Clinical Neurophysiology 126 (2015) 763–771
Contents lists available at ScienceDirect
Clinical Neurophysiology journal homepage: www.elsevier.com/locate/clinph
Dynamic spectral indices of the electroencephalogram provide new insights into tonic pain M. Gram a, C. Graversen a,b, S.S. Olesen a, A.M. Drewes a,c,⇑ a
Mech-Sense, Department of Gastroenterology and Hepatology, Aalborg University Hospital, Aalborg, Denmark Mech-Sense, Department of Radiology, Aalborg University Hospital, Aalborg, Denmark c Center for Sensory-Motor Interactions (SMI), Department of Health Science and Technology, Aalborg University, Denmark b
a r t i c l e
i n f o
Article history: Accepted 17 July 2014 Available online 21 August 2014 Keywords: Cold pressor EEG Theta Pain Tonic Reliability Dynamics
h i g h l i g h t s The cold pressor test provides reliable EEG spectral indices. Tonic pain induced widespread changes in the EEG spectrum, with theta, beta3 and gamma bands
correlating to the overall perceived pain. Dynamic EEG analysis revealed that theta activity is associated to both inter-individual and dynamic
perception of tonic pain.
a b s t r a c t Objective: This study aimed to investigate reliability of electroencephalography (EEG) during rest and tonic pain. Furthermore, changes in EEG between the two states as well as dynamics and relation to pain ratings were investigated. Methods: On two separate days EEG was recorded in 39 subjects during rest and tonic pain (cold pressor test: left hand held in 2 °C water for 2 min.) while pain intensity was rated continuously. Dynamic spectral analysis was performed on the EEG. Between-day reliability of spectral indices was assessed and correlations to pain ratings were investigated. Results: EEG reliability was high during both states. The relative spectral indices increased in delta (1–4 Hz; P = 0.0002), beta3 (18–32 Hz; P < 0.0001) and gamma (32–70 Hz; P < 0.0001) bands during tonic pain, and decreased in theta (4–8 Hz; P < 0.0001), alpha1 (8–10 Hz; P < 0.0001), alpha2 (10–12 Hz; P < 0.0001) bands. Theta, beta3 and gamma bands correlated significantly to the area-under-curve of pain ratings, but only theta was dynamic and correlated to the pain ratings (R = 0.88, P < 0.0001). Conclusions: EEG assessed during tonic pain is a valid experimental pain model both in terms of reliability between days and in connection between cortical activity and pain perception. Significance: EEG during tonic pain is more pain-specific and should be used in future basic and pharmacological studies. Ó 2014 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
1. Introduction Experimental pain models are widely used in the pain research community and have recently also gained increasing popularity in the clinical setting. They are used to unravel pain mechanisms in ⇑ Corresponding author at: Mech-Sense, Department of Gastroenterology and Hepatology, Aalborg University Hospital, Mølleparkvej 4, 9100 Aalborg, Denmark. Tel.: +45 99326243; fax: +45 99326507. E-mail address: [email protected] (A.M. Drewes).
acute and chronic pain conditions, and to evaluate the mechanisms of action underlying analgesic treatments. Experimental pain models are also commonly used in phase II studies where new analgesics are used as proof of concept in healthy individuals before initiation of clinical studies, as they keep down expenses and allows for fast and efficient evaluation of new analgesic compounds (Mitchell et al., 2004; Olesen et al., 2012). Most experimental pain models are based on recordings of subjective psychophysical responses to stimuli of controlled intensity. In addition, various methods have been used to objectively
http://dx.doi.org/10.1016/j.clinph.2014.07.027 1388-2457/Ó 2014 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
764
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
characterize evoked pain responses. Measurements of cortical electrical activity following rapid phasic noxious stimuli have been used to quantify pain perception as time-locked electroencephalogram (EEG) potentials (Frøkjær et al., 2011; Gram et al., 2013). However, it has been speculated that such brief stimuli are not suited to simulate natural and clinical pain conditions. Hence, increased focus has been given to experimental tonic pain which simulates chronic pain better (Rainville et al., 1992; Nir et al., 2010, 2012). Thermal stimulation with hand immersion in either hot or cold water for a set amount of time are common methods to induce experimental tonic pain, In particular immersion of the hand in cold water, the cold pressor test (CP), has been widely used to induce tonic pain in experimental pain studies (Lowery, 2003). It is generally agreed that the CP test mimic clinical pain due to its high levels of unpleasantness (Rainville et al., 1992). However, EEG recorded following tonic stimulations lack the time-locked waveforms seen in evoked potentials (Frøkjær et al., 2011). Therefore more advanced methods such as spectral analysis are required to explore and visualize the EEG activity. Spectral analysis uses time–frequency methods to compute standardized spectral EEG indices (Graversen et al., 2012a). The algorithm most commonly used is the Fast Fourier-transform (FFT), but more advanced methods such as the continuous wavelet transform (CWT) can be utilized to achieve superior time–frequency resolution and less sensitivity to noise (Akin, 2002; Gram et al., 2013). Despite growing use of continuous EEG recorded during tonic pain, there is to date no published data on the reliability of EEG during CP pain. This is probably due to the complex nature of measuring reliability and lack of well-defined guidelines for analysis (Atkinson and Nevill, 1998; Bruton et al., 2000). However reliability measures are of importance in power estimations for future studies and also indicate whether a specific measurement is of value (Atkinson and Nevill, 1998; Bruton et al., 2000; Hanneman, 2008). We hypothesized that: (I) pain scores and EEG spectral indices during CP pain are reliable and (II) dynamics of spectral EEG indices are associated with subjective pain perception. Thus the aims of this study were to: (a) investigate reliability of pain scores during CP, (b) to investigate reliability of EEG spectral indices during resting state and tonic pain, (c) to investigate differences in EEG spectral indices between resting state and CP (i.e. static EEG indices), (d) to investigate dynamics of EEG spectral indices during resting state and CP (i.e. dynamic EEG indices) and finally (e) explore if the subjective perception of pain (psychophysics) are associated with spectral indices of the EEG. 2. Methodology The study was conducted between November 2010 and April 2012 in the research laboratory at Mech-Sense, Department of Gastroenterology and Hepatology, Aalborg University Hospital, Denmark. The local ethics committee approved the study (reference No. N-20100046).
followed by EEG recordings during the CP test. Experimental procedures were done on the same time of the day. Prior to the first study day, a training session was conducted in order to introduce the subject to the laboratory environment and verify that subjects were able to tolerate the cold pressor pain. Experimental tests were carried out by well-trained experimenters in a quiet room and each participant was tested by the same experimenter on both study days. Testing in female subjects was standardized with regard to phase of the menstrual cycle in order to control for variations in hormone levels. 2.3. Cold pressor test The CP test was performed using circulated water bath (Grant, Fischer Scientific, Slangerup, Denmark). The water was cooled to 2 °C and the subjects immersed their left hand up to the wrist for 2 min while water was circulated. Subjects rated their perceived pain on a handheld electronic device with a visual analogue scale (VAS) going from no feeling of pain to the worst imaginable pain. The pain ratings were continuously sampled electronically with a frequency of 10 Hz on a scale from 0 to 10. Subjects commonly rate their pain in discrete steps using this method, which combined with the absence of verbal communication for pain rating is thought to reduce influence on EEG recordings. 2.4. EEG recordings On both days, EEG was recorded in a dimly lit room, first during a resting period and then during CP. During the resting state recordings, subjects were instructed to keep their eyes open while minimizing eye blinking during the 2.5 min. period. EEG was recorded during CP by starting the recording simultaneously as the subject submerges their hand in cold water. EEG was recorded from a standard 62-channel cap (Quick-Cap International, Neuroscan, El Paso, TX, USA), amplified digitally on a Synamps 2 system (Neuroscan Compumedics, El Paso, TX, USA) and recorded for later analysis (Neuroscan 4.3.1, Neuroscan, El Paso, TX, USA). 2.5. Pre-processing The data were first pre-processed in the Neuroscan EEG software in the following steps: (1) zero-phase shift notch filtering (49–51 Hz) using a finite impulse-response filter with a slope of 24 dB/octave; (2) zero-phase shift band-pass filtering (1–70 Hz) using a finite impulse-response filter with a slope of 12 dB/octave; (3) blinded visual inspection of data quality for all channels. Channels with abnormal signals were discarded and replaced by signals interpolated from neighboring electrodes; (4) re-referencing to the average electrode; (5) finally, resting EEG was cleaned by selecting 2 min of artefact-free EEG from the 2.5 min recording where the investigator was blinded to the origin of the recording. 2.6. Spectral analysis of EEG dynamics
2.1. Study subjects This study included 39 healthy subjects who fulfilled the following inclusion criteria: (1) aged between 20 and 65 years; (2) no previous diseases or psychiatric disorders causing pain. A medical doctor conducted a routine health examination in order to rule out any diseases before enrolment. All subjects provided written informed consent and were compensated for participation in the study. 2.2. Experimental protocol For each subject two experimental sessions were performed separated by 7 days. Each session consisted of resting state EEG
Spectral analysis of EEG amplitudes was carried out using Matlab 2012a (The Matworks, Inc., Natick, MA, USA) to retain information about dynamics over time. The continuous wavelet transform was applied to the EEG signals from all channels. The complex Morlet wavelet was chosen for analysis with a bandwidth of 10 Hz and a center frequency of 1 Hz. Scales for the transform were chosen to match frequencies ranging from 1 to 70 Hz, with a 0.5 Hz between-scale frequency interval. The absolute value of the obtained wavelet coefficients were used in the following analysis. To assess the static EEG spectral indices the wavelet coefficients were split into the following eight standardized bands: delta
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
(1–4 Hz), theta (4–8 Hz), alpha1 (8–10 Hz), alpha2 (10–12 Hz), beta1 (12–18 Hz), beta2 (18–24 Hz), beta3 (24–32 Hz) and gamma (32–70 Hz). The wavelet coefficients were averaged over time and then scales contained within each frequency band were summed together to yield the absolute activity within each frequency band. The dynamics of the EEG signals were analyzed by splitting the wavelet coefficients in time into epoch windows of 15 s, with a 7.5 s overlap. Each epoch was analyzed as described for the static EEG to track changes in the spectral indices over time. The first epoch (equivalent to 7.5 s of EEG recording) was discarded due to movement artefacts caused by the subjects positioning their hand in the water. The absolute EEG activity is affected by several factors such as scalp thickness and connection between the electrode and scalp (Law, 1993). Relative EEG activity accounts for these individual differences and has been shown to be closely correlated to e.g., brain perfusion assessed by positron emission tomography (Cook et al., 1998). However, frequency bands can affect each other when using relative EEG activity, which can yield different results compared to absolute activity. Relative and absolute EEG activity cannot be compared directly, making comparisons between studies difficult in some cases (Malver et al., 2014). However, both methods reveal complimentary information regarding cerebral activity (Leuchter et al., 1993). Hence, while relative EEG indices are the main endpoint of the study, results using absolute EEG indices are also reported as an additional endpoint. The relative activity was calculated separately for each channel by dividing each epoch by the average energy content for the entire recording and multiplying by 100. The values then represent the percentage of total amplitude contained in each frequency band. 2.6.1. Statistics Intra-class correlations coefficients (ICCs), coefficient of variations (CVs) and limits of agreement were calculated for the area under the VAS curves (AUCs), and average amplitude of spectral EEG indices. The ICC was calculated by a two-way random model (Weir, 2005). The CV is reported as the standard deviation of the measurement expressed as a percentage of the mean. Bootstrapping was utilized to determine 95% confidence intervals for both CV and ICC based on 1000 bootstrap samples. The limits of agreement was calculated as the 95% random error component of the data, and is reported as bias ± random error (Atkinson and Nevill, 1998). To reduce dimensionality of data, the results from the two days were averaged after calculating reliability measures. A mixed model was used with conditions (resting state and CP pain) and time as factors. Also, mixed model were used to analyze differences in topographic distributions of the EEG indices. Results from these analyses are reported as the retrieved Z-values projected to corresponding topographic maps. For correlation analysis a two-step procedure was employed. First, EEG indices showing differences between resting state and CP conditions were correlated to pain rating AUCs. Second, EEG indices showing statistically significant dynamics during CP and correlations to pain score AUCs (step one) were analyzed for ‘‘dynamic correlations’’ between the continuous VAS curve and dynamic spectral EEG indices. For this analysis, pain recordings were down-sampled by a factor of 75 to match the frequency of EEG epochs of one epoch per 7.5 s. Pearson’s linear correlation coefficient or Spearman’s rho was used for correlation analysis as appropriate. 3. Results Thirty-nine subjects were enrolled in the study (18 females and 21 males; average age: 26.9 ± 6.5 years). Five subjects were
765
excluded from analysis as they could not tolerate the CP test for two minutes and four subjects were excluded due to poor EEG data quality. Hence, 30 subjects were included in the final analysis. 3.1. Reliability of experimental pain scores and EEG indices The area under the VAS curve (AUC) was highly reliable with a CV of 3% (Table 1). Reliability measures of spectral EEG indices are reported in Table 1. Taken together, the EEG indices were reliable for the resting state condition (CV, 4.2–12.7%) and CP condition (CV, 3.9–13.7%). Reliability measures from the absolute EEG indices are reported in Table 2. Variation between days was slightly higher than for relative EEG indices for resting state condition (CV, 6.7–15.8%) and CP condition (CV, 7.9–18.4%). 3.2. Static spectral EEG indices The average amplitude of the relative spectral EEG indices during resting state and CP conditions are shown in Fig. 1a (the average across all electrodes is shown). During the CP condition increased delta (P < 0.0001), beta3 (P < 0.0001) and gamma (P < 0.0001) activities were seen, while activities of theta (P < 0.0001), alpha1 (P < 0.0001), alpha2 (P < 0.0001) decreased during CP. The topographical distributions of the spectral indices during resting state and CP conditions are shown in Fig. 2 along with the Z-values from the post hoc analysis for each channel to indicate where the differences were statistically strongest. The changes were different when using absolute power as shown in Fig. 1b. Hence, there was an increase in the theta band (P = 0.03), beta1 (P < 0.0001) and beta2 (P < 0.0001) bands. As for the relative power increases were seen in the delta (P < 0.0001), beta3 (P < 0.0001) and gamma (P < 0.001) bands, and a decrease in power was seen in the alpha2 (P = 0.04) band. 3.3. Dynamics of spectral EEG indices Fig. 3 illustrates the dynamics of spectral EEG indices during resting state and CP conditions. During resting state, the EEG showed a gradual increase over time for the theta (P = 0.008), alpha1 (P < 0.0001), alpha2 (P < 0.0001), beta1 (P = 0.002) and beta2 (P = 0.01) bands. In contrast, delta, beta3 and gamma bands were not dynamic (all P > 0.1). During CP condition, a gradual increase over time was seen for the delta (P = 0.03), theta (P = 0.0005), alpha1 (P < 0.0001), alpha2 (P < 0.0001) and beta1 (P = 0.0001) bands, while beta2, beta3 and gamma bands were not dynamic (all P > 0.2). The appearance of the dynamic curves of absolute EEG indices was almost identical to the relative EEG indices. Absolute EEG indices during resting state showed gradual increases in for the theta (P = 0.02), alpha1 (P < 0.0001), alpha2 (P < 0.0001), beta1 (P = 0.006) and beta2 (P = 0.01) bands. For CP condition a gradual increase was present in the delta (P = 0.03), theta (P = 0.001), alpha1 (P < 0.0001), alpha2 (P < 0.0001) and beta1 (P = 0.0001) bands. 3.4. Correlations between CP pain and spectral EEG indices Correlation analysis of the spectral EEG indices (selected if they showed significant differences between resting state and CP conditions) revealed correlations between the AUC and the theta (r = 0.43, P = 0.02), beta3 (r = 0.45, P = 0.01) and gamma (r = 0.43, P = 0.04) bands (Fig. 4A–C). However, as stated above only the theta band showed significant dynamics during CP and
766
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
Table 1 Intraclass correlation coefficient (ICC), coefficient of variation (CV) and limits of agreement for area-under-curve of pain ratings and relative spectral EEG indices (delta (1–4 Hz), theta (4–8 Hz), alpha1 (8–10 Hz), alpha2 (10–12 Hz), beta1 (12–18 Hz), beta2 (18–24 Hz), beta3 (24–32 Hz) and gamma (32–70 Hz)) during resting state and cold pressor pain. ICC and CV are reported with 95% confidence intervals computed from 1000 bootstrap samples. The limits of agreement are reported as bias ± 95% random error component. ICC
CV (%)
Limits of agreement
Pain ratings Area-under-curve (AUC)
0.95 [0.88–0.98]
3.4 [2.4–4.7]
15.37 ± 99.74
Resting state EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.72 0.74 0.93 0.89 0.82 0.74 0.55 0.57
[0.50–0.87] [0.36–0.92] [0.84–0.96] [0.79–0.95] [0.66–0.92] [0.49–0.86] [0.09–0.79] [0.22–0.82]
9.6 [7.3–12.2] 4.2 [3.1–6.4] 8.1 [6.3–10.3] 8.3 [6.1–10.9] 4.4 [3.3–5.9] 6.9 [5.0–9.2] 9.9 [7.5–13.9] 12.7 [9.2–17.7]
0.14 ± 7.67 0.22 ± 3.41 0.07 ± 3.60 0.51 ± 3.25 0.03 ± 2.48 0.11 ± 2.42 0.10 ± 2.41 0.09 ± 1.92
Cold pressor EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.80 0.71 0.87 0.83 0.74 0.70 0.51 0.51
[0.69–0.88] [0.44–0.86] [0.80–0.94] [0.70–0.91] [0.52–0.89] [0.52–0.83] [0.32–0.68] [0.30–0.69]
7.1 [5.2–9.2] 3.9 [2.9–5.4] 6.6 [4.8–8.5] 6.8 [4.9–8.7] 4.1 [2.9–5.7] 6.6 [5.2–8.5] 9.4 [7.7–11.9] 13.4 [11.2–16.1]
1.08 ± 6.70 0.15 ± 2.64 0.39 ± 2.21 0.22 ± 1.95 0.01 ± 2.33 0.34 ± 2.22 0.30 ± 2.53 0.30 ± 2.37
Table 2 ICC, CV and limits of agreement for absolute spectral EEG indices during resting state and cold pressor pain. ICC and CV are reported with 95% confidence intervals computed from 1000 bootstrap samples. The limits of agreement are reported as bias ± 95% random error component. ICC
CV (%)
Limits of agreement
Resting state EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.63 0.82 0.93 0.96 0.93 0.77 0.59 0.55
[0.35–0.79] [0.64–0.91] [0.87–0.97] [0.87–0.99] [0.80–0.97] [0.48–0.92] [0.07–0.85] [0.14–0.84]
15.0 [11.4–20.3] 8.2 [6.0–11.3] 11.1 [8.4–14.0] 8.5 [5.9–12.3] 6.7 [4.8–10.6] 9.0 [6.1–13.8] 11.8 [7.8–18.6] 15.8 [11.2–23.5]
2.10 ± 35.66 2.78 ± 17.18 2.22 ± 15.56 0.57 ± 10.19 1.41 ± 9.81 1.69 ± 10.69 1.23 ± 9.65 0.90 ± 7.44
Cold pressor EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.78 0.81 0.86 0.93 0.81 0.67 0.54 0.54
[0.69–0.86] [0.65–0.90] [0.76–0.93] [0.86–0.97] [0.55–0.92] [0.32–0.85] [0.21–0.76] [0.20–0.72]
14.1 [10.3–17.5] 8.4 [6.2–11.5] 7.9 [5.6–10.7] 7.9 [6.0–10.6] 8.4 [6.3–11.2] 11.7 [9.0–16.1] 14.6 [11.1–19.8] 18.4 [14.5–23.6]
1.29 ± 53.72 2.12 ± 20.07 2.46 ± 9.81 1.54 ± 6.38 2.10 ± 15.51 2.84 ± 16.36 2.26 ± 17.13 1.96 ± 13.58
consequently, only this band was considered for correlation to the continuous CP pain ratings. The analysis showed a close correlation between theta dynamics and corresponding VAS ratings (r = 0.88, P < 0.0001) (Fig. 4D). No correlations were found between absolute EEG indices and AUC (all P > 0.1). A correlation was present between the dynamics of absolute EEG in the theta band and the continuous CP pain ratings (r = 0.81, P = 0.001). 4. Discussion In the present study we explored the relationship between tonic pain perception and cortical processing, as assessed by EEG during a tonic experimental pain model (the cold pressor test). First, we investigated the reliability between days of cold pressor induced pain and concurrent electrical brain activity (EEG) and found that both were reliable between days. Next, differences in static and dynamic spectral EEG indices between resting state and cold
pressor conditions were analyzed. Overall, most frequency ranges of the EEG showed distinct patterns of changes in static EEG indices measured as an average during the two min of tonic pain. Here theta, beta3 and gamma bands were correlated to the areaunder-curve of pain ratings. Finally, we explored whether these changes in EEG indices were associated with pain perception by comparing the dynamics of pain ratings with EEG dynamics. Interestingly, we found no dynamics in beta3 and gamma EEG bands, indicating that these are not directly related to the temporal sequence of neuronal pain processing. However, dynamics in the theta range were closely correlated to pain ratings, showing a strong inter-individual and intra-recording relationship to pain perception. 4.1. Reliability High reliability has earlier been reported for the conditioned pain modulation during CP, but to date no results have been
767
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
(a)
30
*
EEG amplitude [%]
25 Resting state Cold pressor pain
** 20
**
15
**
10
** **
5
0
(b)
120
Delta
Theta
Alpha 1
Alpha 2
Beta 1
Beta 2
Beta 3
Gamma
**
EEG amplitude
100 Resting state Cold pressor pain
80
* **
60
**
*
40
** **
20
0
Delta
Theta
Alpha 1
Alpha 2
Beta 1
Beta 2
Beta 3
Gamma
Fig. 1. Average EEG activity during resting state and cold pressor conditions for (a) relative EEG indices and (b) absolute EEG indices. Error bars indicate 95% confidence intervals. ⁄ = P < 0.05; ⁄⁄ = P < 0.0001.
published reliability of EEG recorded during CP. The analysis presented here looks primarily to the limits of agreement for assessment of reliability, aided by CV as it provides complimentary information with regards to variability of the mean. Results showed high reliability, with most frequency bands showing only a few percent of variation in limits of agreement with negligible bias. In both static and dynamic conditions the delta, beta3 and gamma bands showed the lowest reliability which might be explained by contamination from eye blinks and eye movements for the delta band, and contamination from muscle contractions in the neck for the high frequency bands (Dowman et al., 2008). When comparing ICC with CV and limits of agreement in this study some frequency bands exhibit very low ICC values, while both CV and limits of agreement indicating acceptable variability. This is in accordance with other reports, where ICC had inaccurately low values (Costa-Santos et al., 2011). However, ICC values are still reported here due to the complex nature of reliability and lack of standardized guidelines. Interestingly, reliability for EEG during CP pain seems to be slightly better than resting EEG for all frequency bands except gamma. This might be due to increased attention level in the subjects due to the severe CP pain. Furthermore, during this focusing on tonic pain the contribution from various processes dealing with conscious and non-conscious information may be suppressed. Reliability for absolute EEG indices was slightly worse than for relative EEG, but still within acceptable limits for most frequency bands.
4.2. Static spectral EEG indices Previous studies on EEG during tonic cold pain have generally found decreased alpha activity and increased beta activity (Backonja et al., 1991; Chang et al., 2002; Dowman et al., 2008). This is in agreement with our results where alpha activity decreased while beta2 and beta3 activities increased. Theta band activity has previously been reported to increase during CP, which is in line with results from this study, where absolute theta activity increases. However, the relative theta activity decreased. These findings appear contradictory, but are caused by use of different methods which can complicate comparison of results from different studies (Malver et al., 2014). Hence, previous studies reporting increased theta activity have performed the analysis using the absolute spectral indices while this study also reported relative values. Therefore the discrepancy is caused by interrelations between the frequency bands, where differences in other bands overpower the changes in theta, thus causing the relative theta value to decrease despite the absolute theta increases. This was revealed by performing the identical analysis with absolute EEG indices and reporting both results. Absolute and relative power provides different and complimentary information on cerebral activity, and it is of general belief that no one measure is more correct than the other (Leuchter et al., 1993, 1999). Absolute EEG indices do not have problems with interrelations between frequency bands, as seen in this study regarding the theta band. On the other hand the use of this measure complicates comparisons between individuals since the absolute EEG indices are vulnerable to
768
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
state
Cold pressor
Difference
Delta
Theta
Alpha 1
Beta 1
Z-values
Alpha 2
Beta 2
Beta 3
Gamma Fig. 2. Topographical representation of relative EEG activity during resting state and cold pressor conditions. Z-values from post hoc comparison of individual electrodes are plotted on corresponding topographic maps with contour lines added at ±2.6 (significance level P
Contents lists available at ScienceDirect
Clinical Neurophysiology journal homepage: www.elsevier.com/locate/clinph
Dynamic spectral indices of the electroencephalogram provide new insights into tonic pain M. Gram a, C. Graversen a,b, S.S. Olesen a, A.M. Drewes a,c,⇑ a
Mech-Sense, Department of Gastroenterology and Hepatology, Aalborg University Hospital, Aalborg, Denmark Mech-Sense, Department of Radiology, Aalborg University Hospital, Aalborg, Denmark c Center for Sensory-Motor Interactions (SMI), Department of Health Science and Technology, Aalborg University, Denmark b
a r t i c l e
i n f o
Article history: Accepted 17 July 2014 Available online 21 August 2014 Keywords: Cold pressor EEG Theta Pain Tonic Reliability Dynamics
h i g h l i g h t s The cold pressor test provides reliable EEG spectral indices. Tonic pain induced widespread changes in the EEG spectrum, with theta, beta3 and gamma bands
correlating to the overall perceived pain. Dynamic EEG analysis revealed that theta activity is associated to both inter-individual and dynamic
perception of tonic pain.
a b s t r a c t Objective: This study aimed to investigate reliability of electroencephalography (EEG) during rest and tonic pain. Furthermore, changes in EEG between the two states as well as dynamics and relation to pain ratings were investigated. Methods: On two separate days EEG was recorded in 39 subjects during rest and tonic pain (cold pressor test: left hand held in 2 °C water for 2 min.) while pain intensity was rated continuously. Dynamic spectral analysis was performed on the EEG. Between-day reliability of spectral indices was assessed and correlations to pain ratings were investigated. Results: EEG reliability was high during both states. The relative spectral indices increased in delta (1–4 Hz; P = 0.0002), beta3 (18–32 Hz; P < 0.0001) and gamma (32–70 Hz; P < 0.0001) bands during tonic pain, and decreased in theta (4–8 Hz; P < 0.0001), alpha1 (8–10 Hz; P < 0.0001), alpha2 (10–12 Hz; P < 0.0001) bands. Theta, beta3 and gamma bands correlated significantly to the area-under-curve of pain ratings, but only theta was dynamic and correlated to the pain ratings (R = 0.88, P < 0.0001). Conclusions: EEG assessed during tonic pain is a valid experimental pain model both in terms of reliability between days and in connection between cortical activity and pain perception. Significance: EEG during tonic pain is more pain-specific and should be used in future basic and pharmacological studies. Ó 2014 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
1. Introduction Experimental pain models are widely used in the pain research community and have recently also gained increasing popularity in the clinical setting. They are used to unravel pain mechanisms in ⇑ Corresponding author at: Mech-Sense, Department of Gastroenterology and Hepatology, Aalborg University Hospital, Mølleparkvej 4, 9100 Aalborg, Denmark. Tel.: +45 99326243; fax: +45 99326507. E-mail address: [email protected] (A.M. Drewes).
acute and chronic pain conditions, and to evaluate the mechanisms of action underlying analgesic treatments. Experimental pain models are also commonly used in phase II studies where new analgesics are used as proof of concept in healthy individuals before initiation of clinical studies, as they keep down expenses and allows for fast and efficient evaluation of new analgesic compounds (Mitchell et al., 2004; Olesen et al., 2012). Most experimental pain models are based on recordings of subjective psychophysical responses to stimuli of controlled intensity. In addition, various methods have been used to objectively
http://dx.doi.org/10.1016/j.clinph.2014.07.027 1388-2457/Ó 2014 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.
764
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
characterize evoked pain responses. Measurements of cortical electrical activity following rapid phasic noxious stimuli have been used to quantify pain perception as time-locked electroencephalogram (EEG) potentials (Frøkjær et al., 2011; Gram et al., 2013). However, it has been speculated that such brief stimuli are not suited to simulate natural and clinical pain conditions. Hence, increased focus has been given to experimental tonic pain which simulates chronic pain better (Rainville et al., 1992; Nir et al., 2010, 2012). Thermal stimulation with hand immersion in either hot or cold water for a set amount of time are common methods to induce experimental tonic pain, In particular immersion of the hand in cold water, the cold pressor test (CP), has been widely used to induce tonic pain in experimental pain studies (Lowery, 2003). It is generally agreed that the CP test mimic clinical pain due to its high levels of unpleasantness (Rainville et al., 1992). However, EEG recorded following tonic stimulations lack the time-locked waveforms seen in evoked potentials (Frøkjær et al., 2011). Therefore more advanced methods such as spectral analysis are required to explore and visualize the EEG activity. Spectral analysis uses time–frequency methods to compute standardized spectral EEG indices (Graversen et al., 2012a). The algorithm most commonly used is the Fast Fourier-transform (FFT), but more advanced methods such as the continuous wavelet transform (CWT) can be utilized to achieve superior time–frequency resolution and less sensitivity to noise (Akin, 2002; Gram et al., 2013). Despite growing use of continuous EEG recorded during tonic pain, there is to date no published data on the reliability of EEG during CP pain. This is probably due to the complex nature of measuring reliability and lack of well-defined guidelines for analysis (Atkinson and Nevill, 1998; Bruton et al., 2000). However reliability measures are of importance in power estimations for future studies and also indicate whether a specific measurement is of value (Atkinson and Nevill, 1998; Bruton et al., 2000; Hanneman, 2008). We hypothesized that: (I) pain scores and EEG spectral indices during CP pain are reliable and (II) dynamics of spectral EEG indices are associated with subjective pain perception. Thus the aims of this study were to: (a) investigate reliability of pain scores during CP, (b) to investigate reliability of EEG spectral indices during resting state and tonic pain, (c) to investigate differences in EEG spectral indices between resting state and CP (i.e. static EEG indices), (d) to investigate dynamics of EEG spectral indices during resting state and CP (i.e. dynamic EEG indices) and finally (e) explore if the subjective perception of pain (psychophysics) are associated with spectral indices of the EEG. 2. Methodology The study was conducted between November 2010 and April 2012 in the research laboratory at Mech-Sense, Department of Gastroenterology and Hepatology, Aalborg University Hospital, Denmark. The local ethics committee approved the study (reference No. N-20100046).
followed by EEG recordings during the CP test. Experimental procedures were done on the same time of the day. Prior to the first study day, a training session was conducted in order to introduce the subject to the laboratory environment and verify that subjects were able to tolerate the cold pressor pain. Experimental tests were carried out by well-trained experimenters in a quiet room and each participant was tested by the same experimenter on both study days. Testing in female subjects was standardized with regard to phase of the menstrual cycle in order to control for variations in hormone levels. 2.3. Cold pressor test The CP test was performed using circulated water bath (Grant, Fischer Scientific, Slangerup, Denmark). The water was cooled to 2 °C and the subjects immersed their left hand up to the wrist for 2 min while water was circulated. Subjects rated their perceived pain on a handheld electronic device with a visual analogue scale (VAS) going from no feeling of pain to the worst imaginable pain. The pain ratings were continuously sampled electronically with a frequency of 10 Hz on a scale from 0 to 10. Subjects commonly rate their pain in discrete steps using this method, which combined with the absence of verbal communication for pain rating is thought to reduce influence on EEG recordings. 2.4. EEG recordings On both days, EEG was recorded in a dimly lit room, first during a resting period and then during CP. During the resting state recordings, subjects were instructed to keep their eyes open while minimizing eye blinking during the 2.5 min. period. EEG was recorded during CP by starting the recording simultaneously as the subject submerges their hand in cold water. EEG was recorded from a standard 62-channel cap (Quick-Cap International, Neuroscan, El Paso, TX, USA), amplified digitally on a Synamps 2 system (Neuroscan Compumedics, El Paso, TX, USA) and recorded for later analysis (Neuroscan 4.3.1, Neuroscan, El Paso, TX, USA). 2.5. Pre-processing The data were first pre-processed in the Neuroscan EEG software in the following steps: (1) zero-phase shift notch filtering (49–51 Hz) using a finite impulse-response filter with a slope of 24 dB/octave; (2) zero-phase shift band-pass filtering (1–70 Hz) using a finite impulse-response filter with a slope of 12 dB/octave; (3) blinded visual inspection of data quality for all channels. Channels with abnormal signals were discarded and replaced by signals interpolated from neighboring electrodes; (4) re-referencing to the average electrode; (5) finally, resting EEG was cleaned by selecting 2 min of artefact-free EEG from the 2.5 min recording where the investigator was blinded to the origin of the recording. 2.6. Spectral analysis of EEG dynamics
2.1. Study subjects This study included 39 healthy subjects who fulfilled the following inclusion criteria: (1) aged between 20 and 65 years; (2) no previous diseases or psychiatric disorders causing pain. A medical doctor conducted a routine health examination in order to rule out any diseases before enrolment. All subjects provided written informed consent and were compensated for participation in the study. 2.2. Experimental protocol For each subject two experimental sessions were performed separated by 7 days. Each session consisted of resting state EEG
Spectral analysis of EEG amplitudes was carried out using Matlab 2012a (The Matworks, Inc., Natick, MA, USA) to retain information about dynamics over time. The continuous wavelet transform was applied to the EEG signals from all channels. The complex Morlet wavelet was chosen for analysis with a bandwidth of 10 Hz and a center frequency of 1 Hz. Scales for the transform were chosen to match frequencies ranging from 1 to 70 Hz, with a 0.5 Hz between-scale frequency interval. The absolute value of the obtained wavelet coefficients were used in the following analysis. To assess the static EEG spectral indices the wavelet coefficients were split into the following eight standardized bands: delta
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
(1–4 Hz), theta (4–8 Hz), alpha1 (8–10 Hz), alpha2 (10–12 Hz), beta1 (12–18 Hz), beta2 (18–24 Hz), beta3 (24–32 Hz) and gamma (32–70 Hz). The wavelet coefficients were averaged over time and then scales contained within each frequency band were summed together to yield the absolute activity within each frequency band. The dynamics of the EEG signals were analyzed by splitting the wavelet coefficients in time into epoch windows of 15 s, with a 7.5 s overlap. Each epoch was analyzed as described for the static EEG to track changes in the spectral indices over time. The first epoch (equivalent to 7.5 s of EEG recording) was discarded due to movement artefacts caused by the subjects positioning their hand in the water. The absolute EEG activity is affected by several factors such as scalp thickness and connection between the electrode and scalp (Law, 1993). Relative EEG activity accounts for these individual differences and has been shown to be closely correlated to e.g., brain perfusion assessed by positron emission tomography (Cook et al., 1998). However, frequency bands can affect each other when using relative EEG activity, which can yield different results compared to absolute activity. Relative and absolute EEG activity cannot be compared directly, making comparisons between studies difficult in some cases (Malver et al., 2014). However, both methods reveal complimentary information regarding cerebral activity (Leuchter et al., 1993). Hence, while relative EEG indices are the main endpoint of the study, results using absolute EEG indices are also reported as an additional endpoint. The relative activity was calculated separately for each channel by dividing each epoch by the average energy content for the entire recording and multiplying by 100. The values then represent the percentage of total amplitude contained in each frequency band. 2.6.1. Statistics Intra-class correlations coefficients (ICCs), coefficient of variations (CVs) and limits of agreement were calculated for the area under the VAS curves (AUCs), and average amplitude of spectral EEG indices. The ICC was calculated by a two-way random model (Weir, 2005). The CV is reported as the standard deviation of the measurement expressed as a percentage of the mean. Bootstrapping was utilized to determine 95% confidence intervals for both CV and ICC based on 1000 bootstrap samples. The limits of agreement was calculated as the 95% random error component of the data, and is reported as bias ± random error (Atkinson and Nevill, 1998). To reduce dimensionality of data, the results from the two days were averaged after calculating reliability measures. A mixed model was used with conditions (resting state and CP pain) and time as factors. Also, mixed model were used to analyze differences in topographic distributions of the EEG indices. Results from these analyses are reported as the retrieved Z-values projected to corresponding topographic maps. For correlation analysis a two-step procedure was employed. First, EEG indices showing differences between resting state and CP conditions were correlated to pain rating AUCs. Second, EEG indices showing statistically significant dynamics during CP and correlations to pain score AUCs (step one) were analyzed for ‘‘dynamic correlations’’ between the continuous VAS curve and dynamic spectral EEG indices. For this analysis, pain recordings were down-sampled by a factor of 75 to match the frequency of EEG epochs of one epoch per 7.5 s. Pearson’s linear correlation coefficient or Spearman’s rho was used for correlation analysis as appropriate. 3. Results Thirty-nine subjects were enrolled in the study (18 females and 21 males; average age: 26.9 ± 6.5 years). Five subjects were
765
excluded from analysis as they could not tolerate the CP test for two minutes and four subjects were excluded due to poor EEG data quality. Hence, 30 subjects were included in the final analysis. 3.1. Reliability of experimental pain scores and EEG indices The area under the VAS curve (AUC) was highly reliable with a CV of 3% (Table 1). Reliability measures of spectral EEG indices are reported in Table 1. Taken together, the EEG indices were reliable for the resting state condition (CV, 4.2–12.7%) and CP condition (CV, 3.9–13.7%). Reliability measures from the absolute EEG indices are reported in Table 2. Variation between days was slightly higher than for relative EEG indices for resting state condition (CV, 6.7–15.8%) and CP condition (CV, 7.9–18.4%). 3.2. Static spectral EEG indices The average amplitude of the relative spectral EEG indices during resting state and CP conditions are shown in Fig. 1a (the average across all electrodes is shown). During the CP condition increased delta (P < 0.0001), beta3 (P < 0.0001) and gamma (P < 0.0001) activities were seen, while activities of theta (P < 0.0001), alpha1 (P < 0.0001), alpha2 (P < 0.0001) decreased during CP. The topographical distributions of the spectral indices during resting state and CP conditions are shown in Fig. 2 along with the Z-values from the post hoc analysis for each channel to indicate where the differences were statistically strongest. The changes were different when using absolute power as shown in Fig. 1b. Hence, there was an increase in the theta band (P = 0.03), beta1 (P < 0.0001) and beta2 (P < 0.0001) bands. As for the relative power increases were seen in the delta (P < 0.0001), beta3 (P < 0.0001) and gamma (P < 0.001) bands, and a decrease in power was seen in the alpha2 (P = 0.04) band. 3.3. Dynamics of spectral EEG indices Fig. 3 illustrates the dynamics of spectral EEG indices during resting state and CP conditions. During resting state, the EEG showed a gradual increase over time for the theta (P = 0.008), alpha1 (P < 0.0001), alpha2 (P < 0.0001), beta1 (P = 0.002) and beta2 (P = 0.01) bands. In contrast, delta, beta3 and gamma bands were not dynamic (all P > 0.1). During CP condition, a gradual increase over time was seen for the delta (P = 0.03), theta (P = 0.0005), alpha1 (P < 0.0001), alpha2 (P < 0.0001) and beta1 (P = 0.0001) bands, while beta2, beta3 and gamma bands were not dynamic (all P > 0.2). The appearance of the dynamic curves of absolute EEG indices was almost identical to the relative EEG indices. Absolute EEG indices during resting state showed gradual increases in for the theta (P = 0.02), alpha1 (P < 0.0001), alpha2 (P < 0.0001), beta1 (P = 0.006) and beta2 (P = 0.01) bands. For CP condition a gradual increase was present in the delta (P = 0.03), theta (P = 0.001), alpha1 (P < 0.0001), alpha2 (P < 0.0001) and beta1 (P = 0.0001) bands. 3.4. Correlations between CP pain and spectral EEG indices Correlation analysis of the spectral EEG indices (selected if they showed significant differences between resting state and CP conditions) revealed correlations between the AUC and the theta (r = 0.43, P = 0.02), beta3 (r = 0.45, P = 0.01) and gamma (r = 0.43, P = 0.04) bands (Fig. 4A–C). However, as stated above only the theta band showed significant dynamics during CP and
766
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
Table 1 Intraclass correlation coefficient (ICC), coefficient of variation (CV) and limits of agreement for area-under-curve of pain ratings and relative spectral EEG indices (delta (1–4 Hz), theta (4–8 Hz), alpha1 (8–10 Hz), alpha2 (10–12 Hz), beta1 (12–18 Hz), beta2 (18–24 Hz), beta3 (24–32 Hz) and gamma (32–70 Hz)) during resting state and cold pressor pain. ICC and CV are reported with 95% confidence intervals computed from 1000 bootstrap samples. The limits of agreement are reported as bias ± 95% random error component. ICC
CV (%)
Limits of agreement
Pain ratings Area-under-curve (AUC)
0.95 [0.88–0.98]
3.4 [2.4–4.7]
15.37 ± 99.74
Resting state EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.72 0.74 0.93 0.89 0.82 0.74 0.55 0.57
[0.50–0.87] [0.36–0.92] [0.84–0.96] [0.79–0.95] [0.66–0.92] [0.49–0.86] [0.09–0.79] [0.22–0.82]
9.6 [7.3–12.2] 4.2 [3.1–6.4] 8.1 [6.3–10.3] 8.3 [6.1–10.9] 4.4 [3.3–5.9] 6.9 [5.0–9.2] 9.9 [7.5–13.9] 12.7 [9.2–17.7]
0.14 ± 7.67 0.22 ± 3.41 0.07 ± 3.60 0.51 ± 3.25 0.03 ± 2.48 0.11 ± 2.42 0.10 ± 2.41 0.09 ± 1.92
Cold pressor EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.80 0.71 0.87 0.83 0.74 0.70 0.51 0.51
[0.69–0.88] [0.44–0.86] [0.80–0.94] [0.70–0.91] [0.52–0.89] [0.52–0.83] [0.32–0.68] [0.30–0.69]
7.1 [5.2–9.2] 3.9 [2.9–5.4] 6.6 [4.8–8.5] 6.8 [4.9–8.7] 4.1 [2.9–5.7] 6.6 [5.2–8.5] 9.4 [7.7–11.9] 13.4 [11.2–16.1]
1.08 ± 6.70 0.15 ± 2.64 0.39 ± 2.21 0.22 ± 1.95 0.01 ± 2.33 0.34 ± 2.22 0.30 ± 2.53 0.30 ± 2.37
Table 2 ICC, CV and limits of agreement for absolute spectral EEG indices during resting state and cold pressor pain. ICC and CV are reported with 95% confidence intervals computed from 1000 bootstrap samples. The limits of agreement are reported as bias ± 95% random error component. ICC
CV (%)
Limits of agreement
Resting state EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.63 0.82 0.93 0.96 0.93 0.77 0.59 0.55
[0.35–0.79] [0.64–0.91] [0.87–0.97] [0.87–0.99] [0.80–0.97] [0.48–0.92] [0.07–0.85] [0.14–0.84]
15.0 [11.4–20.3] 8.2 [6.0–11.3] 11.1 [8.4–14.0] 8.5 [5.9–12.3] 6.7 [4.8–10.6] 9.0 [6.1–13.8] 11.8 [7.8–18.6] 15.8 [11.2–23.5]
2.10 ± 35.66 2.78 ± 17.18 2.22 ± 15.56 0.57 ± 10.19 1.41 ± 9.81 1.69 ± 10.69 1.23 ± 9.65 0.90 ± 7.44
Cold pressor EEG Delta Theta Alpha1 Alpha2 Beta1 Beta2 Beta3 Gamma
0.78 0.81 0.86 0.93 0.81 0.67 0.54 0.54
[0.69–0.86] [0.65–0.90] [0.76–0.93] [0.86–0.97] [0.55–0.92] [0.32–0.85] [0.21–0.76] [0.20–0.72]
14.1 [10.3–17.5] 8.4 [6.2–11.5] 7.9 [5.6–10.7] 7.9 [6.0–10.6] 8.4 [6.3–11.2] 11.7 [9.0–16.1] 14.6 [11.1–19.8] 18.4 [14.5–23.6]
1.29 ± 53.72 2.12 ± 20.07 2.46 ± 9.81 1.54 ± 6.38 2.10 ± 15.51 2.84 ± 16.36 2.26 ± 17.13 1.96 ± 13.58
consequently, only this band was considered for correlation to the continuous CP pain ratings. The analysis showed a close correlation between theta dynamics and corresponding VAS ratings (r = 0.88, P < 0.0001) (Fig. 4D). No correlations were found between absolute EEG indices and AUC (all P > 0.1). A correlation was present between the dynamics of absolute EEG in the theta band and the continuous CP pain ratings (r = 0.81, P = 0.001). 4. Discussion In the present study we explored the relationship between tonic pain perception and cortical processing, as assessed by EEG during a tonic experimental pain model (the cold pressor test). First, we investigated the reliability between days of cold pressor induced pain and concurrent electrical brain activity (EEG) and found that both were reliable between days. Next, differences in static and dynamic spectral EEG indices between resting state and cold
pressor conditions were analyzed. Overall, most frequency ranges of the EEG showed distinct patterns of changes in static EEG indices measured as an average during the two min of tonic pain. Here theta, beta3 and gamma bands were correlated to the areaunder-curve of pain ratings. Finally, we explored whether these changes in EEG indices were associated with pain perception by comparing the dynamics of pain ratings with EEG dynamics. Interestingly, we found no dynamics in beta3 and gamma EEG bands, indicating that these are not directly related to the temporal sequence of neuronal pain processing. However, dynamics in the theta range were closely correlated to pain ratings, showing a strong inter-individual and intra-recording relationship to pain perception. 4.1. Reliability High reliability has earlier been reported for the conditioned pain modulation during CP, but to date no results have been
767
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
(a)
30
*
EEG amplitude [%]
25 Resting state Cold pressor pain
** 20
**
15
**
10
** **
5
0
(b)
120
Delta
Theta
Alpha 1
Alpha 2
Beta 1
Beta 2
Beta 3
Gamma
**
EEG amplitude
100 Resting state Cold pressor pain
80
* **
60
**
*
40
** **
20
0
Delta
Theta
Alpha 1
Alpha 2
Beta 1
Beta 2
Beta 3
Gamma
Fig. 1. Average EEG activity during resting state and cold pressor conditions for (a) relative EEG indices and (b) absolute EEG indices. Error bars indicate 95% confidence intervals. ⁄ = P < 0.05; ⁄⁄ = P < 0.0001.
published reliability of EEG recorded during CP. The analysis presented here looks primarily to the limits of agreement for assessment of reliability, aided by CV as it provides complimentary information with regards to variability of the mean. Results showed high reliability, with most frequency bands showing only a few percent of variation in limits of agreement with negligible bias. In both static and dynamic conditions the delta, beta3 and gamma bands showed the lowest reliability which might be explained by contamination from eye blinks and eye movements for the delta band, and contamination from muscle contractions in the neck for the high frequency bands (Dowman et al., 2008). When comparing ICC with CV and limits of agreement in this study some frequency bands exhibit very low ICC values, while both CV and limits of agreement indicating acceptable variability. This is in accordance with other reports, where ICC had inaccurately low values (Costa-Santos et al., 2011). However, ICC values are still reported here due to the complex nature of reliability and lack of standardized guidelines. Interestingly, reliability for EEG during CP pain seems to be slightly better than resting EEG for all frequency bands except gamma. This might be due to increased attention level in the subjects due to the severe CP pain. Furthermore, during this focusing on tonic pain the contribution from various processes dealing with conscious and non-conscious information may be suppressed. Reliability for absolute EEG indices was slightly worse than for relative EEG, but still within acceptable limits for most frequency bands.
4.2. Static spectral EEG indices Previous studies on EEG during tonic cold pain have generally found decreased alpha activity and increased beta activity (Backonja et al., 1991; Chang et al., 2002; Dowman et al., 2008). This is in agreement with our results where alpha activity decreased while beta2 and beta3 activities increased. Theta band activity has previously been reported to increase during CP, which is in line with results from this study, where absolute theta activity increases. However, the relative theta activity decreased. These findings appear contradictory, but are caused by use of different methods which can complicate comparison of results from different studies (Malver et al., 2014). Hence, previous studies reporting increased theta activity have performed the analysis using the absolute spectral indices while this study also reported relative values. Therefore the discrepancy is caused by interrelations between the frequency bands, where differences in other bands overpower the changes in theta, thus causing the relative theta value to decrease despite the absolute theta increases. This was revealed by performing the identical analysis with absolute EEG indices and reporting both results. Absolute and relative power provides different and complimentary information on cerebral activity, and it is of general belief that no one measure is more correct than the other (Leuchter et al., 1993, 1999). Absolute EEG indices do not have problems with interrelations between frequency bands, as seen in this study regarding the theta band. On the other hand the use of this measure complicates comparisons between individuals since the absolute EEG indices are vulnerable to
768
M. Gram et al. / Clinical Neurophysiology 126 (2015) 763–771
state
Cold pressor
Difference
Delta
Theta
Alpha 1
Beta 1
Z-values
Alpha 2
Beta 2
Beta 3
Gamma Fig. 2. Topographical representation of relative EEG activity during resting state and cold pressor conditions. Z-values from post hoc comparison of individual electrodes are plotted on corresponding topographic maps with contour lines added at ±2.6 (significance level P
