SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Objective

People with fibromyalgia (FM) report a number of physical, cognitive, and psychological symptoms. The purpose of the current study was to determine whether people with FM differed based on the type and severity of symptoms, and if so, whether subgroups differ with respect to health care utilization, functional ability, and work status.

Methods

Symptom, health care utilization, work, and physical data were available for 2,182 female responders to an Internet survey. Factor analysis was conducted on the physical and cognitive/psychological symptoms, and resulting factor scores were utilized in a cluster analysis to identify subgroups based on symptoms. Cluster groups were compared on a set of variables (e.g., health care utilization, coping).

Results

Factor analyses resulted in 3 symptom factor scores: musculoskeletal, non-musculoskeletal, and cognitive/psychological symptoms. The optimal cluster solution to the cluster analysis revealed 4 clusters. Group 1 was high on all 3 symptom domains, group 2 was moderate on the 2 physical symptom domains and high on cognitive/psychological symptoms, group 3 was moderate on the 2 physical symptom domains and low on cognitive/psychological symptoms, and group 4 was low on all symptom domains. The more symptomatic groups reported the greatest amount of health care utilization and difficulty in coping with symptoms.

Conclusion

The FM population is heterogeneous with regard to symptom reporting. Additional research is needed to better understand differential symptom experience among people with FM. Clarification of these differences may increase understanding of the mechanisms involved in FM and provide guidance for treatment decisions.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

The diagnosis of fibromyalgia syndrome (FMS) is based on 2 criteria: 1) chronic widespread pain, defined as pain of at least 3 months' duration in at least 3 quadrants of the body and axial, and 2) report of pain on palpation of at least 11 of 18 specified locations (tender points) (1). Although pain and fatigue are the core symptoms of FMS, a plethora of concomitant symptoms (e.g., muscle spasms, headaches, problems with memory and concentration, and depression) are frequently present (2). Conceptually, these symptoms may be divided into 2 broad groups: physical and cognitive/psychological. The average number of symptoms reported by patients with FMS varies (3, 4), and the variability is partially associated with the method used (e.g., spontaneous report, response to formal questionnaires) and number of symptoms presented in the questionnaires. The diversity and severity of comorbid symptoms associated with FMS probably contribute to the lengthy and expensive process patients undergo in an attempt to obtain a diagnosis and treatment. The broader a set of symptoms are, the greater the number of procedures that tend to be performed to rule out objective diseases prior to a diagnosis of FMS. In an Internet survey of more than 2,500 people reporting FMS, more than 75% reported that they had seen more than 3 health care providers prior to obtaining their diagnosis, and ∼25% had seen more than 6 health care providers prior to diagnosis (2). In a study of 2,260 newly diagnosed patients with FMS, the average number of visits to health care professionals during the year prior to diagnosis was 25 (5).

At a theoretical level, the diversity of symptoms reported by FMS patients is consistent with the view held by some experts that FMS is a multisystem disorder rather than a disorder that primarily involves musculoskeletal symptoms (6). Moreover, the observation of wide variability in symptom reporting suggests that patients with FMS may be a heterogeneous group (7, 8).

The primary purpose of the present study was to examine symptom reporting patterns in a large cohort of FMS patients. More specifically, the first goal was to identify whether patients varied in either patterns and/or severity of symptom reporting. In other words, do all patients report both psychological and physical symptoms, or do some only report physical symptoms? In addition, do all patients report that symptoms are very troubling, or do some symptoms have a more minimal impact? Second, if groups of patients are identified based on patterns of symptom reporting, do these groups differ with respect to indices of health care utilization, functional ability, and work status?

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Sample.

In February 2006, the National Fibromyalgia Association (NFA), a layperson support and advocacy organization, posted a comprehensive survey on their Web site. The survey was completed by a cohort of 2,583 people in a 3-day period. A total of 86 responders were men, and thus were not included in the current study. Of the 2,497 female responders, 315 failed to provide information about symptoms and therefore were excluded from the analyses of symptoms, leaving a final sample of 2,182. Using a random sample generator, SPSS for Windows, version 14 (SPSS, Chicago, IL), the sample was divided into 2 approximately equal-sized groups. Sample 1 (n = 1,071) was used for the primary cluster analysis, and sample 2 (n = 1,111) was used to validate the results from the primary analysis.

Questionnaire.

The questionnaire was developed by a multidisciplinary task force of FMS experts consisting of physicians, physical therapists, psychologists, nurses, and patient advocates who attended a 2-day meeting organized by the NFA. The initial questionnaire included items regarding socioeconomic, functional, treatment, and health history, along with physical, cognitive, and emotional symptoms (2). The initial version was circulated among the panel attending the meeting and several FMS experts who had been invited to the meeting but who had been unable to attend. A final version of the questionnaire was created based on feedback from the initial version. The specific items utilized in the current analysis were selected from the final NFA questionnaire and are described in detail below.

Symptom checklist.

Participants were asked to rate on an 11-point scale the degree to which each of a set of 18 physical, psychological, and cognitive symptoms was a problem (where 0 = no problem and 10 = extreme problem) in the past week. The symptoms included pain, fatigue, musculoskeletal symptoms (e.g., morning stiffness, spasms, spasticity, swelling), other physical symptoms commonly reported by FMS patients (e.g., headache/migraine, restless legs, abdominal pain, bladder problems, postural instability, dizziness, rashes), emotional symptoms (e.g., anxiety, depression, anger), and cognitive problems (e.g., concentration problems, forgetfulness).

Impact on physical ability.

Participants were asked to respond to a set of 12 questions regarding their ability to perform activities associated with daily living (e.g., bathing, climbing stairs, shopping). The scale ranged from 0–4, where 0 = cannot do at all and 4 = can do. Because some respondents did not provide answers about each activity, an average physical impact score was calculated by summing across all activities and dividing by the total number of items answered.

Statistical analyses.

All data analyses were conducted using SPSS for Windows, version 14 (SPSS). Based on our interest in patterns on physical and psychological symptoms, symptoms were first grouped as either physical or cognitive/psychological. Principal components factor analyses were then performed separately on each of the 2 groups of variables: physical and cognitive/psychological. Oblique rotations were utilized to aid in interpretation, and the number of factors chosen was based on evaluation of the scree plot. Regression scores from the 2 factor analyses were then used in the subsequent cluster analysis procedure. Factor analyses of all physical symptoms resulted in 2 distinct factors: musculoskeletal symptoms (i.e., pain, stiffness, spasticity, spasms) and non-musculoskeletal symptoms (i.e., abdominal pain, bladder problems, rashes). Cross-loading was defined as a ≤0.15 difference among factor loadings, and a variable was considered to load with ≥0.39 loading (9). Five symptoms had high cross-loading between the 2 factors (i.e., restless leg, swelling, dizziness, postural instability, and fatigue), and migraine did not load significantly on either factor; therefore, these were eliminated. The total model accounted for 52.07% of the variance (musculoskeletal symptoms = 37.24%, non-musculoskeletal symptoms = 14.83%). The factor analysis was performed on the cognitive and psychological symptoms (i.e., depression, anger, anxiety, concentration, and forgetfulness), all of which loaded on a single factor (cognitive/psychological) that accounted for 57.51% of the total variance. Factor scores for each of the 3 factors (musculoskeletal, non-musculoskeletal, and cognitive/psychological) were used in the subsequent clustering procedure.

Cluster analysis was performed to identify symptom profiles among respondents (10). The k-means clustering procedure, which allocates data points into a specified number of clusters based on the centroids of each data point, was utilized to classify respondents into unique clusters. The number of clusters retained was determined based on 2 criteria: stability (i.e., reproducibility) and interpretability. A solution was considered stable if the centroids produced in sample 2 were not significantly different from the centroids produced in sample 1. The cluster groups must also be interpretable, which refers to the alignment of the clusters with clinical reports and experience of working with FMS patients.

The cluster analysis was first performed on sample 1 (n = 1,071), and then cross-validated on sample 2 (n = 1,111). Following determination of cluster groups, results of the cluster analysis were validated on sample 1 through significance tests that compared groups defined by the cluster solution on a set of relevant clinical variables (10). Chi-square tests of significance were utilized on categorical variables, and an analysis of variance was performed on continuous variables.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Demographic data for responders in the initial (sample 1) and confirmation (sample 2) samples are shown in Table 1. Sample 1 was predominantly white (90.9%), with a mean ± SD age of 46.88 ± 10.81 years. Nearly the entire cohort (98%) reported they had been diagnosed with FMS by a health care professional and >76% of the sample indicated that they had been diagnosed with FMS for ≥3 years. More than two-thirds of the sample was married (67.4%), and the majority (83%) had at least some college education. Approximately 31% of responders reported filing a disability claim due to the impact of their FMS symptoms. Approximately half of the sample (53.2%) reported believing they were able to work at an income-producing job. Statistical analyses comparing sample 1 and sample 2 revealed no significant differences on any of the demographic or health care variables, suggesting that the 2 samples were representative of the total cohort.

Table 1. Demographic characteristics for female responders to the National Fibromyalgia Association questionnaire for sample 1 and sample 2
VariableSample 1 (n = 1,071)*Sample 2 (n = 1,111)*P
  • *

    Exact number for each variable may vary slightly due to missing data.

Age, mean ± SD years46.88 ± 10.8146.78 ± 10.710.83
Length of diagnosis, mean ± SD  0.27
 Not officially diagnosed19 ± 1.819 ± 1.7 
 0–6 months67 ± 6.380 ± 7.2 
 7–12 months68 ± 6.446 ± 4.1 
 1–2 years93 ± 8.7101 ± 9.1 
 3–4 years215 ± 20.1237 ± 21.4 
 >4 years608 ± 56.8626 ± 56.4 
Length of symptoms, mean ± SD  0.17
 0–6 months2 ± 0.210 ± 0.9 
 7–12 months19 ± 1.819 ± 1.7 
 1–2 years39 ± 3.751 ± 4.6 
 3–4 years151 ± 14.2153 ± 13.9 
 >4 years854 ± 80.2869 ± 78.9 
Race, no. (%)  0.65
 Hispanic4 (0.4)6 (0.5) 
 White973 (90.9)1,021 (91.9) 
 African American18 (1.7)20 (1.8) 
 Asian5 (0.5)4 (0.4) 
 Native American46 (4.3)33 (3.0) 
 Other24 (2.2)27 (2.4) 
Marital status, no. (%)  0.82
 Never married129 (12.1)120 (10.8) 
 Divorced/separated193 (18.1)205 (18.5) 
 Widowed26 (2.4)29 (2.6) 
 Married721 (67.4)756 (68.1) 
Living arrangement, no. (%)  0.47
 Living with someone915 (85.4)939 (86.5) 
 Living alone134 (12.7)146 (13.3) 
Education, no. (%)  0.34
 Grade school2 (0.2)1 (0.1) 
 High school178 (16.8)154 (14.0) 
 Some college397 (37.4)437 (39.8) 
 College274 (25.8)299 (27.3) 
 Graduate/professional school210 (19.8)206 (18.8) 
Income, mean ± SD dollars45,000–50,000 ± 4,00045,000–50,000 ± 4,0000.99

The frequency and average rating of symptoms for responders that provided symptom information (n = 2,182) are shown in Table 2. The most frequently reported symptoms were fatigue (n = 2,175 [99.7%]), pain (n = 1,990 [91.2%]), and stiffness (n = 1,954 [89.6%]). Respondents reported their most problematic symptoms were stiffness (mean ± SD rating 7.17 ± 2.49), fatigue (7.05 ± 2.07), and pain (6.45 ± 1.97). The mean ± SD number of symptoms endorsed was 10.59 ± 1.39 out of a possible 18.

Table 2. Frequency and average rating of symptoms among female responders to the National Fibromyalgia Association questionnaire (n = 2,182)
VariableFrequency >3, no. (%)Average rating, mean ± SD
Core physical  
 Stiffness1,954 (89.6)7.17 ± 2.49
 Pain1,990 (91.2)6.45 ± 1.97
 Spasticity1,522 (69.8)5.13 ± 2.98
 Muscle spasms1,359 (62.3)4.82 ± 3.23
Extraneous physical  
 Abdominal pain1,030 (47.2)3.60 ± 2.89
 Bladder problems666 (30.5)2.50 ± 2.98
 Rashes471 (21.6)1.89 ± 2.96
Cognitive/psychological  
 Forgetfulness1,721 (78.9)5.88 ± 2.67
 Concentration1,750 (80.2)5.87 ± 2.6
 Anxiety1,274 (58.4)4.52 ± 3.07
 Anger1,042 (47.8)3.87 ± 2.96
 Depression1,229 (56.3)4.37 ± 3.14
Other symptoms  
 Fatigue2,175 (99.7)7.05 ± 2.07
 Migraine1,230 (56.4)4.36 ± 3.09
 Balance981 (45.0)3.48 ± 2.87
 Dizziness800 (36.7)2.90 ± 2.82
 Swelling860 (39.4)3.20 ± 3.09
 Restless leg1,001 (45.9)3.58 ± 3.44

Development of participant profiles.

The k-means clustering procedure was conducted with 3 factor scores (musculoskeletal, non-musculoskeletal, and cognitive/psychological symptoms) as the clustering variables, with iterations of 2, 3, 4, 5, and 6 cluster solutions. The cluster analyses were repeated on sample 2 for cross-validation, with successful replication with up to 4 groups (Figure 1). Statistical comparisons between centroids indicated the mean factor scores were not different for 11 of 12 comparisons between samples 1 and 2. The scores on the non-musculoskeletal factor were different for the group with moderate physical/low cognitive/psychological symptoms (P < 0.001); however, the total amount of variance accounted for by the relationship was only 2.8%, suggesting the difference was not meaningful. Additionally, differences were identified on external variables of interest with 4 groups (Tables 3 and 4), and therefore the 4 cluster solution was retained.

thumbnail image

Figure 1. A, Final cluster centers (mean standardized scores) for symptom domains by cluster group for sample 1 (n = 1,071). B, Final cluster centers for symptom domains by cluster group for sample 2 (n = 1,111). MS = musculoskeletal symptoms; NMS = non-musculoskeletal symptoms; CP = cognitive/psychological symptoms; HH = high physical/high CP group; MH = moderate physical/high CP group; ML = moderate physical/low CP group; LL = low physical/low CP group.

Download figure to PowerPoint

Table 3. Demographic characteristics for responders by cluster group for sample 1 (n = 1,071)*
VariableHH (n = 202)MH (n = 304)ML (n = 274)LL (n = 291)P
  • *

    HH = high physical/high cognitive/psychological symptoms group; MH = moderate physical/high cognitive/psychological symptoms group; ML = moderate physical/low cognitive/psychological symptoms group; LL = low physical/low cognitive/psychological symptoms group.

  • Exact number for each variable may vary slightly due to missing data.

  • P < 0.05.

Age, mean ± SD years46.80 ± 9.4647.43 ± 10.2947.96 ± 10.6246.38 ± 12.270.20
Length of symptoms, mean ± SD    0.18
 0–6 months0 ± 00 ± 00 ± 02 ± 0.7 
 7–12 months4 ± 2.04 ± 1.36 ± 2.25 ± 1.7 
 1–2 years3 ± 1.513 ± 4.38 ± 3.015 ± 5.2 
 3–4 years34 ± 16.947 ± 15.528 ± 10.342 ± 14.5 
 4 years160 ± 79.6239 ± 78.9229 ± 84.5226 ± 77.9 
Length of diagnosis, mean ± SD    0.28
 Not diagnosed4 ± 2.04 ± 1.37 ± 2.64 ± 1.4 
 0–6 months14 ± 6.913 ± 4.315 ± 5.525 ± 8.6 
 7–12 months12 ± 5.923 ± 7.611 ± 4.022 ± 7.6 
 1–2 years21 ± 10.426 ± 8.617 ± 6.229 ± 10.0 
 3–4 years38 ± 18.867 ± 22.051 ± 18.659 ± 20.3 
 4 years113 ± 55.9171 ± 56.3173 ± 63.1151 ± 52.1 
Race, no. (%)    0.86
 Hispanic/Latino1 (0.5)1 (0.3)1 (0.4)1 (0.3) 
 White179 (88.6)282 (92.8)247 (90.1)265 (91.4) 
 African American3 (1.5)3 (1.0)7 (2.6)5 (1.7) 
 Asian1 (0.5)2 (0.7)0 (0)2 (0.7) 
 Native American12 (5.9)12 (3.9)10 (3.6)12 (4.1) 
 Other6 (3.0)4 (1.3)9 (3.3)5 (1.7) 
Marital Status, no. (%)    0.005
 Never married14 (7.0)48 (15.8)31 (11.3)36 (12.4) 
 Divorced/separated53 (26.4)54 (17.8)47 (17.2)39 (13.4) 
 Widowed3 (1.5)7 (2.3)10 (3.6)6 (2.1) 
 Married131 (65.2)195 (64.1)186 (67.9)209 (72.1) 
Living status, no. (%)    0.41
 Living with someone179 (89.9)256 (86.2)233 (86.3)247 (86.7) 
 Living alone18 (9.1)41 (13.8)37 (13.7)38 (13.3) 
Education, no. (%)    < 0.001
 Grade school0 (0)1 (0.3)0 (0)1 (0.3) 
 High school53 (26.2)54 (17.9)47 (17.3)24 (8.4) 
 Some college91 (45.0)120 (39.7)100 (36.9)86 (30.1) 
 College30 (14.9)87 (28.8)65 (24.0)92 (32.2) 
 Graduate/professional school28 (13.9)40 (13.2)59 (21.8)83 (29.0) 
Income, mean ± SD dollars39,000 ± 3,00045,000 ± 3,00052,000 ± 3,00059,000 ± 3,000< 0.001
Table 4. Health and disability variables for responders by cluster group for sample 1 (n = 1,071)*
VariableHH (n = 202)MH (n = 304)ML (n = 274)LL (n = 291)P
  • *

    Values are the number (percentage) unless otherwise indicated. Scale items are on a 0–10 scale, where 0 = no problem. ER = emergency room; FM = fibromyalgia. See Table 3 for additional definitions.

  • Exact number for each variable may vary slightly due to missing data.

  • P < 0.001.

Filed workers compensation claim15 (7.6)18 (6.0)21 (7.9)14 (4.9)0.48
Filed disability claim94 (48.0)113 (37.8)76 (28.3)42 (14.9)< 0.001
Times seen in ER in past year    < 0.001
 097 (48.87)186 (61.6)185 (68.5)224 (78.6) 
 1–486 (43.2)109 (36.1)79 (29.3)60 (21.1) 
 5–810 (5.0)5 (1.7)2 (0.7)1 (0.4) 
 9–123 (1.5)1 (0.3)4 (1.5)0 (0) 
 >123 (1.5)1 (0.3)0 (0)0 (0) 
Times seen provider in past year    < 0.001
 04 (2.0)14 (4.6)7 (2.6)11 (3.8) 
 1–457 (28.2)113 (37.2)132 (48.2)175 (60.3) 
 5–848 (23.8)74 (24.3)65 (23.7)66 (22.8) 
 9–1247 (23.3)57 (18.8)32 (11.7)19 (6.6) 
 >1246 (22.8)46 (15.1)38 (13.9)19 (6.6) 
Providers seen for diagnosis    < 0.001
 None3 (1.5)6 (2.0)2 (0.7)5 (1.7) 
 1–231 (15.5)63 (20.9)74 (27.0)99 (34.1) 
 3–458 (29.0)105 (34.8)88 (32.1)102 (35.2) 
 5–635 (17.5)47 (15.6)40 (14.6)36 (12.4) 
 >673 (36.5)81 (26.8)70 (25.5)48 (16.6) 
Believe able to work63 (31.7)131 (43.4)166 (61.3)204 (70.8)< 0.001
Believe provider treats FM legitimately    < 0.001
 Not at all17 (8.6)28 (9.5)11 (4.1)7 (2.5) 
 Somewhat60 (30.3)69 (23.3)66 (24.6)43 (15.4) 
 Legitimate57 (28.8)88 (29.7)78 (29.1)92 (33.0) 
 Very64 (32.3)111 (37.5)113 (42.2)137 (49.1) 
Lack of social support, mean ± SD5.19 ± 3.224.62 ± 3.174.53 ± 3.233.82 ± 3.20< 0.001
Ability to cope with symptoms, mean ± SD6.40 ± 2.245.68 ± 2.144.87 ± 2.323.79 ± 2.26< 0.001
Sleep symptoms, mean ± SD     
 Fall asleep6.8 ± 3.216.47 ± 3.05.7 ± 3.213.95 ± 3.1< 0.001
 Stay asleep7.05 ± 2.916.86 ± 2.626.49 ± 2.935.01 ± 2.99< 0.001
 Awake rested7.64 ± 0.647.29 ± 2.646.63 ± 2.655.55 ± 2.54< 0.001
Physical impact score, mean ± SD25.28 ± 9.5330.47 ± 9.5131.60 ± 9.3737.47 ± 7.94< 0.001

Cluster 1: high physical/high cognitive/psychological symptoms group.

Final cluster centers (centroids: the mean scores of the clustering variables for each cluster) are shown in Figure 1. The scores are based on the z distribution of the factor scores with a mean ± SD of 0 ± 1; thus, the greater a cluster centroid deviates from 0, the greater the magnitude of difference on that variable with respect to the other clusters. Examination of Figure 1 suggests that cluster 1, which included 202 respondents (18.86%), is highest on all 3 symptom domains. All scores were greater than mean ± SD 0 ± 0.5 (musculoskeletal = 0.91, non-musculoskeletal = 1.51, cognitive/psychological = 1.04), suggesting that they were different from the average symptom scores. With respect to all other clusters, the first cluster scored at least 50% higher on all symptom domains, and was thus labeled the high physical/high cognitive/psychological symptoms (HH) group.

Cluster 2: moderate physical/high cognitive/psychological symptoms group.

The second cluster was comprised of 304 respondents (28.38%), and scores on the physical factors were less than mean ± SD 0 ± 0.5 (musculoskeletal = 0.26, non-musculoskeletal = −0.35), whereas the cognitive/psychological (0.76) factor was more than 0 ± 0.5. Therefore, the second cluster was considered moderate on the 2 physical factors, and high on the cognitive/psychological factor (MH).

Cluster 3: moderate physical/low cognitive/psychological symptoms group.

Approximately 25.58% (n = 274) of the sample was included in the third cluster group. Similar to the second cluster, the scores on both physical factors were less than 0.5 SD from the mean, and were considered moderate (musculoskeletal = 0.35, non-musculoskeletal = 0.07). However, the scores on the cognitive/psychological factor were greater than 0.5 SD from the mean and were considered low (−0.59), and thus the group was labeled as the moderate physical/low cognitive/psychological symptoms (ML) group.

Cluster 4: low physical/low cognitive/psychological symptoms group.

The remaining 291 respondents (27.17%) of sample 1 comprised the fourth group. These respondents scored relatively lower (at least 0.5 SD from the mean) on all 3 symptom factors. Cluster centroids were approximately twice as low as compared with the HH group on the musculoskeletal (−1.16), non-musculoskeletal (−0.70), and cognitive/psychological factors (−0.87); therefore, the fourth group was labeled the low physical/low cognitive/psychological symptoms (LL) group.

Validation of cluster solution.

The cluster solution was validated by conducting identical k-means clustering procedures with iterations of 2, 3, and 4 cluster solutions on a second sample randomly selected from the total set of sample 2. Inspection of Figure 1 reveals that the fourth cluster solution identified was virtually identical for both samples.

In addition to cross-validating the cluster solution on a separate sample, significance tests that compared the obtained clusters in sample 1 on variables of clinical importance were conducted (11). Results are described below.

Demographic variables.

Table 3 shows demographic characteristics for the 4 cluster profiles in sample 1. Statistical comparisons indicated that there were no significant differences among the 4 clusters in age (F[3,1049] = 1.05, P = 0.20), length of diagnosis (χ2 = 17.69, P = 0.28), race (χ2 = 9.31, P = 0.86), or living arrangement (χ2 = 2.91, P = 0.41). A significant difference among cluster profiles was detected in marital status (χ2 = 23.37, P = 0.005). The HH group was most likely to have been divorced (26.4%), followed by the MH (17.8%), ML (17.2%), and LL (13.4%) groups. Analysis also revealed an overall effect for education (χ2 = 70.61, P < 0.001), with 29% of the LL group, 21.8% of the ML group, 13.9% of the HH group, and 13.2% of the MH group reporting attending graduate or professional school. A main effect for income suggested differences among profiles (F[3,1046] = 20.46, P < 0.001), with post hoc results indicating the HH group < MH group < ML group < LL group (P < 0.009 for all).

Health care, function, and work variables.

Health care, functional status, and disability/work beliefs for the cluster groups are shown in Table 4. Overall, the HH group was characterized by the greatest amount of health care utilization, the worst physical function, and the least favorable work characteristics. Conversely, the LL group was characterized by the least amount of health care utilization and the best physical function and work characteristics of all 4 groups.

Physical impact scores were significantly different among groups (F[3,946] = 66.03, P < 0.001), and discriminated between all but the 2 moderate physical groups (HH group < MH group ≤ ML group < LL group; P < 0.001 for all), with the HH group reporting the worst physical function (mean ± SD 25.28 ± 9.53), and the LL reporting group the best (37.47 ± 7.94). Lack of social support also varied among clusters (F[3,1063] = 7.59, P < 0.001), with post hoc tests indicating that the HH group differed from all other groups (HH group ≥ MH group ≥ ML group > LL group; P < 0.001 for all), and the ML group differed from the HH group (ML group < HH group; P < 0.03). A main effect for the amount of difficulty in managing symptoms indicated differences among the symptom profiles in coping (F[3,1056] = 6.13, P < 0.001), and all groups significantly differed from one another, with the greatest amount of difficulty reported by the HH group (HH group > MH group > ML group > LL group; P < 0.001 for all).

Symptom groups differed in their belief that they are able to work at an income-producing job (χ2 = 91.82, P < 0.001), with the LL group the most likely to agree (70.8%), followed in descending order by the ML (61.3%), MH (43.4%), and HH (31.7%) groups. The number of group members filing a disability claim also varied among symptom clusters (χ2 = 67.48, P < 0.001), with 48.0% of the HH, 37.8% of the MH, 28.3% of the ML, and 14.9% of the LL groups reporting they had filed a claim.

The number of health care providers seen in the past year varied by symptom cluster (χ2 = 87.89, P < 0.001), as did the number of times they visited the emergency room (χ2 = 70.43, P < 0.001) and the number of health care providers they had seen before obtaining a diagnosis of FMS (χ2 = 43.78, P < 0.001), with the greatest amount of health care utilization in the HH group, followed by the MH, ML, and LL groups, respectively.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

The concept of subgroups within FMS was considered even prior to the development of the formal American College of Rheumatology (ACR) diagnostic criteria (1, 12, 13). Since the publication of the ACR criteria in 1990, subgroups have been described based on psychosocial parameters (8), tender point thresholds, sensory thresholds, and psychological variables (7). To our knowledge, our study is the first that identifies subgroups of people with FMS based on patterns of reporting musculoskeletal symptoms, other physical symptoms, and cognitive/psychological symptoms. Although several symptoms commonly associated with FMS (e.g., fatigue) were not included in the cluster analysis due to cross-loading, the intent of the subgrouping was to describe patterns based on types of symptoms, and not for diagnostic purposes. The withdrawal of these symptoms does not suggest that they do not play an important role in FMS and should be considered in assessment. Four types of FMS patients were identified based on their reporting of a set of these 3 symptom domains: 1) HH, 2) ML, 3) MH, and 4) LL. The analyses on external variables provided evidence that the 4 clusters differed significantly on several important variables, including health care utilization, degree of physical impairment, and ability to cope with symptoms. This, combined with the cross-validation on the test set, supports the validity of the methods used to identify patient clusters.

With regard to the physical symptom patterns, there was a tendency for FMS patients within each of the 4 clusters to report similar degrees of musculoskeletal and non-musculoskeletal physical symptom severity, as evidenced by the relatively flat lines between musculoskeletal and non-musculoskeletal factor scores in Figure 1. The comparable prevalence of musculoskeletal and non-musculoskeletal symptoms within clusters suggests that FMS may not be an exclusively musculoskeletal problem, which is consistent with the research by Yunus (6, 14, 15).

Approximately 50% of the FMS patients (both the HH and MH groups) reported major problems with psychological and cognitive symptoms. The HH and LL groups reported that their physical symptoms (musculoskeletal and non-musculoskeletal) and their psychological symptoms were comparable in severity. In contrast, the MH group reported that psychological symptoms were more severe than physical symptoms, whereas the ML group showed the opposite pattern. The 2 groups that reported greater problems with the cognitive/psychological symptoms reported a relative inability to manage their symptoms as compared with the LL and ML groups. Based on their personal recognition of an inability to cope with their symptoms, these patients might benefit the most from psychological treatment aimed at improving coping skills (16–18).

There are a number of possible explanations to account for the differences in symptom reporting among patients with FMS. One possibility is that as FMS progresses, symptom severity increases and it evolves into a multisystem syndrome. However, this hypothesis was not supported in the current study, since duration of symptoms and time elapsed since diagnosis were equivalent among all cluster groups. Another possibility is that patients with greater severity of symptoms have more extensive physical pathology, as a result, for example, of comorbid medical conditions. Third, it is possible that high symptom reporters experience more distressing symptoms because they have developed central nervous system sensitization, increased levels of inflammatory neurotransmitters such as substance P, or hypothalamic–pituitary–adrenal axis dysfunction (for review, see refs.6,19). These 2 latter explanations warrant additional research to determine their relative contributions to patterns of symptom reporting.

One notable difference among the subgroups was in education level, with the LL group > ML group > MH group > HH group reporting lower levels of education. Thus, there was an inverse relationship between participants' education levels and the number and severity of the symptoms they reported. This relationship raises the possibility that responses of participants were influenced, at least to some extent, by an education-related response set, such that what is really being measured is the degree to which a person will agree or disagree with an item, regardless of content (20). It has been suggested that the acquiescence is associated with item ambiguity (21), so that when respondents are unsure of an item, they will tend to answer in the affirmative. A study that alters the scale presentation and either balances the number of positive and negative symptom items or presents a negative scale to a subset and a positive symptom scale to a subset may provide insight into the role cognitive bias and response set may play in response to self-report questionnaires in general and, in particular, symptom reporting.

This study has several limitations. All of the data were derived from self-reports of a sample of participants who responded to an online survey conducted by the NFA. We did not have independent information about whether they actually met the diagnostic criteria for FMS, or about any other medical data that they provided. Moreover, they were not a clinical sample and reported high levels of education. Thus, the results may not generalize to other samples such as people with FMS or clinical populations. The percentage included in the HH group is likely an underestimate compared with what may be observed in clinical settings, since the sample may not have been specifically seeking treatment at the time of the survey.

Additionally, multiple analyses were performed to identify associations between the 4 clusters identified in this study and variables such as perceived work capacity and utilization of health resources. As a result, it is possible that some of the associations described above may be spurious. Since this was an exploratory study, we decided not to make any adjustment for the performance of multiple analyses. However, observation of the results reported does indicate that levels of statistical significance usually exceeded P < 0.001, and those may be reasonably valid.

The diagnosis of FMS remains controversial, in part due to the failure to identify clear pathologic mechanisms to aid in diagnosis. The current recommended classification criteria, positive tender point evaluation and chronic widespread pain of at least 3 months' duration, are broad and nonspecific. Application of these criteria has resulted in a diverse group of people being diagnosed with FMS. Results from the current study confirm there are several subgroups within the FMS population. The symptom patterns reported by patients with FMS may reflect differences in the mechanisms underlying their conditions. For example, it is possible that musculoskeletal symptoms dominate the experiences of some patients with FMS, whereas cognitive/psychological symptoms dominate the experiences of others. Research is needed to explore the associations among the subgroups identified in previous research (7, 8, 12, 13) and the subgroups identified in the current study. Understanding the differences within FMS patients may guide future revisions of the diagnostic criteria for FMS, serve as a basis for better understanding the mechanisms involved in FMS, and inform treatment decision making.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

Dr. Turk had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Wilson, Robinson, Turk.

Acquisition of data. Wilson, Robinson, Turk.

Analysis and interpretation of data. Turk.

Manuscript preparation. Wilson, Robinson, Turk.

Statistical analysis. Wilson, Robinson, Turk.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES

The authors would like to thank Peter G. Waldo for his insightful comments regarding data analyses.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES