Potential causal factors of CFS/ME: a concise and systematic scoping review of factors researched

Chronic fatigue syndrome/myalgic encephalomyelitis (CFS/ME) is understood as a complex condition, likely triggered and sustained by an interplay of biological, psychological, and social factors. Little oversight exists of the field of causal research. This systematic scoping review explores potential causal factors of CFS/ME as researched by primary studies. We searched eight databases for primary studies that examined potential causal factors of CFS/ME. Based on title/abstract review, two researchers independently sorted each study’s factors into nine main categories and 71 subordinate categories, using a system developed with input given during a 2018 ME conference, specialists and representatives from a ME patient advocacy group, and using BMJ Best Practice’s description of CFS/ME etiology. We also extracted data related to study design, size, diagnostic criteria and comparison groups. We included 1161 primary studies published between January 1979 and June 2019. Based on title/abstract analysis, no single causal factor dominated in these studies, and studies reported a mean of 2.73 factors. The four most common factors were: immunological (297 studies), psychological (243), infections (198), and neuroendocrinal (198). The most frequent study designs were case–control studies (894 studies) comparing CFS/ME patients with healthy participants. More than half of the studies (that reported study size in the title/abstract) included 100 or fewer participants. The field of causal hypotheses of CFS/ME is diverse, and we found that the studies examined all the main categories of possible factors that we had defined a priori. Most studies were not designed to adequately explore causality, rather to establish hypotheses. We need larger studies with stronger study designs to gain better knowledge of causal factors of CFS/ME.

The vastness of the research field on etiological factors reflects the variety of causal hypotheses [19], and in the past two years alone, numerous systematic reviews have been published examining distinct causal factors such as immunological [12,16,29], metabolic [15,23], circulatory [27], and epigenetic factors [3]. In 2016, the Research Council of Norway selected the CFS/ME field as the pilot field for a research priority setting partnership strategy, in which patients, next of kin, and related organizations prioritized research questions for commissioned research [14]. When topics for CFS/ME-related research proposals were requested [31], patients and next of kin submitted more than 700 submissions. Sixty percent of their submissions involved questions about causal factors. At the suggestion of the Research Council of Norway, we conducted a systematic scoping review of primary studies that examined predisposing, triggering and maintaining factors to CFS/ME.

Methods
A systematic scoping review presents an overview of research relating to a specific topic or question, without producing a summary answer to that question or assessing the methodological quality of primary studies [4,33]. This method was well-suited to this review's exploratory and descriptive aims. Following Arksey and O'Malley's framework for scoping reviews, we collaborated with a working group from the Norwegian ME Association to develop the research question as well as inclusion and exclusion criteria, and we published a protocol in Norwegian [25]. We included all primary studies, of any design besides case reports, that examined potential predisposing, triggering, and maintaining factors in patients with CFS/ME. We searched eight databases in June 2019 with free text words (in title and summary) and controlled terms for the patient group, combined with free text words and terms for cause/risk. We had no restrictions relating to year of publication or language. Additional file 1: Appendix S1 contains an overview of all databases searched as well as the search strategy for OVID. The PRISMA Checklist for Scoping Reviews is available in Additional file 2: Appendix S2.
Two researchers independently screened titles/ abstracts for inclusion using Covidence software [34]. In the event of disagreement or uncertainty, we consulted a third researcher. Using EPPI Reviewer [32], one researcher extracted data from titles/abstracts on diagnostic criteria, comparison group, study design, number of participants, and the region where the study was conducted (Table 1). We did not retrieve the full texts of the included studies, and we based the data extraction on the titles and abstracts only.
Two researchers independently categorized every potential causal factor using a system of nine major categories, 48 sub-categories, and 23 further subordinate categories, as specified in a codebook. We developed this coding system based on the description of CFS/ME in BMJ Best Practice [6], and supplemented with input from an ME conference in 2018, as well as from specialists and representatives from the ME Association. Figure 1 is a screenshot of an abstract in EPPI Reviewer, annotated to display coding. The nine causal categories were infections, immunological, neuroendocrinal/hormonal/metabolic, genetic/ epigenetic, circulatory, gastrointestinal, neurobiological, psychological/psychosocial/socioeconomic and other factors. If a study reported multiple factors, we coded all factors, but each factor was mutually exclusive (that is, two categories were never required to describe one factor). Each main category had up to nine sub-categories. For example, we divided infections into bacteria, viruses, fungi, parasites, other and not reported. Most sub-categories had further subordinate categories; for example, viruses were further categorized human herpesvirus, Epstein-Barr virus, xenotropic murine leukemia virus, hepatitis virus, other, and not reported (Additional file 3: Appendix S3). We developed and refined the codebook continually during the first half of the project, with particular attention to patterns within other subordinate categories, such that common factors coded therein would be extracted and receive their own codes. Figure 2 displays the results of the search and screening process. In total, 1161 studies were included (Additional file 4: Appendix S4).

Characteristics of studies
The number of published studies per year increased from one in 1979 to 40 from January to June 2019 (Fig. 3). We extracted all data from the studies' titles and abstracts. Of 1161 studies, 68 were missing abstracts. Only 217 of the studies reported which diagnostic criteria were used in the titles/abstracts; 169 of these used the 1994 CDC/Fukada criteria [18]. Regarding study design, most studies (889) were case-control studies, in which CFS/ME patients were compared to some type of control or comparison group, at one point in time. There were 115 cross-sectional studies without a comparison or control group. A total of 67 studies were prospective or retrospective cohort studies. Fourteen studies fell into the category of "other" study design, which included twin studies, registry studies, and population-based screening studies. In 69 cases, study design was not reported in titles/abstracts.
A total of 679 prospective, retrospective, case-control, or "other" designs included comparison groups, and most used healthy/non-patient comparisons (61 to 66% of studies within each study design). Only 226 studies used other patient groups as comparisons. 181 studies used a comparison group but did not specify which type, and 73 did not report on comparison groups in titles/abstracts. Study size varied considerably: in 117 studies, there were fewer than 30 participants, 440 had between 30-100 participants, 289 contained 101-1000 participants, and 31 contained more than 1000 participants. Cohort and "other" study designs were largest. Finally, only 170 studies stated in titles/abstracts the country in which they were conducted. Nearly half (87) were European, while 56 were conducted in North or South America.
No single causal factor has dominated research in the field (Fig. 4). The number of studies on infection factors varied the most: it was the category with the largest number of studies before 1995 as well as

Discussion
This systematic scoping review included 1161 primary studies on possible causal factors of CFS/ME, and we extracted data from titles and abstracts. Immunological, psychological, infections, and neuroendocrine factors have been most frequently studied. In most categories, there was a steady increase in the number of studies from 1994 and a slightly stronger increase from 2010 to 2019. Many studies were not limited to a single factor or causal hypothesis, but explored numerous factors, from environmental factors to genes. We did not find one dominant causal factor studied, and included studies reported a mean of nearly three different factors, which we interpret as a healthy diversity in the research field. As the number of published articles has increased steadily from 1979-2019, the variation of explored factors has also remained high. This indicates a great deal of uncertainty in the field, but the broad interest in diverse possible causal factors is positive. A 2008 review included 11 studies that examined multiple possible "risk factors or prognostic flags" [20]. Our scoping review included 1161 studies, and although our scope was larger than the 2008 review, there is a clear increase in research in the field. However, our findings echo the conclusions of numerous recent reviews [3,12,15,16,23,27,29], namely, that our confidence in the evidence is too low to draw conclusions about causal factors. This is due to the preponderance of relatively weak study designs. For most questions regarding etiology, we must rely on observational studies, as experimental trials randomly exposing people to hypothesized causal factors are rarely possible to conduct, and even if technically possible, would be unethical. Cross-sectional studies with or without comparisons or controls can provide evidence on correlations and help develop theories or hypotheses of causality, but are ill-suited to verify them. Most of the studies included were designed so that they were not suitable to detect causality, although many claimed this as their purpose. Even though studies might identify a correlation between a factor and a diagnosis, we cannot know if it is a signal of a causal factor, or a consequence of the diagnosis. We found no increase in the number of prospective, longitudinal studies being published over time. Prospective cohort studies following individuals that are exposed versus not exposed will provide stronger evidence on etiology than case control studies and retrospective cohort studies, but unmeasured confounding will inevitably lead to risk of bias, as in all observational studies.
In addition, more than half of the studies (that reported study size) included fewer than 100 participants, and around 60% of all studies compared CFS/ME patients with healthy controls-potentially inappropriate comparisons. If effect sizes are small, such small studies likely lack the statistical power to reveal differences between CFS/ME patients and other groups. However, determining statistical power of primary studies will only be possible with a full systematic review that examines studies in full-text. While it may be difficult to recruit enough patients with CFS/ME for studies, researchers need to prioritize prospective cohort studies with a sufficient number of participants. When enough data is collected to allow for advanced statistical techniques, sufficient comparison of individuals with different risks, exposure, and genetic variables is possible [26,30]; propensity score matching, for example, has been recently utilized in studies of other rare diseases and exposures [2]. An additional research challenge is the unambiguous diagnosis of CFS/ME using standard criteria. A systematic review from 2014 identified 20 different sets of diagnostic criteria [7]. CDC 1994/Fukuda criteria [18] were the most commonly used in that review, as in ours. This review did not find any evidence that some of the diagnostic criteria sets could identify conditions more likely with e.g. a specific organic etiology. In our sample based on titles/ abstracts, only 18% of studies reported which diagnostic criteria had been used. Some countries have registers of CFS/ME patients that are voluntarily organized and use combinations of self-report and symptom-based criteria (e.g. [8], while Norway and others use diagnostic codes (ICD-10, G.93.3) reported in broader nationwide registries [5]. As it is unclear if the type of diagnostic criteria or case definitions used influences the findings of primary studies, the way the patients are diagnosed should be clearly stated and, ideally, standardized in future studies [17].

Strengths and limitations
We searched in eight databases and used a broad and sensitive search strategy. One strength of the study is that a CFS/ ME patient advocacy group helped formulate the research questions and provided substantial input to the data extraction form and the causal factor categorization system. An overall strength of scoping reviews is that they provide a descriptive overview of existing research and knowledge gaps. A clear limitation of this review is that we extracted data from titles/abstracts only, and not from full-texts. While this was a pragmatic decision based on resources available, reading full texts would likely have provided us more information about study size, region, and comparison groups. It is important for readers to note that much of the data we report as "missing" from the title/abstract level may have been reported in full text. As we did not aim to synthesize the results of the studies, we think that the data extracted from the titles/abstracts provides the intended descriptive overview of the categories of causal factors that have been studied, and the study designs that have been used.

Conclusion
We saw a large variety of causal factors explored, although a precipitous decline in research on infections as causal factors in the past five years. It is problematic that research into causal factors of CFS/ME used different study designs, diagnostic criteria, control groups, and methods in general. Our scoping review shows that larger studies with stronger designs are needed. We hope to see more well-designed prospective cohort studies in the future, particularly as post-covid-19 fatigue-and the potential risk for developing CFS/ME after infection with Sars-CoV-2-is explored.