Vitamin and mineral status in chronic fatigue syndrome and fibromyalgia syndrome: A systematic review and meta-analysis

Background Many chronic fatigue syndrome (CFS) and fibromyalgia syndrome (FMS) patients (35–68%) use nutritional supplements, while it is unclear whether deficiencies in vitamins and minerals contribute to symptoms in these patients. Objectives were (1) to determine vitamin and mineral status in CFS and FMS patients as compared to healthy controls; (2) to investigate the association between vitamin and mineral status and clinical parameters, including symptom severity and quality of life; and (3) to determine the effect of supplementation on clinical parameters. Methods The databases PubMed, EMBASE, Web of Knowledge, and PsycINFO were searched for eligible studies. Articles published from January 1st 1994 for CFS patients and 1990 for FMS patients till March 1st 2017 were included. Articles were included if the status of one or more vitamins or minerals were reported, or an intervention concerning vitamins or minerals was performed. Two reviewers independently extracted data and assessed the risk of bias. Results A total of 5 RCTs and 40 observational studies were included in the qualitative synthesis, of which 27 studies were included in the meta-analyses. Circulating concentrations of vitamin E were lower in patients compared to controls (pooled standardized mean difference (SMD): -1.57, 95%CI: -3.09, -0.05; p = .042). However, this difference was not present when restricting the analyses to the subgroup of studies with high quality scores. Poor study quality and a substantial heterogeneity in most studies was found. No vitamins or minerals have been repeatedly or consistently linked to clinical parameters. In addition, RCTs testing supplements containing these vitamins and/or minerals did not result in clinical improvements. Discussion Little evidence was found to support the hypothesis that vitamin and mineral deficiencies play a role in the pathophysiology of CFS and FMS, and that the use of supplements is effective in these patients. Registration Study methods were documented in an international prospective register of systematic reviews (PROSPERO) protocol, registration number: http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42015032528.


Methods
The databases PubMed, EMBASE, Web of Knowledge, and PsycINFO were searched for eligible studies. Articles published from January 1 st 1994 for CFS patients and 1990 for FMS patients till March 1 st 2017 were included. Articles were included if the status of one or more vitamins or minerals were reported, or an intervention concerning vitamins or minerals was performed. Two reviewers independently extracted data and assessed the risk of bias.

Results
A total of 5 RCTs and 40 observational studies were included in the qualitative synthesis, of which 27 studies were included in the meta-analyses. Circulating concentrations of vitamin E were lower in patients compared to controls (pooled standardized mean difference (SMD): -1.57, 95%CI: -3.09, -0.05; p = .042). However, this difference was not present when restricting the analyses to the subgroup of studies with high quality scores. Poor study quality and a substantial heterogeneity in most studies was found. No vitamins or minerals have been repeatedly or consistently linked to clinical parameters. In addition, RCTs testing supplements containing these vitamins and/or minerals did not result in clinical improvements. PLOS

Methods
We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (S1 Table) [19]. Prior to start of article inclusion, we documented study methods in an international prospective register of systematic reviews (PROSPERO) protocol, registration number: CRD42015032528, http://www.crd.york.ac.uk/PROSPERO/display_ record.asp?ID=CRD42015032528.

Data sources and searches
The databases PubMed, EMBASE, Web of Knowledge, and PsycINFO were systematically searched. Articles published between January 1 st 1994 and 1990, for CFS and FMS respectively, and March 1 st 2017 were included. We focused on the most recent diagnostic guidelines, namely the International Center of Disease Control (CDC) diagnostic criteria for CFS that was established in 1994 [1], and the American College of Rheumatology (ACR) criteria for FMS in 1990 [2]. To retrieve relevant articles from PubMed, we formulated a search string (S1 Appendix) that consisted of CFS, FMS, and synonyms, vitamins, minerals, micronutrients and synonyms, while excluding systematic reviews or animal studies. This search string was adapted according to the thesaurus of the databases EMBASE, Web of Knowledge, and PsycINFO. All included studies were screened for potential references that were not included in the first search. Duplicates were removed, as well as studies including pediatric participants. There were no language restrictions; included non-English articles were translated (French, Italian, Polish, and Turkish articles) by native speakers.

Study selection
Title and abstract were screened by two independent reviewers (M.L.J. and I.M.) for the following criteria: (1) CFS or FMS patients; (2) vitamin or mineral status; and (3) study design. Studies which were in agreement with the eligibility criteria were retrieved as full text. Discrepancies between the two researchers were resolved by consensus, and when needed a third assessor was consulted (J.G.M.R.). Reasons for exclusion and percentage of agreement, as Cohen's kappa, between the assessors were documented. Participants of the included studies had to be adults (i.e. !18 years) suffering from CFS or FMS according to the official diagnostic criteria [1][2][3]. Studies that involved patients with a combination of CFS and FMS or other comorbid medical conditions were excluded. Furthermore, the vitamin or mineral status had to be assessed or reported in the article, or there had to be an intervention concerning vitamins or minerals. Patients were compared with healthy controls in observational studies, or vitamin and mineral supplementation were compared with placebo in intervention studies. Lastly, cross-sectional studies comparing cases and controls, cohort studies and randomized controlled trials (RCTs) were included. Case reports, clinical cohorts without appropriate controls (e.g. controls with musculoskeletal pain or fatigue), (systematic) reviews, expert opinion, and other study designs were excluded.

Data extraction
Two reviewers (M.L.J. and I.M.) independently extracted data and assessed the risk of bias for each study. The first ten articles were screened together to pilot the data extraction and risk of bias form. Reasons for exclusion and percentage of agreement between the assessors were documented.
From the included articles, the following information was extracted: name first author, publication year, type FSS, number and age of the participants, and vitamin or mineral status. In addition, data on smoking habits or alcohol use, diet (and assessment tool used), BMI (or waist circumference, waist-hip ratio), physical activity (assessment tool), socioeconomic status, ethnicity, severity of illness (assessment tool), duration of illness, co-morbidities (somatic and psychiatric), medication use, clinical parameters including symptom severity and quality of life, and in case of RCTs the relevant co-intervention(s) were also extracted.

Quality assessment
To assess quality of RCTs, the Cochrane Collaboration's tool for assessing risk of bias was employed [20]. For observational studies, literature indicates lack of a single methodological assessment tool [21,22]. Therefore, we adjusted a previously developed quality tool for observational studies in this field [23], for use in studies that focus specifically on the association between vitamin and mineral status and CFS or FMS. Eight of the nine items in this original quality tool originated from guidelines or tools for either reporting or appraising observational research [24][25][26]. These items were adjusted to the specific question on vitamins and minerals and classified into three key domains: appropriate selection of participants (validated disorder, representative controls, in-and exclusion criteria, disease characteristics), appropriate quantification of vitamin and mineral status (duplicate quantification, appropriate outcome), and appropriate control for confounding (assessed confounders, analyses adjusted). The item: "Is the assessor blind for disease status", was excluded since from the original quality tool since it is not applicable in the current review. Furthermore, we added the item "Are methods for assessment of vitamin and mineral status clearly stated", based on the adapted Newcastle Ottawa scale for cross-sectional studies (S2 Appendix) [27]. RCTs that contained relevant observational data (n = 4/5), were assessed with both the Cochrane tool and the observational studies quality tool. For both quality tools, items were rated as (0) low risk, (1) medium risk, and (2) high risk of bias. The maximum attainable quality score was 14 for RCTs, and 18 for observational studies.

Data synthesis and analysis
We first constructed an overview of available data on the different vitamins and minerals. Characteristics of the included studies were systematically listed to generate a clear overview of the current literature on vitamins and minerals in CFS and FMS patients. For those vitamins and minerals with more than five studies available, we did quantitative syntheses on aggregated data. For these syntheses, data was pooled with the random effects model of metaanalysis, using Stata statistical software, version 14 (Statacorp LP, Texas). To allow pooling across studies that used different outcomes of vitamin or mineral plasma or serum levels, we calculated the standardized mean difference (SMD). For proportions of deficiencies, the odds ratio (OR) was calculated and pooled. Subsequently, the SMD and OR for each study were weighted by their inverse variance and the corresponding 95%CI were calculated. The existence of heterogeneity among studies was assessed by Q-tests, and the degree of the heterogeneity was quantified by calculating the I-squared (I 2 ) value. Publication bias was inspected visually by a funnel plot, and an Egger's test was conducted to quantify funnel plot asymmetry [28]. The Tweedie's Trim and Fill test was performed as an additional sensitivity analysis to identify and correct for funnel plot asymmetry arising from publication bias [29]. When the Trim and Fill test was performed, and additional studies were added to the analyses, contourenhanced funnel plots were used instead of regular funnel plots to examine whether asymmetry in the funnel plots was due to publication bias [30]. Subgroup analyses were performed including studies with more than half of the maximum study quality score (>9 quality points), if more than three studies with a sufficient quality score were available. Furthermore, vitamin and mineral status of CFS and FMS patients were investigated separately if more than three studies were available. Findings were considered statistically significant if P<0.05.

Study inclusion
Results of the systematic review and meta-analysis are presented in a flow diagram (Fig 1). Cohen's kappa's for the abstract and full text selection were 0.96 and 0.89 respectively, indicating very good consistency of agreement [31]. Out of 108 studies included for the full text review, 45 studies were included in the current review.
Characteristics of the included studies are presented in Table 1, and results of the quality assessment in Tables 2 and 3. Most studies involved FMS patients (n = 35/45); 4 of the 5 RCTs also contained relevant observational data. Vitamin and mineral status was mainly assessed in plasma or serum (n = 40/45). Furthermore, quality scores revealed poor study quality (i.e. equal or less than half of the maximum study quality score) in the vast majority of observational studies (n = 27/44; range 4-14 points) and RCTs (n = 3/5; range 5-12 points). Only few observational studies defined all described in-and exclusion criteria for the investigated population, including medication use, somatic morbidity, and psychiatric morbidity (n = 10/44). The CFS or FMS diagnostic criteria were often described in observational studies, but researchers failed to state whether or not the syndromes were diagnosed by a physician (n = 40/44). Disease characteristics were frequently not fully presented (n = 15/44), or were completely absent (n = 18/44) in observational studies. Almost all observational studies did not assess vitamin or mineral in duplicate (n = 38/44). Most studies that assessed vitamin or mineral status did not clearly state the methods for assessment of vitamin and mineral status (n = 27/44). Furthermore, most observational studies did not adjust their analyses for any potential confounders (n = 43/44). Lastly, most RCTs had a medium to high risk of bias for random sequence generation (n = 3/5), allocation concealment (n = 3/5), blinding of outcome assessment (n = 4/5), incomplete data (n = 4/5), selective reporting quantification (n = 3/5), and other bias (n = 5/5).

Systematic review
Studies that were not included in the meta-analyses are presented in Table 4.
Interventions. Five RCTs were included. The first RCT determined the effect of magnesium citrate treatment in combination with amitriptyline versus amitriptyline only, on FMS symptoms, over a period of 8 weeks [34]. They found that amitriptyline and magnesium supplementation was more effective on all measured outcomes than amitriptyline alone. The second RCT investigated the effect of a polynutrient supplement (containing several vitamins (including A, B, C, D, E), minerals (including calcium, magnesium) and (co)enzymes), on fatigue and physical activity of patients with CFS, over a period of 10 weeks [37]. They found no significant difference between the placebo and treatment group on any of the outcome measures. A third RCT examined vitamin C and E treatment combined with exercise versus exercise only, in FMS patients, over a period of 12 weeks [57]. Although both interventions lead to significantly higher vitamin A, C, and E serum levels, the FMS symptoms did not improve in both groups. Furthermore, the most recent RCT investigated the effect of vitamin D, on symptoms in CFS patients, over a period of 6 months [75]. Despite a statistically significant increase in vitamin D, they found no evidence of improvement in symptoms of fatigue or depression. Lastly, in the fifth RCT, cholecalciferol was administered for 20 weeks in FMS patients, with the dosage depending on patients calcifediol levels [74]. A significant treatment effect on intensity of pain was found in the treatment group versus placebo. No changes in somatization, depression and anxiety, physical and mental health, and FMS symptom severity were observed in both the treatment and placebo group.

Meta-analysis
Vitamin C, vitamin D, vitamin D deficiency (<20ng/ml), vitamin E (Fig 2), and the minerals calcium, and magnesium status, and were reported in more than five studies and were therefore investigated using meta-analysis (Fig 3). Meta-analysis revealed that circulating  Vitamins and minerals in chronic fatigue syndrome and fibromyalgia syndrome     Subgroup analyses. Subgroup analyses were performed including studies with more than half the maximum study quality score (>9 quality points), if more than three studies with a sufficient quality score were available. The additional analysis was not possible for magnesium, since only two studies achieved more than half of the maximum quality score. No differences in circulating concentrations of vitamin C (patients n = 93, controls n = 102, pooled SMD:- Subgroup analyses were performed separately for the syndromes, when more than three studies were available per syndrome. Since vitamin D, vitamin D deficiency and calcium were only determined in FMS patients, additional subgroup analyses were possible for vitamin C, vitamin E and magnesium. No statistically significant difference between patients and controls was found in the three studies investigating circulating concentrations of vitamin C in FMS patients (patients n = 101, controls n = 100; pooled SMD:0.14, 95%CI:-0.16,0.44; p = .32). However, the heterogeneity was substantially lower (I 2 = 13.3% versus 88.5% in the overall analysis including CFS patients), indicating a high consistency of studies' results. The significant difference in circulating concentrations of vitamin E between patients and controls disappeared when the single CFS study was excluded (patients n = 141, controls n = 120; pooled SMD:-0.95, 95%CI:-2.41,0.50; p = .20. Lastly, no considerable differences were found in analyses of the five studies investigating circulating concentrations of magnesium in FMS patients (patients n = 203, controls n = 133; pooled SMD:-0.51, 95%CI:-1.34,0.32; p = .23).
Publication bias. Finally, we tested whether publication bias could have affected the results. Corresponding funnel plots can be found in Fig 4. Egger's test showed that there was significant funnel plot asymmetry in vitamin E (p = .039), with no significant asymmetry among the other analyses. Trimming was performed in the calcium studies using the Trim and Vitamins and minerals in chronic fatigue syndrome and fibromyalgia syndrome Fill test, and the contour-enhanced funnel plot revealed two added studies in the statistically significant areas. No studies were trimmed or filled among the vitamin C, vitamin D, vitamin D deficiency, vitamin E, and magnesium studies, indicating absence of substantial publication bias. Vitamins and minerals in chronic fatigue syndrome and fibromyalgia syndrome

Discussion
We found little evidence to support our hypothesis that vitamin and mineral deficiencies play a role in the pathophysiology of both CFS and FMS, or that the use of nutritional supplements is effective in these patients. Poor study quality and considerable heterogeneity in most studies was found, which makes it difficult to reach a final conclusion. Consistent significant lower circulating concentrations were found repeatedly and in the majority of studies for vitamin A and vitamin E in patients compared to controls. However, the significant difference in circulating concentrations of vitamin E between patients and controls disappeared when excluding low quality studies. None of these or other vitamins and minerals have been repeatedly or consistently linked to clinical parameters. In addition, RCTs testing supplements containing these vitamins and/or minerals did not result in clinical improvements.
This review has several strengths. First, this is the first review focusing on vitamin and mineral deficiencies among CFS and FMS patients. We were able to give a clear overview of the current knowledge existing in literature. Second, we included only studies that examined CFS and FMS patients according to the official diagnostic criteria. We therefore have included relatively homogeneous groups of patients. Third, because we defined strict in-and exclusion criteria, e.g. patients should meet the official diagnostic criteria, or clinical cohorts must have an appropriate control group, poor quality studies were filtered out. Nevertheless, the vast majority of the included studies scored a quality score below a reasonable study quality. Fourth, enough studies that investigated similar vitamins or minerals were available, which made it possible to conduct six meta-analyses. Lastly, we had no language restrictions for the included abstracts or full text articles, which enabled us to include all relevant articles.
We must acknowledge that this study also has its limitations, which are mostly due to limitations in original studies on which this review was based. First, most studies were observational in nature. In general, observational studies have a lower validity than RCTs, and they are more susceptible to bias (e.g. selection and information bias) and confounding factors. Potential confounders were assessed in about half of the studies, but almost no studies adjusted their analyses for potential confounders. Consequently, the results of the current review may be affected by the methodological weaknesses that are accompanied by the observational study designs. Second, quality assessment revealed a poor study quality in the majority of studies. This demonstrates that substantial improvements can be made in terms of study quality, especially in specification of in-and exclusion criteria, presenting disease characteristics of the participants, making use of validated methods to assess vitamin and mineral status, to perform the vitamin and mineral assessments in duplicate, and, as mentioned earlier, to adjust analyses for potential confounders. Furthermore, a quality issue in research on CFS and FMS patients is that of careful selection of control groups. Our quality assessment showed that many included studies fell short because of the selection of the controls, which could result in inaccurate study results. Third, a problem that affects the validity of meta-analyses is the presence of publication bias. Funnel plots indicated the absence of publication bias in the majority of the meta-analyses. Trimming was performed among the calcium studies, and two "missing" studies were added, while no significant funnel plot asymmetry was present. However, trimming was performed in the statistically significant areas, which argues against the presence of publication bias. Although Egger's test is preferred for more than 10 studies, it revealed significant funnel plot asymmetry in vitamin E, while no trimming was performed. It is therefore possible that the significant outcomes of vitamin E in patients are influenced by publication bias. Lastly, a substantial to considerable heterogeneity in most studies was found, which makes it difficult to reach a final conclusion about vitamin status in CFS and FMS patients. This review reveals that very few RCTs have investigated the effect of vitamin and mineral supplementation versus placebo in CFS and FMS patients. Most published RCTs found no treatment effect of vitamin and mineral supplementation on clinical parameters. So, the evidence for beneficial effects of supplementation in CFS and FMS patients is not proportional to the large quantity of supplements that are used by these patients. Nevertheless, the industry of vitamin and minerals supplements is increasing, for example, Americans spend an estimated $36.7 billion each year on supplements [77]. This is important information, since the vitamins and minerals in these products are sometimes supplemented in doses high enough to cause side effects, for example gastric discomfort, insomnia, dizziness or weakness [17]. The vast majority of available studies concerned FMS patients. Several FMS studies investigated vitamin D, whereas most CFS studies have focused on vitamin E. Only one CFS study that investigated vitamin E was suitable for inclusion in the meta-analysis. It is remarkable that the significant difference of vitamin E between patients and controls disappeared when the single CFS study was excluded in the sensitivity analysis, while the studies that were not suitable for inclusion in the meta-analysis reported significant lower vitamin E concentrations in particularly CFS patients versus controls. Further research is needed to determine whether this may indicate that vitamin E levels are lower in CFS patients, but not in FMS patients. This systematic review and meta-analysis provides no further insights in whether the remaining vitamins and minerals differ between these two medical conditions.
We conclude that there is little evidence to support the hypothesis that vitamin and mineral deficiencies play a role in the pathophysiology of both CFS and FMS. Furthermore, the current literature on vitamins and minerals in CFS and FMS is of poor quality and stresses the need for well-performed intervention research, and large population-based and age-matched prospective studies in CFS and FMS, in order to gain more insight in the role of vitamins and minerals in the pathophysiology of CFS and FMS. According to our results, potential vitamins and minerals that should be further examined include vitamin A and vitamin E.
Supporting information S1 Table.