This longitudinal study is a secondary analysis of data from a randomized controlled trial (RCT) on patients with fibromyalgia.29 In that trial, data were collected at baseline and in three post-treatment periods: eight-week and six- and 12-month follow-ups (Fig. 1). The RCT was previously approved by the Research Ethics Committee of Universidade Cidade de São Paulo, Brazil (CAAE: 51328215.1.0000.0064) and was registered in Clinical Trials Database (number: NCT03050606). All patients signed a consent form to participate in the study. Details of the RCT protocol were previously published.29

Figure 1.

Flow diagram of the study.

(0.14MB).

The aim of the RCT was to evaluate the clinical and cost-effectiveness of Pilates compared to aerobic exercises in the treatment of patients with fibromyalgia. In this study, which included 97 randomly assigned patients with fibromyalgia, the main results showed that, except for the EQ-5D at 12-month follow-up, there was no difference between Pilates and aerobic exercises for the impact of fibromyalgia and health-related quality of life (HRQoL) at eight-week and six- and 12-month follow-ups. Based on these results, we decided to combine both groups in a unique group for the analysis in this study.

The study was conducted at the Center of Excellence in Clinical Research in Physical Therapy at Universidade Cidade de São Paulo, Sao Paulo, Brazil.

The study included 97 patients of both sexes, aged 20–75 years, who met the 2010 fibromyalgia classification criteria of the American College of Rheumatology,30 and with pain intensity of three or more points on the Pain Numerical Rating Scale (PNRS).31 Exclusion criteria were any contraindication to physical activity, pregnancy, uncontrolled systemic disease, any neurological, cardiorespiratory, and musculoskeletal condition that could interfere with treatment, advanced joint disease, severe lower limb joint injury or pain, partial or total muscle tear, amputations, recent surgery, and inability to speak or read Brazilian-Portuguese.

Patients were assessed by a blinded assessor for eligibility and demographic and anthropometric data. The following clinical outcomes were measured at baseline and at the eight-week and six- and 12-month follow-ups: HRQoL (utility indexes), pain intensity, impact of fibromyalgia, kinesiophobia, sleep quality, and depression. Baseline and eight-week follow-up were face to face, while the six- and 12-month follow-ups were conducted via telephone by two blinded assessors, trained to administer the questionnaires used in the study. All questionnaires used in this study were translated and validated to Brazilian-Portuguese.9,31–40

The evaluation instruments used in this study included the: SF-6D, EQ-5D, EQ-visual analog scale (EQ-VAS), Fibromyalgia Impact Questionnaire (FIQ), PNRS, Tampa Scale for Kinesiophobia (TSK-11), Pittsburgh Sleep Quality Index (PSQI), and Beck Depression Inventory I (BDI-I).

The SF-6D measures HRQoL with six domains: physical functioning, role limitation, mental health, social functioning, pain, and vitality.9,33,34 The score ranges from 0 (worst health state) to 1 (best health state).9,33,34 The EQ-5D measures HRQoL with five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each dimension has three levels of severity: no problems, some problems, and extreme problems.39,41 The score ranges from 0 (worst health state) to 1 (best health state).39,41 The EQ-5D questionnaire contains a 20-cm VAS, which is not included in the calculation of the final EQ-5D score and can be independently assessed. The EQ-VAS is a direct measure of the patient's health state and the score ranges from 0 (worst health state) to 100 (best health state).41 A numerical scale from 0 to 100 points was used instead of the original VAS to allow for follow-ups via telephone, based on the good correlation between the VAS and the Numerical Rating Scale (NRS).31 The FIQ assesses the impact of fibromyalgia on QoL with 10 questions. Question 1 contains a 10-item subscale related to physical function. In items 2 and 3, the participant marks the number of days they felt well and the number of days they were unable to work due to fibromyalgia in the last seven days. Items 4–10 are numerical scales on which the patient rates work difficulty, pain, fatigue, morning tiredness, stiffness, anxiety, and depression. A NRS from 0 to 10 was used instead of the original VAS to allow for follow-ups via telephone, based again on the good correlation between VAS and NRS.31 The FIQ score ranges from 0 to 100, with higher values indicating greater impact of fibromyalgia on QoL.38

The PNRS is an 11-point scale (0–10) that assesses the patient's mean pain intensity over the past seven days, where 0 represents “no pain” and 10 represents “pain as bad as it could be.”31 The TSK is composed of 17 questions addressing kinesiophobia. The score of each question ranges from 1 to 4. The total score varies between 17 and 68, with higher scores representing a higher degree of kinesiophobia.35,40 The PSQI consists of 19 questions distributed into seven components (subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleeping medication, and daytime dysfunction), which are rated on a score ranging from 0 to 3. The sum of the scores from the seven components ranges from 0 to 21, and higher scores indicate worst sleep quality.32 The BDI-I is a 21-item self-report rating inventory, with answers generating a score that varies from 0 to 3. The individual item scores are added up to calculate the final score. Scores below 10 correspond to no depression, scores between 10 and 16 correspond to mild to moderate depression, scores from 17 to 29 correspond to moderate to severe depression, and scores above 30 correspond to severe depression.36,37,42

Both groups received an educational intervention in the first session.29 Additionally, patients randomized to the Pilates group received Pilates-based exercises including exercises on a mat with accessories, and on apparatus, with activation of the core muscles while exhaling. Patients randomized to the aerobic exercise group received aerobic exercises on a stationary bike or treadmill, with training heart rate of 57 to 76% of the maximum heart rate, corresponding to mild to moderate intensity of exercise.29 Both groups received 16 one-hour individualized exercise sessions over eight weeks.29 Due to the nature of the intervention, patients and therapists were not blinded to group allocation.

All statistical analyses were conducted with SPSS software version 19 for Windows (SPSS Inc., Chicago, Illinois). Baseline characteristics were presented descriptively. The COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) initiative considers a sample size of at least 50 patients as very good for the analysis of construct validity and responsiveness. Thus, the sample size of this study (n = 97 patients) may be considered relevant for both analyzes.43 Construct validity was analyzed based on the correlation between the impact of fibromyalgia and the utility indexes instruments SF-6D, EQ-5D, and EQ-VAS at baseline using Spearman’s correlation coefficient (r) due to the ordinal nature of the instruments’ scores. The hypotheses established for this measurement property are described in the first column of Table 1. Construct validity was considered high when less than 25% of the hypotheses are rejected, moderate when 25 to 50% of the hypotheses are rejected, and poor when more than 50% of the hypotheses are rejected.44

Responsiveness was assessed based on the correlation between the change scores of the impact of fibromyalgia and each utility index instrument at baseline and in the eight-week and six- and 12-month follow-ups, also using Spearman’s correlation coefficient (r) due to the ordinal nature of the instruments’ scores. Responsiveness was also analyzed using the hypothesis and constructing a receiver operating characteristic (ROC) curve using the FIQ as the change criterion. For this last analysis, the FIQ scores were converted into categorical outcomes considering the minimum clinically relevant improvement of 14%.45 Patients who achieved 14% or greater improvement in the FIQ at eight-week and six and 12-month follow-ups were categorized as “improved” and patients who did not achieve 14% improvement were categorized as “not improved.” The hypothesis test established to assess responsiveness is described in the first column of the Table 1. Responsiveness was considered high when less than 25% of hypotheses are rejected, moderate when 25 to 50% of hypotheses are rejected, and poor when more than 50% of hypotheses are rejected.46 Additionally, for a scale to be considered responsive, the area under the curve (AUC) must be ≥0.70.43 To classify the magnitude of the correlations for construct validity and responsiveness, r<0.30 indicate weak correlation, r ≥ 0.30 and <0.60 indicate moderate correlation, and r ≥ 0.60 indicate good correlation.47

Linear regression models were also performed to explore clinical and sociodemographic factors associated with the utility indexes instruments in the Brazilian population with fibromyalgia at baseline. Thus, for each of the three dependent continuous variables (SF-6D, EQ-5D, and EQ-VAS), the following independent variables at baseline were considered: educational level expressed as low and high level of education (categorical variable), age expressed in years, income expressed in number of minimum wages, symptom duration expressed in months, and pain intensity, impact of fibromyalgia, kinesiophobia, sleep quality, and depression (continuous variables). All variables were forced in the multiple linear regression model using the backward method, and the variables with p < 0.05 remained in the model (entry criteria of p ≤ 0.05 and removal criteria of p ≥ 0.10). The assumptions of normality, multicollinearity, and linearity were not violated in all models and were considered present in the occurrence of tolerance <0.10. A test of normal distribution of the linear regression models was conducted by plotting both residuals and normal distribution.

The main objectives of this study were to determine whether the different instruments used to assess health state utility indexes are comparable, and which clinical variables are associated with the index of each instrument. In the construct validity analysis, the correlation between the EQ-5D and the SF-6D was moderate. The EQ-VAS showed a moderate correlation with the SF-6D and was not correlated with the EQ-5D. The correlations between the SF-6D, the EQ-5D, and the EQ-VAS with the FIQ were moderate. In the responsiveness analysis, the EQ-5D was more responsive in the assessment of clinical change at eight-week and six-month follow-up, while the SF-6D was responsive only in the eight-week follow-up. The EQ-VAS was not responsive in any of the follow-ups, and no instrument was responsive in the 12-month follow-up. In the multiple regression analysis, the impact of fibromyalgia was associated with the EQ-5D and the SF-6D health state utility indexes. In contrast, symptom duration and depression were associated with the utility indexes measured by the EQ-VAS. Although the beta coefficients were statistically significant, the effect sizes seem to be small and not clinically important.

The most common health state utility indexes instruments used to calculate QALY are the EQ-5D and SF-6D.21 The health state utility indexes instruments should evaluate the same construct. However, some studies show that these questionnaires present different results at the same willingness-to-pay threshold per QALY.14,15 Thus, these instruments seem to measure different aspects of HRQoL,10 which explains the differences found in the measurement properties in the present study.

A study13 tested the Norwegian versions of the SF-6D and EQ-5D in patients with chronic low back pain and the results for construct validity were similar to those of our study. The SF-6D and EQ-5D presented a moderate correlation with the Oswestry disability index (r = 0.38 and r = 0.58, respectively). However, the authors did not investigate the correlation between the two health state utility indexes instruments. Unlike the present study, the responsiveness values for these instruments were higher for patients with chronic low back pain (in the AUC, the possibility of correctly discriminating a “responder” or “non-responder” patient was 90% for the SF-6D and 83% for the EQ-5D). In addition, the same study13 also found a low agreement between the SF-6D and EQ-5D.

Another study10 compared the health state utility indexes instruments in participants with different clinical conditions and different levels of QoL. The results showed a low agreement between the health state utility indexes instruments when analyzing the intraclass correlation coefficient, which ranged from 0.28 to 0.55. Furthermore, the study10 also showed that the SF-6D is more indicated in the assessment of mild health conditions, while the EQ-5D is more indicated for severe health impairments. In other words, these health state utility indexes instruments measure different aspects and depend on the severity of the clinical condition evaluated. Thus, we believe that the EQ-5D would be the most appropriate instrument to assess patients with fibromyalgia, due to the high impact on patients' QoL caused by this health condition.48

The comparison of the measurement properties of the EQ-VAS was not yet performed. EQ-VAS is a direct method of health state utility evaluation and this may explain the poor correlation with the SF-6D and EQ-5D, which are indirect methods. The use of a direct method based on the EQ-VAS seems to be limited due to its weak economic basis, as it does not involve an individual choice or decision regarding uncertainty.49,50 In addition, this method also has a scale bias because patients are averse to answer health states near to the extreme ends of a scale.51 Regarding the identification of clinical and demographic variables that may be associated with the health state utility indexes instruments, no studies with this approach were found for any health condition.

As the EQ-5D and SF-6D generate different scores under various clinical conditions,50 these two instruments should not be used interchangeably to measure treatment preference, as they tend to favor different interventions, which may confuse the decision-maker, patient, and clinician.13,15 Ideally, data related to economic evaluation should be grouped according to the utility index instrument, rather than evaluating QALY independently of the instrument used, as usually occurs in some systematic reviews of economic evaluation. The present study showed that the EQ-5D and the SF-6D have a moderate correlation with the FIQ, however the EQ-5D showed short- and medium-term (eight-week and six-month follow-ups, respectively) responsiveness for patients with fibromyalgia, while the SF-6D only showed short-term (eight-week follow-up) responsiveness. These findings may favor the use of the EQ-5D in Brazilian patients with fibromyalgia. Future studies should test the measurement properties of the Brazilian versions of these utility indexes instruments under different health conditions to determine which instrument is the most appropriate.

The Brazilian-Portuguese versions of the utility instruments had not been compared previously, and all studies that performed these comparisons had been conducted in high-income countries.13–15 Thus, our results show the reality of another cultural and socioeconomic context, which is especially important in these evaluations. In addition, this study analyzed the measurement properties according to the COSMIN recommendations.9,33,34,39,43 This evaluation was performed with a relevant sample size and classifications for each measurement property evaluated.43 The present study also included regression analyses, which may help to better understand the components most involved with the constructs of each instrument.

A potential limitation of this study is that some questions of the FIQ and the EQ-VAS were evaluated numerically in all assessments. However, this form of assessment did not seem to affect the responsiveness outcome, given that in the eight-week face-to-face assessment, the EQ-VAS no longer showed good responsiveness either by correlation with other health state utility indexes instruments nor by ROC curve analysis. Another limitation is that the final score of EQ-5D, SF-6D, and FIQ is calculated based on ordinal data. In a final score derived from ordinal data, the distances between the categories of the instruments are separated by unknown quantities of the measured variable and can vary across the instrument.52,53 Thus, the final score of EQ-5D, SF-6D, and FIQ may not be appropriate to calculate change scores, and the results of responsiveness analyses should be interpreted with caution, although previous studies also assessed the responsiveness of the EQ-5D and SF-6D using correlation between change scores of the instruments.13,15