A randomized controlled trial was performed. We used a between-subjects experimental design with repeated measures (Montero & León, 2007). The study was guided by the CONSORT statement (Moher et al., 2010).

Women with FM aged between 24 and 62 years were recruited from the Rheumatology Service and Pain Unit of Virgen de las Nieves University Hospital and from AGRAFIM (a local FM association), both in Granada, Spain. Selected patients were referred to the Clinical Psychology Unit of our university, where the clinical assessment and treatment took place. Patients were eligible for participation if they met the following inclusion criteria: (1) having been diagnosed with FM according to the American College of Rheumatology criteria (Wolfe et al., 1990); (2) having significant self-reported insomnia according to well-established diagnostic criteria, i.e., when the chief complaint is nonrestorative sleep or difficulty in initiating or maintaining sleep and the complaint continues for at least one month (the disturbance must occur at least three times a week for a month) (American Psychiatric Association, 2000); and (3) following a stable medication regime over the past month. Exclusion criteria were: (1) being pregnant; (2) having major medical conditions (e.g., inflammatory rheumatic diseases, uncontrolled endocrine disorders, cancer) including a clinical history of significant head injury or neurological disorder; (3) having a severe psychopathology such as major depression with suicide ideation, schizophrenia, personality disorder (American Psychiatric Association, 2000); (4) suffering from other sleep disorders that better explained insomnia (e.g., having an apnea-hypopnea index of 15 or more, having a periodic limb movement-related arousal index of 15 or more, or other previously diagnosed significant sleep disorders); (5) having severe dependence on hypnotic drugs, suggested by the use of a hypnotic agent in a higher than recommended dosage or repeated episodes of rebound insomnia on withdrawal; and (6) being enrolled in another physical or psychological treatment during the study period.

One hundred and seven candidates (FM women) who had received health care in the Rheumatology Service and Pain Unit in the past two years and current active members of the FM association were screened for eligibility via a brief telephone survey conducted by a psychologist. Eligible participants (52 FM women) underwent a medical examination by a rheumatologist and a psychological assessment by a psychologist in two sessions. The first session focused on a physical examination and collection of data on socio-demographic characteristics, onset and time course of symptoms, history of current insomnia and FM, past medical history, drug intake and psychological status. Three participants were ruled out during the interview because they did not meet the eligibility criteria for the study. After the interview, the remaining 49 FM women were given a set of questionnaires about pain, sleep, emotional distress and health status to be completed at home in one week. They received detailed instructions to complete the questionnaires and a psychologist was available over the telephone to answer any questions about completing the self-reported measures. The second session was devoted to completing additional data about clinical and psychological features and collecting the self-reported measures. FM participants who met the inclusion criteria and had undergone the evaluation sessions (n=49) were enrolled in a single night of ambulatory PSG within the next two weeks before therapy. Considering the inclusion criteria in the PSG analysis, 39 eligible participants were blindly randomized (1:1) to either the CBT-P (n=19) or the CBT-C (n=20) group using a computerized number generator. Of the 39 randomized participants, 15 in the CBT-P group and 17 in the CBT-C group completed the treatment and post-therapy assessment and were included in the analyses. A per protocol analysis was carried out in the present study, thus analysis was restricted to participants who finished the study. The participant flow diagram can be seen in Fig. 1.

Fig. 1.

Flow diagram of participants through the phases of the clinical trial.

Note. CBT-C=Cognitive-behavioral therapy for insomnia and pain, CBT-P=Cognitive-behavioral therapy for pain, PSG=Polysomnography.

(0.3MB).

Baseline and post-treatment assessments were performed between one and two weeks before and after therapy, respectively. Primary outcomes measures included perceived sleep quality measured by the Pittsburgh Sleep Quality Index (PSQI), and objective SE obtained by PSG. The remaining PSG variables were considered as secondary outcome measures.

This study was approved by the Ethics Committee of the University of Granada. All participants signed informed consent before engaging in the research protocol.

Perceived sleep quality was measured by PSQI. This instrument comprises 19 items that assess seven dimensions of sleep quality. Scores of these seven components are added to obtain a global score ranging from 0 to 21. Higher scores in this global measure indicate worse sleep quality. The PSQI has shown acceptable internal consistency (ranging between .67 and .81), sensitivity and specificity in the Spanish general population (Royuela & Macías, 1997).

Eligible participants were enrolled in a full-night PSG study at home that was recorded and digitized using the SomnoScreen TM plus (SomnoMedics, Germany). Gold electrodes were placed according to the International 10-20 system for electroencephalogram (EEG) recording. PSG signals included the following EEG channels: F3-A2, F4-A1, C3-A2, C4-A1, O1-A2, O2-A1, which were sampled at 128Hz; two electrooculogram channels; chin and bilateral anterior tibialis surface electromyography; two electrocardiographic leads; nasal and oral airflow (nasal pressure and thermistor); respiratory effort of thorax and abdomen (inductance plethysmography), body position; and finger oximetry. Two trained researchers blinded to group allocation hooked up the portable device and sensors used for PSG recording. The start and end of recordings were programmed according to each patient's sleep schedule to ensure higher ecological validity.

Sleep stages were blindly scored offline with DOMINO Software (DOMINO 2.2.0., supplied with the SomnoScreen TM plus) by a researcher trained in PSG scoring (G.P.). The criteria of Rechtschaffen and Kales (1968) were followed, based on 30 s epochs for hypnograms. We analyzed key sleep architectural variables derived from the PSG analyses included total time in bed (TIB, the period of time between bedtime and awakening in the morning), TST, which is the period of time between sleep onset and sleep offset, excluding all wake stages, and SE, that is TST/TIB. The following variables were also analyzed and expressed as a percentage of TST: time spent in Stages 1, 2, 3 and 4 (S1, S2, S3 and S4, respectively), non-rapid eye movement sleep (NREM), rapid eye movement sleep (REM). In addition, wake time was defined as the percentage of time spent awake from bedtime to final wake up time. We also considered the following variables in minutes: REML and slow-wave sleep latency (SWSL).

Starting within one or two weeks after the baseline PSG and questionnaire assessment, participants in both therapy groups (CBT-P and CBT-C) received nine weekly sessions of therapy that lasted approximately 90minutes each (the contents of the CBT are shown on Table 1). The therapy was delivered by three female postdoctoral researchers in clinical psychology with expertise in CBT tailored to patients suffering from sleep disturbances and chronic pain conditions. E.M. was in charge of unique group in each treatment condition (i.e., CBT-P and CBT-C), the same applies for the other two therapists (i.e., M.P.M. and A.I.S.). This scheme of delivering therapy allowed us to control inherent personal features of each therapist when applying therapy. Treatment integrity was ensured through the use of a specific manual for each therapy modality that included detailed information about each session. These guides were created, designed and piloted by the three aforementioned researchers (i.e., E.M., M.P.M. and A.I.S.) under the umbrella of a research project with public funding. In addition, there were regular clinical meetings between the therapists and the research group, and video recordings to monitor the implementation of the intervention. Sessions were conducted in groups of five to seven participants. Self-reported measures were administered by a psychologist who was blinded to group assignment. FM women were instructed to avoid making changes in their usual medication regimen or enrolling in other non-pharmacological interventions during the study.

CBT-P program. This therapy modality focused on psychoeducation on pain physiology, physiological deactivation, goal setting, increasing activity level and social skills, problem-solving and coping with dysfunctional beliefs and rumination related to pain, based on the pain-related fear and avoidance model (Vlaeyen, Crombez, & Linton, 2016).

CBT-C program. This therapy included many components of CBT-P (i.e., relaxation training, development of social skills, problem-solving and retraining the cognitive processes that maintain chronic pain). In addition, based on the strong evidence supporting the application of CBT models to insomnia in patients with chronic pain (Tang, 2018), this intervention was designed to target the factors that perpetuate sleep problems, considering Harvey's cognitive model of insomnia (Harvey, 2002) and the evidence-based practice parameters for the psychological treatment of the insomnia (Morin & Benca, 2017).

Analyses were performed using the SPSS-20.0 statistical package (SPSS, Inc., Chicago, IL, USA). Probabilities less than .05 were used as the level of significance. Data were summarized as the mean, standard deviation and percentages. Differences between both treatment groups in socio-demographic and clinical characteristics and outcome measures at baseline were compared using t-tests or Chi-square statistics. The effects of each treatment condition were analyzed by examining changes in PSG parameters and perceived sleep quality using 2 (Group; CBT-P vs. CBT-C)×2 (Time; Pre-treatment vs. Post-treatment) ANOVAs. Greenhouse-Geiser correction was computed. Additionally, unpaired and paired two-sample Student's t tests were computed. Effect sizes were calculated for statistical significance using partial η2 and Cohen's d. According to Cohen's guidelines (Cohen, 1998), d=0.20 is a small effect, 0.50 is a medium effect and 0.80 is a large effect; and η2=0.01 is a small effect, 0.06 is a medium effect, and 0.14 is a large effect.

Given an anticipated average effect size of Cohen's d=1.26 in sleep parameters (see review by Martínez, Miró, & Sánchez, 2014), a p-value<.05, and a desired statistical power of .90, the minimum required sample size for the study was 30 patients (n=15 subjects per group).

At baseline, the CBT-P and CBT-C groups were statistically similar in all the socio-demographic and clinical variables measured (all p's>.16, see Table 2). The mean age of FM participants was 49.97 years (SD=7.91) and 93.8% of participants in both groups were married or cohabited with a partner. The percentage of participants with non-compulsory secondary or higher education was 71.9%. More than half of participants (58.8%) had an inactive work situation (i.e., unemployed, temporary disability or permanent disability). Concerning clinical aspects, the mean duration of FM symptoms was 11.34 years (SD=7.52) in the total sample, and approximately 50% of participants were taking a stable dose of antidepressants, hypnotics and analgesics. The rest of patient did not take any antidepressants or hypnotics and analgesics were rarely consumed in this group.

As can be seen on Table 3, ANOVA for the self-reported measure indicated effects close to significance of Group×Time in sleep quality (p=.08). The CBT-C group showed a small but significant decrease in the PSQI score from 15.56±4.05 to 13.18±4.60 (p<.05).

At baseline, the CBT-P and CBT-C groups did not show any statistically significant differences regarding PSG variables (all p's>.05). As shown in Table 3, ANOVAs for objective sleep variables showed a significant effect of Group in TIB, TST and percentage of S4. Significant effect of Time was found in REML and close to significant in TST, SE, percentage of REM and S2. Significant effect of Group×Time was observed in TST and percentage of NREM. In the case of TIB and S2, that kind of effect was marginally significant.

Participants in CBT-P condition showed a greater and moderate to large significant increase in TIB and TST from pre- to post-treatment according Cohen's d (-0.75 and -0.80, respectively). In addition, within-group comparisons showed significant changes through these parameters (p<.05; Table 3). In contrast, participants in the CBT-C group did not show any significant changes in either measure. With regard to SE, patients in the CBT-C condition showed statistically significant improvement in SE according to within-subject comparisons (i.e., SE increased from 86.11±7.58% to 91.14±4.70%; p<.01).

As regards the variables related to the hypnogram structure, we did not observe any significant effect according ANOVA in percentage of wake time in participants. However, within-group comparisons revealed a significant reduction in this variable through CBT-C group (i.e., wake time decreased from 13.80±7.54% to 8.86±4.70%; p<.01).

Significant effects of Group×Time were found in percentage of NREM and were related to a decrease from pre- to post-treatment in the CBT-P group (p<.05).

Light sleep significantly decreased from pre- to post-treatment in the CBT-P group but not in the CBT-C group. This change was due to a moderate reduction of S2 from 51.26±10.48% to 43.94±9.58% (p<.05), although ANOVA only showed significant effect close to significance of Time and Group×Time in this parameter.

Concerning deep sleep a significant effect of Group was found in S4, but we only observed a marginally significant within-subject improvement in CBT-C participants in this restorative sleep stage (p=.08).

As shown on Table 3, scarce significant effects were found in the ANOVA of SWSL and REML. Thus, only an effect of Time was observed in the latter. However, within-subject comparisons showed that latencies in SWS and REM onset exhibited a moderate reduction in the CBT-C condition (SWSL: 52.10±19.28% to 37.02±18.11%; p<.05 and REML: 156.67±74.61% to 120.28±69.31; p<.05). This improvement in sleep architecture was also observed in the CBT-P group, although it did not reach statistical significance.