Autonomic and muscular responses and recovery to one-hour laboratory mental stress in healthy subjects

Forty-four healthy subjects participated in the study (Table 1). The participants were recruited as controls for a group of pain patients with a female predominance, and therefore comprised thirty-five women and nine men. They were recruited from public institutions and private companies in Trondheim. Subjects were excluded if they fulfilled all of the three following criteria: (1) headache or musculoskeletal pain for more than one day per month, and (2) had visited a physician, and (3) took medication for the complaint (all three conditions to be fulfilled). In addition, subjects considering their headache or pain to be more than "unpleasant" (i.e. a higher degree of pain) were excluded if (1) they experienced the pain more than one day per month, or (2) had visited a physician for the pain, or (3) took medication for the pain (i.e. any of the three conditions fulfilled). No participants took drugs with a possible interaction with neural, vascular or muscular function (e.g. antiepileptics, β-blockers, and antidepressants).

All subjects answered a questionnaire on biographical data (marital status, weight, medication, and stimulants), exercise habits, and the neuroticism index of the Eysenck Personality Questionnaire (EPQ-N)[16]. The questionnaire further included an index of symptoms concerning the autonomic nervous system ("autonomic symptom index"). For this purpose a subset of ten questions were chosen (No. 26–35) from the Composite Autonomic Symptom Profile [17]. The questions assessed different domains of autonomic symptoms (orthostatic, sudomotor, gastrointestinal, visual, vasomotor, reflex syncope). Sub-indexing different autonomic domains was not done due to the limited number of questions. The answers were graded. A serious extent of a symptom was given a higher value than a less serious. E.g. the answer to the questions: "In the last year, to what extent have you been in a cold sweat?", were graded as: "have not had" (value 0), mild (value 1), moderate (value 2), severe (value 3). The highest possible sum score was 30.

All potential participants went through a short telephone interview to exclude those not fulfilling inclusion criteria. Subjects not excluded by the initial screening received the questionnaire by post within two weeks of the test day. On the morning of the test day the subjects first went through a short interview controlling the answers from the questionnaire. Afterwards venous blood was sampled from the right cubital fossa. Subjects were instructed to empty their bladder before starting the test. Brassieres were removed and subjects wore only a light shirt on the upper part of the body. The laboratory temperature was regulated to 24.5 ± 1.0°C and was recorded every ten minutes during the experiment.

The subject was seated in an office chair with the lower arms resting on the table top before, during and after the test. Subjects got acquainted to the work-task by performing a mini-trial with instructions before the test started. The mini-trial was performed without introducing stress-imposing feedback on reaction time and was used to determine the subjects' habitual, non-stressed reaction time. Short maximal voluntary contractions were performed on each pair of muscles twice (frontalis muscle – raising eyebrows, temporalis – clenching teeth, neck – pushing head back against resistance, trapezius – pushing extended arms upwards against resistance at 45° angle out from the body). The maximal contractions were carried out in order to normalize the muscle activity during test to a percent of maximal force. However, the variability between the two maximal muscle contractions in the frontalis muscle was too large to make a reliable estimate of the maximal muscle force and thus none of the muscle activity measurements were normalized. In order to measure the subjects habitual level of physiological activation, the laboratory experiment started with a five minute period which served as a baseline period for the physiological variables. The subjects were alone in the room and were not given any instructions other than to find a comfortable position with their arms resting on the table in front of them. To ensure that all subjects had the same low level of muscle activity before the test started a five minute feedback period with muscle activity visualized on a screen followed. The subject experienced how it was possible to influence the level of muscle activity by adopting different postures and thereafter concentrated on minimising any muscle activity. The stressful task [18] was then performed: a two-choice reaction-time test on a monitor, lasting one hour. An open ("frame") and a solid ("brick") quadrangle were placed in a square pattern, and a written suggestion on how to move the brick to superimpose on the frame was given. The subject responded by pressing one of two keys ("correct" or "wrong") with the right middle or index finger. The task was to be carried out as quickly and correctly as possible. The PC program provided feedback on whether an answer was correct or wrong, and on the response time (very slow, slow, normal, fast, very fast) related to the subjects performance in the mini-trial carried out before the experiment started. Together with the feedback a new task was presented. After the end of the stressful task, all measurements continued for thirty minutes. The test person was instructed to sit still and relax during the rest period. Pain, perceived tension and fatigue was reported immediately before (baseline) and every ten minutes during and after the test by scoring on a 100 mm visual analogue scale (VAS) with the endpoints marked no pain/tension/fatigue and worst imaginable pain/tension/fatigue. The subjects were asked to assess pain in locations corresponding to the electromyography electrode positions; in the shoulders, neck, temples and forehead on both sides. The subjects were not allowed to see previous records when scoring.

A second blood sample was drawn during 5 min immediately after the test, before the 30 minute recovery period. Blood analysis was not a major aim of the study and these results are reported elsewhere (Nilsen et al., submitted).

Muscle activity was quantified by bilateral bipolar recording of surface electromyography (SEMG) (electrode diameter 6 mm, inter-electrode distance 20 mm). The system noise level was less than 1.5 μV root mean square (RMS). The signals were bandpass-filtered (10–1250 Hz) and stored on a digitizing recorder (Earth Data 128). Data were subsequently fed into an A/D converter (Powerlab 16S; ADInstruments Pty Ltd, Sydney, Australia; sampling rate 2 kHz) for calculation of the RMS values (100 ms running time window). Sharp transients and electrical activity from the heart in the SEMG signals were removed with a median filter (Matlab ver 6, The MathWorks inc.).

The following electrode sites were used: (1) Frontalis muscle; both electrodes placed on a vertical line crossing the pupil, 10 mm and 30 mm above the upper border of the eyebrow. (2) Temporal muscle; the lower electrode 10 mm posterior to the lateral canthus of the orbit, and the second electrode 20 mm above. (3) Splenius muscle; upper electrode 35 mm lateral to the spinous process of C2, and the second electrode 20 mm below. (4) Trapezius muscle; medial electrode 10 mm lateral to the midpoint of a line connecting the acromion and the spinous process of C7, and the second electrode 20 mm lateral to the first electrode. The ground electrode was placed on the spinous process of C7.

Activity in the autonomic nervous system was assessed by measurements of continuous non-invasive finger blood pressure (Portapres)[19], measurements of skin blood flow with Laser-Doppler flowmetry (Moorlab, 4 channels, time constant 0.02 s, low-pass filter 22 kHz), and measurements of the respiration pattern with a thermistor (Flaga, Embla S-AF-010) below the nose with active elements in each nostril and in front of the mouth. The blood pressure cuffs were mounted on the intermediate phalanx at the left middle and ring fingers. Finger skin blood flow was measured bilaterally with the electrodes (fibre separation 0.5 mm) placed on the volar side of the distal phalanx (pulp) of the thumb. Signals were sampled at 200 Hz.

Respiration frequency was calculated by the Chart 4.2 software (ADInstruments Pty Ltd, Sydney, Australia). Heart rate and blood pressure were calculated with the Beatscope 1.0 software (TNO, Amsterdam, the Netherlands). One blood pressure recording could not be analyzed due to technical difficulties.

Technical difficulties resulted in exclusion of seven subjects from analysis of respiration frequency and exclusion of two subjects from analysis of heart rate and blood pressure responses.

Mean values for each 10-minute period were calculated for all physiological recordings. Muscular activity and finger blood flow values are reported as the average of the left and right side for each region because ANOVA repeated measures analysis (rANOVA) revealed no side differences for the finger skin blood flow and muscle activity except for the frontalis muscle SEMG (left side (10.9 μV) > right side (9.2 μV); F(43) = 8.0, p = 0.007). However, performing all subsequent tests separately for right and left frontalis muscles did not give deviant results from those reported. Pain scores are reported from the side with the highest response (there were no side differences in neither pain level (side effect) nor pain development (side × time effect) for any of the four regions (rANOVA, Fs ≤ 3.2, p ≥ 0.08).

ANOVA with repeated measurements was used for evaluation of subgroup effects (sex, marital status, employment status, regular exercise, smokers, and alcohol drinking introduced sequentially one at a time as between-subject factors) with ten time intervals. For subgroup analysis of the recovery period we calculated a recovery variable (the difference between the mean of the last 10 minutes of rest (85–95 min) and the baseline period mean), a measure considered to be more meaningful than the absolute level when comparing groups [9]. Feedback data is displayed in figures, but feedback was not included in ANOVAs because we intended to study responses related to stress in this study. Recovery variables were analysed with one-way ANOVA tests.

For evaluation of the total response to the test we first performed repeated measures ANOVA tests (no between-subjects factors, evaluating the within-subject effect of time) with the same time intervals as in the subgroup analysis. For further post-hoc exploration of the response and recovery time-course we performed a series of paired-sample tests (Student's t-tests for physiological variables (Gauss-distributed) and Wilcoxon signed rank test for subjective variables (not Gauss-distributed)): We first evaluated the early response to the stressful task by comparing the first part (0–10 min) of the stressful task to baseline (immediately before the stressful task for the subjective variables). Secondly, changes during the stressful task (adaptation/summation effects) were investigated by a comparison of the first (0–10 min) and the last (50–60 min) part of the stressful task. Thirdly, we evaluated the recovery by comparing the change from the end of the stressful task (50–60 min) with the first part of the recovery period (65–75 min) and the first (65–75 min) and last (85–95 min) part of the recovery period with baseline.

Physiological responses (the difference between the average of the whole stressful task (0–60 min) and the average of the baseline period) and subjective responses to the stressful task (the difference between the maximal value during the 60 minute stress period and the value reported immediately before starting the test) were calculated as summary-variables for correlation analysis. Subjects with a pain response larger than zero were defined as pain responders. For each subject the location with the highest pain response during the task was identified (i.e. only one location for each subject). The pain response in this location (maximal pain location) was treated as a separate summary variable in the analysis (and it is the pain scores in this specific location we have displayed graphically).

Possible associations between variables were investigated by correlating the muscular responses (trapezius, splenius, temporalis, frontalis) with the autonomic responses (systolic and diastolic blood pressure, heart rate, respiration frequency and finger skin blood flow), and by correlating physiological responses (as above) with subjective responses (maximal pain, tension and fatigue), and finally by correlating the subjective responses with each other (i.e. maximal pain, tension and fatigue). The correlation coefficients between pain and muscular responses were calculated separately for each localisation (i.e. left temple pain with left temporalis muscle activity). Furthermore, as post-hoc analysis we searched for possible correlations between blood pressure/finger skin blood flow recovery variables and physiological responses, subjective responses and other recovery variables. We used Pearson correlation (r_p) for physiological variables (Gauss-distributed) and Spearman's rank order correlation (r_s) when subjective data were involved (not Gauss-distributed).

Because Mauchly's test of sphericity was significant in all ANOVA repeated measures tests with time as a within-subject effect we used Huynh-Feldt correction of degrees of freedom for these results. Two-tailed p-values less than 0.05 were considered to be significant. Because the hypotheses testing in this study involved several autonomic subsystems with insufficient a priori knowledge on possible relation to pain, we did not correct for multiple comparisons.

For transport expenses and the inconvenience (total time expenditure for each participant was 4 hours) participants received NOK 500 (USD 75). The Regional Committee for Medical Ethics approved the protocol, and all participants gave written informed consent before volunteering. Experiments were performed according to the Helsinki Declaration.

The development of all physiological variables is illustrated in Figure 1, 2 and 3.

The stressful task induced a clear response evident in all physiological variables (Table 2 and 3; baseline vs. 0–10 min, p ≤ 0.006) except for the splenius (p = 0.28) and temporalis muscle SEMG (p = 0.96).

Furthermore, age correlated negatively with the average respiration frequency response (r_p = -0.44, p = 0.006) and height correlated negatively with the average systolic blood pressure response (r_p = -0.41, p = 0.008). None of the other physiological responses (Table 2) correlated with age, height or weight.

Comparing the last ten minutes of the stressful task to the first ten minutes of the stressful task (Table 2 and 3) revealed a fall in heart rate with 2.5 beats/min (p = 0.001) and a reduced respiration frequency with 0.89 breaths/min (p = 0.04), indicating adaptation to the task for these two variables only. However, in the same time interval temporalis muscle activity increased with 0.82 μV (p = 0.03) and finger skin blood flow showed a trend towards lower values (p = 0.09). The other physiological variables were stable throughout the stressful task (p ≥ 0.33).

The heart rate response correlated with the trapezius muscle response (r_p = 0.44, p = 0.004) and the temporalis muscle response (temporalis vs. heart rate, r_p = 0.41, p = 0.008). The other correlations in the SEMG vs autonomic response matrix were non-significant (p > 0.06).

Upon cessation of the stressful task, heart rate (p < 0.001), respiration frequency (p < 0.001) and muscle activity in the trapezius (p < 0.003) and the frontalis (p < 0.002) decreased significantly (50–60 min vs. 65–75 min). Trapezius and frontalis SEMG recovered to the baseline level (baseline vs. 65–75 min, p ≥ 0.10) while heart rate and respiration frequency recovered to a level lower than baseline (baseline vs. 65–75 min, p ≤ 0.03). However, systolic and diastolic blood pressure, finger skin blood flow and muscle activity in the splenius and temporalis muscles did not change significantly upon cessation of the stressful task (50–60 min vs. 65–75 min and 50–60 min vs. 85–95 min, p > 0.10). The systolic and diastolic blood pressure level remained elevated and finger skin blood flow was reduced during the whole recovery period (baseline vs. 85–95 min p ≤ 0.001).

The finger skin blood flow recovery variable (Table 2) correlated negatively the systolic and diastolic blood pressure recovery variables (r_p = -0.52, p = 0.001 and r_p = -0.40, p = 0.01 respectively). This means that a high blood pressure at the end of the recovery period was associated with a small finger skin blood flow at the same time. The finger skin blood flow and blood pressure recovery variables did not correlate with other physiological (HR, muscle, respiration) response or recovery variables (r ≤ 0.25, p ≥ 0.11).

Development of tension, fatigue and pain scores in the maximal pain location is illustrated in Figure 4. Subjects reported increased tension (p = 0.02) and increased pain in the temples (p = 0.03) and forehead (p = 0.01) already ten minutes into the stressful task (0 min vs. 10 min), while fatigue (p = 0.52) and pain in the shoulder and neck (p > 0.52) did not increase during the first ten minutes (Table 2 and 3). All subjective variables increased further during the stressful task (10 min vs. 60 min; p < 0.008 except for a trend in temple pain (p = 0.06)), and were significantly reduced ten minutes into the recovery period (60 min vs. 75 min, p < 0.008). However, fatigue and pain in neck (and maximal pain) did not recover to baseline (0 min vs. 95 min; p < 0.04). Pain in the shoulders showed a trend towards non-recovery ten minutes into the recovery period (p = 0.08) but recovered to baseline after 30 minutes (p = 0.20), while tension and pain in temples and forehead returned to baseline ten minutes into the recovery period (p > 0.48).

Thirty subjects (68.2%) reported an increase in pain in at least one location during the test and twenty-eight subjects (63.6%) had an increase in pain VAS score of more than 30 mm during the test (Table 4). The pain response was most evident in the neck and/or shoulder (Table 4).

Pain responses did not correlate with tension and fatigue responses (r_s ≤ 0.19, p ≥ 0.20), however, fatigue and tension responses were correlated (r_s = 0.48, p = 0.001).

Pain, tension and fatigue responses did not correlate significantly with physiological responses (r_s ≤ 0.28, p ≥ 0.071, correlation coefficients between pain and muscular responses were calculated separately for each localisation). However, the fatigue response correlated with systolic (r_s = 0.34, p = 0.03, Figure 5) and diastolic blood pressure recovery (r_s = 0.31, p = 0.047) indicating a larger fatigue response during the stressful task for those subjects who recovered less during the rest period. However, no significant correlations were found between the blood pressure recovery and the pain and tension response variables (r_s ≤ 0.16, p ≥ 0.31) and finger skin blood flow recovery was not correlated to subjective responses (r_s ≤ 0.16, p ≥ 0.29).

Except for a correlation between the autonomic symptom index (Table 1) and the blood pressure response (Table 2, r_s = 0.38, p = 0.014), the physiological responses were not correlated to the Nevroticism index or the "autonomic symptom index". The Nevroticism index (EPQ-N, Table 1) correlated with pain and fatigue responses (r_s ≥ 0.36, p ≤ 0.016).

Subgroup analyses with the dichotomized variables in Table 1 (sex, marital status, employment status, regular exercisers, smokers, and alcohol drinking) revealed that women had lower respiratory frequency (15.2 vs. 17.1 breaths/min, rANOVA; sex effect F(1,35) = 4.5, p = 0.04) and higher frontalis SEMG (11.1 vs. 6.1 μV, rANOVA; sex effect F(1,42) = 6.7, p = 0.01). Moreover, smokers had higher blood systolic blood pressure level (130 mmHg vs. 120 mmHg, rANOVA; smoking effect F (1,36) = 4.7, p = 0.04) and we found a time × marital status interaction for maximal pain (rANOVA; F(3.0,141.2) = 2.6, p = 0.048) with higher maximal pain response for those living alone compared to cohabitants (17.7 vs 14.5 mm VAS). Subgroup analysis of the recovery variables did however not reveal any differences (One-way ANOVA Fs ≤ 2.9, p ≥ 0.097). It must be noted that some subgroups had few cases (Table 1), and were not ideal for subgroup effect analysis.