The effects of chiropractic spinal manipulation on central processing of tonic pain - a pilot study using standardized low-resolution brain electromagnetic tomography (sLORETA)

The objectives of the study were to investigate changes in pain perception and neural activity during tonic pain due to altered sensory input from the spine following chiropractic spinal adjustments. Fifteen participants with subclinical pain (recurrent spinal dysfunction such as mild pain, ache or stiffness but with no pain on the day of the experiment) participated in this randomized cross-over study involving a chiropractic spinal adjustment and a sham session, separated by 4.0 ± 4.2 days. Before and after each intervention, 61-channel electroencephalography (EEG) was recorded at rest and during 80 seconds of tonic pain evoked by the cold-pressor test (left hand immersed in 2 °C water). Participants rated the pain and unpleasantness to the cold-pressor test on two separate numerical rating scales. To study brain sources, sLORETA was performed on four EEG frequency bands: delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz) and beta (12–32 Hz). The pain scores decreased by 9% after the sham intervention (p < 0.05), whereas the unpleasantness scores decreased by 7% after both interventions (p < 0.05). sLORETA showed decreased brain activity following tonic pain in all frequency bands after the sham intervention, whereas no change in activity was seen after the chiropractic spinal adjustment session. This study showed habituation to pain following the sham intervention, with no habituation occurring following the chiropractic intervention. This suggests that the chiropractic spinal adjustments may alter central processing of pain and unpleasantness.


Methods
The study was conducted according to the Declaration of Helsinki. The North Denmark Region Committee on Health Research Ethics approved the study (N-20150033). The study was one-way blinded; therefore, the participants did not know which type of intervention they received. This study was registered retrospectively on 24 th August 2018 with the Australian New Zealand Clinical Trial Registry (trial registration number ACTRN12618001420235). The study design is given in Supplementary Figure S1.
Subjects. Fifteen subclinical pain subjects (10 males, (mean ± SD) age = 32.1 ± 7.2 years, BMI = 24.02 ± 5.94 kg/m 2 ) participated in the study. Subclinical pain refers to recurrent spinal dysfunction such as mild pain, ache or stiffness for which treatment is not yet sought, and most importantly no pain on the day of experimental assessment, to avoid the confounding effects of altered resting pain levels. All subjects gave their written informed consent to participate in the study.
Before entering the study, the subjects were introduced to the lab environment and were assessed by a chiropractor with 15 years of experience to assure that the subjects pass the eligibility criteria to participate in the study. Subjects were included if they were aged between 18 and 50 years, had a history of recurring spinal dysfunction such as mild pain, ache or stiffness without a history of known trauma. Subjects were ineligible to participate if they exhibited actual pain on the day(s) of the experiment, had no evidence of spinal dysfunction, had absolute contraindications to chiropractic spinal adjustments, had experienced previous significant adverse reactions to chiropractic care or spinal manipulation, or if they had sought treatment for their pain symptoms. The subjects were also required to be fluent in the understanding of written and spoken English to participate in the study.
Experimental protocol. The subjects participated in two experimental sessions; sham and chiropractic, separated by 4.0 ± 4.2 days at Aalborg University Hospital. Each session consisted of a 1-minute resting state EEG recording, followed by a CP test during which EEG was also recorded. The resting EEG was recorded in order to observe the central processing of tonic pain itself (i.e. difference between pre-intervention CP-EEG and pre-intervention resting EEG) and this would subsequently aid in interpreting the changes in the central processing of tonic pain following the interventions. Figure 1 shows the overview of the experiment.
Interventions. The chiropractic spinal adjustment intervention and sham intervention were similar to those used in previous studies that have investigated the neurophysiological effects of chiropractic spinal adjustments 23,30,31 . The two researchers who carried out the interventions made a concerted effort to maintain the same levels of language and professional dealing with the participants. We advertised about the recruitment on a Facebook page. All the participants (upon questioning) were completely naïve to chiropractic care, so had no previous experience with a chiropractor, and therefore did not know what to expect. Participants were blinded to nature of intervention (sham or adjustment intervention). At the end of the second session, before telling them which session a real chiropractic intervention was, and which one was a sham intervention, the participants were asked whether they thought one session could have been a sham. Out of the 15 subjects, three subjects said that one session felt like a sham. Out of these three, two were correct and one was wrong, as this third person thought the sham was real and the real session was a sham.
Chiropractic spinal adjustment intervention. The chiropractic spinal adjustments carried out were high-velocity, low-amplitude thrusts to the spine or pelvic joints, which is a standard adjustment technique used by chiropractors and is also referred to as spinal manipulation. The sites chosen for the spinal adjustments were based on the clinical indicators of spinal and pelvic joint dysfunction 19 , which were: tenderness to palpation of the relevant joints; manual palpation for restricted intersegmental range of motion; palpable asymmetric intervertebral muscle tension, and any abnormal or blocked joint play and end-feel of the joints. These indicators have been shown to be reliable when used in combination to identify level of subluxation in the spine 48 . Multiple levels of the spine were adjusted in each participant by a registered chiropractor during the chiropractic adjustment session.
Sham intervention. The sham intervention acted as a physiological control. During the sham intervention, the chiropractor simulated a spinal adjustment session, which included passive and active movements of the participants head, spine, and body, similar to what is done during the chiropractic adjustment session. Care was www.nature.com/scientificreports www.nature.com/scientificreports/ taken during the sham session to not take any joints to their end range of motion or to cause a cavitation in the spine. This was done to limit the afferent barrage from large diameter afferents in the paraspinal muscles to the central nervous system, while controlling for the body movements, touch and vestibular input associated with setting up to provide chiropractic adjustments.
Cold-pressor test. The CP test was performed using a circulating water bath (Grant, Fischer Scientific, Slangerup, Denmark). The water was cooled to 2 °C and the subjects immersed their left hand in the water up to the wrist for 80 seconds (Fig. 1A).
Pain and unpleasantness scores. The participants rated their pain and unpleasantness on two separate numerical rating scales after their hand had been in the water for 60 seconds (Fig. 1C). The two scales ranged from 0 (no pain/unpleasantness) to 10 (maximum pain/unpleasantness), and the experimenter noted down the scores.
EEG. EEG was recorded at a sampling rate of 1000 Hz in a dimly lit room using a 61-channel cap (MEQNordic A/S, Jyllinge, Denmark) and Synamp system (Neuroscan Compumedics, El Paso, TX, USA). The reference electrode was just above AFz. During the recordings, the participants lay in a supine position and were instructed to relax, keep their focus on a point and reduce eye blinking.
The EEG was preprocessed offline. The following preprocessing steps were performed using Neuroscan 4.3.1 (Neuroscan, El Paso, TX, USA): (1) noisy channels were interpolated using their neighboring channels; (2) afterward, a 50 Hz notch filter was applied, (3) followed by a band-pass filter of 1 to 70 Hz. For the remaining preprocessing steps, MATLAB 2015b (The MathWorks, Inc., Natick, MA, USA.) was used: (4) EEG was truncated for further analysis: the resting EEG from 2 to 58 s was taken, and for the CP part, 72 s of EEG were taken, starting at 8 s from the onset of the stimulus (Fig. 1B). The first 8 seconds of CP-EEG data were removed to avoid artifacts including muscle contractions caused by the immediate unpleasantness after immersing the hand into the cold water 38 ; (5) to reduce the computational load in the sLORETA matrix calculations, the filtered EEG was downsampled by a factor of 4; and finally, (6) for obtaining smooth power spectral density, the EEG was divided into epochs with length of 8 s to facilitate the averaging procedure in sLORETA.
sLORETA. The underlying sources of the EEG were estimated using the sLORETA software package, version 20151222 44 (available at http://www.uzh.ch/keyinst/loreta). The sLORETA was done in the frequency domain to localize neural oscillators on the average referenced EEG. The EEG was average referenced in the sLORETA software. Cross-spectral matrices for each subject were computed in sLORETA software for four frequency bands: (i) delta (1-4 Hz), (ii) theta (4-8 Hz), (iii) alpha (8)(9)(10)(11)(12), and (iv) beta . The cross-spectral matrices for each participant were then averaged as the input for sLORETA source analysis. The sLORETA software was used to estimate the statistical differences in brain activity (in the four EEG frequency bands) between: 1. The baselines of both experimental sessions for (i) resting state EEG and (ii) CP-EEG to make sure there were no unexpected differences, 2. The baseline CP-EEG and baseline resting state EEG to observe how the brain processes tonic pain ( Fig. 1D), 3. The post-intervention and baseline EEG for (i) resting state and (ii) CP to find the effect of each intervention on neural activity (Fig. 1D).
The cortical gray matter was divided into 6239 voxels with a resolution of 5 mm 3 . The head model and electrode coordinates according to Montreal Neurological Institute average MRI brain map (MNI-152) 49 were used. Spectral analysis. The EEG power spectral analysis was performed using FieldTrip toolbox 50 to assist the sLORETA results. To estimate the differences in the power spectrum of the four EEG frequency bands between the conditions mentioned in the section above, the power spectra between 1 and 32 Hz of the EEG were calculated using Fourier basis with an Hanning window of 1 s, which was followed by computation of the average power of each frequency band.
Statistics. The data are presented as mean ± SD unless otherwise indicated. The statistical significance threshold was p < 0.05.
Two-way repeated measures analysis of variance (ANOVA) was performed to identify the changes in the pain and unpleasantness scores with time (before and after) and intervention (sham and chiropractic) as the two factors. If the overall significance in the ANOVA test was found, all pairwise multiple comparisons procedures (Student-Newman-Keuls Method) were performed in order to assess where the differences were. The software package SigmaStat version 3.0 (SPSS Inc. Chicago, IL, USA) was used for the above statistical analysis.
The statistical analysis for source localization was done using the sLORETA software's built-in statistics tool using statistical non-parametric mapping 51 which adjusted for multiple comparisons by utilizing Fisher's random permutation test with 5000 randomizations. To compare current sources in different frequency bands, paired two-tailed Student's t-test was used to compare the baselines of chiropractic and sham sessions; the baseline CP and baseline resting state; and the post-sessions and baselines.
Non-parametric cluster-based permutation test 52 was used to identify the differences in the EEG power spectrum between the baselines of chiropractic and sham sessions; the baseline CP and baseline resting state; and the post-sessions and baselines. The clusters were defined as two or more continuous channel-power pairs each with p < 0.05 from the paired two-tailed t-test with respect to the conditions. The t-values within each cluster were added to get the cluster-level statistics and the maximum of cluster-level statistics was used as the test statistic. www.nature.com/scientificreports www.nature.com/scientificreports/ A cluster was considered significant if its Monte Carlo probability for each tail exceeded the threshold of 0.025 compared to the reference distribution approximated by Monte Carlo method with 5000 permutations.

Results
All fifteen of the enrolled subjects successfully completed the experiment and data from all subjects were used for the analysis.
Pain and unpleasantness scores. The pain and unpleasantness scores of the CP test are summarized in Table 1. The time effects on the pain scores (F 1,14 = 6.7, p < 0.05) and unpleasantness scores (F 1,14 = 9.6, p < 0.05) were significant. The posthoc test revealed that the pain scores decreased after the sham intervention (p < 0.05), whereas the unpleasantness scores decreased after both interventions (both p < 0.05). There were no significant interactive effects present between the time and intervention.  Table 2. sLORETA localized EEG cortical sources with significant differences between the baseline coldpressor and baseline resting state. The number of voxels with significant power changes (p < 0.05) is listed.
www.nature.com/scientificreports www.nature.com/scientificreports/ Effects on source location. The sLORETA analysis showed no differences in both the resting state EEG baselines and the CP-EEG baselines in all frequency bands (all p > 0.05).
The comparison between baseline CP-EEG and baseline resting state EEG showed a widespread increase in cortical activity during pain compared to resting state in all frequency bands. The results from the sLORETA analysis are summarized in Table 2 and Fig. 2.
Neither of the two interventions changed the resting state EEG (all p > 0.05). The brain activity underlying the CP test decreased following the sham intervention ( Fig. 3 and Table 3) but showed no differences following the spinal manipulation. Following the sham session, the most significant decreases (p < 0.05) in activity due to CP were seen in the delta (Brodmann area 32, cingulate gyrus, limbic lobe) and alpha (Brodmann area 42, transverse temporal gyrus, temporal lobe) bands, whereas marginally significant decrease (p = 0.05) were seen in the theta (Brodmann area 9, medial frontal gyrus, frontal lobe) and beta (Brodmann area 22, superior temporal gyrus, temporal lobe) bands. The complete list of brain regions showing changes in activity can be seen in Table 3. www.nature.com/scientificreports www.nature.com/scientificreports/ Effects on power spectrum. Similar to the sLORETA results, the spectral analysis showed no differences in both the resting state EEG baselines (Fig. 4A) and the CP-EEG baselines (Fig. 4B) in all frequency bands (all p > 0.05).
The comparison between baseline CP-EEG and baseline resting state EEG showed a widespread increase in EEG power during pain as compared to resting state in all frequency bands (Fig. 4C).
Neither of the two interventions changed the resting state EEG power (Fig. 4D,E) (all p > 0.05). The EEG power underlying the CP test decreased following the sham intervention (Fig. 4F), with significant decrease in the beta band. Although non-significant for the delta, theta, alpha band, the trend of decreased brain activity was similar to the sLORETA results. The EEG power underlying the CP test showed a slight (non-significant) increase following the spinal manipulation (Fig. 4G) in all frequency bands except the alpha band.

Discussion
In this study, we investigated the effects of altering sensory input from dysfunctional areas of the spine (by spinal adjustments) on pain perception and the neural activity (in terms of source localization of EEG) obtained during the cold-pressor test. The pain scores due to the cold-pressor test decreased by 9% after the sham session, whereas the unpleasantness scores decreased by 7% after both interventions. The brain activity associated with the cold-pressor test following the sham session decreased significantly in the delta and alpha bands; and showed a marginally significant decrease in theta and beta bands, whereas there were no changes in EEG source localization following the chiropractic spinal adjustment session. EEG activity during tonic pain. This study found increased neural activity in all frequency bands when CP was compared with resting state. The increase in cortical activity due to tonic pain is supported by other studies using EEG [53][54][55] , PET 56 and fMRI 57,58 . In this study, the brain regions with a change in activity included the insula and anterior cingulate cortex, which are among the most often reported active regions during pain perception in many neuroimaging studies using fMRI and PET 59,60 . Hence, the insula and anterior cingulate cortex likely have an important role in the processing of pain. We found increased delta, theta and beta activities, which is www.nature.com/scientificreports www.nature.com/scientificreports/ consistent with findings of studies utilizing tonic pain in healthy volunteers 40,[53][54][55]61 . The increase of the delta, theta and beta activities in the anterior cingulate and insula cortices, among other regions in this study, likely implies negative feelings to the pain induced by CP, as it has been reported by many studies 58,60,62,63 that these regions are associated with emotional aspects of pain processing.
There was an increase in alpha activity during the CP test. Alpha oscillations have been found to be increased over frontal or parieto-occipital regions during tonic pain using the CP test 53 and hypertonic saline injection 61 . On the other hand, alpha power and underlying neural activity have also been shown to decrease following tonic pain 37,40,64 . The type or intensities of stimulus used in these studies may be the reason for inconsistency with the current study. For example, Shao et al. used a 10 °C CP test 40 , Babiloni et al. used CO2-laser stimulation 64 and heat stimulation was used by Nir et al. 37 . The changes in the alpha EEG band are associated with attention processes 65 and anticipation to pain 64 . Therefore, the increase of source activity underlying the alpha band in this study is likely due to the subjects' attention to pain.

Sham intervention.
After the sham intervention there was a decrease in pain scores during the tonic pain. This is not an unexpected result, since habituation to pain is a normal reaction in humans and animals to continuous or repetitive painful stimuli, which decreases the perceived pain and pain-related responses [66][67][68] . There was also a decrease in the neural activity underlying EEG during tonic pain following the sham intervention, indicating the pain scores related to the underlying activity in these specific brain regions. The regions which showed the most significant decrease were: cingulate gyrus, limbic lobe (delta) and transverse temporal gyrus, temporal lobe (alpha). Marginally significant decrease was seen in the medial frontal gyrus, frontal lobe (theta) and superior temporal gyrus, temporal lobe (beta). The areas that showed a decrease due to the CP test are a subgroup of brain areas, which were also activated due to the CP test itself. Hence, the decrease of activity in these brain areas likely has a role in the inhibition of pain, as seen in the decreased pain scores. In this study, there were about 10 minutes between the first and second EEG recording during tonic pain. The after-effects of tonic pain can last up to 30 minutes in humans 69,70 and therefore, the decreased cortical activity, in combination with the decreased pain/ unpleasantness perception of the CP stimulus after the sham intervention, can most likely be attributed to central habituation to the CP-induced pain. It is possible that there is some form of placebo effect occurring as well, as most subjects in this study were novices to chiropractic and therefore did not know what to expect.
Although habituation to pain is a normal reaction in humans and animals to continuous or repetitive pain stimuli [66][67][68] , it is a multifactorial event that is not well understood 71 . It has been hypothesized that habituation actually involves dual competing processes of depression (habituation) and facilitation (sensitization), that combine to give a behavioral or perceptual outcome 72 . Habituation to pain may lead to maladaptive plastic changes in the neural system. Maladaptive neuroplasticity has been reported to be induced by chronic pain 73 . These maladaptive neuroplastic changes cause individual sufferers to experience symptoms and functional disturbance, rather than the pain itself 74-76 . Spinal adjustment intervention. In the present study, there were no differences in neural activity due to CP-induced pain following the spinal adjustment session. This may be due to the spinal adjustments having an impact on pain habituation. It is possible that the altered afferent input from the spine following spinal adjustments modulates the interplay between the dual competing processes of depression (habituation) and facilitation (sensitization). Previous studies that have investigated the hypoalgesic effects of spinal manipulation on temperature-induced pain have produced conflicting results 14 . Millan et al. suggested that this may be related to the type of pain fibers (C-fiber or A-delta fibers) that are stimulated during different testing protocols, with C-fiber mediated pain being more influenced by a spinal manipulation intervention than A-delta fiber mediated pain 14 .  Table 3. sLORETA localized EEG cortical sources during the cold-pressor test with significant differences after sham session compared to baseline activity. The number of voxels with significant power changes (p < 0.05) is listed for delta and alpha bands; and with marginal significant changes (p = 0.05) for the theta and beta bands.
www.nature.com/scientificreports www.nature.com/scientificreports/ Previous studies have also suggested that spinal manipulation may reduce the central sensitization of pain 16,25 . This seems somewhat paradoxical based on the findings of the present study because a lessening of sensitization would be expected to result in an increase in habituation, as opposed to a decrease 72 . It is possible that instead of simply decreasing central sensitization, spinal adjustments may in some cases 'reset' the facilitatory and inhibitory processes associated with habituation. This may be due to spinal adjustments resulting in altered afferent paraspinal tissue input that affects the manner in which the somatosensory cortex integrates subsequent afferent information, as has been previously hypothesized 10,23,25,27 . For example, it has been shown that spinal adjustments alter processing in the prefrontal cortex 31 . The prefrontal cortex has been shown to impact the degree to which www.nature.com/scientificreports www.nature.com/scientificreports/ the insular cortex is activated during repeated cold pressor stimulation, thus can alter the way in which a person's brain habituates to a cold-pain stimulus 77 . Gaining a greater understanding of these potential mechanisms and processes may be important when considering the effects of chiropractic care on both acute and chronic pain. Study considerations. The sham and experimental interventions were separated by up to 10 days. This may induce some time effects in the data. Furthermore, this was a pilot study with 15 subjects where the neural response to tonic pain was assessed after a single session of spinal adjustments. We recognize that the number of subjects was not high. However, due to the little variation in response and the high sensitivity of the experimental models, the numbers are within the normal range for such explorative studies 10,[78][79][80][81][82] . To further validate these findings, future studies should look at the effects of chiropractic care on central processing of tonic pain in a larger population and over a longer period of chiropractic care.
It would also be worth considering to include a non-intervention session, as mind-set has been shown to impact habituation to cold-pain stimuli 83 . From the current study design, it is therefore hard to make firm conclusions about the sham only effects. It is impossible to be sure whether our results following the sham were due to the habituation, placebo effect, or both. Having a separate no-intervention control session would have helped elucidate this.
Finally, it would also have been interesting to analyze the gamma band for a study such as this. However, to do that, longer EEG recordings are required to facilitate the removal of the artifacts, especially those related to muscles as the spectrum of EMG overlaps that of the gamma band. The independent component analysis (ICA) can be used for this purpose but it requires approximately 20 to 30 times squared number of channels amount of data points, and it is not easy from a practical (pain) point of view to record this amount of EEG during the CP-test as most of the subjects cannot tolerate pain this long.

Conclusion
This study showed a habituation to pain response following the sham intervention, with no changes in the neural processing of tonic pain following the chiropractic spinal adjustment session. Changes to spinal function with chiropractic spinal adjustments appears to affect the way in which the central nervous system responds to repeated pain stimuli. However, this needs to be further explored before concrete conclusions can be made. Future studies should investigate the long-term effects of chiropractic care on pain processing in sub-clinical pain populations as well as in patients suffering from acute and chronic pain.