risk factors and findings were extracted in a concise tabular overview and studies synthesized narratively. Results. Eleven studies were identified that met inclusion criteria: two case-control studies, four cohort studies, three studies combining a cohort with a case-control study design, one case-control and twin study and one cross-sectional survey. The studies looked at a variety of demographic, medical, psychological, social and environmental factors to predict the development of CFS/ME. The existing body of evidence is characterized by factors that were analysed in several studies but without replication of a significant association in more than two studies, and by studies demon- strating significant associations of specific factors that were not assessed in other studies. None of the identified factors appear suitable for the timely identification of patients at risk of developing CFS/ME within clinical practice. Conclusions. Various potential risk factors for the development of CFS/ME have been assessed but definitive evidence that appears meaningful for clinicians is lacking. Key words: Chronic fatigue syndrome, myalgic encephalomyelitis, risk, review. Introduction Chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) is characterized by a range of symptoms, namely fatigue (often triggered by minimal activity), malaise, headaches, sleep disturbances, difficulties with concentration and muscle pain. The symptoms can fluctuate in their intensity. The severity of the disease varies considerably and CFS/ME can be a disabling condition, placing a substantial burden on patients, their carers and health-care professionals. The population prevalence is estimated at about 0.4 % for adults in the UK, making it a relatively common condition (Department of Health, 2002). Widely accepted criteria for case definitions are the Oxford criteria (Sharpe et al. 1991) and the Centre for Disease Control (CDC) definition (Fukuda et al. 1994). The condition has been the topic of much discussion over the past two decades (Prins et al. 2006). Research continues to investigate potential causes (Afari & Buchwald, 2003; White, 2004), how to diagnose the condition (Bagnall et al. 2005) and what its treatment and management (Chambers et al. 2006) should involve. Children and severely affected patients are of special interest to clinicians as not all treatments can be applied and very little is known about the effec- tiveness of treatment in these groups (Bagnall et al. 2005). Given the potential impact of the condition on patients, their carers and health-care professionals, the ability to identify risk factors involved with the development of the condition would be a crucial breakthrough. Specifically, in people presenting with early suspected CFS/ME (e.g. before the 6 months' illness duration necessary for a formal diagnosis), it would be useful to know what the risk factors or prognostic flags are that might be linked with progression to CFS/ME and possibly indicate early interventions. We therefore conducted a systematic scoping review, bringing together all available studies in the literature that use an appropriate research design in order to identify risk factors for the development of CFS/ME. This was part of a larger project to find evidence to support the development of guidelines on the diagnosis and management of CFS/ME in adults and children. We systematically searched the literature for studies that met strict inclusion criteria. Studies were considered that explicitly had the syndrome CFS/ME according to recognized case definitions as the outcome and used multiple factors to predict the progression to the development of the syndrome. The review aims at documenting which potential risk factors have been assessed and which factors appear to be significantly associated with the development of the syndrome. Method Literature search We searched the databases Medline (1966 to May 2005), EMBASE (1980 to May 2005), PsycINFO (1872 to April 2005), CENTRAL (May 2005), Social Science Citation Index (1945-2005), Science Citation Index (1945-2005), Index to Scientific and Technical Proceedings (1982-2005), PASCAL (May 2005), Inside Conferences (May 2005), AMED (1985-January 2005) and HEED (June 2005). For each database, individual search strategies were developed (these are available from the authors). The search was broad with the objective of identifying all studies of CFS/ME and possible synonyms. We applied no language restriction. Further references were sought by scanning the reference lists of retrieved articles and submissions of study references from stakeholders involved in the development of the UK CFS/ME guidelines for the National Institute for Health and Clinical Excellence (NICE). On peer review of this publication, referees identified three more papers that met inclusion criteria and had been published after the search dates. Study selection Two reviewers independently assessed all titles and abstracts identified from the literature searches for potential relevance to the review questions. Potentially relevant papers were retrieved in full and assessed by two independent reviewers, applying the specified inclusion criteria. Two reviewers are commonly used in systematic reviews to minimize the risk of introducing bias to the results of the review. If the two reviewers cannot agree, a third reviewer is consulted to resolve the differences. Inclusion criteria Study type. Any study aimed at identifying potential risk factors or prognostic flags for the development of CFS/ME. Studies focusing on fatigue rather than the defined syndromes CFS or ME were not eligible for inclusion in the review. Studies that only assessed factors affecting the progression of the disease severity in the course of the condition were also excluded. Population. Adults and/or children aged 5 years ormore. Study design. Any study reporting a multivariate or regression analysis. Studies only reporting rates of occurrences of individual sample characteristics or only reporting correlations with single risk factors were not eligible for inclusion in the review because of the high potential for bias associated with these study types. Study characteristics, data extraction and study quality The study information regarding participants, setting, study design, statistical method, analysed predictor variables and findings was extracted into standard- ized forms. Studies were formally classified (e.g. case-control study, cohort study) and the data collection methods regarding the predictors of CFS/ME were classified as retrospective (the diagnosis was established first and characteristics were assessed later), concurrent (the diagnosis and the predictor variables were assessed at the same time) or prospective (the study was designed prospectively, the risk factors were assessed at baseline and the diagnosis was established at follow-up). The data extraction and assessment were carried out by one reviewer and checked by a second reviewer. Discrepancies were resolved by reference to the original study. Where necessary, arbitration was by a third reviewer. Data synthesis The findings are synthesized in a narrative review and summarized in a concise table to facilitate the comparisons of individual study results (see Table 1). The replication of results and discrepancies were investigated. Data relating to the pre-specified subgroups of children and patients severely affected by CFS/ME were considered separately. Results Search results The literature search identified 10768 publications. Their titles and abstracts were screened: 643 full paper copies were retrieved and assessed for inclusion in the review (see Fig. 1). Eleven studies met the inclusion criteria (Bell et al. 1991; Wessely et al. 1995; Cope et al. 1996; White et al. 2001; Chalder et al. 2003; Huibers et al. 2004a,b; Viner & Hotopf, 2004; Heim et al. 2006; Hickie et al. 2006; Kato et al. 2006). Study characteristics and quality One cross-sectional survey (Chalder et al. 2003), one nested case-control and twin study (Kato et al. 2006), two case-control studies (Bell et al. 1991; Heim et al. 2006), four cohort studies (White et al. 2001; Huibers et al. 2004 a,b; Viner & Hotopf, 2004) and three studies that combined a cohort study with a case-control study design (Wessely et al. 1995; Cope et al. 1996; Hickie et al. 2006) were identified. The majority of studies used a prospective design (Wessely et al. 1995; Cope et al. 1996; White et al. 2001; Huibers et al. 2004a,b; Viner & Hotopf, 2004; Hickie et al. 2006; Kato et al. 2006), one prospective study (Huibers et al. 2004b) combined prospective and concurrent data (measurements at baseline and at follow-up) to predict CFS/ME, one study combined retrospective and concurrent data collection (Bell et al. 1991) and one cross-sectional study used a concurrent data collection (Chalder et al. 2003). In nine studies the focus of the publication was on predicting CFS/ME in adults, although the birth cohort study of Viner & Hotopf (2004) used the prediction of ever having had or currently suffering from CFS/ME. Two studies examined the development of CFS/ME in children (Bell et al. 1991; Chalder et al. 2003). The quality of the reporting varied; several publications failed to present detailed descriptions of their statistical methods and not all studies reported the goodness of fit of the final prediction model. Cope et al. (1996). Sixty-four adult patients with chronic fatigue (3 months' duration and a score of 9 or more on a fatigue questionnaire) and 64 matching controls were identified from a cohort of primary-care patients recruited 6 months previously with a clinically diagnosed viral illness. Twenty-three of the cases fulfilled criteria for CFS. Heim et al. (2006). This case-control study compared self-reported childhood trauma and psychopathology in participants with clinically confirmed CFS and a matched control group. Forty-three cases met the 1994 CDC criteria for CFS. Hickie et al. (2006). Identified patients with Epstein-Barr virus, Coxiella burnetii or Ross River virus (n=253) were followed from the time of acute infection. At 6 months, 28 patients met diagnostic criteria for CFS and these were followed for a further 6 months and compared to a matched control group. Huibers et al. (2004a). Employees (n=151) who were on sick leave because of fatigue were followed for 12 months. Of these, 66 were CFS-like cases (met research criteria for CFS without a formal diagnosis) at baseline. All persons with medical conditions that could explain fatigue were excluded. At follow-up, 28 participants met CFS criteria. Huibers et al. (2004b). Employees (n=1143) with unexplained fatigue (excluding somatic conditions that could explain fatigue) were followed prospectively for 44 months. At follow-up, 94 participants met research criteria for CFS without a formal diagnosis (CFS-like caseness), 457 were non-CFS fatigue cases and 592 were no longer fatigue cases. Kato et al. (2006). This study linked data on personality traits and stress obtained previously in a large twin sample (n=19150) and a diagnosis of CFS or CFS-like illnesses established in a telephone interview. In this sample, 447 participants met the 1994 CDC criteria, all participants who reported fatigue at the time of the personality and stress assessment were excluded. Viner & Hotopf (2004). Data obtained from a national birth cohort (babies born in England, Scotland and Wales, 5-11 April 1970) followed up at 5, 10, 16 and 29-30 years were analysed. At the last follow-up, 93 of 11 261 participants (0.8 %) reported ever having CFS/ME and 48 (0.4 %) reported having the condition currently. Wessely et al. (1995). Questionnaires to assess fatigue and psychiatric morbidity were sent to over 2000 adult patients in participating general practices. Prevalence of chronic fatigue and CFS was subsequently assessed in patients who attended the surgery with a symptomatic infection (exposed cohort) or for other reasons (non-exposed cohort). Most (84 %) were followed up at 6 months. White et al. (2001). Primary-care patients, including a cohort of students, with infectious mononucleosis or an upper respiratory tract infection were followed until 6 months after onset. The sample included patients meeting the Oxford criteria, the 1994 CDC criteria and patients with not otherwise specified fatigue or idiopathic chronic fatigue at follow-up. Subgroups Children and adolescents Bell et al. (1991). All 914 students in a school district were sent a questionnaire regarding CFS symptoms. Thirty-three children with symptoms were interviewed, 21 had a physician-confirmed diagnosis of CFS. These were each matched with two randomly selected asymptomatic controls from the sample. Numerous risk factors and exposures were tested. Chalder et al. (2003). Mothers (n=10438) of children aged 5-15 in a survey of families living in private households in England, Scotland and Wales in 1999 were asked whether their child had ME or CFS. They were then asked to complete a General Health Questionnaire (GHQ). A sample (n=4240) of children aged 11-15 was interviewed. The prevalence of CFS/ME was eight out of 4240 (0.19 %). Severely affected. No study was identified that predicted severe CFS/ME seperately. Identified cases with CFS/ME and prevalences The number of patients with CFS/ME who were identified in the individual studies and formed the basis for the outcome prediction varied from eight (Chalder et al. 2003) to 447 (Kato et al. 2006). The outcome CFS/ME was defined as meeting published criteria in three of the studies (Cope et al. 1996; Huibers et al. 2004a,b), in one study the self-report of having CFS or having had CFS was used (Viner & Hotopf, 2004), in four studies the participants seem to have been assessed in a diagnostic interview (Wessely et al. 1995; White et al. 2001; Chalder et al. 2003; Kato et al. 2006) and in three studies the participants had a diagnosis explicitly confirmed by a physician (Bell et al. 1991; Heim et al. 2006; Hickie et al. 2006). The prevalence of CFS/ME was 0.19 % in the social survey (Chalder et al. 2003), 0.43 % (0.83 % for lifetime prevalence) in the birth cohort study (Viner & Hotopf, 2004), 6.8 % in the study of patients with infectious mononucleosis or upper respiratory infections (White et al. 2001), 8.2 % in the sample of fatigued people within the working population of The Netherlands (Huibers et al. 2004b) and 18.5 % in the group of fatigued employees (Huibers et al. 2004a). Identified risk factors The individual studies varied greatly in number and type of risk factors they analysed (see Table 1). These ranged from psychological characteristics (e.g. Huibers et al. 2004a) to very specific exposure incidences such as raw shellfish ingestion (Bell et al. 1991). Factors that were found to be significantly associated with the development of CFS/ME in the final predictive models included: older age (Chalder et al. 2003; Huibers et al. 2004b), being female (Huibers et al. 2004b; Viner & Hotopf, 2004), low or middle rather than high educational level (Huibers et al. 2004b), father in professional/managerial occupation in childhood (Viner & Hotopf, 2004), the presence of an anxiety disorder (Chalder et al. 2003), mood disorder (pre-morbid or 2 months post-infection, White et al. 2001), personality trait emotional instability (Kato et al. 2006), stress (Kato et al. 2006), history of allergies or asthma (Bell et al. 1991), a long-standing limiting medical condition aged 10 years (Viner & Hotopf, 2004), musculoskeletal pain (Hickie et al. 2006), sick certification after viral illness (Cope et al. 1996), low fitness 2 months post-infection (White et al. 2001), no sport in spare time at 10 years old (Viner & Hotopf, 2004), lower physical functioning at baseline assessment (Huibers et al. 2004a), exhaustion (Huibers et al. 2004b), the presence of fatigue at time of viral illness (Cope et al. 1996; Hickie et al. 2006), fatigue severity (Huibers et al. 2004b), days spent in bed at the onset of an infection (White et al. 2001), visits to the general practitioner (GP) or occupational physician (Huibers et al. 2004 b), poor self-rated health (Huibers et al. 2004b), higher scores on malaise inventory at 30 years (Viner & Hotopf, 2004), childhood trauma (Heim et al. 2006), psychological symptom attribution (Cope et al. 1996), other family members with symptoms of CFS (Bell et al. 1991), recent ingestion of raw milk (Bell et al. 1991) and belonging to a general rather than a student primary-care sample (White et al. 2001). All studies assessed the age of the participants and in most studies the association of age and the development of CFS/ME appeared to have been analysed. However, age was only found to be a significant predictor in two of the 11 studies: one a study predicting CFS/ME in children in England, Scotland and Wales (Chalder et al. 2003) and one a sample of the working population of The Netherlands (Huibers et al. 2004b). Gender was a predictor of CFS/ME in the final model in two out of the 11 studies, one following the working population sample (Huibers et al. 2004b) and the other the birth cohort study by Viner & Hotopf (2004). All included studies assessed the gender of the participants and several of the studies tested the association of gender and CFS/ME but in no other study was gender reported as a predictor in the final predictive model. The educational level of the participant or the parents of the participating child was assessed in several studies. This variable was reported as significantly related to the development in CFS/ME in the final prediction model in two studies but conflicting effects were found; while a birth cohort study (Viner & Hotopf, 2004) found that growing up with a father in professional/managerial occupation increased the chances of developing CFS/ME slightly, Huibers et al. (2004 b) reported that those employees with a higher educational level had lower odds of developing CFS/ME. Several studies identified a medical or psychological vulnerability in the history of the participants who suffered from CFS/ME; that is, history of allergies or asthma in children (Bell et al. 1991), presence of an anxiety disorder in children (Chalder et al. 2003), sick certification related to a viral illness and presence of fatigue at the time of viral infection (Cope et al. 1996), a long-standing limiting medical condition at 10 years of age and scoring high on a malaise assessment inventory (Viner & Hotopf, 2004), high stress levels previous to diagnosis (Kato et al. 2006), musculoskeletal pain (Hickie et al. 2006) or a mood disorder (White et al. 2001). Childhood trauma was also associated with an elevated risk of CFS (Heim et al. 2006). Studies that followed up fatigued individuals seem to indicate that higher scores on defining or related CFS/ME criteria at the start of the study increases the likelihood of developing or continuing to suffer from CFS/ME (Huibers et al. 2004a,b). In some studies a defining characteristic of CFS/ME was used, such as fatigue (e.g. Huibers et al. 2004b), to predict CFS/ME, or the differentiation between risk factor and defining characteristic or cause of CFS/ME was not made. Discussion For the presented systematic scoping review we searched the existing international literature thoroughly, applied strict inclusion and exclusion criteria, and took measures to reduce errors and bias to provide a comprehensive and clear overview of the existing multiple predictor studies on risk factors for the development of CFS/ME. This overview demonstrates clearly which potential risk factors have been assessed in the literature, which factors have shown a significant association with the development of CFS/ME, which results have been replicated and where there are inconsistencies in the existing literature. The review identified 11 studies that met the specified inclusion criteria. These used a variety of methodological designs such as case-control study, cross-sectional survey, cohort studies including a study following a birth cohort, a combined cohort and case-control study design and a nested case-control study using data from a twin registry. Several studies used a prospective data collection design. In most studies the focus of the publication was on predicting CFS/ME in adults. No study was identified that predicted severe CFS/ME separately. The quality of the reporting varied but was poor overall, especially with regard to the statistical methods, given the dependency of the results on the methods used. The existing literature shows that a large number and a great variety of potential risk factors for CFS/ME have been assessed. However, the included studies did not appear to reveal risk factors that are evidently useful for clinicians in assisting them to establish a diagnosis for patients presenting with potential symptoms of CFS/ME. Significantly associated with the development of CFS/ME in the final predictive model were: older age, being female, low or middle rather than high educational level, father in professional/managerial occupation in childhood, presence of an anxiety disorder, mood disorder (pre-morbid or 2 months post-infection), emotional instability, childhood trauma, history of allergies or asthma, long-standing limiting medical condition aged 10 years, sick certification after viral illness, low fitness 2 months post-infection, no sport in spare time at 10 years old, lower physical functioning at baseline assessment, exhaustion, presence of fatigue at time of viral illness, fatigue severity, days spent in bed at onset of an infection, visits to GP or occupational physician, musculoskeletal pain, poor self-rated health, higher scores on malaise inventory at 30 years, stress, psychological symptom attribution, other family members with symptoms of CFS, recent ingestion of raw milk and belonging to a general rather than a student primary-care sample. The included studies varied in the number and types of potential risk factors that they tested, ranging from psychological characteristics (e.g. Huibers et al. 2004a) to very specific exposure incidences such as raw shellfish ingestion (Bell et al. 1991). It stands to reason that only when potential factors have been assessed can an association be tested. Not all studies seem to have excluded CFS/ME-defining factors from the prediction of CFS/ME, which makes the studies difficult to compare. The factors age and gender were kept in the final predictive model in two studies indicating that older age (in a sample of children and a working population sample of adults) and female sex increased the risk of developing CFS/ME. In other included studies these factors seem to have been assessed but the influence of the variable was negligible in the presence of other risk factors; for example, Chalder et al. (2003) explicitly reported that being female was not a significant risk factor. In the literature the prevalence of CFS/ME is reported to be approximately twice as common in women (e.g. Jason et al. 1999; Department of Health, 2002; Reyes et al. 2003). Educational level was also kept as a predictor in two of the included studies but the effects were in opposite directions. Even though CFS/ME was initially linked to a higher socio-economic status ('yuppie flu '), overall the studies included in the review do not suggest this, nor do other existing empirical studies (Department of Health, 2002). Several predictors were clearly study specific (e.g. the sample affiliation the participant stemmed from; see White et al. 2001) or had not been assessed in other review eligible studies (e.g. raw milk ingestion; Bell et al. 1991). Some identified risk factors have been reported previously in studies predicting chronic ill health (e.g. being less fit or active after infectious mononucleosis has been replicated as a predictor of prolonged fatigue and state of illness; see Candy et al. 2002). Several studies identified a medical or psychological vulnerability in the history of the person that will develop CFS/ME (e.g. the presence of an anxiety disorder in children or a history of allergies or asthma). However, the identified characteristics seem to be either common occurrences in people with or without CFS/ME or not sufficiently specific to CFS/ME, so it remains unclear whether health-care professionals could use these as a prognostic flag in clinical practice. Acute physical or psychological stress has been discussed in the literature as a precipitating factor that might trigger the onset of CFS/ME (see Prins et al. 2006). While it does appear that the identified factors cannot be used easily in clinical practice to specifically identify those people presenting with early suspected CFS/ME symptoms who will progress to develop CFS/ME, it is noteworthy that several characteristics could be addressed clinically and do not represent perpetuating or untreatable factors. Similarly, Candy et al. (2003) also concluded, regarding predictors of fatigue following the onset of infec- tious mononucleosis, that the individual identified psycho-behavioural predictors were characterized by being amenable to clinical interventions. Our review was restricted to studies that used a specific type of analysis considered to be a suitable source of evidence to answer questions about the empirical evidence of risk factors for CFS/ME. Only studies that predicted the syndrome CFS/ME rather than fatigue or similar single characteristics of the syndrome were considered. Additionally, we set out to review studies that used a methodological design that allowed the simultaneous assessment of multiple potential risk factors, that is the syndrome had to be predicted in a multivariate analysis, a multiple regression or logistic regression. There are several other approaches to determine risk factors, one example being studies that count the prevalence of CFS/ME in a group that has been exposed to a discussed risk factor such as glandular fever (e.g. White et al. 1998), studies comparing the prevalence of CFS/ME in particular subgroups such as patients with or without co-morbid psychological disorders (e.g. Wessely et al. 1997) or studies listing the prevalence of exposures or characteristics comparing a case and control group for CFS/ME (e.g. Reyes et al. 1996). The studies included in this review, by contrast, present a more focused, with regard to the outcome, and more advanced analysis with regard to the field of potential risk factors. The studies varied greatly in the number of participants they identified that fulfilled agreed case definitions of the syndromes, with prevalences ranging from 8 to 447. From this, it follows that some of the studies had only a very small sample of patients with CFS/ME (e.g. Chalder et al. 2003), which provided the basis for finding characteristics that differ- entiate CFS/ME cases from non-cases that weaken the validity of the studies. The prevalences in the included cohort studies varied depending on the composition of the analysed sample of participants. Some studies were based on unselected samples such as a birth cohort study (Viner & Hotopf, 2004) and social survey data on children (Chalder et al. 2003) while others stemmed from selected populations such as a sample of fatigued employees absent from work (Huibers et al. 2004a,b) or a follow-up of patients diagnosed with viral illnesses (e.g. Cope et al. 1996). Large sample sizes are required in unselected samples, given that the prevalence of CFS/ME is an estimated 0.4 % for adults in the UK for instance (Department of Health, 2002; replicated in the included birth cohort study by Viner & Hotopf, 2004). The quality of the evidence depends on the quality of the study design and analysis and whether these allow correct inferences. However, the quality of the research depends also on the questions that are being asked, that is the selection of potential risk factors that are assessed and used as predictors for CFS/ME. The existing body of evidence is characterized by factors that were analysed in several studies but a significant association was not replicated in more than two studies, or by studies that showed significant associations of factors that no other study analysed. Further studies should answer the question of risk factors for CFS and ME with sufficiently large samples and by taking the existing body of evidence on risk factors into account. Acknowledgements We thank Vickie Orton for performing the literature searches, members of the NICE Guideline Development Group for the provision of additional literature, and Paul Wilson, Jeremy Miles and two anonymous reviewers for valuable comments on earlier drafts of the manuscript. The project was funded by the National Institute for Health and Clinical Excellence who commissioned the National Collaborating Centre for Primary Care (part of the Royal College of General Practitioners) to produce guidelines for 'The Diagnosis and Management of Chronic Fatigue Syndrome/Myalgic Encephalomyelitis (or Encephalopathy) in Adults and Children'. The work forms part of the independent synthesis of research evidence to support the development of these guidelines. Declaration of Interest None. The views expressed in this paper are those of the authors and not necessarily those of the NCC-PC, RCGP or National Institute for Health and Clinical Excellence. Table Table 1. Study characteristics and results ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Study Participants Setting Design Method and analysed variables Findings ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Bell et al. n=63 children (21 USA School Case-control Method: Linear/multiple regression Twenty-one children met 1988 CDC criteria (1991) cases, 42 controls), study Predictor variables: Raw milk (at any time, recently), (physician-confirmed diagnosis). % female=48 Retrospective raw eggs, raw shellfish, raw cheese, other family The best logistic regression model included the Age: mean 12.8 years and concurrent members with CFS symptoms, allergies/asthma, variables other family members with symptoms of CFS (cases), 12.3 (controls); data collection private well (at present, at any time), outdoor (RR 35.9, 95 % CI 2.84-488.5, p=0.007), recent median 13; range 6-17 camping, proximity to orchards or farmland, ingestion of raw milk (RR 44.3, 95 % CI 3.21-606.5, exposure to animals in the house (dogs, cats, fish, p=0.005) and history of allergies or asthma (RR 23.3, mice, birds, hamsters, others), home heating 95 % CI 1.67-327.3, p=0.019). RRs adjusted for source (oil, electricity, hot air, wood, kerosene, the other variables natural gas), exposure to animals on property (cattle, horses, sheep, swine, cats, dogs, ducks, goats, chickens, others), tick bite, blood transfusion, appendicitis Chalder et al. n=4240 children UK Cross-sectional Method: Logistic regression Eight children met 1994 CDC criteria. Risk factors for (2003) Age: range 11-15 years Community survey Predictor variables: Age, sex, mother's score on CFS (CDC, 1994) were older age (OR 1.9; 95 % CI Concurrent data GHQ, presence of any anxiety disorder and any 1.0-3.7, p=0.03) and presence of an anxiety disorder collection depressive disorder (OR 8.8; 95 % CI 1.8-43.5, p=0.008). Female sex was not a significant risk factor (p=0.8) Cope et al. n=128 (64 cases of UK Primary Cohort and Method: Logistic regression Twenty-three patients met the Oxford criteria. CFS (1996) chronic severe fatigue, care Case-control Predictor variables: Age, sex, GHQ-3 score 6 months after viral illness was predicted by sick 64 controls), study (psychiatric morbidity), symptom attributional certification (OR 8.5, 95 % CI 4.2-17.2, p=0.002), a % female=78 Prospective data style (mainly psychological, equally psychological symptom attribution (OR 2.1, 95 % CI Age: range 18-45 years; collection psychological/physical, mainly physical), sick 1.6-2.7, p=0.007) and presence of fatigue at the time mean 30.5 (cases), 31.4 certification and the presence of fatigue (all of viral illness (OR 6.4, 95 % CI 2.5-16.4, p=0.05) (controls); S.D.=6.5 recorded at the time of the viral illness), and (cases), 7.3 (controls) past psychiatric history Heim et al. n=103 (43 cases, USA Case-control Method: Logistic regression Forty-three participants met 1994 CDC criteria. (2006) 60 controls), Community study Predictor variables: Exposure to childhood trauma Childhood trauma was associated with an elevated % female=82 Retrospective adjusted for: age, sex, race; sample split by CFS risk even in the presence of low levels of Age: mean 50.5 years data collection low versus high psychopathology psychopathology; risk increased in the presence of high levels of psychopathology (ranging from OR 3.39 when predicting CFS and low level anxiety to OR 8.07 when predicting CFS and high level depression). Analyses adjusted for age, sex and race Hickie et al. n=253, % female=43 Australia Cohort and Method: Linear/multiple regression Twenty-eight participants met 1994 CDC criteria. (2006) Age: mean 34 years Primary case-control Predictor variables: Age, sex, education, acute Fatigue at baseline predicted CFS post-infection at care study sickness, irritability, musculoskeletal pain, mood 6 (p<0.001) and 12 months (p<0.05), musculoskeletal Prospective data disturbance, neurocognitive disturbance, fatigue, pain at baseline predicted CFS at 6 months (p<0.05) collection pre-morbid psychiatric disorder, intercurrent psychiatric disorder, neuroticism, locus of control, confirmed Epstein-Barr virus, confirmed Ross River virus, confirmed Q fever Huibers et al. n=151 (138 at The Cohort study Method: Logistic regression Sixty-six participants met 1994 CDC criteria. Lower (2004a) follow-up), Netherlands Prospective data Predictor variables: Age, sex, education, group baseline scores on physical functioning predicted % female=55 Workplace collection allocation (therapy or control), fatigue (Checklist CFS in the whole group (OR per S.D.=0.27, 95 % CI Age: mean 43 years Individual Strength), duration of fatigue, physical 0.15-0.46) and in those not CFS-like cases at functioning (SF-36), pain (SF-36), and self- baseline (OR per S.D.=0.15, 95 % CI 0.06-0.46) at reported health (SF-36), depression (SCL-90), 12 months' follow-up. cognitive difficulties (SCL-90), somatization In the CFS-like employees at baseline, CFS was (SCL-90), exhaustion (MBI-GS), professional predicted by higher fatigue (OR per S.D.=2.25, 95 % efficacy (MBI-GS), self-efficacy (SES), somatic CI 1.06-5.31) and pain scores (OR per S.D.=2.56, attribution (CAL), psychological attributions 95 % CI 1.37-4.76) (CAL), absence from work at baseline Huibers et al. n=1143, % female=35 The Cohort study Method: Logistic regression Ninety-four participants met 1994 CDC criteria at (2004b) (CFS-like cases at Netherlands Prospective and Predictor variables: Age, sex, low/middle/high follow-up. Baseline factors that predicted CFS-like follow-up); 27 Community concurrent data educational level, fatigue severity (Checklist caseness compared with non-CFS fatigue at 44 (non-CFS fatigue (Sample of collection Individual Strength), burnout, need for recovery, months follow-up were older age (OR per S.D.=1.36, cases); 33 (non-fatigue the working emotional exhaustion, psychological distress, 95 % CI 1.08-1.84), exhaustion (OR 1.33, 95 % CI cases) population) anxious mood, depressed mood, physical 1.04-1.67) female sex (OR 0.41, 95 % CI 0.24-0.70), Age: mean 43.8 years attribution of fatigue, psychological attribution low educational level (v. high, OR 3.82, 95 % CI (CFS-like cases); 41.2 of fatigue, no specific attribution of fatigue, 1.92-7.61), middle educational level (v. high, OR (non-CFS fatigue self-rated health, health complaints, absent from 2.48, 95 % CI 1.35-4.51) and self-reported visits to cases); 41.3 (non- work, impairment in work, impartment in the GP (OR 1.98, 95 % CI 1.19-3.29). Factors that fatigue cases); activities, pregnancy, shocking life events in predicted CFS-like caseness compared with noxxx S.D.=7.5 (CFS-like past 12 months, visit to GP or occupational fatigue were fatigue severity (OR 1.37, 95 % CI cases); 7.9 (non-CFS physician, sleep disturbances 1.11-1.86), exhaustion (OR 1.66, 95 % CI 1.28-2.18), fatigue cases); 7.9 low educational level (v. high, OR 2.61, 95 % CI (non-fatigue cases) 1.29-5.24), middle educational level (v. high, OR 2.38, 95 % CI 1.27­4.45) self-reported visits to the GP (OR 3.06, 95 % CI 1.71-5.50) and occupational physician (OR 0.46, 95 % CI 0.22-0.95) and poor self-rated health (OR 0.56, 95 % CI 0.33-0.95) Kato et al. n=19 150, Sweden Nested case- Method: Conditional logistic regression Four hundred and forty-seven participants met 1994 (2006) % female=53 Community control study Predictor variables: Extraversion, emotional CDC criteria. Age: range 42-64 years (twin and twin study instability, stress, interaction between the There was a 72 % CFS risk increase with each standard registry) Prospective data variables deviation increase in emotional instability and a 64 % collection increase for stress. When genetic influences were controlled (by including only monozygotic twins), the association between emotional instability and chronic fatigue was no longer significant; however, the impact of stress became more pronounced suggesting some genes may serve as a buffer and others may increase susceptibility to stress Viner & n=11 261, UK Cohort study Method: Logistic regression Ninety-three participants reported ever having Hotopf % female=51 Community Prospective data Predictor variables: Sex, father in professional/ CFS/ME (with 48 having it currently). CFS/ME (2004) Age: 29-30 years collection managerial occupation, mother achieved A levels was associated with having a long-standing limiting at follow-up or equivalent/degree/diploma, ethnicity, medical condition at 10 years (OR 2.2, 95 % CI 1.3-3.8), birthweight, birth order, presence of long- female sex (OR 2.4, 95 % CI 1.4-4.3), father in standing medical condition significantly limiting professional/managerial occupation in childhood (OR home and school life at 10 years, history of atopy 2.5, 95 % CI 1.4-4.3), high scores on malaise inventory by 10 years; obesity at 10 years; sport played in at 30 years (OR 2.6, 95 % CI 1.6-4.3) and negatively spare time and at school at 10 years, days of with sometimes or often played sport in spare time school missed for health or emotional reasons in rather than never at 10 years (OR 0.5, 95 % CI past year at 10 years, score on British ability scales 0.3-1.0). Mother's educational status and at 10 years; significant illness in either parent socio-economic status at 30 years were a priori before 10 years, child behaviour problems included in the model (Rutter scale) and maternal malaise at 5 years; self-esteem, Rutter score, conduct/impulsive/ hyperactive score and maternal malaise at 10 years, score on adolescent GHQ and maternal malaise inventory score at 16 years, malaise and occupation at 30 years Wessely et al. n=2366 (1199 patients UK Primary Cohort and Method: Linear/multiple regression Thirty-six participants met CDC 1994, 33 Oxford, (1995) with symptomatic care case-control Predictor variables: Age, sex, social class, number 20 Australian and 16 CDC 1988 criteria. There were infections, 1167 other study of visits to GP, fatigue questionnaire, GHQ scores no significant differences between the exposed and patients), Prospective data before presentation (from community screening) non-exposed cohorts in the proportion of CFS % female=68 collection and at presentation; fatigue questionnaire and cases (by Oxford, Australian or CDC criteria) at (exposed cohort), 70 GHQ scores at 6-month follow-up; results of CFS 6 months. Adjustment for number of visits to GP (non-exposed cohort) checklist, psychological assessment, functional during the year before recruitment did not affect Age: range 18-45 years; impairment (MOS-SF), psychiatric morbidity, the OR for CFS. Multivariate analyses for the mean 32.7 (exposed), somatic symptoms for fatigue, coping, life events, syndrome were not reported. The authors concluded 33.5 (non-exposed); anxiety and depression (HADS) that the study provides no evidence that common S.D.=7.5 (exposed); infections are related to the development of chronic 7.5 (non-exposed) fatigue or CFS White et al. n=250; % female=51 UK Primary Cohort study Method: Logistic regression Thirty-eight participants met Oxford, 17 the 1994 (2001) Age: range 16-65 years; care Prospective data Predictor variables: Age, sex, sample (GP v. student), CDC criteria (18 patients with 'idiopathic chronic median=22 collection vocabulary IQ, socio-economic class, father's fatigue' were added to this group). Significant socio-economic class, infectious mononucleosis, predictors of CFS (Oxford) 6 months after an infection positive Monospot at onset, Epstein-Barr virus were belonging to a general rather than student primary- IgM positive, cervical lymphadenopathy, atopy, care sample (OR 3, 95 % CI 1.63-5.49, p<0.001), mood biochemical markers, fatigue at onset, time in bed disorder at 2 months (OR 2.32, 95 % CI 1.22-4.42, at onset, exercise power, fitness (1-min step test), p<0.01) and fitness at 2 months (OR 0.35, 95 % CI GP attendances in years before onset, pre-morbid 0.22-0.90, p<0.05). When predicting CDC defined psychiatric disorder (various times and GP CFS, a pre-morbid mood disorder (OR 1.82, 95 % CI record), pre-morbid psychiatric treatment, 1.15-2.89, p<0.01) rather than the mood disorder at pre-morbid mood disorder, anxiety and 2 months was predictive, as was days in bed (OR depression (HADS at 1 and 2 months), social 1.07, 95 % CI 1.02-1.12, p<0.01). There was a positive adversity, extroversion, emotionality (self- and interaction between emotional personality and peer rating) bed rest ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ CFS, Chronic fatigue syndrome; ME, myalgic encephalomyelitis; S.D., standard deviation; OR, odds ratio; CI, confidence interval; RR, relative risk; GP, general practitioner; GHQ, General Health Questionnaire; SF-36, Short Form Health Survey; SCL-90, Symptom Checklist 90; MBI-GS, Maslach Burnout Inventory - General Survey; SES, Self-Efficacy Scale; CAL, Causal Attributions List; MOS-SF, Medical Outcome Study 20-item questionnaire; HADS, Hospital Anxiety and Depression Scale; IgM, immunoglobulin M. A detailed data extraction table can be obtained from the authors. Fig. 1. Study flow. +-------------------------------------------+ |10768 publications identified with search | |strategy, relevance screening by title | |and abstracts (plus keywords where | |available) | +---------------------+---------------------+ | +-----------------------------------+ | | 10125 studies excluded (reason: | +------------------------> | not about needs, obviously no | | | multiple predictors, obviously | V | discussion paper only, obviously | +-------------------------------------------+ | not CFS/ME, obvious duplicate | |643 potentially relevant studies retrieved,| | publication) | |inclusion assessment by full text, 3 | +-----------------------------------+ |recent studies identified upon peer review | +---------------------+---------------------+ | +-----------------------------------+ | |635 studies excluded (reason: not | +------------------------> |about risk, not CFS/ME, no multiple| | |predictors, no empirical study/ | | |discussion paper only, duplicate | V |publications) | +-------------------------------------------+ +-----------------------------------+ |11 studies met inclusion criteria and are | |included in the review | +-------------------------------------------+ References Afari N, Buchwald D (2003). Chronic fatigue syndrome: a review. American Journal of Psychiatry 160, 221-236. Bagnall A-M, Hempel S, Chambers D, Orton V, Forbes C (2005). The Investigation and Management of Chronic Fatigue Syndrome (CFS)/Myalgic Encephalomyelitis (ME) in Children and Adults - Systematic Review. University of York, Centre for Reviews and Dissemination: York. Bell KM, Cookfair D, Bell DS, Reese P, Cooper L (1991). Risk factors associated with chronic fatigue syndrome in a cluster of pediatric cases. Reviews of Infectious Diseases 13 (Suppl. 1), S32-S38. Candy B, Chalder T, Cleare AJ, Peakman A, Skowera A, Wessely S, Weinman J, Zuckerman M, Hotopf M (2003). Predictors of fatigue following the onset of infectious mononucleosis. Psychological Medicine 33, 847-855. Candy B, Chalder T, Cleare AJ, White PD, Hotopf M (2002). Recovery from infectious mononucleosis: a case for more than symptomatic therapy? A systematic review. British Journal of General Practice 52, 844-851. Chalder T, Goodman R, Wessely S, Hotopf M, Meltzer H (2003). Epidemiology of chronic fatigue syndrome and self reported myalgic encephalomyelitis in 5-15 year olds: cross sectional study. British Medical Journal 327, 654-655. Chambers D, Bagnall A-M, Hempel S, Forbes C (2006). Interventions for the treatment, management and rehabilitation of patients with chronic fatigue syndrome/ myalgic encephalomyelitis (CFS/ME): an updated systematic review. Journal of the Royal Society of Medicine 99, 506-520. Cope H, Mann A, Pelosi A, David A (1996). Psychosocial risk factors for chronic fatigue and chronic fatigue syndrome following presumed viral illness: a case-control study. Psychological Medicine 26, 1197-1209. Department of Health (2002). A Report of the CFS/ME Working Group: Report to the Chief Medical Officer of an Independent Working Group. Department of Health: London. Fukuda K, Straus SE, Hickie I, Sharpe MC, Dobbins JG, Komaroff AL, and the International CFS Study Group (1994). The chronic fatigue syndrome: a comprehensive approach to its definition and study. Annals of Internal Medicine 121, 953-959. Heim C, Wagner D, Maloney E, Papanicolaou DA, Solomon L, Jones JF, Unger ER, Reves WC (2006). Early adverse experiences and risk for chronic fatigue syndrome. Archives of General Psychiatry 63, 1258-1266. Hickie I, Davenport T, Wakefield D, Vollmer-Conna U, Cameron B, Vernon SD, Reeves WC, Lloyd A; Dubbo Infection Outcomes Study Group (2006). Post-infective and chronic fatigue syndromes precipitated by viral and non-viral pathogens: prospective cohort study. British Medical Journal 333, 575. Huibers MJ, Bleijenberg G, van Amelsvoort LG, Beurskens AJ, van Schayck CP, Bazelmans E, Knottnerus A (2004a). Predictors of outcome in fatigued employees on sick leave: results from a randomised trial. Journal of Psychosomatic Research 57, 443-449. Huibers MJ, Kant IJ, Knottnerus JA, Bleijenberg G, Swaen GM, Kasl SV (2004b). Development of the chronic fatigue syndrome in severely fatigued employees: predictors of outcome in the Maastricht cohort study. Journal of Epidemiology and Community Health 58, 877-882. Jason LA, Richman JA, Rademaker AW, Jordan KM, Plioplys AV, Taylor RR, McCready W, Huang C-F, Plioplys S (1999). A community-based study of chronic fatigue syndrome. Archives of Internal Medicine 159, 2129-2137. Kato K, Sullivan PF, Evengard B, Pedersen NL (2006). Premorbid predictors of chronic fatigue. Archives of General Psychiatry 63, 1267-1272. Prins JB, van der Meer JWM, Bleijenberg G (2006). Chronic fatigue syndrome. Lancet 367, 346-355. Reyes M, Dobbins JG, Mawle AC, Steele L, Gary Jr. HE, Malani H, Schmid S, Fukuda K, Stewart J, Nisenbaum R, Reeves WC (1996). Risk factors for chronic fatigue syndrome: a case-control study. Journal of Chronic Fatigue Syndrome 2, 17-33. Reyes M, Nisenbaum R, Hoaglin DC, Unger ER, Emmons C, Randall B, Stewart JA, Abbey S, Jones JF, Gantz N, Minden S, Reeves WC (2003). Prevalence and incidence of chronic fatigue syndrome in Wichita, Kansas. Archives of Internal Medicine 163, 1530-1536. Sharpe MC, Archard LC, Banatvala JE, Borysiewicz LK, Clare AW, David A, Edwards RHT, Hawton DM, Lambert HP, Lane RJM, McDonald EM, Mowbray JF, Pearson DJ, Peto TEA, Preedy VR, Smith AP, Smith DG, Taylor DJ, Wessely S, White PD (1991). A report - chronic fatigue syndrome: guidelines for research. Journal of the Royal Society of Medicine 84, 118-121. Viner R, Hotopf M (2004). Childhood predictors of self reported chronic fatigue syndrome/myalgic encephalomyelitis in adults: national birth cohort study. British Medical Journal 329, 941. Wessely S, Chalder T, Hirsch S, Pawlikowska T, Wallace P, Wright DJ (1995). Postinfectious fatigue: prospective cohort study in primary care. Lancet 345, 1333-1338. Wessely S, Chalder T, Hirsch S, Wallace P, Wright D (1997). The prevalence and morbidity of chronic fatigue and chronic fatigue syndrome: a prospective primary care study. American Journal of Public Health 87, 1449-1455. White PD (2004). What causes chronic fatigue syndrome? British Medical Journal 329, 928-929. White PD, Thomas JM, Amess J, Crawford DH, Grover SA, Kangro HO, Clare AW (1998). Incidence, risk and prognosis of acute and chronic fatigue syndromes and psychiatric disorders after glandular fever. British Journal of Psychiatry 173, 475-481. White PD, Thomas JM, Kangro HO, Bruce-Jones WD, Amess J, Crawford DH, Grover SA, Clare AW (2001). Predictions and associations of fatigue syndromes and mood disorders that occur after infectious mononucleosis. Lancet 358, 1946-1954. -------- (c) 2007 Cambridge University Press