Longitudinal outcomes among patients with fibromyalgia, chronic widespread pain, or localized chronic low back pain

Study design and participants

This retrospective cohort study included participants from the Pain Registry for Epidemiological, Clinical, and Interventional Studies and Innovation (PRECISION) from August 2019 through July 2023 [14]. The registry recruited participants from the 48 contiguous states primarily utilizing social media advertising. Eligible registry participants were 21–79 years of age and had chronic LBP (≥3 months) and sufficient language proficiency to complete case report forms in English. Persons who reported being pregnant or residing at institutional facilities were not eligible. Study participants completed case report forms at enrollment and subsequent quarterly encounters for up to 12 months. This research was approved by the North Texas Regional Institutional Review Board (IRB protocol 2015-169), and the participants gave written informed consent prior to providing data.

Classification of the FM, CWP, and LBP groups

Registry participants were asked at enrollment and the 12-month follow-up encounter if they were currently diagnosed with FM. Participants with a current FM diagnosis at both encounters comprised the FM group. Similarly, CWP was measured at both encounters utilizing the bothersomeness of widespread pain item on the NIH TF Minimum Dataset [7]. It queried participants about how much they were bothered by widespread pain (or pain in most of the body) during the past 4 weeks, with response options including “not at all,” “a little bit,” or “a lot.” Participants who were not classified as having FM, but who were bothered a lot by widespread pain at both encounters, comprised the CWP group. Participants who did not have a current FM diagnosis at both encounters, and were also not at all bothered by widespread pain at both encounters, comprised the LBP group. The remaining participants without a group classification were excluded from the study.

Outcome measures

The primary outcomes included pain and function. A numerical rating scale ranging from 0 to 10 measured typical LBP intensity in the 7 days prior to an encounter. The Roland-Morris Disability Questionnaire measured function [15]. It consists of 24 items that measured back-related disability on the encounter date, with scores ranging from 0 to 24. The HRQOL outcomes were assessed with scales from the Patient-Reported Outcomes Measurement Information System with 29 items (PROMIS-29) that measured physical function, anxiety, depression, fatigue, sleep disturbance, participation in social roles, and pain interference [16]. Scale scores were normed according to the US general population to have a mean (standard deviation [SD]) of 50 (10) [16]. As an exception, the sleep disturbance scale was normed with a calibration sample enriched for chronic illness. All pain, function, and HRQOL outcomes were measured at enrollment and at quarterly encounters for up to 12 months. Higher scores represent worse outcomes on all measures, except on the HRQOL scales for physical function and participation in social roles.

Baseline and longitudinal covariates

Comprehensive data pertaining to chronic LBP were collected at enrollment [14], and a series of baseline variables were selected to characterize participants and control for potential confounding. Sociodemographic characteristics included age, gender, race, ethnicity, and educational level. Health history included cigarette smoking status, musculoskeletal comorbidities (herniated disc, sciatica, osteoarthritis, and osteoporosis), and general medical comorbidities (hypertension, heart disease, diabetes mellitus, asthma, and depression). Psychological factors included pain catastrophizing and pain self-efficacy, measured with the Pain Catastrophizing Scale [17] and Pain Self-Efficacy Questionnaire [18], respectively. The chronic LBP history was characterized by ongoing duration (<1 year, 1–5 years, and >5 years). Data on treatments included nonpharmacological treatments ever utilized (exercise therapy, yoga, massage therapy, spinal manipulation, acupuncture, and cognitive behavioral therapy [CBT]), opioid therapy, and lumbar spine surgery. Current opioid use and prior lumbar spine surgery were measured at each encounter.

Statistical analysis

The utility of the bothersomeness of widespread pain item on the NIH TF Minimum Dataset [7] as an FM screening test was measured utilizing sensitivity, specificity, and positive and negative predictive values, and depicted with a receiver operating characteristic curve. Baseline participant characteristics were described according to chronic pain group (CPG), utilizing number (%) or mean (SD). Group differences were assessed utilizing the chi-square test for categorical variables and analysis of variance for continuous variables. Post-hoc comparisons were performed if omnibus p values were ≤0.05 to identify significant differences between the FM and CWP groups utilizing 2x2 subtables for chi-square analyses or the least significant difference for analysis of variance.

Generalized estimating equation (GEE) models were utilized to measure longitudinal outcomes over 12 months, initially utilizing a partially adjusted model that included a time variable. These GEE models utilized an autoregressive AR(1) correlation structure, fixed effects, and a CPG x time interaction term. The GEE analyses were repeated utilizing a comprehensive multivariable model that included baseline and longitudinal covariates to adjust for potential confounding in addition to measuring time trends. The clinical relevance of differences in outcomes among CPGs were assessed utilizing thresholds for the magnitude of Cohen’s d statistic (small effect, d=0.2; medium effect, d=0.5; large effect, d=0.8) [19]. Differences involving magnitudes of d<0.2 were not considered clinically relevant. All Cohen’s d statistics were transformed so that larger values represented worse outcomes compared with referents (i.e., for FM or CWP groups compared with the LBP group, or for the FM group compared with the CWP group).

Statistical power was estimated with the General Linear Mixed Model Power and Sample Size program for repeated-measures designs [20], and involved hypothesized mean between-group differences in outcomes over 12 months=0 (i.e., no differences in outcomes between groups). The sample size was sufficiently large to exceed 80 % statistical power in a variety of scenarios involving the base correlation between pain and function and the decay rate of the base correlation. The study was not designed to detect time interaction effects owing to uncertainties about the potential interactions (reversed, fully attenuated, or partially attenuated) and thresholds for clinical relevance [21]. Data were managed and analyzed utilizing the IBM SPSS Statistics Software (Version 29). Hypotheses were assessed at an alpha level of 0.05 utilizing two-sided tests.

Participant characteristics

The 1,531 participants with chronic LBP ranged in age from 21 to 79 years at baseline, with a mean of 53.0 (SD, 13.2) years, and 1,132 (73.9 %) were female. Among 1,358 (88.7 %) participants who completed 12 months of follow-up, 312 (23.0 %) were initially classified as having FM (n=66), CWP (n=56), or LBP (n=190). A total of 64 participants with FM tested positive on the bothersomeness of widespread pain item (sensitivity, 0.97). Conversely, 190 of 246 participants without FM tested negative on this item (specificity, 0.77). The positive and negative predictive values were 0.53 and 0.99, respectively (Supplemental Figure 1). The area under the receiver operating characteristic curve was 0.87 (95 % confidence interval [CI], 0.83–0.91) (Supplemental Figure 2). Two participants with FM but without CWP were excluded from further analysis on the theoretical grounds that patients with FM should have CWP.

Baseline participant characteristics according to chronic pain group

The mean age of the 310 remaining participants was 52.3 (SD, 13.6) years, and 238 (76.8 %) were female. These included 64 (20.6 %) participants with FM, 56 (18.1 %) with CWP, and 190 (61.3 %) with LBP. Characteristics that differed across CPGs included gender, educational level, cigarette smoking status, all medical comorbidities except osteoporosis and heart disease, pain catastrophizing, pain self-efficacy, and use of spinal manipulation, CBT, and opioid therapy (Table 1). However, the differences between the FM and CWP groups were limited to gender and the use of four treatments (spinal manipulation, acupuncture, CBT, and opioid therapy). The FM group was more likely to include females and to have utilized the nonpharmacological treatments, whereas the CWP group currently utilized opioid therapy more frequently.

Partially adjusted outcomes

A total of 1,535 encounters were completed over 12 months, including 316 (20.6 %), 276 (18.0 %), and 943 (61.4 %) in the FM, CWP, and LBP groups, respectively. The FM and CWP groups both had greater LBP intensity than the LBP group (Figure 1A). Mean (95 % CI) pain intensity scores were 6.6 (6.3–6.9), 6.9 (6.5–7.3), and 4.9 (4.7–5.1) in the FM, CWP, and LBP groups, respectively, adjusted for time and CPG x time interaction effects (Supplemental Table 1). Compared with the LBP group, there were large differences in pain intensity in the FM group (d=0.83) and CWP group (d=0.98). However, the difference in pain intensity between the FM and CWP groups was not clinically relevant (d=−0.15). Pain intensity decreased over time (β, −0.145; 95 % CI, −0.213 to −0.078; p<0.001). There were FM x time (β, 0.130; 95 % CI, 0.015 to 0.245; p=0.03) and CWP x time (β, 0.191; 95 % CI, 0.081 to 0.301; p<0.001) interaction effects, each indicating that the differences in pain intensity between the respective CPGs and the LBP group increased over time.

Figure 1:

Pain and function outcomes over time. Low Back Pain intensity was measured with a numerical rating scale ranging from 0 to 10 for the typical pain intensity in the 7 days prior to an encounter. Back-related disability was measured with the Roland-Morris questionnaire on the encounter date, with scores ranging from 0 to 24. Error bars represent 95 % confidence intervals. p values were derived from generalized estimating equations for the overall outcomes over 12 months of follow-up. p values were <0.001 for both outcomes.

The FM and CWP groups both had greater back-related disability than the LBP group (Figure 1B). Mean (95 % CI) disability scores were 18.0 (16.9–19.0), 18.3 (17.3–19.2), and 10.4 (9.6–11.2), in the FM, CWP, and LBP groups, respectively, adjusted for time and time interaction effects (Supplemental Table 1). Disability decreased over time (β, −0.295; 95 % CI, −0.442 to −0.148; p<0.001). There was a CWP x time interaction effect (β, 0.406; 95 % CI, 0.159 to 0.654; p=0.001), indicating that the difference in disability between the CWP and LBP groups increased over time.

Figure 2:

Health-related quality-of-life (HRQOL) outcomes over time. Each aspect of HRQOL was measured with the Patient-Reported Outcomes Measurement Information System with 29 items. Error bars represent 95 % confidence intervals. p values were derived from generalized estimating equations for the overall outcomes over 12 months of follow-up. p values were <0.001 for all outcomes.

The FM and CWP groups both reported worse HRQOL than the LBP group on all scales (Figure 2). The differences between the FM and LBP groups ranged from d=0.75 for depression to d=1.24 for fatigue, whereas the differences between the CWP and LBP groups ranged from d=0.87 for sleep disturbance to d=1.05 for pain interference (Supplemental Table 2). The FM group had worse HRQOL than the CWP group on five scales; however, only the difference in fatigue was clinically relevant (d=0.21). Conversely, The FM group had better HRQOL pertaining to depression (d=−0.23). Physical function (β=0.228; 95 % CI, 0.048 to 0.408; p=0.01) and participation in social roles (β=0.362; 95 % CI, 0.102 to 0.622; p=0.006) both increased over time, and pain interference decreased over time (β=−0.528; 95 % CI, −0.742 to −0.315; p<0.001). There was an FM x time interaction effect for depression (β=−0.583; 95 % CI, −1.141 to −0.025; p=0.04), indicating that the difference in depression between the FM and LBP groups decreased over time. There was also a CWP x time interaction effect for pain interference (β=0.487; 95 % CI, 0.061 to 0.914; p=0.03), indicating that the difference in pain interference between the CWP and LBP groups increased over time.

Fully adjusted outcomes

There were no differences in LBP intensity among the CPGs in the fully adjusted model that included a wide array of potential confounders (Table 2). Pain intensity decreased over time (β=−0.143; 95 % CI, −0.210 to −0.076; p<0.001). There were FM x time (β=0.129; 95 % CI, 0.016 to 0.243; p=0.03) and CWP x time (β=0.190; 95 % CI, 0.081 to 0.299; p<0.001) interaction effects, each indicating that the respective differences in pain intensity between these groups and the LBP group increased over time. Characteristics associated with greater pain intensity were: age; being non-White, Hispanic, or a current smoker; longer duration of LBP; and pain catastrophizing; whereas greater educational level and pain self-efficacy were inversely associated with pain intensity (Supplemental Table 3).

Mean (95 % CI) back-related disability scores were 15.3 (13.7–17.0), 16.2 (14.8–17.7), and 12.7 (11.4–14.1), in the FM, CWP, and LBP groups, respectively (Table 2). Disability decreased over time (β=−0.291; 95 % CI, −0.433 to −0.149; p<0.001). There was a CWP x time interaction effect (β=0.424; 95 % CI, 0.175 to 0.673; p<0.001), indicating that the difference in disability between the CWP and LBP groups increased over time. Characteristics associated with disability were the number of medical comorbidities, pain catastrophizing, and current use of opioid therapy, whereas educational level and pain self-efficacy were inversely associated with disability (Supplemental Table 3). There were no clinically relevant differences in pain or disability between the FM and CWP groups (Table 2).

The FM group had worse HRQOL than the LBP group on all scales except anxiety and depression, whereas the CWP group had worse HRQOL on all scales (Table 3). Physical function (β=0.234; 95 % CI, 0.060 to 0.408; p=0.008) and participation in social roles (β=0.374; 95 % CI, 0.121 to 0.628; p=0.004) increased over time, whereas pain interference (β=−0.523; 95 % CI, −0.730 to −0.317; p<0.001) decreased over time. There was an FM x time interaction effect for depression (β=−0.563; 95 % CI, −1.109 to −0.018; p=0.04), indicating that the difference in depression between the FM and LBP groups decreased over time. There was also a CWP x time interaction effect for pain interference (β=0.484; 95 % CI, 0.063 to 0.904; p=0.02), indicating that the difference in pain interference between the CWP and LBP groups increased over time. In comparison with the LBP group, there were small to medium differences in HRQOL in the FM group (ranging from d=0.32 for physical function to d=0.59 for fatigue) and in the CWP group (ranging from d=0.33 for depression to d=0.58 for fatigue). The only clinically relevant differences between the FM and CWP groups involved anxiety (d=−0.32) and depression (d=−45), wherein the FM group had better outcomes (Table 3). Females reported worse physical function and greater pain interference than males (Supplemental Table 4). Pain self-efficacy was consistently associated with better outcomes on all HRQOL scales, whereas pain catastrophizing increased anxiety, depression, fatigue, and pain interference.