Fibromyalgia 2016 criteria and assessments: comprehensive validation in a Norwegian population

1 Introduction

The diagnosis of fibromyalgia (FM) is based on criteria set by the American College of Rheumatology (ACR) [1]. In 2010 new provisional ACR criteria for diagnosing FM and measuring symptom severity were published [2], as a result of a series of objections to the former criteria [1]. The 1990 criteria defined FM based on chronic widespread pain (CWP) and tenderness in 11 out of 18 predefined tender points (TP) [1], but were later criticized in both clinical practice [3] and research [2], [4].

Many physicians did not know how to examine for tender points or refused to do it, and consequently, a correct evaluation of tender points was rarely performed [5], [6]. In addition, tender points are often interpreted as muscle (peripheral) pathology, while contemporary research suggest that changes in central nervous mechanisms are the likely pathogenic processes [7]. Another limitation of the ACR 1990 criteria is that they do not consider symptoms beyond pain, even though it is clear that the fibromyalgia phenotype often expresses other symptoms like fatigue, sleep disorders, cognitive dysfunction, and somatic symptoms [2], recognized previously [8], [9]. An additional problem is that the ACR 1990 criteria do not grade the severity of FM symptoms or monitor quantitative changes in FM outcomes since they were simply dichotomous, i.e. “whether you have fibromyalgia or not”. Thus, researchers have criticized the ACR1990 criteria [7], [10], [11], [12], [13] and raised questions as whether «fibromyalgia falls foul of a fallacy» [14]. The shift in the conceptualization of fibromyalgia that occurred in the clinic and in research studies, however, provided no clear new case definition. The new ACR 2010 FM diagnostic criteria (FDC) shifted the fibromyalgia definition somewhat toward other important symptoms, not only pain. Thus, the 2010 criteria were not aiming for perfect agreement with 1990 criteria because the FM definition was changed to some extent [2].

The ACR 2010 FM diagnostic criteria were based on a Widespread Pain Index (WPI) (range 0–19) and a Symptom Severity Scale (SSS) score (range 0–12). One of the items of the SSS required an assessment from a physician, and in 2011 the criteria were modified so that all items could be obtained by patient self-administration using the Fibromyalgia Survey Questionnaire (FSQ) for research [15]. The 2011 Fibromyalgia survey diagnostic criteria (FSDC) were satisfied if the following three conditions are met: (1) the WPI ≥7 and the SSS score ≥5, or WPI is 3–6 plus SSS score ≥9; (2) symptoms have been present at a similar level for at least 3 months; and (3) the patient does not have a disorder that would otherwise explain the pain [15].

Based on research published in 2010–2016 the FSDC were updated in 2016. These criteria were evaluated to adequately serve as diagnostic criteria when used in the clinic, but also as classification criteria when used for research [4]. The 2016 criteria are satisfied if the following three conditions are met: (1) the WPI ≥7 and the SSS score ≥5, or WPI is 4–6 with SSS score ≥9; (2) generalized pain, defined as pain in at least four of five regions, is present; and (3) symptoms have been present at a similar level for at least 3 months. The previous requirement that the patient could not have other conditions that could explain the pain was removed (4). The sum of the WPI and the SSS is termed the Fibromyalgia Severity (FS) score (range 0–31), but also the term Fibromyalgianess score has been used [2], [4], [15], [16]. The FSQ has been translated into some languages, including German [17], Spanish [18], Persian [19], Japanese [20], Turkish [21], French [22] and Korean [23], and validated in these populations. The aim of this project is to validate the Norwegian version of FSQ according to the 2011 and 2016 FSDC against the ACR 1990 FM criteria.

2 Methods

3 Statistics

The data management and the analysis were conducted with the SPSS version 25 (IBMSPSS, Chicago, IL, USA). Descriptive analyses of demographic and clinical characteristics between persons with fibromyalgia and controls were performed with the chi-square statistic for categorical variables and with nonparametric Mann–Whitney U test for continuous variables since the distribution was not normal. p-Values below 0.01 were considered statistically significant.

The evaluation of the FSQ included internal consistency reliability for the FS score (sum-score of the entire scale) and separately for SSS and WPI by the coefficient alpha (Cronbach). Test retest intra-reliability analyses of participants answering the FS 1 week apart were performed using intra-class correlation (ICC) (Model Mixed and Type Absolute).

Construct validity included analyses of convergent and discriminant (divergent) validity. To assess convergent validity the Spearman correlation (rho) between the self-report FSQ, tender points (TPs) counting and FIQ was performed. A correlation was strong if r was higher than 0.6, moderate if r was between 0.3 and 0.6 and low if r was below 0.3. To assess discriminant validity the Mann-Whitney U test was used to compare the FS score and subscales of FSQ among those with and without a FM diagnosis.

Receiver Operating Characteristics (ROC) analysis was performed using FS score, and the Area Under the Curve (AUC) was calculated using the ACR1990 as the reference. The sensitivity, the specificity, the positive and negative likelihood ratios (LR+ and LR−) and the accuracy were used to calculate the best cut-off points for the FS score. In addition, the sensitivity, specificity, LR+, LR− and accuracy were established for the 2011 and 2016 FSDC using the ACR1990 criteria as reference. Accuracy is defined as the proportion of the study group correctly classified as positive or negative.

4 Results

One-hundred and twenty-four persons with FM and sixty-four without FM, aged 18–70 years, were invited to the study. Of these, four FM-patients and two controls were excluded due to age >70 and missing data. This left 120 FM patients (119 women) and 62 controls (59 women) to be enrolled in the study, mean age 53.1 (SD 10.9) and 48.4 (SD 13) years, respectively. The socio-demographic and clinical characteristics are shown in Table 1. Age, education and working status revealed statistically significant differences between FM and the control group (p<0.01). For the clinical characteristics, significant differences were found for all variables (p<0.001), except depression and anxiety.

To evaluate the internal consistency of the FSQ, the Cronbach’s alfas were measured for FS, WPI and SSS, which revealed 0.904, 0.888 and 0.793, respectively. The Cronbach’s alfas for the FM-patients and the controls are shown in Table 2. Test-retest reliability was run on a subsample of the FM group (n=27) performing intraclass correlations (ICC) for FS, WPI, SSS, which showed 0.862, 0.835 and 0.867, respectively. The ICC for the 25 single FS items are shown in Table 3.

Spearman’s correlations were calculated in order to test the construct, convergent validity. FSQ (including FS, WPI and SSS) correlated significantly with FIQ total and all its subscales (FIQ function, FIQ overall, FIQ symptoms and the FIQ single items, except item two items [anxiety and depression]). The FS, WPI and SSS correlation coefficients with FIQ total were 0.74, 0.59, 0.85, respectively. TPC also correlated significantly with FIQ total and the major subscales. Results are presented in Table 4.

The percentages of those who met the FSDC 2011 and 2016 criteria in the FM group were significantly greater than in the non-FM group (89.9 vs. 17.7 and 83.2 vs. 14.5, respectively, p<0.01). The sensitivity, specificity, positive and negative likelihood ratio (LR+, LR−) of both the 2011 and 2016 FSDC with ACR1990 FM diagnosis as the reference were good, comparing FM-patients and controls. The accuracy rate for both criteria was 84%. A receiver operator characteristic (ROC) analysis showed that the optimal cutoff-point for the FS score was 12.5 revealing a 93.9% sensitivity and 68.4% specificity. The AUC was very good (0.860). Results are presented in Table 5.

5 Discussion

We found the Norwegian version of the fibromyalgia survey questionnaire (FSQ-N) to be a valid instrument to diagnose fibromyalgia and applicable in surveys for chronic widespread pain. The 2011/2016 FSDC allow the use of merely self-reporting, while the ACR2010 also needs a physician confirmation on one item. The intention behind the original ACR2010 and its 2011 revision was to develop simpler tools for clinical diagnosis and research in fibromyalgia without performing tender point examinations, as well as adding a graded severity scale for extra fibromyalgia symptoms [15]. The 2016 FSDC revision highlighted and re-introduced the widespread pain criterion plus the not-exclusive criterion, which states that a “diagnosis of fibromyalgia is valid irrespective of other diagnoses” [4]. It also intended to merge and optimize the clinical diagnostic and research perspectives from the 2010/2011 criteria. Thus, we wanted to examine and validate the Fibromyalgia Survey Questionnaire (FSQ) and its FSDCs for both the 2011 and 2016 versions, in addition to examine the optimal cut-off point of the FS score for the fibromyalgia-diagnosis in Norwegian.

In this study we found the fibromyalgia patients to be a little older than the controls, they had lower education levels and often were working less. The clinical examinations acknowledged differences in tender points counting (TPC), pain thresholds for mechanical pressure (KPa) and fibromyalgia impact (FIQ), as well as WPI, SSS and FS (from the FSQ), indicating more pain and co-morbid symptoms in fibromyalgia than controls, i.e. the same significant contrasts as shown in, e.g. the Persian, Korean, and Spanish validation studies [19], [23], [32]. Our findings indicate that FSQ-N can differentiate between persons with fibromyalgia as defined by the ACR 1990, 2010, 2016 criteria and controls and how to distinguish between FM and non-FM.

The Cronbach’s alfa of FS, WPI and SSS, were good (0.904, 0.888 and 0.793, respectively), as well as the test-retest (intra-rater) reliability analysis of the 25 FSQ-N single items, disclosing an acceptable internal consistency of the FSQ-N in line with other validation studies from France, Germany, Iran, Korea and Turkey [22] [17], [19], [21], [23]. The test retest reliability showed an intraclass correlation coefficient (ICC) range from 0.435 to 0.879, i.e. comparable numbers with e.g. Turkish (0.383–0.818) [21] and French (0.600–0.891) [22] numbers for the same items. This supports the assumption that FSQ-N is a reliable questionnaire.

We also performed correlation analyzes to test the construct, convergent validity of FSQ-N (including its FS, WPI and SSS subscales), versus the well-known and valid Fibromyalgia Impact Questionnaire (FIQ) [28]. For these analyses we used Spearman’s correlation as a conservative approach and because some subcategories of FIQ were not normally distributed. FS score correlated significantly and strongly with FIQ total, FIQ symptoms and most FIQ single items (>0.70), except anxiety and depression, while FIQ function and FIQ overall correlated moderately with the FS score. Our FS, WPI and SSS correlation coefficients with FIQ total were 0.74, 0.59 and 0.85, respectively. To compare, the Persian had 0.45, 0.33, 0.45 and the Spanish 0.65, 0.50, 066, respectively [19], [32], while the Korean publicized very good WPI and SSS vs FIQ correlations (0.815, 0.854), and the French a single, modest FS versus FIQ correlation (=0.67) [22], [23]. The German validation study only ran FS with anxiety and depression correlations [17]. As in the comparable Persian study [19], our FS and SSS scores correlated much better with the FIQ total and symptom scores than with FIQ function, probably because of overlapping sleep, fatigue and pain and depression symptoms, but no function scores.

Our TPC measures correlated somewhat weakly with FS, WPI and SSS compared with the respective Spanish (0.63, 0.55, 0.61 vs. 0.71, 0.69, 0.65) and Korean studies (0.71, 0.69, 0.65) [23], [32] but much better than the Persian (0.11, 0.07, 0.14). In the original FM 2011 study, the TPC versus WPI and SSS correlations were 0.77 and 0.68, respectively [15]. Our TPC correlation with FIQ was 0.59, while, e.g. the Spanish was weaker (Spearman’s r=0.24), from which they concluded that an increasing number of tender points does not indicate fibromyalgia impact [33]. However, in our study FS score correlated well with both FIQ score and number of TPC, thus supporting the presumption that FSQ-N has good construct, convergent validity for fibromyalgia diagnosis and research according to both fibromyalgia impact and tender points counting.

The number of persons who met the 2011 and 2016 FSDC criteria in the a priori FM group (i.e. those who answered “yes” to the fibromyalgia question in the HUNT3 survey) was significantly greater than in the corresponding a priori non-FM group. These findings are in line with other studies [15], [20], [32].

The ACR1990 criteria were used as a reference versus the two new FM criteria to reveal sensitivity and specificity measures for our a priori FM and non-FM subjects by contingency table analyses. The FSQ-N showed good sensitivity for detecting fibromyalgia using the 2011 and 2016 FSDC (93.9 and 88.8, respectively), as well as moderate specificity (71.3 and 77.5) and positive likelihood ratios (3.27 and 3.95) with acceptable negative likelihood ratios (0.09 and 0.15). These measures showed quite similar sensitivity values as the original 2011 FSDC as well as the Spanish and Turkish versions, and better than, e.g. the Japanese and Persian studies [19], [20]. Nevertheless, a recent Korean study has probably the best measures until now (sensitivity and specificity with 95% CI, for 2011 and 2016 criteria, respectively): 90.8 (84.5–95.1) and 85.1 (79.2–89.1); 93.1 (87.3–96.7) and 90.7 (85.5–94.0) [23].

Our positive likelihood ratio levels indicated that a positive FSQ-N could occur in fibromyalgia versus non-fibromyalgia with an OR of about 4, which is acceptable, but it is poorer than, e.g. the Spanish results (OR=10.8) [32], while the negative likelihood ratio was quite similar as in our 2011/2016 FSDC findings (0.13 vs. 0.09).

We also measured accuracy, which is defined as “the extent to which a test-value” (e.g. from the FSQ of the 2011/2016 criteria) “reflects or agrees with the reference value” (e.g. ACR1990), i.e. the proportion of correct outcomes of a method [34]. The accuracy scores for both the 2011 and 2016 FSDC vs. ACR1990 were 84% in our contingency table analyses. This is a better outcome than, e.g. the German (72.7%) and Persian studies (75.7), but poorer than the Spanish (89%). Our ROC analyses also indicate that the best cut-off for the FS-scale (WPI + SSS) in classifying fibromyalgia is 12.5 (range 0–31). This replicates the Spanish validation study [32] and confirms the original studies which suggested 13 as an optimal FS cut-off [15], [35], but in contrast to the Persian and Japanese ROC-analyses, with their low 8.5 and 9.5 points, respectively [19], [20]. The “accuracy” term is often used interchangeably with “concordance”, which is not quite correct. While accuracy is measured at a certain cut-off point used for dichotomizing the probability distribution given by logistic model, the “concordance” measures the “overall accuracy” over a range of cutoff points expressed the equivalent term=AUC (Area Under Curve). The AUC in our ROC (receiver operating characteristic) curve was 0.86, i.e. “very good”, but still poorer than the excellent Korean AUC=0.97 [23] (0.9–1.0=excellent, 0.8–0.9=very good, 0.7–0.8=good). Our AUC ROC curve is shown in Fig. 1.

Figure 1:

Receiver Operating Characteristics (ROC) curve using FS score with the 1990 ACR diagnosis.

The 2011/2016 FSDC allow self-reports for both individual diagnostics and epidemiological research. A combined ACR1990 + FSDC approach has been tested, but did not add any benefits to single approaches, revealing only a modest sensitivity (75.6%) [32].

6 Conclusion

Our study shows that the Norwegian version of FSQ (FSQ-N) is useful and a valid tool to assess fibromyalgia including its co-morbid core-symptoms such as chronic widespread pain and symptom severity on a continuous scale. The 2016 revision of the 2010/2011 Fibromyalgia Survey Diagnostic Criteria (FSDC) can operate as both a diagnostic instrument for clinical practice and to be applicable for classification in surveys and research in a Norwegian context.