The role of symptoms severity, heart rate, and central sensitization for predicting sleep quality in patients with fibromyalgia

Abstract

Background

Clinical predictors of sleep quality in patients with fibromyalgia syndrome (FMS) are still unknown. By identifying these factors, we could raise new mechanistic hypotheses and guide management approaches. We aimed to describe the sleep quality of FMS patients and to explore the clinical and quantitative sensory testing (QST) predictors of poor sleep quality and its subcomponents.

Methods

This study is a cross-sectional analysis of an ongoing clinical trial. We performed linear regression models between sleep quality (Pittsburgh Sleep Quality Index [PSQI]) and demographic, clinical, and QST variables, controlling for age and gender. Predictors for the total PSQI score and its seven subcomponents were found using a sequential modeling approach.

Results

We included 65 patients. The PSQI score was 12.78 ± 4.39, with 95.39% classified as poor sleepers. Sleep disturbance, use of sleep medications, and subjective sleep quality were the worst subdomains. We found poor PSQI scores were highly associated with symptom severity (FIQR score and PROMIS fatigue), pain severity, and higher depression levels, explaining up to 31% of the variance. Fatigue and depression scores also predicted the subjective sleep quality and daytime dysfunction subcomponents. Heart rate changes (surrogate of physical conditioning) predicted the sleep disturbance subcomponent. QST variables were not associated with sleep quality or its subcomponents.

Conclusions

Symptom severity, fatigue, pain, and depression (but no central sensitization) are the main predictors of poor sleep quality. Heart rate changes independently predicted the sleep disturbance subdomain (the most affected one in our sample), suggesting an essential role of physical conditioning in modulating sleep quality in FMS patients. This underscores the need for multidimensional treatments targeting depression and physical activity to improve the sleep quality of FMS patients.

Introduction

Fibromyalgia syndrome (FMS) is a chronic musculoskeletal syndrome characterized primarily by generalized pain that varies in severity and localization but also includes a myriad of multisystem dysfunctions involving cognitive and sleep complaints. According to the American College of Rheumatology updated guideline, sleep alterations (“waking unrefreshed”) are included in the symptom severity score recommended for FMS diagnosis. Indeed, sleep disturbances (eg, non-restorative sleep) are highly common in FMS patients, with prevalence ranging from 60 to 80%.¹

Growing evidence suggests a bidirectional relationship between sleep disorders and FMS, since they are not only a consequence of FMS but are also involved as comorbidity in the pathogenesis and disease exacerbation.² One large longitudinal study reported that previously healthy women experiencing poor quality of sleep at baseline, were at higher risk for developing FMS. The risk ratio was 3.4 and presented a dose-dependent effect based on how severe the sleep problems were.²

The relationship of sleep alterations and FMS was further explored in a meta-analysis of sleep quality and efficiency, including case-control studies comparing those outcomes in FMS patients versus healthy controls.³ They found lower sleep quality (lower efficiency and larger latencies) and altered polysomnography recordings (less sleep time, larger percentage of stage 1, and smaller percentage of slow-wave sleep) in FMS. The authors concluded that due to the observational nature of the study, more research of sleep disturbances within FMS populations is needed to understand the underlying factors driving this relationship.

Only a few studies have examined the clinical factors that influence sleep quality in FMS patients and most of them did not use standardized clinical assessments^4,5 or adjusted analyses to reduce confounding.⁶ Similarly, to the best of our knowledge, no study has explored the factors influencing sleep quality subcomponents (such as sleep efficiency, latency, disturbances, etc.). Although disease severity and pain intensity have been associated with sleep disturbance in FMS,^6,7 little is known about the contribution of other factors associated with sleep quality in other chronic pain^8,9 and rheumatologic conditions,¹⁰ such as depression, anxiety, physical conditioning, and central sensitization.

Therefore, there is a need for identifying potential determinants of sleep quality (and its subcomponents) in FMS patients. Relevant clinical and sensorial variables (quantitative sensory testing [QST]) need to be tested to further raise new mechanistic hypothesis on the sources and consequences of sleep alterations in FMS (eg, pain inhibition impairment versus emotion-related dysfunctions). Likewise, new treatment approaches can be developed by identifying and modulating such predictors.

Hence, we aimed to describe the sleep quality of FMS patients using a validated self-reported scale (the Pittsburgh Sleep Quality Index [PSQI]), and to explore the clinical and QST predictors of poor sleep quality and its subcomponents (PSQI subscales) in patients with FMS. The subcomponents of the PSQI are needed to understand the multidimensionality of the quality of sleep. The global score alone would be insufficient to disentangle the factors influencing the quality of life of FMS patients.

We hypothesize that FMS patients would have poor sleep quality (PSQI >5 points).¹¹ Regarding predictors, we hypothesize that disease severity, pain intensity, mood alterations, and static and dynamic QST (surrogates of peripheral and central sensitization) would be highly associated with poor quality of sleep.

Methods

Study design and inclusion criteria

This study is a cross-sectional analysis of the baseline data from an ongoing randomized, double-blind clinical trial investigating transcranial direct current stimulation (tDCS) and aerobic exercise on fibromyalgia patients (NCT03371225). The data presented in this study was collected from the baseline visit, before the intervention, from May 13, 2019, to June 22, 2022. All the subjects signed consent forms before starting the protocol, and the study was previously approved by the Institutional Review Board of Mass General Brigham’s ethical committee (protocol number 2017P002524). Further details can be found in the published protocol.¹²

We included patients between 18 and 64 years old, with a fibromyalgia diagnosis based on the criteria from the 2010 American College of Rheumatology (ACR):¹³ widespread pain rated with a minimum average of 4 on a 0 to 10 visual analog scale (VAS) that lasted at least 6 months, without any other comorbid chronic pain conditions diagnosis, and that is pain resistant to analgesic and pain medication. The exclusion criteria comprised clinically significant or unstable medical or psychiatric disorder, self-report of substance abuse in the last 6 months, history of significant neurological deficit, previous neurological procedures involving craniotomy, severe depression (higher than 30 on the Beck’s Depression Scale), pregnancy, large opioid usage (usage of over 30 mg of oxycodone/hydrocodone or 7.5 mg of hydromorphone (Dilaudid) or equivalent), and any increased risks associated with physical exercises.

Demographic and clinical variables

Self-reported demographic variables were assessed from all the subjects, for instance: age, gender, race, and education level. We also used questionnaires to evaluate clinical variables, such as VAS pain level,¹⁴ VAS anxiety level,¹⁵ VAS depression level, VAS sleepiness level,¹⁶ Beck Depression Inventory (BDI),¹⁷ Quality of Life Scale (QOLS),¹⁸ revised fibromyalgia impact questionnaire (FIQR),¹⁹ brief pain inventory (BPI)²⁰—with regard to pain and interference subscale -, the patient-reported outcomes measurement information system (PROMIS-29 v2.0) for anxiety, fatigue and fibromyalgia,²¹ use of antidepressants, and smoking and drinking habits. We also evaluated their baseline and maximum heart rate—participants walked on a treadmill over 30 minutes and the investigator sequentially increased the treadmill speed by 0.1 mph every 5 seconds, until the participant reached 60%–70% of age-predicted maximal heart rate following the formula HRmax = 208 − (0.7 × age).^22–27 We also propose to use the relative increase of the HR max compared to the baseline HR as a surrogate marker of physical conditioning—calculated as HR max/HR baseline—meaning the higher this increase, the poorer the physical conditioning.²⁸ Finally, we also assessed their body mass index (BMI).

PSQI assessment

This scale was developed by Buysse et al. in 1988²⁹ and aims to evaluate the overall quality of sleep. The result is based on 19 self-rated items, that lead to seven total components. The components are: (1) subjective sleep quality (0 is very good, 1 is good, 2 is fairly bad and 3 is very bad); (2) sleep latency (the cut points are 15 minutes, 16 to 30 minutes, 31 to 60 minutes and over 60 minutes, respectively graded from 0 to 3); (3) sleep duration (the cut points are over 7 hours, from 6 to 7 hours, from 5 to 6 hours and less than 5 hours, respectively graded from 0 to 3); (4) habitual sleep efficiency (correlation between the total number of hours asleep and the total number of hours spent in bed); (5) sleep disturbances (sum of number of disturbances such as waking up at night or early morning, getting up to use the bathroom, etc.); (6) use sleeping medications (amount of sleep medications, either prescribed or over the counter, that the patient used on the last month); and (7) daytime dysfunction (overall difficulty of staying awake during daytime for activities such as driving and eating, and having enthusiasm to perform general activities on the last month). The higher the score, the worse is the quality of sleep (ranging from 0 to 21 points, each component scored from 0 to 3). Previous studies suggested that scores higher than 5 are considered poor sleep quality compared to normative data.^11,29 There are no stablished cut-points for the subscales of the PSQI test, nevertheless we decided to use the median as a reference point. Since we took extreme values for the cut-points this allowed us to identify which subcomponents were the most affected.

QST variables

The sensory profile was measured during the baseline assessment using QST. The measurements used in our current analysis were conditioned pain modulation (CPM), temporal slow pain summation (TSPS) and pain-60. CPM is used to evaluate subjects’ capacity to inhibit the pain,³⁰ indirectly evaluating the descending pain modulatory system (DPMS). By using a painful stimulus associated with a conditioning pain stimulus, there is the activation of a cortically regulated spinal-bulb-spinal loop by the diffuse noxious inhibitory control (DNIC) mechanism, leading to the “pain Inhibits pain” phenomenon.³¹ TSPS was used to assess central sensitivity with repetitive stimuli pulses ³² following Staud et al. suggested protocol;³³ Pain-60 is a measurement of the temperature that leads to a pain level 60 on a 0–100 numerical pain scale (NPS) after a series of heat stimulations—this measurement is mostly used as a reference for the CPM test. However, this value itself is not used directly to calculate the CPM score, and the values of both scores are independent from each other.

Statistical analysis

We calculated the minimal sample size needed to detect alterations in PSQI scores of FMS patients. A previous meta-analysis³ reported three statistically significant effect sizes (global scores: g = 2.19; sleep latency: g = 1.75; and sleep efficiency: g = 1.08) when comparing the PSQI scores of FMS patients and healthy controls. Considering an alpha of 0.05 and power of 80%, the sample size needed to detect the reported mean differences are 14, 20, or 48 patients, respectively. Therefore, our sample size provided a 98% power to detect alterations in PSQI scores of FMS patients.

The descriptive statistics of the demographic and clinical characteristics of the study population were performed using the mean, standard deviation, or frequency. At first, we performed a univariate linear regression analysis between sleep quality (PSQI) and clinical/demographic and QST variables as independent variables to identify potential predictors. Then we performed a multivariate analysis to adjust the associated factors by age and gender. Due to the expected multicollinearity between pain, depression, and disease severity (measured by FIQR), we fit separate multiple regression models following a sequential modelling approach with predictors sets.³⁴ This is the regression strategy of choice when the aim is to determine the importance of a set of predictor variables (usually those measuring a related aspect) in a regression equation, one set at the time, under the assumption of high multicollinearity between sets (thus, a model including all important predictors is not recommended).³⁴ Before conducting each regression model, normality was inspected visually using histograms and statistically with the Shapiro Wilk test. We also verified the other assumptions needed for linear regression such as homoscedasticity, and normality of the residuals. Each regression model determined whether the variance explained by the specific set contributed significantly to the total variance in sleep quality, after controlling for age and gender (well-known confounders between chronic pain conditions and sleep disturbances).^35,36 We defined the following predictors sets: (1) Demographics (age and gender); (2) Mental health (BDI and VAS anxiety level); (3) pain intensity (BPI total score); (4) QST profile (CPM, Pain-60, TSPS); (5) HR profile (HR relative change); (6) FMS-specific variables (FIQR, disease duration, PROMIS-29 v2.0 fatigue). The variable selection for each set was based on theoretical relevance, pattern of correlation with the outcome variable and other potential predictor variables, and the assumptions underlying multiple regression analysis (normality and homoscedasticity). Finally, we used the same procedure for each subcomponent of the PSQI. We used the adjusted R-squared (aR²) as a measure of goodness-of-fit for our multiple models. Statistical significance was set at P < .05. All statistical analyses were performed using a standard software package (Stata, version 15.0; StataCorp).

Results

Sample characteristics

We included 65 patients, with 87.69% of the sample composed of female participants. The mean duration of the disease was 10.53 years, and the mean age was 48.1 years. Also, the mean FIQR score was 54.63 ± 19.33, suggesting mild to severe disease activity. Table 1 summarize all the included demographic and clinical characteristics of the patients with FMS (n = 65).

Table 1.

Demographic and clinical variables.

Measurements	Mean ± SD or %
Age	48.15 ± 11.29
Gender (female, %)	57 (87.69)
BMI	28.83 ± 7.97
Race (White, %)	48 (73.84)
HR baseline	84.89 ± 11.61
HR max	174.29 ± 7.90
HR relative change	1.97 ± 0.25
Duration of fibromyalgia	10.63 ± 8.2
Education level (Higha, %)	56 (90.76)
Number of diseases	7.2 ± 3.51
Smoking (yes, %)	14 (21.53)
Alcohol (yes, %)	27 (41.53)
Pain level (VAS)	5.88 ± 1.8
Anxiety level (VAS)	4.42 ± 2.72
Stress level (VAS)	5.38 ± 2.92
Depression level (VAS)	3.73 ± 2.87
BDI total score	16.33 ± 9.04
Quality of life score	68.58 ± 14.56
FIQR	54.63 ± 19.33
BPI total	5.2 ± 1.68
BPI interference	5.34 ± 2.15
BPI pain	5.15 ± 1.62
Use of antidepressants (yes, %)	31 (47.69)
PROMIS pain	2.13 ± 1.80
PROMIS anxiety	1.64 ± 1.48
PROMIS fatigue	2.38 ± 1.92

Abbreviations: BDI: Beck depression scale. BMI: body mass index. BPI: brief pain inventory. FIQR: fibromyalgia impact questionnaire. HR baseline: heart rate measure before the patient walked on the treadmill. HR max: 60%–70% of age-predicted maximal heart rate after the patient walked on the treadmill. HR relative change: percentage increase of the HR max compared to the baseline HR (HR max/HR baseline). Number of diseases: number of diseases that the patient has besides Fibromyalgia. PROMIS: patient-reported outcomes measurement information system. PSQI: Pittsburgh sleep quality index. VAS: visual analog scale.

^aHigh education is defined as having a graduate school degree.

Sleep quality

Table 2 shows the global and subcomponent scale scores for the PSQI. The mean ± SD global PSQI score was 12.78 ± 4.39 (range 3–19), with 62 patients (95.39%) classified as poor sleepers (global PSQI > 5). The most affected subcomponents were sleep disturbance, use of sleep medications, and subjective sleep quality.

Table 2.

Sleep quality total and subcomponents scores (n = 65).

Measurements	Mean ± SD
Subjective sleep quality	1.93 ± 0.86
Sleep latency	1.92 ± 0.98
Sleep duration	1.52 ± 1.38
Habitual sleep efficiency	1.49 ± 1.31
Sleep disturbance	2.12 ± 0.54
Use of sleep medications	1.96 ± 1.34
Daytime dysfunction	1.81 ± 0.84
Total score	12.78 ± 4.39

Univariate analyses

Our results from the univariate analyses are presented in Table 3. As shown, BDI, FIQR, BPI interference, BPI total, BPI pain, VAS depression, and PROMIS pain are positively correlated to the PSQI index, with P values less than .05 for all the variables mentioned. The variables with the highest beta coefficients were BPI total score and its subscales (pain interference and intensity).

Table 3.

Univariate models for overall quality of sleep (PSQI total score).

Variable	ß-coefficient	Unadjusted P value	95% confidence interval
BDI (0–63)	0.207	<.001	0.096; 0.317
FIQR (0–90)	0.107	<.001	0.057; 0.158
BPI interference (0–10)	0.773	.002	0.297; 1.249
BPI total (0–10)	0.948	.003	0.338; 1.559
BPI pain (0–10)	0.766	.022	0.112; 1.420
VAS depression (0–10)	0.397	.036	0.026; 0.768
PROMIS pain (0–10)	0.601	.046	0.009; 1.193
Bilateral pain	−2.086	.055	−4.217; 0.045
VAS pain intensity (0–10)	0.525	.100	−0.103; 1.154
Gender (female)	−2.668	.108	−5.941; 0.603
Duration of fibromyalgia (months)	0.103	.124	−0.029; 0.235
PROMIS fatigue (0–10)	0.411	.151	−0.154; 0.977
BMI	0.091	.188	−0.045; 0.227
Race (width, %)	−1.486	.233	−3.955; 0.982
Quality of life (0–112)	−0.043	.247	−0.119; 0.0312
Level of education (Higha, %)	1.684	.289	−1.464; 4.833
Antiepileptics	1.298	.302	−1.195; 3.791
Use of alcohol	1.128	.311	−1.080; 3.337
TSPS	−0.335	.312	−0.993; 0.321
Antidepressants	−1.129	.314	−3.354; 1.09656
VAS anxiety (0–10)	0.199	.327	−0.203; 0.602
VAS stress (0–10)	0.180	.347	−0.200; 0.560
PROMIS anxiety (0–10)	0.273	.463	−0.466; 1.012
Pain-60	−0.084	.814	−0.801; 0.632
CPM response	−0.044	.888	−0.680; 0.591
Age (years)	−0.005	.909	−0.103; 0.092
Baseline Heart Rate	0.004	.920	−0.090; 0.100
HR relative change	0.2727273	.899	−0.276; 5.456

Abbreviations: BDI: Beck depression scale. BMI: body mass index. BPI: brief pain inventory. FIQR: fibromyalgia impact questionnaire. HR baseline: heart rate measure before the patient walked on the treadmill. HR max: 60%–70% of age-predicted maximal heart rate after the patient walked on the treadmill. HR relative change: percentage increase of the HR max compared to the baseline HR (HR max/HR baseline). Number of diseases: number of diseases that the patient has besides Fibromyalgia. PROMIS: patient-reported outcomes measurement information system. PSQI: Pittsburgh sleep quality index. VAS: visual analog scale.

^aHigh education is defined as having a graduate school degree.

Multivariate analyses

The results of the multiple regression models are shown in Table 4. The first model that tested the contribution of demographic variables (age and gender) to sleep quality was not statistically significant (F_{2, 62}= 1.36, aR² = 0.01, P = .26). The models that accounted for the QST profile (set 4) and heart rate profile (set 5) variables were not significant statistically significant either (P > .05). The set of mental health variables (set 2) was statistically significant (F_4,61 = 3.87, aR² = 0.15, P = .007), with BDI having a significant and positive correlation with sleep quality (the higher the depression score, the poorer the sleep quality). The third set with a pain variable was statistically significant as well (F_3,60 = 3.38, aR² = 0.10, P = .02), with BPI total score having a positive association (higher pain intensity, poorer sleep quality). Finally, the set 6 (FMS-specific variables) was the 1 that more explained the variance of the PSQI index (F_5,59 = 6.65, aR² = 0.31, P < .0001), with FIQR and PROMIS fatigue having a positive statistically significant association (higher the FIQR score and fatigue, poorer the sleep quality).

Table 4.

Multivariate models for overall quality of sleep (PSQI total score).

Variable	Demographics		Mental health		Pain		QST profile		Heart rate profile		FMS-specific variables
	(Set 1)		(Set 2)		(Set 3)		(Set 4)		(Set 5)		(Set 6)
	${\mathrm{R}}_{=}^2$ 0.01		R2 = 0.15		R2 = 0.10		R2 = 0.19		R2 = 0.04		R2 = 0.31
	ß	p	ß	p	ß	p	ß	p	ß	p	ß	p
Age	0.162	0.748	0.003	0.948	0.011	0.829	0.036	0.493	0.015	0.765	0.004	0.936
Gender	−2.814	0.105	1.935	0.234	−1.58	0.360	−5.11	0.011	−2.86	0.107	−2.63	0.076
BDI total score			0.208	0.001
VAS anxiety			0.097	0.636
BPI total score					0.87	0.009
CPM							0.098	0.756
Pain-60							0.191	0.614
TSPS							0.465	0.193
HR relative change									0.356	0.871
FIQR											0.067	0.020
Duration of fibro (months)											0.134	0.052
PROMIS Fatigue											1.71	0.027

Finally, we performed the multivariate analysis using each PSQI component as a dependent variable (models are shown in the Supplementary Material). We found significant predictors for subcomponent 1 (subjective sleep quality), 5 (sleep disturbance), and 7 (daytime dysfunction). For subcomponent 1, only the set 6 (FMS-specific variables) was predictive (F_5,59 = 3.98, aR² 0.19, P = .004), where PROMIS fatigue was a positive predictor. In subcomponent 5, the set that best explained the variance was the set 5 (HR profile) (F_3,61 = 3.00, aR²= 0.09, P = .03), where the relative change of HR was the main predictor. Moreover, in the subcomponent 7, two sets were predictive the set 2 (F_4,60 =2.79, aR²= 0.10, P = .03) and 6 (F_5,59 =5.03, aR²= 0.24, P = .007), with the FMS-specific variables (PROMIS fatigue from set 6) explaining more variance than the mental health variables (depression from set 2). Finally, for the subcomponents 2 (sleep latency), 3 (sleep duration), 4 (habitual sleep efficiency) and 6 (use of sleep medications), we found no associated factors. Interestingly, the QST variables (set 4) were not predictive of PSQI total score nor its subcomponents.

Discussion

Our results support the fact that FMS patients have reduced sleep quality. Specifically, sleep disturbance, use of sleep medications, and subjective sleep quality were the subdomains with the lowest scores in FMS patients. These values were lower than those reported as the score cutoff of the PSQI in general population (score of four in general population compared to 12.78 in our study).²⁹ In our sample, the prevalence of poor sleep quality was 95%. Furthermore, we identified clinical predictors of sleep quality and its subcomponents. We found that poor overall sleep quality in FMS patients is highly associated with symptom severity (high FIQR score and PROMIS fatigue), pain severity (high BPI total score), and higher depression levels (BDI scores). These clinical factors also influenced 2 PSQI subcomponents; subjective sleep quality (subcomponent 1) was associated with higher fatigue scores, and daytime dysfunction (subcomponent 7) was influenced by higher depression and fatigue levels. Interestingly, the most altered domain (sleep disturbance, subcomponent 5) was only associated with HR relative changes (an indirect measure of physical conditioning), suggesting that poor fitness status is an independent predictor of sleep disturbance in FMS. The rest of the PSQI subcomponents (sleep latency, duration, efficiency, and use of sleep medications) remained unexplained by our clinical and sensorial variables, requiring further exploration. Finally, despite the growing evidence indicating that central sensitization is a determinant of sleep quality in chronic pain conditions,⁸ we found no association between the QST profile (Pain-60, TSPS, and CPM) and sleep quality nor its subcomponents. These results could suggest that sleep dysfunction in FMS is less correlated to the mechanisms of sensory dysfunction, hinting that the affective and physical conditioning components might explain the sleep dysfunction in population.

Our findings of poor sleep quality in FMS patients were aligned with previous literature assessing polysomnography recordings. For instance, several studies have shown that patients with FMS have reduced slow-wave sleep,^37–39 1 of the stages of non-REM sleep, which is essential for restoration, influencing homeostatic parameters like heart rate, blood pressure, cerebral glucose consumption, and cortisol levels. Studies using self-reported scales are also in line with these findings. Osorio et al. found a global PSQI score 3 times higher in patients with FM compared to controls. Similarly, the prevalence of poor sleep quality (characterized as a PSQI global score bigger than five) in our sample was 95%. Moreover, we found that “sleep disturbances” (fragmented sleep) was the most affected subdomain, which was also reported by Munguía-Izquierdo et al. who reported analogous findings.⁷ Although the nature of the events that disrupted sleep in FMS is unclear, we hypothesize that pain episodes during night and thoughts rumination (associated with depressive symptoms) might explain the constant sleep interruptions in these patients.⁴⁰

We found in the multivariate analysis that poor sleep quality in patients is related to the severity of fibromyalgia-related symptoms. We expected this association not only because the FIQR measures the severity of the disease but also because it encompasses a variety of FMS symptoms, such as pain, depression, anxiety, tenderness, and functionality. This also agrees with Munguía-Izquierdo et al., who found a positive relationship between FIQR and poor sleep quality.⁷ Similar results were found in other chronic pain conditions such as osteoarthritis.⁹ Thus, disease activity and severity critically influenced sleep quality in a transdiagnosis manner. In our study we found that fatigue level was the most predictive factor from the FMS-specific symptoms. Hence, we suggest exploring specific clinical expressions to gain a more granular understanding of sleep quality in chronic conditions instead of only using overall symptomatology.

Furthermore, we described a relationship between the PSQI index, Beck Depression Inventory (BDI), and Brief Pain Inventory (BPI). Our findings are aligned with current evidence, although the real “causal direction” is still unclear. Bigatti, S. M et al. examined 492 patients with FMS during one year of assessment, and sleep quality (assessed by the PSQI scale) was a significant predictor for pain at the end of the study, meaning that greater sleep disturbance in baseline predicts greater pain at one year.⁴¹ Moreover, several studies have shown that sleep plays an essential role in pain processing—as sleep disruption can reduce activity in descending inhibitory pain pathways, which could explain the direction of the relationship between pain and poor quality of sleep. Studies also suggest a bidirectional relationship between sleep disturbances, and psychiatric problems such as anxiety and depression. Our exploratory findings give support to the current evidence that the more depressed the FMS patients are, poorer their sleep quality. Future longitudinal studies including these three variables are needed to control for reverse causality.

Moreover, we found no associations between central sensitization and sleep quality. Sleep deprivation and less sleep hours have already been associated with greater pain and pain sensitization in chronic pathologies.⁴² Affleck et al. showed that the same happens on fibromyalgia patients.⁴³ Also, TSPS have been found to be increased in women who have insomnia.⁴⁴ However, our exploratory results indicates that the components of QST (eg, sensory profile) are not correlated either with the PSQI index or the subcomponents. We hypothesize that self-reported sleep quality is more influenced by emotion-related processes compared to pain inhibition processes; thus, highlighting the urgent need of managing mental health of FMS patients together with pain symptoms. An alternative explanation is that our sample is not representing the whole FMS spectrum, since our inclusion criteria are narrow and severe patients (likely with more central sensitization) are not included. Therefore, more studies are needed, since a better understanding of the interaction between sleep and central pain mechanisms could improve the treatment options for FMS patients.

Interestingly, we found that sleep disturbance (subcomponent 5) was inversely correlated to the HR relative change—meaning that the higher the HR value, less sleep disturbance. Although this is a secondary result, it might suggest that better physical conditioning is a protective factor against poor sleep quality. A meta-analysis of 14 studies that analyzed the effects of exercise in sleep quality (using the PSQI scale) and the pooling results had a statistically significant effect on sleep quality (a larger reduction in PSQI score) compared to the control with a pooled mean difference of −2.19 (95% CI −2.96 to −1.41, P < .00001). These results suggest that improving sleep quality in patients with FMS could potentially improve their pain and fatigue.

The clinical implications of the relationship between depression, fatigue, pain severity, and poor quality of sleep in FMS, highlights the need for providing FMS patients with multidimensional treatment, not only focusing on pain reduction but also attempting management approaches targeting those frequent comorbidities. For instance, sleep hygiene interventions in FMS improved additionally fatigue and pain,⁴⁵ and exercise interventions to improve fatigue had a small but beneficial effect on sleep.⁴⁶ Another potential alternative could include digital and telehealth interventions, as smartphone-based sleep tracking or wearable devices to measure sleep patterns and behaviors. In this manner, taking advantage of the bidirectional relationship and treating those disorders in parallel, ideally in an integrated management program, might contribute to a more comprehensive improvement in the wellbeing of FMS patients.

This study presents strengths and limitations that should be noted. The results found here are in line with most of the evidence regarding quality of sleep and FMS. Also, to the best of our knowledge this is the first article that aims to evaluate the main predictors of each PSQI subcomponent and including QST variables. However, sleep variables were self-reported, which may lead to some level of recall bias and are not as precise as objective measurements—in despite that the chosen dependent variable has been previously validated on FMS patients ⁴⁷ and is correlated to polysomnographic findings.⁴⁸ Also, our inclusion criteria include only FMS patients that could tolerate aerobic exercise; thus, severe patients could have potentially been excluded from the study. Together with the small sample size, our study results cannot be generalized to a broader population. Therefore, the findings shown here are exploratory and further studies are necessary to validate them. Finally, as this study has an observational cross-sectional design, we cannot determine the direction of the relationship found, which implies that we cannot determine whether poor sleep quality is a consequence of the disease or part of the pathophysiological process that composes it. However, our results are important for generating hypotheses and help with the design of future well-powered studies.

Conclusion

In conclusion, our results confirm that FMS patients have moderate to severe sleep disturbances and identify that FMS symptom severity (represented by FIQR and PROMIS fatigue), pain, and depression are significant predictors of poor overall sleep quality in this population. On the contrary, central sensitization was not associated with PSQI total scores or its subcomponents. The clinical factors only predicted two subcomponents of PSQI (subjective quality and daytime dysfunction), underscoring the multifactorial etiology of sleep alterations in FMS patients, which requires further exploration. Interestingly, heart rate changes independently predicted the sleep disturbance subdomain (the most affected one in our sample), suggesting an essential role of physical conditioning in modulating sleep quality in FMS patients. This highlights that multidimensional treatments targeting depression (eg, remote psychotherapy) and improving physical activity (eg, home-based exercise programs) need to be integrated into the management approach of FMS patients, which potentially can enhance the quality of sleep and symptom severity.

Supplementary material

Supplementary material is available at Pain Medicine online.

Funding

This work is supported by National Institutes of Health (NIH) grant R01 AT009491-01A1.

Conflicts of interest: The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.