Abstract
Objective. To compare outcome-predictor relationships in fibromyalgia (FM) and rheumatoid arthritis (RA), to provide information regarding the competing hypotheses that FM is a continuum or a discrete disorder.
Methods. We studied 3 outcome variables (work disability, opioid use, depression) and 12 clinical predictor variables in 2,046 patients with FM and 20,374 with RA. We determined whether outcome-predictor relationships were stronger in FM or RA by measuring the areas under the receiver-operating curves. We used fractional polynomial logistic regression to create graphic models for the outcome-predictor relationships.
Results. All measures of status and outcome were more abnormal in FM than in RA. Depression was reported in 33.4% of patients with FM compared with 15.1% of those with RA. The predictor-outcome relationship was significantly stronger in RA in 28 of the 36 tests, and not different in the remainder. The relationship between outcome and predictor variables was generally similar in patients with FM and RA. However, unmodeled depression that was not explained by study variables was noted in FM.
Conclusion. Our data are consistent with the hypothesis that FM is the end of a severity continuum, but that additional psychological factors are an integral part of the syndrome.
There are 3 conceptualizations of fibromyalgia (FM)1. The first considers it to be a distinct disorder with an underlying biologic basis (a disorder of central processing with neuroendocrine/neurotransmitter dysregulation). The second model conceptualizes FM as an artificial, if sometimes useful, social construct that describes the end of a pain-distress continuum. A third view sees FM as a kind of unhappiness mixed with medicalization. Similar arguments have surrounded depression with respect to it being a unitary (continuum) or a binary (biologic vs non-biologic) or a medicalized disorder, with current opinion favoring the unitary approach. Recently, regulatory authorities have approved treatments for FM, offering some de facto support, although no proof, for FM as a distinct disorder. A model of the key FM variable, widespread pain, based on observed data, is shown in Figure 1.
We hypothesized that if FM was a distinct disorder rather than a continuum of symptoms, the relation between outcome variables and severity variables might be distorted compared with a similar relationship in recognized conditions, such as rheumatoid arthritis (RA). In addition, exploration of outcome and predictor relationships might shed light on how variables perform in FM, regardless of how it is conceptualized.
We chose 3 outcomes to investigate: work disability, uses of opioids, and depression; and we investigated whether a series of 12 predictor variables would perform similarly or differently in FM compared with RA.
MATERIALS AND METHODS
Patient population
We studied participants in the National Data Bank for Rheumatic Diseases (NDB) longitudinal study of rheumatic disease outcomes. NDB participants are diagnosed by United States rheumatologists and are recruited from their practices. Patients are followed prospectively with semiannual, detailed, 28-page questionnaires, as described2–4. This report utilized NDB data in a longitudinal cohort analysis of 22,420 adult participants (aged 18–103 yrs), of whom 20,374 had RA and 2,046 had FM. Patients were enrolled continuously beginning in 1999 and ending in January 2008. Rheumatic disease diagnoses were made or confirmed by the patient’s rheumatologist. Study variables were assessed at entry into the NDB and at every subsequent semiannual questionnaire. In our report we analyzed one randomly selected observation from each patient.
Study variables
Demographic variables included age and sex. Work and disabled status variables were based on patient self-report. We used self-reported disability rather than government social security disability (SSD) awards in order to identify disability status in patients who were not eligible for SSD awards for administrative reasons. Validation studies have demonstrated the reliability of the work and disability assessments5.
Questions relating to depression and mood included self-reported depression and the Medical Outcomes Study Short-Form 36 (SF-36) mood scale6. We assessed self-reported depression by a single question in each semiannual survey (paraphrasing from a table of questions), “Have you had a problem with depression in the last 6 months?”.
Comorbidity was measured by a patient-reported composite comorbidity score (range 0–9) comprising 11 present or past comorbid conditions including pulmonary disorders, myocardial infarction, other cardiovascular disorders, stroke, hypertension, diabetes, spine/hip/leg fracture, depression, gastrointestinal (GI) ulcer, other GI disorders, and cancer7,8. For the purposes of our study, self-reported depression was omitted from the scale so that it would only assess nondepressive comorbid conditions. As a consequence, its range in our study was 0–8.
Questions related to illness severity included the Health Assessment Questionnaire Disability Index (HAQ)9,10, visual analog scales (VAS) for pain and patient global severity, the Patient Activity Scale (PAS)11, the Regional Pain Scale (RPS)12, and the Symptom Intensity (SI) Scale13. The PAS is formed by multiplying the HAQ by 3.33 and then dividing the sum of the VAS pain, VAS global, and HAQ by 3. This yields a 0–10 scale. The PAS is a composite patient measure of RA activity. The RPS is a self-administered count of the number of painful nonarticular regions. The RPS score ranges from 0 to 19. The SI Scale measures the combination of fatigue and pain extent. Derived from the fatigue and RPS, the SI Scale combines these 2 measures in continuous form according to the following formula: [VAS fatigue + (RPS/2)]/2. This yields a scale with a 0 to 9.75 range.
We defined 3 outcome variables: disabled, as a measure of work status; opioid use, as a measure of severe pain outcome; and self-reported depression, as a measure of clinical depression. All of the other clinical variables were used as predictor covariates. Analyses of disabled status were restricted to patients < 62 years of age to avoid confusion in identifying disabled patients who were retired. The sample sizes for these analyses were 10,662 for RA and 1,351 for FM.
Statistical analyses
T-tests and chi-squared tests were used to test whether the groups differed for the study variables (Table 1). Figure 1 is hypothetical model of the distribution of the RPS in FM and non-FM subjects. In this figure, FM subjects are the actual patients with FM in the study. Non-FM subjects are the study’s patients with RA who have SI Scale scores < 8 (95% of RA subjects).We imposed this restriction in order to try to remove patients with RA who had FM from the figure.
Graphical inspection of the relation between predictor and outcome variable revealed it to be nonlinear. Therefore, we modeled the relation using fractional polynomial logistic regression, as shown for selected variables in Figures 2–4. To compare the strength of the relation in the groups, we compared the area under the receiver-operation characteristic (ROC) curves derived from the logistic model for the FM and RA patient groups (Table 2).
To be certain that the observed results were not an artifact of the difference in RA and FM sex prevalence (Table 1), we reran all of the study analyses after excluding men from analyses. The results were essentially unchanged.
RESULTS
Covariates and outcomes are more abnormal in FM than RA
Table 1 shows that all clinical variable scores were substantially more abnormal for those with FM compared with RA (p < 0.001). We selected 3 binary variables to represent the illness outcomes of depression, pain, and work disability, and we used the outcomes as dependent variables in the analyses of Figures 2–4. For these variables the results were self-reported depression 33.4% versus 15.0%, opioid use 32.0% versus 22.9%, and disabled status 30.5% versus 21.2%, for FM versus RA, respectively. In addition, lifetime depression was 63.9% in FM and 33.0% in RA.
The relation between outcome variables and their predictors in FM compared with RA
We tested whether the modeled relationship between predictor and outcome variables was stronger in RA compared with FM by comparing the respective areas under the ROC curves. As shown in Table 2, the predictor-outcome relationship was significantly stronger in RA in 28 of the 36 tests. In the remaining tests, there were no significant differences.
In Figures 2–4, we modeled these relationships using fractional polynomial logistic regression. With respect to work disability, Figure 2, the curves are similar, as are the intercepts, except for the sleep scale. Taken as a whole, these data indicate that patients with FM and RA respond similarly with respect to work disability; however, the association is weaker (Table 2) in FM than RA.
The curves are also similar for opioid use, a measure of severe pain outcome (Figure 3). However, there is a slight separation of the curves at the lower ranges of mood, HAQ, fatigue, and sleep.
Figure 4 is concerned with depression. Once again the curves are similar in shape, but the intercepts are different and the curves for FM are shifted upwards substantially, except for symptom count. We interpret the curves of Figure 4 to indicate that unmodeled depression is greater in FM than RA. The data of Figure 3 also suggest that unmodeled opioid use is somewhat greater in FM.
DISCUSSION
We found that illness severity measures are substantially more abnormal in FM than RA. Although this is not new information, the very large sample sizes and range of variables studied make these results robust and informative. Our study also provides information about depression, an area of controversy in FM, and about the interaction of depression with other clinical variables. Specifically, with respect to prevalence, we noted that self-reported depression occurred in 33.4% and lifetime depression in 63.9% compared with 15.0% and 33.0% in RA, respectively. Previous studies of FM have shown depression to be increased in clinic populations14 and in the general population15. Depression exists along a continuum of severity16, and the prevalences reported here do not distinguish between major depression and subsyndromal and minor depression (subthreshhold depression). However, subsyndromal depression is a medically significant disorder17.
An important study result was the demonstration that the shape (slope) of the outcome-predictor relationships was similar in FM and RA (Figures 2–4). One can conclude, with respect to outcome-predictor models, that FM has more abnormal outcomes and predictors, as shown in Table 1, but that FM curves are almost superimposable on RA curves. Still, we found that there are some important differences. First, approximately 78% of the associations shown in Table 2 are weaker in FM compared with RA. There are 2 possible explanations for this finding. The first is that FM assessments are “noisier,” in effect made with greater error. The literature supports this interpretation with respect to differences between observed performance and stated performance18, and in overreporting of symptoms19. By itself, this suggests that the continuum approach to FM is not fully sufficient, and that other factors partially differentiate FM from other painful conditions.
The relation between depression and predictor variables shown in Figure 4 offers some further insights. The FM curves are shifted upward, with the exception of the symptom-count curve. This suggests that unmodeled depression may be an important feature of FM. By unmodeled depression we mean depression that is present irrespective of covariate status, suggesting perhaps that a key feature of FM is depression that is unrelated to symptom severity. There are several possible explanations for depression that is unrelated to covariate severity. The first, and we believe most likely, is that for some proportion of patients with FM, FM is to some extent causally related to depression. The literature suggests an increase in depression in families where the proband has FM20,21. The second possibility is that observed FM represents healthcare-seeking behavior and that, in turn, is associated with depression22. In addition, the physician may be more likely to consider the diagnosis of FM in those who are depressed.
We noted that there are 2 contrasting hypotheses about FM, the first that it is a distinct disorder, and the second that it represents the end of a continuum of pain and distress. The data of this report are consistent with FM primarily following the continuum theory, but with additional psychological factors an integral part of the syndrome. Wessely and Hotopf carefully, and at length, summarized the issues regarding distinct disorder versus continuum disorder and concluded “that fatigue and myalgia syndromes are arbitrarily created syndromes that lie at the extreme end of the spectrum of polysymptomatic distress”23.
One should be cautious about extrapolating the result of our study to all patients with FM. No model of FM, whether it is behavioral, genetic, or a disorder of central processing with neuroendocrine/neurotransmitter dysregulation, is capable of explaining more than a small fraction of the FM phenotype. In addition, although this is unlikely, diagnostic differences for FM among referring rheumatologists may have led to heterogeneity among study subjects.
One limitation of our study relates to the definition of depression, which is based on patient self-report rather than formal examinations or a standard scale. However, the area under ROC curve for mood in RA was 0.827 and, in data not shown, was 0.823 for the SF-36 mental component scale, levels that are satisfactory for research.
Despite having more abnormal scores for clinically relevant outcome and predictor variables, the relationship between outcome and predictor variables was generally similar in patients with FM and RA. However, unmodeled depression that was not explained by study variables was noted in FM, and the strength of outcome-predictor relationships was weaker in FM. The data of our report are consistent with the theory that FM is the end of a severity continuum, but that additional psychological factors are an integral part of the syndrome.
Acknowledgments
We thank Prof. J.J. Rasker for his helpful thoughts and manuscript review.
Footnotes
- Accepted for publication November 19, 2008.