# Heart rate variability responses to cognitive stress in fibromyalgia are characterised by inadequate autonomous system stress responses: a clinical trial

We recruited 51 female FM patients aged 18 to 65 years, who fulfilled the American College of Rheumatology 1990 diagnostic criteria for FM1, through Helsinki University Hospital (HUH) outpatient clinics, City of Vantaa Health Centre, and the private clinic of one of the authors (RM). Thirty-one healthy, age- and gender-matched controls were also recruited. We recruited the maximum number of patients and matched volunteers available during the funding period of the study and the resulting number was similar to previous studies11,12,13,14. The exclusion criteria were: diabetes, heart disease, uncontrolled hypertension, peripheral atherosclerotic disease, neurological, neuromuscular or muscle disease, severe psychiatric disorders, continuous use of beta-blockers, beta-agonists, or statins, any musculoskeletal condition that would prevent participation in cycle ergometry (which was to be conducted at a later stage), and poor Finnish language skills that would affect the ability to answer the questionnaires. The patient selection process has been described in detail earlier16.

Between November 2015 and June 2018, the subjects visited the HUH Pain Clinic where the diagnosis of FM was confirmed for the patients and excluded for the controls through interview and clinical examination by the same author (TZ). The questionnaires used were completed by the subjects before the measurement protocol described below. We registered our study retrospectively to ClinicalTrials.gov (NCT03300635) on 03/10/2017.

Table of Contents

### Background data and questionnaires

We collected data on the subject’s medical background and lifestyle factors. Leisure time physical activity was rated with a four-point Likert scale for frequency (from “None” to “Several times per week”) and for intensity (from “Walking” to “Brisk running”), and subjective physical fitness was rated with a three-point Likert scale (“Worse than average”, “Average”, or “Better than average”). From these, we summed a physical activity score from three to seven. Sleep quality was assessed with a dichotomous question (yes/no) about subjective sleep disturbance and with a four-point Likert scale for waking during sleep (from “Not usually” to “Five or more times per night”). Body mass index (BMI) was calculated from height and weight, and smoking status (smoker/non-smoker) was recorded.

The Fibromyalgia Impact Questionnaire (FIQ) assesses the severity of FM symptoms and their impact on daily functioning17. We used the validated Finnish language version of the FIQ, which consists of ten items. These include ability to conduct daily activities; number of days of wellbeing; number of sick leave days; and impact of FM on ability to work during the previous week. The other six items assess the impact of FM on pain intensity, fatigue, feeling refreshed in the mornings, stiffness, anxiety or feeling of tension, and depression or sadness, using visual analogue scales (VAS). The FIQ score ranges from 0 to 100, with higher scores indicating greater FM impact18.

The Perceived Stress Scale (PSS) assesses the amount of stress experienced over the previous month, scored from 0 to 40, with higher scores indicating greater stress and less feeling of control19. The Pain Catastrophizing Scale (PCS) measures pain-related catastrophizing on a scale from 0 (no catastrophizing) to 52 (most catastrophizing). The State-Trait Anxiety Inventory (STAI) consists of two sub-inventories evaluating immediate feelings of anxiety (state anxiety, STAI-A) and stable, long-term susceptibility to anxiety (trait anxiety, STAI-B). Both STAI-A and STAI-B are scored from 20 to 80, with higher scores indicating more anxiety or anxiety traits respectively.

### Measurement protocol

We modelled our protocol after Thieme et al.20, but used only one form of cognitive stress repeatedly, instead of different stressors, to allow evaluation of the effect of repetitive stress. The measurement consisted of five four-minute phases of alternating relaxation and cognitive stress: first relaxation, first stress, second relaxation, second stress, and final relaxation phase.

At the start of the measurement, between each phase, and at the end of the measurement, subjects were asked to rate their subjective stress and pain intensities on a 0 to 10 numeric rating scale (NRS: 0 = no stress or pain; 10 = worst stress or pain imaginable). During the relaxation phases, subjects were asked to relax as best they could while listening to calm classical music or in silence, based on individual preference. During both stress phases, participants were played recordings of 14 series of ten numbers between zero and nine and asked to mentally sum the numbers and state the result. Subjects were told whether their answers were correct or false but were unaware that in four cases they were told that their answer was incorrect, regardless of the actual answer. Throughout the stress phases, background noise of 60 dB white noise was applied.

During the measurements, subjects sat comfortably in a chair in a small room with only one subject and one researcher present. Room temperature was between 20 and 24 °C and humidity between 40 and 60%. As the protocol included speaking, breathing rate was spontaneous and not controlled.

### Heart rate variability measurement and signal processing

Heart rate (HR) as beats per minute was recorded with a heart rate monitor strap (Polar T31, Polar Electro, Finland) placed around the chest at the level of the lowest third of the sternum.

Surface electromyography (sEMG) readings were also recorded bilaterally from the trapezius, biceps, and the erector spinae muscles at L4 level, as published previously16. To complement the HR data, we extracted electrocardiogram (ECG) signals from the sEMG recording. We used Matlab R2017b for signal processing. The raw sEMG readings were first detrended to mean zero amplitude. The signals’ power spectra were then inspected visually to identify sharp peaks of alternating current (AC) noise caused by various electronic devices, most commonly at 50 Hz and multiples thereof. Noise peaks were removed using interpolation to flatten the power of the noisy frequency spectrum range of 1 Hz to the mean power of frequency range 0.5 Hz above and below the noisy range.

The best ECG signal was obtained from the right erector spinae (FM n = 24; healthy controls (HC) n = 20) or the left trapezius (FM n = 27; HC n = 11). ECG signals were visually inspected to determine which channel to use for optimum results. A 3-min sample starting at 20 s of each recording was used. From these channels, the QRS complexes were located by filtering the EMG signal with a third-order Butterworth bandpass filter to the 10 to 40 Hz range containing most of the ECG signal. The ECG signal was isolated with principal component analysis (PCA).

HRV analysis was performed using KUBIOS HRV Premium 3.2.0 software. As measures of HRV, we used heart rate (HR), mean interval between heart beats (measured between successive R peaks of the QRS complex in ms [RR]) (RRmean), root mean squared interval differences of successive beats (RMSSD), and the standard deviation of intervals between normal heart beats (SDNN). These measures are usable in ultra-short ECG samples, such as the three-minute samples in our case9.

For the baseline values, we used the mean HR, RRmean, RMSSD, and SDNN for the first 30 s of the recording.

### Statistical methods

For comparison between two groups, the Mann–Whitney U-test was used for continuous variables and the χ2-test for categorical variables. Comparisons between three groups or clusters were done with analysis of variance (ANOVA) with post-hoc testing using the Tukey test.

We used linear modelling with generalized least squares (GLS) to test how HRV variables were affected by the FM status, the stress-relaxation protocol, and whether the reactions of FM patients and controls differed during the protocol, i.e. group-time interaction. To evaluate reactivity, the base model also needed to include the baseline value of the outcomes. We also tested a second model adjusting for BMI, smoking status, and leisure time physical activity (LTPA). To improve interpretability, LTPA was dichotomised into physically active (physical activity score of seven or more) and inactive. Formulae for HR are shown below:

$$HR \, = \, Group \, \times \, Time \, + \, Baseline,$$

$$HR \, = \, Group \, \times \, Time \, + \, Baseline \, + \, BMI \, + \, Smoking \, + LTPA.$$

Significant group differences in HRV or HRV reactivity were evaluated for suitability as clustering variables for cluster analysis within the FM subgroup post hoc. We conducted k-means clustering on z-transformed variables, with the number of variables and clusters restricted to n = 10 × d × k (d = number of clustering variables, k = number of clusters)21.

Statistical analyses were done with R version 4.0.0 (The R Foundation for Statistical Computing 2020). GLS was done with the nlme R package version 3.1-14722.

### Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki and the study protocol was approved by the Ethics Committee of the Helsinki and Uusimaa Hospital District. All subjects provided written informed consent.

Source link