The data used in this study came from the Lifelines study, a multi-disciplinary prospective population-based cohort study examining in a unique three-generation design the health and health-related behaviors of 167,729 individuals living in the north of Netherlands (22). The study assessed a broad range of biomedical, socio-demographic, behavioral, physical and psychological variables which may contribute to future health outcomes (22). People with low educational attainment and smokers are somewhat under-represented in the Lifelines cohort but otherwise, it is representative of the total Dutch population (23). Exclusion criteria were: severe psychiatric mental disorders (e.g., schizophrenia), severe physical illness, insufficient fluency in the Dutch language to complete questionnaires, inability to visit the general practitioner and limited life expectancy (<5 years). The participants were recruited between 2006 and 2013 and followed up twice during the subsequent 3 years, on average 17 months and 29 months after the baseline. At each follow-up, abbreviated versions of the baseline questionnaire were administered. Written informed consent was obtained from all participants. The study followed the guidelines of the Declaration of Helsinki and all procedures involving human subjects were approved by the UMCG Medical Ethical Committee under number 2007/152.

The present study included respondents who were 18 or more years of age at baseline and had completed the questionnaire items relevant to fibromyalgia or CFS at baseline and follow-up. Not all participants completed the detailed interview to diagnose psychiatric disorder (the MINI neuropsychiatric interview (see below).

The baseline variables were chosen from the comprehensive Lifelines questionnaire if they had been reported as predictors in previous articles or our previous analysis (3, 13, 21). They included: socio-demographic variables, general medical illnesses, psychiatric disorders, stress, Body Mass Index (BMI), health behaviors, healthcare use, and medication use.

Socio-demographic features included sex, age, duration of formal education, marital and employment status and level of income.

Somatic symptoms were measured using the somatization scale of the Symptom Check List-90 questionnaire (24). It has been shown to be suitable for epidemiological studies (25). It has 12 items and each has five response categories. The total score was used in the logistic regression analyses.

The analysis used the self-reported onset of fibromyalgia (or CFS) at one or both of the two follow-up assessments (at 17 months or 29 months) as the outcome variable. A new onset of fibromyalgia was recorded when, at either follow-up questionnaire, the respondent answered yes to the question “have you had fibromyalgia since the last questionnaire”? Similarly, onset of CFS was recorded when the participant answered yes to the question “have you had CFS since the last questionnaire”?

All analyses were performed on SPSS statistics 25. For the fibromyalgia analysis, participants were allocated to one of the following groups at baseline: (i) no current or prior self-reported fibromyalgia or current muscle pain, (ii) no current or prior fibromyalgia but bothered “very much” or “extremely” by pain in their muscles, (iii) has current or prior fibromyalgia. Groups (i) and (ii) were then divided according to whether or not respondents reported fibromyalgia at follow-up. This yielded five groups:

No self-reported fibromyalgia or current marked muscle pain at baseline ➔ (a) no self-reported fibromyalgia at follow-up, (b) self-reported fibromyalgia at follow-up.

No self-reported fibromyalgia but marked muscle pain at baseline ➔(c) no self-reported fibromyalgia at follow-up, (d) self-reported fibromyalgia at follow-up, (e) Self-reported fibromyalgia at baseline.

The same procedure was followed for the CFS analysis:

No prior CFS or current persistent fatigue ➔ (a) no CFS at follow-up, (b) self-reported CFS at follow-up.

No prior CFS but feeling tired “most” or “all of the time” ➔ (c) no CFS at follow-up, (d) self-reported CFS at follow-up, (e) self-reported existing/prior CFS.

The Lifelines database included 148,614 participants aged 18 or over who had completed the questions regarding fibromyalgia both at baseline and follow-up (Figure 1). Of these, 4,710 reported previous fibromyalgia and a further 7,481 reported marked muscle pain at baseline. Of the 136,423 participants without fibromyalgia or marked muscle pain, 685 (0.5%) reported a new onset of fibromyalgia during the follow-up for 2.4 years. Of the 7,481 participants with marked muscle pain, 281 (3.75%) reported new-onset fibromyalgia during follow-up. A further 28 participants reported onset of both CFS and fibromyalgia during the follow-up period and were excluded from this study (Figure 2).

Of 136,423 participants with complete data regarding CFS, 1983 reported CFS at baseline and a further 10,217 reported persistent fatigue. Of 124,223 participants without prior CFS or persistent fatigue at baseline, 292 (0.23%) reported CFS onset during follow-up. Of the 10,217 with persistent fatigue at baseline, 192 (1.9%) reported CFS onset during follow-up.

Overall, 786 participants reported onset of fibromyalgia (with or without prior marked muscle pain) and 146 (18.6%) of these had DSM-IV psychiatric disorders at baseline. Of 334 new onsets of reported CFS (with or without persistent fatigue), 90 (27%) had psychiatric disorder.

Univariable analysis of baseline variables to test their association with self-rated fibromyalgia or CFS.

Potential predictors of self-rated fibromyalgia (except psychiatric disorders) are shown in Table 1 using univariable analysis to compare participants who did, and did not, report onset of fibromyalgia. It can be seen that in participants who were free of muscle pain at baseline (Table 1 columns 1–4), fibromyalgia onset was significantly associated with nearly all potential predictors. By contrast, among participants with marked muscle pain (columns 5–7), only some of the potential predictors of fibromyalgia were significantly associated with subsequent self-reported fibromyalgia onset. These predictors were: female sex, chronic sinusitis, GERD, osteoarthritis, rheumatoid arthritis, IBS, impaired sleep and SCL somatization score. The extreme right hand column of Table 1 shows, purely for comparison purposes, the data for participants who reported fibromyalgia at baseline.

Comparable data for CFS are displayed in Table 2. For participants who did not report persistent fatigue at baseline (columns 1–4), nearly all of the potential predictors were significantly associated with reported CFS onset. In the group with persistent fatigue at baseline (columns 5–7), approximately half of the potential predictors showed a significant association with reported CFS onset (older age, non-single marital status, off sick, low income, few years of education, chronic cystitis, GERD, osteoarthritis, rheumatoid arthritis, IBS, fibromyalgia, paracetamol use, impaired sleep, SCL somatization score and negative health perception). Reported CFS onset was not associated with female sex whether or not there was prior persistent fatigue.

Univariable analysis of psychiatric disorders and subsequent reported fibromyalgia and CFS.

Data concerning psychiatric disorders in these groups are shown in Tables 3, 4. In participants without marked muscle pain at baseline (columns 1–4 in Table 3), psychiatric disorder was clearly associated with subsequent self-reported fibromyalgia with both self-reported prior psychiatric disorders and DSM IV disorders. For participants with marked baseline muscle pain (columns 5–7 of Table 3), the difference is not significant for self-reported prior psychiatric disorders and only just reaches significance for DSM-IV disorders.

The pattern was different in CFS. Whether there was prior persistent fatigue or not, onset of CFS was associated with psychiatric disorders either self-reported or measured using the MINI interview (Table 4).

The last column of Tables 3, 4 show the proportion of prevalent cases of fibromyalgia and CFS who had psychiatric disorder at baseline. The proportions with any DSM-IV disorders (19.4% for fibromyalgia and 27% for CFS) were very similar to the proportion of new-onset cases with psychiatric disorder at baseline (18.6 and 27% respectively).

Logistic regression analyses to adjust for all potential predictors.

In the logistic regression analyses including all potentially relevant variables, the independent predictors of self-reported fibromyalgia were: female sex, impaired sleep quality, negative perception of health, married/cohabiting status and osteoarthritis (Table 5). Psychiatric disorders, either total number or individual diagnosis, were not a predictor (Table 5). If psychiatric disorder yes/no was entered instead of number of psychiatric disorder, OR = 0.96 [95%CI: 0.74–1.24). If number of MINI diagnoses was entered instead of self-rated psychiatric disorders, OR = 1.05 (0.84–1.32). When bodily pain was added as an independent variable to the regression analysis, it was a strong predictor (OR = 1.41 [1.22–1.61], p < 0.001); the other predictors remained the same.

By contrast, the logistic regression analysis identified predictors of self-reported CFS as: negative perception of health, number of psychiatric disorders, impaired sleep quality, older age, fibromyalgia and low income (Table 6). When the individual psychiatric diagnoses were included in the regression analysis, only depression emerged as a predictor (OR = 1.83 [95%CI, 1.33–2.51]). This remained the case when the RAND variable vitality was added to the regression analysis; (lack of) vitality was a strong predictor (OR = 1.50 [1.18–1.92]), but the other predictors were unaffected.

When the regression analyses were repeated using the DSM-IV diagnoses instead of self-reported psychiatric disorders, the DSM IV diagnoses of GAD (OR = 1.67 [1.18–2.38]) and MDD (OR = 1.91 [1.34–2.72]) were significant predictors of CFS. But neither were predictors of fibromyalgia: GAD OR = 1.02 [0.71–1.48]). MDD: OR = 0.89 [0.57–1.37]).

The strengths and limitations of this study must be recognized. The strengths include its prospective design, the wide range of predictors and a sample size large enough to analyze separately participants with and without marked muscle pain (or persistent fatigue for the CFS analysis) at baseline. The study included two measures of psychiatric diagnosis: a lifetime history (self-reported) and an objective measure (MINI interview) at baseline.

The limitations include the predictors, which did not include recent infection, childhood abuse and health anxiety as these are not included in the Lifelines database (3, 13, 18). There were relatively few onset cases compared to the large population studied so the results of the logistic regression analyses must be interpreted with caution. The most important limitation was the reliance on self-report for the main outcomes, which means that inaccurate diagnosis or under-reporting cannot be ruled out though the incidence reported here is similar to comparable studies (3, 18–20).

Most previous studies have concerned prevalent cases and, like the present one, have reported a higher rate of psychiatric disorders in CFS than fibromyalgia (1, 10, 32). In adolescents rates of anxiety and depression in CFS were twice those in chronic widespread pain (17% v 8%); depression was a predictor of CFS but not chronic widespread pain (33). It is notable that the prevalence of depression and anxiety in population-based samples (22–33%) is much lower than in clinical samples (50 or 66%) (2, 4–6, 13, 21, 34–36). There do not appear to be any previous studies comparing the prevalence of psychiatric disorder in incident cases of fibromyalgia and CFS; the current study found it to be lower in self-reported fibromyalgia than in self-reported CFS.

Previous studies have found the following to be associated with the onset of fibromyalgia: GERD, sleep disorder, IBS, somatic symptoms, anxiety/depression osteoarthritis, migraine, allergic rhinitis and back pain (3). Many studies found depression or anxiety were associated with fibromyalgia, however, in univariate but not in multivariate analyses; the latter included impaired sleep, pain, somatic symptoms and illness behavior (3). Impaired sleep and pain were predictors in the current study and the latter two are closely associated with negative perception of health.

In the most comparable previous study of CFS, prior medical consultation with fatigue was the strongest predictor of CFS onset; depressive disorder was one of several other predictors (13). Many studies of CFS onset have examined cohorts following infection and the evidence concerning psychiatric disorder is conflicting (37). Only half of the relevant birth cohort studies found that CFS onset was predicted by prior psychiatric disorder; these cohort studies did not include important covariates, such as general medical illnesses (14, 17–19).

In the current study, the predictors of self-reported fibromyalgia in participants with prior muscle pain were similar, though less numerous, to those reported previously for fibromyalgia in participants without marked muscle pain (21). In contrast, the predictors of self-reported CFS in this study were more numerous than in the previous one which excluded participants with baseline fatigue (21). Psychiatric disorders, older age, fibromyalgia, osteoarthritis, chronic cystitis and low income were additional predictors so a more complete picture is developed when studying CFS onset in participants who have pre-existing persistent fatigue. It is notable that 40% of all onsets of self-reported CFS occurred in participants with prior fatigue; the proportion with prior muscle pain was only 29% of new onsets of self-reported fibromyalgia.

The evidence from sound population-based studies identifying the risk factors for fibromyalgia and CFS is still limited. Our previous study excluded from the penalized regression analysis participants who reported marked muscle pain or persistent fatigue at baseline on the basis that these participants may have already had unrecognized fibromyalgia or CFS (21). Although that study included the majority of new onsets of self-reported fibromyalgia and CFS, the current study has highlighted the other important groups of new onsets—those with prior muscle pain or persistent fatigue; the risk factors for these onsets appear to differ somewhat from those in the original study. In particular, psychiatric disorder was a predictor of self-reported CFS which develops in the presence of pre-existing persistent fatigue.

It appears that severe muscle pain or persistent fatigue at baseline resolves in some participants and is maintained or worsens in others (Figures 3, 4). This is compatible with the notion that the predictors of onset of fibromyalgia or CFS in people with prior pain or fatigue are actually predictors of worsening of the symptom, or its becoming chronic, rather than predictors of the original pain or fatigue; other general medical disorders as well as psychiatric disorders may play a part in this onset of chronicity (38, 39). Another possibility is that these are predictors of the recognition, probably by a doctor, that the pain or fatigue is diagnosed as fibromyalgia or CFS.

There is a similarity between CFS and IBS; both are seen in post-infective forms, although this seems to be a minority (40). A subgroup of IBS develops in the context of prior psychiatric disorder (41). The current study suggests that the same may be true of CFS. Both viral and bacterial infections have been associated with CFS onset and it is interesting that among participants developing CFS in the current study chronic cystitis was a predictor (13, 37, 42). This reminds us that infective and psychiatric predictors may occur together and are not mutually exclusive.

There is clear evidence that depression and anxiety are common in established cases of fibromyalgia and CFS and these form an important clinical problem (5, 43). Only a minority have psychiatric disorder prior to the development of fibromyalgia or CFS, however, only in CFS were depression and anxiety predictors in the current study. This is one of several differences in the risk factors for developing fibromyalgia and CFS; another recent study found that patients with fibromyalgia have a relatively unique family genetic risk score profile which is distinct from IBS and CFS (44).

Taken together, these studies suggest that greater consistency of findings concerning such mechanisms as metabolomic dysregulation or changes in the immune system in fibromyalgia and CFS might be achieved if subgroups of patients with fibromyalgia or CFS were studied (45, 46). As it is increasingly recognized that CFS and FM are heterogeneous conditions, the presence or absence of psychiatric disorders or multiple somatic symptoms prior to onset of the “functional somatic syndromes” might provide an alternative to the subtypes derived from cluster analysis of established cases (47–49).

Further studies are needed to confirm the findings of this study, firstly to assess whether or not anxiety and depressive disorders seen in established fibromyalgia and CFS were present before the syndrome started and, secondly, to confirm that only in CFS does psychiatric disorder contribute to the development of the syndrome.