An observational, quantitative, cross-sectional study was carried out through a non-random selection of subjects with a diagnosis of FM. The participants were selected and invited in a sequential assignation, through the Rheumatology outpatient clinic from the Yucatán's Peninsula High Specialties Regional Hospital. The inclusion criteria were: female sex, age older than 18 years, having been diagnosed with FM according to the published ACR 2010 classification criteria,11 performed by a certified rheumatologist. Subjects with conditions that could modify body composition resulting in physiological wasting (cancer, chronic infections, poorly absorbed digestive disorders, etc.), besides individuals who withdrew their informed consent and/or could not be interviewed or were unable to be evaluated for anthropometric measurements were excluded.

Study subjects underwent a clinical interview where all relevant socio-demographic and clinical data, including date of birth, marital status, educational status, current occupational status, menopausal status, pharmacological treatment, physical activity, gravidity, and parity were collected.

The evaluation of the clinical FM indicators was carried out using three scores: Widespread Pain index (WPI); Symptoms Severity Scale (SSS); and the Fibromyalgia Impact Questionnaire (FIQ), as established in the National Clinical Practice Guidelines (MEX).12 In the case of WPI, we used the version that comprises 19 anatomical areas (shoulders, arms, forearms, jaws, neck, buttocks, legs, calves, upper back, lower back, thorax, and abdomen) as a continuous scale count of painful regions.13 A complete SSS questionnaire was applied to participants, scoring the levels of severity and other somatic symptoms (sections A and B, respectively). The result was the sum of the elements as reported by the modified ACR 2010 criteria, and referenced from an updated report.14 To assess the FM-related symptoms on the physical and mental health of patients we used the Spanish validated and updated version of the FIQ,15 where the total integrates the sum of all subscales, with higher scores indicating a negative impact (0–100).

The body composition assessment included weight, height, body fat percentage (FM%), lean mass percentage (MM%), visceral fat, waist, arm, and hip girths, and skinfold thickness (biceps and triceps). Also, other five calculated scores were recorded: BMI, waist/hip ratio (W/H), Relative Fat Mass (RFM),16 waist-to-height ratio (W/Ht), and Muscle Mass estimation (MMest).17 The measurements were taken according to the Official Mexican Norm by a single researcher (LOC), who is a qualified expert for Clinical Nutrition, trained in anthropometry procedures. All anthropometric measurements were obtained the same day as the clinical interview, using a portable device body composition monitor and scale (OMRON Healthcare Co. LTD. Japan) without shoes and light clothes (approximated to the nearest 0.1kg). Height was measured using a Harpenden stadiometer (Holtain 602VR®) to the nearest 0.5cm, with participants again not wearing shoes. Determination of skinfold thickness was performed to the nearest 0.1mm at the triceps and on the right side using a Harpenden standard Slim Guide skinfold caliper (TAQ Sistemas Médicos S.A. de C.V, Mexico). All skinfold measurements were performed in triplicate, and the average of the three readings was annotated. They were carried out on the right side of the body on all subjects assuming a relaxed standing position with the arms hanging by the side and according to the validated methods.18

A dimension reduction was carried out with a PCA procedure, obtaining scores (coefficients) as a new variable for each participant that was used to perform comparisons and correlations. Those scores integrated parameters for fat and lean mass content, and FM clinical indicators (WPI, SSS, and FIQ). A correlation matrix was constructed to assess the correlation between anthropometric measures. The Kaiser–Meyer–Olkin test for sampling adequacy (≥ 0.6) and Bartlett's test of sphericity (P value<.05) were verified to validate whether the PCA assumptions were fulfilled. Varimax rotation was applied to obtain orthogonal factors. Fat and lean/muscle mass groups that showed factor loadings greater than 0.3 were considered to have strong associations with that factor. The number of factors that best represent the data was based on the screen plot and eigenvalues which reached a value above 1.5. To establish the internal consistency and the correlation between the variables and loaded components, we ran a Cronbach's alpha, which displayed values>0.7, displaying an acceptable level of internal consistency. The variables comprised for each PCA group (PCA-Fat, PCA-Muscle), are found in Table 1.

Every variable was tested for data distribution, and normality was assessed with Shapiro–Wilk's test. Clinical indicators were categorized as “Low” or “High” groups for each of the FM clinical indicator (WPI, SSS, and FIQ) depending on individual scores, unique sequential ranks, and tied observations. The descriptive data for these categorical variables are specified in Table 2. Additionally, we examined three Binomial Logistic Regression (BLR) models, using those categorical variables (Low and High groups) as dependent variables for each model. The BLRs were accomplished by using the method “enter,” with body composition parameters, age, and other covariates like comorbidities or pharmaceutical treatment as predictor variables. Significant models were tested for multicollinearity and Interactions between included covariates which was based on the rule of ≥10 events per variable and were those of interest to body composition (PCA-Fat, PCA-muscle), and age since it seems to play a role in FM. For all hypothesis tests, P value was considered as significant if it was <.05. The analysis was carried out using IBM® SPSS® Statistics, v. 24 for Windows.

The research protocol was approved by the Research and Bioethics Committees of the participating unit. All participants signed the informed consent letter before entering the study. Subjects in whom alterations in body composition were detected received nutritional counseling and were referred to nutritional care or mental health according to their condition.

Forty-three patients, all women, who were 52.7±11.3 years old (limits: 21–75), had FM diagnosis since 53.6±54 months (limits: 0–192), besides 89±75 months (limits: 4–264) FM duration were included. Most of the patients were taking some form of medication such as NSAIDs (n=30, 65%), opioids (n=13, 28%), and antidepressants or benzodiazepines (n=27, 62% combined). Also, there was a high prevalence of overweight/obesity (90.7%), and most of them had some kind of comorbidity (62.7%). Complete descriptive data can be found in Table 2.

Firstly, the relationship between clinical indicators (WPI, FIQ, SSS) and body composition was determined using Pearson (WPI, FIQ) and Spearman correlation (SSS) according to the data distribution. The scores for fat and muscle were composed of the PCA components that integrated measurements for anthropometry and fat or muscle determinations. The results showed a statistically significant positive correlation between PCA-Fat and two of the clinical parameters: WPI (r=.326, P=.043) and FIQ (r=.325, P=.044) (Fig. 1A, B respectively), but no significance was achieved for SSS. On the contrary, for PCA-muscle the significance was demonstrated only for SSS, with a negative coefficient correlation (r=−.384, P=.013) (Fig. 1C).

Fig. 1.

Correlation analyses between PCA-Fat and WPI scores, FIQ scores, and PCA-Muscle and SSS scores. (A) Preliminary analysis showed the relationship to be linear with both variables normally distributed, as assessed by Shapiro–Wilk's test (P>.05). Bivariate Pearson's correlation established a statistically significant, moderate positive correlation between the scores tested, r=.326, P<.05. (B) The relationship was linear with both variables normally distributed, as assessed by Shapiro–Wilk's test (P>.05). There was a statistically significant, moderate positive correlation between the scores tested, r=.325, P<.05. (C) A Spearman's rank-order correlation was run to assess the relationship between PCA-Muscle and SSS scores. Preliminary analysis showed he relationship to be monotonic, as assessed by visual inspection of a scatterplot. There was a statistically significant, moderate negative correlation between those two variables, rs=−.384, P<.05.

(0.17MB).

We performed t test for independent samples, using the operational groups created from the sequential ranks, and each numerical variable. These groups are namely “Low” and “High” for WPI, SSS, and FIQ. Independent samples t test confirmed significant differences as follows: higher values of triceps sf (P=.038) and PCA-Fat score (P=.009) for High-WPI group (Fig. 2A, B respectively). As for SSS groups, individual significant differences between High-SSS and Low-SSS were observed, with higher values of BMI (P=.007), waist girth (P=.029), RFM (P=.017), W/Ht (P=.001) (Fig. 3A–D), and PCA-muscle scores (P=.001) for the Low-SSS group (Fig. 3E), but no differences were found for the global PCA-Fat score. It is noteworthy to mention that there was also a significant difference in age between High and Low-SSS, being higher for the latter (P=.040) (Fig. 3F). For the groups of High-FIQ and Low-FIQ there were no significant differences.

Fig. 2.

Low and High-WPI group comparisons. Independent-samples t-tests were run to determine if there were differences in body composition parameters between Low-WPI and High-WPI groups. Tricipital skinfold thickness (sf) (A) and PCA-Fat scores (B), were significantly higher in High-WPI group (P=.038 and P=.009, respectively). Data are mean±standard error of the mean.

(0.09MB).

Fig. 3.

Low and High-SSS group comparisons. Independent-samples t-tests were run to determine if there were differences in body composition parameters between Low-SSS and High-SSS groups. Significantly higher values for (A) Body Mass Index (P=.007), (B) Waist girth (P=.029), (C) Relative Fat Mass (P=.017), (D) Waist-to-Height ratio (P=.001), (E) PCA-Fat score (P=.001) were observed in Low-SSS group. (Data are mean±standard error of the mean). (F) The graph displays the distribution of age in both groups, and a significant difference (P=.040) between groups was observed.

(0.28MB).

Two models in which the dependent variables were established as Low/High correspondingly for WPI and SSS tested statistically significant when PCA-muscle, PCA-Fat, and age were entered as covariates (Table 3). These models were statistically significant as follows: model Low/High WPI, X2(3)=7.854, P=.049; model Low/High SSS, X2(3)=16.575, P=.001, explained 24.5%, and 46.2%, of the variance (based on Nagelkerke R2) and correctly classified 66.7%, and 76.9%, of cases for models WPI and SSS, respectively. No multicollinearity was detected in any of the models using variance inflation factor (VIF) and tolerance. This is, none of the variables included in the models showed a VIF less than 3, all tolerance values were higher than 0.2, the condition indexes were smaller than 15, and there were not two or more variables with an eigenvalue greater than 0.90. The two models showed that at least one predictor was significant for the corresponding outcome. In the case of Low/High WPI as the dependent variable, the PCA-Fat score was a significant predictor (OR=2.477, 95% CI 1.052–5.832), in which an increasing PCA-Fat score is associated with an increased likelihood of exhibit high WPI. For the model with Low/High SSS as the dependent outcome, the PCA-Muscle score was the only significant predictor (OR=0.303, 95% CI 0.123–0.745), but in this case, this covariate showed to be a protective factor for High-SSS. The areas under the ROC curve (Fig. 4) were 0.735 (95% CI, 0.580–0.891, P=.013), and 0.850 (95% CI, 0.731–0.969, P<.001), respectively for Low/High-WPI and Low/High-SSS, respectively.

Fig. 4.

ROC curves and areas under the curve (AUC). AUC for A) Low/High-WPI was 0.735 (95% CI, 0.580–0.891, P=.013), and AUC for B) Low/High- SSS was 0.850 (95% CI, 0.731–0.969, P<.001).

(0.24MB).

The comparisons involving all groups of medical treatment, displayed significant differences when the factor was the use of cyclobenzaprine with SSS higher scores for those who registered a continuous use of this medication (P=.025) (Fig. 5A), however that group was composed of only 6 patients. In the case of FIQ, we found significantly higher scores for the group that reported to have used GABA agonists (gabapentin and pregabalin groups, P=.022), and more specifically for pregabalin compared to the group that did not report any of this medication (P=.026) (Fig. 5B).

Fig. 5.

Pharmacological treatment group comparisons. (A) A Mann–Whitney U test was run to determine differences in SSS score between cyclobenzaprine use (yes/no). SSS score was statistically significantly higher in the group under this drug than in those who denied it, U=47.5, z=−2.243, P=.025, using an exact sampling distribution for U. (B) One-way ANOVA with subsequent post hoc tests (Tukey HSD), among different groups under GABAergic medication showed higher scores of FIQ in the patients under pregabalin treatment (P=.026).

(0.17MB).