Modulation effects of different treatments on periaqueductal gray resting state functional connectivity in knee osteoarthritis knee pain patients

Abstract Background The analgesic effect of acupuncture is widely recognized, but the mechanical characteristics of acupuncture for pain relief, compared to non‐steroidal anti‐inflammatory (NSAIDs) and placebo medication, remain unknown. Aims To compare the modulation effects of acupuncture treatment with NSAIDs and placebo medication on descending pain modulation system (DPMS) in knee osteoarthritis (KOA) patients. Methods This study recruited 180 KOA patients with knee pain and 41 healthy controls (HCs). Individuals with KOA knee pain were divided randomly into groups of verum acupuncture (VA), sham acupuncture (SA), celecoxib (SC), placebo (PB), and waiting list (WT), with 36 patients in each group. VA and SA groups included ten sessions of puncturing acupoints or puncturing non‐acupoints acupuncture treatment for two successive weeks. Celecoxib capsules were continuously given orally to patients in the SC group at a dosage of 200 mg daily for 2 weeks. In the PB group, patients received a placebo capsule once a day for 2 weeks at the same dosage as celecoxib capsules. In the WL group, patients did not receive any treatment. Patients underwent a resting‐state BOLD‐fMRI scan pre‐ and post‐receiving the therapy, whereas HCs only underwent a baseline scan. Seed (ventrolateral periaqueductal gray, vlPAG, a key node in DPMS) based resting‐state functional connectivity (rs‐FC) was applied in the data analysis. Results All groups demonstrated improved knee pain scores relative to the initial state. There was no statistical difference between the VA and SA groups in all clinical outcomes, and vlPAG rs‐FC alterations. KOA knee pain individuals reported higher vlPAG rs‐FC in the bilateral thalamus than HCs. KOA knee pain patients in the acupuncture group (verum + sham, AG) exhibited increased vlPAG rs‐FC with the right dorsolateral prefrontal cortex (DLPFC) and the right angular, which is associated with knee pain improvement. In contrast with the SC and PB group, the AG exhibited significantly increased vlPAG rs‐FC with the right DLPFC and angular. Contrary to the WT group, the AG showed greater vlPAG rs‐FC with the right DLPFC and precuneus. Conclusions Acupuncture treatment, celecoxib, and placebo medication have different modulation effects on vlPAG DPMS in KOA knee pain patients. Acupuncture could modulate vlPAG rs‐FC with brain regions associated with cognitive control, attention, and reappraisal for knee pain relief in KOA patients, compared with celecoxib and placebo medication.


| INTRODUC TI ON
Knee osteoarthritis (KOA), the most prevalent type of musculoskeletal disease, is characterized by the degeneration of knee cartilage, bone remodeling, and synovitis. 1,2 KOA is one of the serious global health challenges due to its high prevalence, severe impairment, and enormous societal and financial burden. [3][4][5][6] Knee pain is the main symptom of KOA that causes patients to seek medical attentions. 7 Therefore, the treatment guidelines for KOA emphasized pain control as the main therapeutic concept. [8][9][10][11] The pathophysiology of KOA knee pain is complex. Therefore, understanding the source and mechanism of pain is essential for treatment development. Local indicators of the knee joint and changes in the central nervous system (CNS) play a role in the development of knee pain. Biomechanical variables (including muscle atrophy, overweight, and joint laxity) 1,12 and biochemical parameters (including inflammatory reactions) [13][14][15] are examples of local indicators. Functional or structural unbalance in the upstream conduction and downstream pain modulation systems at the spinal cord and brain level are also linked to knee pain in subjects with KOA. [16][17][18] Treatments for knee pain relief in KOA patients take action through the peripheral, central mechanisms, or both.
The first line recommended medication for treating knee pain caused by KOA in medical practice is NSAIDs medications. 8,19 The most commonly used NSAID is a selective COX-2 inhibitor such as celecoxib. Its administration causes minimal hazard gastrointestinal side effects. 20,21 The peripheral mechanism illustration for celecoxib's analgesia effect is that it reversibly binds to the hydrophilic sac near the activated site of COX-2. It inhibits the transformation of arachidonic acid to prostaglandin H2, resulting in anti-inflammatory and pain-relieving effects. 22 Previous studies have shown that celecoxib might penetrate the blood-brain barrier (BBB) [23][24][25] to exert an potential analgesic effect through descending pain modulation system (DPMS). [26][27][28][29][30] However, side effects and non-responders come with NSAIDs; thus, there is a growing interest in finding alternative treatments. Acupuncture, an ancient treatment technique of traditional Chinese medicine, has shown its safety and efficacy for KOA knee pain. 19,[31][32][33] Acupuncture analgesia efficacy takes action as a compound therapeutic method through multi-levels and multifactors, of which psychological factor is an important component.
Acupuncture has local anti-inflammatory and analgesia effects. [34][35][36][37] Beyond that, acupuncture could also modulate the CNS for pain relief. For instance, substantial evidence has implicated that acupuncture could relieve pain by activating DPMS [38][39][40] and modulating lateral pain pathways (including somatosensory and posterior insula 41 ) and medial pain pathways (including dorsal anterior cingulate and anterior insula 41 ). However, the characteristic of acupuncture analgesia mechanism on DPMS, especially compared with other treatments, remains largely unknown.
Assessing the physical functioning of DPMS in a non-invasive manner continues to be difficult. The resting-state functional magnetic resonance imaging (fMRI) technique allows for the identification of correlations between remote brain regions (resting-state functional connectivity, rs-FC) through their highly correlated lowfrequency spontaneous fluctuations non-invasively. 42 Recently, many research groups have applied rs-FC to evaluate the DPMS in humans. 40,43,44 The periaqueductal gray (PAG) is a mesencephalic brain structure between the third and fourth ventricles. 45 The PAG has several sub-regions, of which ventrolateral PAG (vlPAG) is the key node of the DPMS. 46,47 Furthermore, prior research demonstrated that central sensitization and increased vlPAG activity in osteoarthritis patients are associated with skin irritation in pain areas. 48,49 Past studies noted that impaired DPMS and treatment modulation, using vlPAG as the seed for rs-FC, in patients having chronic pain disorders. 40,50-56 These results indicated that it is possible to evaluate the DPMS and the modulation effects of treatment in KOA knee pain individuals by utilizing vlPAG rs-FC in a non-invasive manner.
In this current work, a comparison of the vlPAG rs-FC was made between individuals who experienced KOA knee pain and the healthy controls (HCs) matched with them. Furthermore, the mechanism by which the vlPAG rs-FC could be modulated by different longitudinal dorsolateral prefrontal cortex (DLPFC) and the right angular, which is associated with knee pain improvement. In contrast with the SC and PB group, the AG exhibited significantly increased vlPAG rs-FC with the right DLPFC and angular. Contrary to the WT group, the AG showed greater vlPAG rs-FC with the right DLPFC and precuneus.
Conclusions: Acupuncture treatment, celecoxib, and placebo medication have different modulation effects on vlPAG DPMS in KOA knee pain patients. Acupuncture could modulate vlPAG rs-FC with brain regions associated with cognitive control, attention, and reappraisal for knee pain relief in KOA patients, compared with celecoxib and placebo medication.
Additionally, the possible different vlPAG rs-FC modulation effects between verum acupuncture and sham acupuncture/celecoxib treatment/placebo medication/waiting list control were compared.
We hypothesized that different interventions have different modulation effects on vlPAG DPMS in KOA knee pain patients, and acupuncture could modulate the pain perception by DPMS, allowing psychological factors such as cognitive control, attention, and reappraisal to influence the pain experiences. It is hoped that this study could enhance the understanding of the mechanical characteristics of acupuncture for KOA pain relief compared with NSAIDs and placebo medication.

| ME THODS
In this study, the clinical and fMRI datasets of KOA knee pain patients were collected from two simultaneous trials. 57,58 Notably, all study operations of the two trials were performed under the same

| Participants
Knee osteoarthritis knee pain patients and HCs were involved in this study. The diagnostic criteria for KOA were derived from the American College of Rheumatology (ACR) criteria and published in their adjusted form in 1991. 59 The inclusion criteria were that each patient: (1) was aged 40-60 years and was right-handed, (2) had continual knee pain throughout the last 3 months, (3) had a mean knee pain score according to the Visual Analog Scale (VAS) ≥3 (range between 0 and 10), (4) had a Kellgren-Lawrence knee joint radiological degree between 0 and 2, according to the scale, 60 (5) received no acupuncture treatment or pain killer medicine in the past 3 months and 1 month, respectively, and (6) had signed the written agreement.
Individuals met the exclusion criteria if they: (1) were pregnant or lactating women, (2) were alcohol or drug abusers, (3) exhibited mental illness, (4) had severe organic diseases, (5) had rheumatoid arthritis, (6) had any other disorders that caused persistent pain, including a history of head injury that resulted in unconsciousness, (7) had a history of allergies to celecoxib, (8) had contraindications to acupuncture or MRI scan (e.g., claustrophobia), or (9) were currently participating in other trials.
The matched HCs were enrolled. The inclusion criteria were that all HCs: (1) were aged from 40-60 years and were right-handed, (2) were free from any pain conditions, (3) had no history of mental or CNS illness, (4) had no history of any systemic illness, like diabetes or hypertension, and (5) provided written permissions to conduct the study. HCs met the exclusion criteria if they: (1) were women who were either pregnant or nursing, (2) were alcohol or drug abusers, or (3) had contraindications to MRI scan.

| Study design
In this trial, the patients experiencing knee pain due to KOA were observed for 4 weeks, of which the first 2 weeks of run-in period checked whether KOA patients were eligible and still willing to participate. There were five patient groups and an HCs group, with 36 participants in each group. The patients were randomly assigned to verum acupuncture (VA), sham acupuncture (SA), celecoxib (SC), placebo (PB), and waiting list (WL) control groups. The celecoxib and placebo groups were masked (double-blinded) to treatment, and the subjects of verum acupuncture and sham acupuncture were kept single-blinded. Two MRI scans were done on KOA knee pain patients before and after treatments, whereas HCs received one scan.

| Intervention
Ten sessions of acupuncture stimulating medication were given to the VA and SA over two successive weeks, each session lasting for 30 min. In the SC group, celecoxib capsules (with an approval number of J20030098 and produced by Pfizer Pharmaceutical Co. Ltd. in Beijing, China) were orally given to patients for 2 weeks of 200 mg.
In the PB group, patients received a placebo capsule once a day for 2 weeks at the same dosage as celecoxib capsules. In the WL group, patients did not receive any treatment, but they were informed that after 2 weeks, they would receive ten sessions of acupuncture or 2 weeks of celecoxib for free.
Two trained and licensed acupuncturists conducted all acupuncture treatments. All acupoints and non-acupoints on both sides of the body were pierced using small needles made of disposable stainlesssteel measuring 0.25 mm to 40 mm in length (Hwato, China). When inserting the needles, the penetrating depth was kept between 0.5 cun and 1.5 cun. The locations in the VA and SA groups were gently manipulated to induce a specific acupuncture sensation (deqi).
All patients refrained from taking their usual medicines intended to treat KOA during the trial. Ibuprofen, in the form of sustainedrelease capsules containing 300 mg, was authorized for use as the rescue therapy in cases of severe pain.

| Outcome measurements
This research focused on the vlPAG rs-FC as its primary endpoint.

| Patients safety
Adverse events caused by acupuncture and celecoxib, such as pain, bleeding, fainting, gastrointestinal reaction, or other severe events, were processed immediately and recorded in details in the case report form.

| Clinical data analysis
All statistical evaluations were performed using SPSS 25.0 (SPSS Inc., Chicago, IL, USA). Normality distribution of the data was assessed, using the visual inspection of histograms and Shapiro-Wilk test. Data that did not follow a normal/Gaussian distribution were analyzed by using non-parametric test. We used the following statistical statistical tests: chi-squared test, Mann-Whitney U test, and Kruska-Wallis H test. Bonferroni's correction was used to account for multiple comparisons. Statistical analysis was conducted using a 2-tailed test, and the significance level was set to 5%. Quantitative statistics were presented as medians (lower quartiles; upper quartiles). Frequency was represented as a percentage (%).

| vlPAG seed-based functional connectivity analysis
Functional BOLD information was pre-processed using DPARSFA To further limit the detrimental impact of head movement, the volumes with FD >0.5 mm, the one time-point preceding and two-time points following the 'poor' time points, were eliminated from the data. 68 The pictures were then detrended, bandpass-filtered between 0.01 Hz and 0.08 Hz, and smoothed with a 6 mm full-width half-maximum Gaussian kernel.
A priori vlPAG seed with the following peak coordinates: 4, −26, −14, and a 3 mm radius was utilized. This seed was already utilized in earlier investigations. 40,43,50,51,69 There are four reasons why this seed was chosen: (1) the study found that an increase in the degree of heat pain resulted in a significant increase in fMRI signal in the same region, 70 (2) its site in the vlPAG was perceived as essential for opioid antinociception, 71,72 (3) celecoxib was discovered to cross the blood-brain barrier (BBB) with the maximum rate of flow accumulated in the midbrain (within PAG), [23][24][25] and (4) prior research declared that the seed was functionally linked to descending modulatory pain sites at the resting status in healthy control subjects, and a significant variation was observed between healthy subjects and chronic pain subjects. 40,51,56 Due to the proximity of the vlPAG seed region to a ventricle with a high pulsatile impact, a seed from a neighboring ventricular aqueduct (peak coordinates: 0, −34, −12, with a 3 mm radius) and seeds from the fourth ventricle (peak coordinates: ±4, 8, 12) 40 were selected as controls.
From the vlPAG seed, the averaged time course was derived, and a voxel-wise correlation analysis was performed. By applying the regression coefficient between all brain voxels and the time sequence of each seed, contrast pictures for each individual were created.
Furthermore, the correlation coefficient map's normality was enhanced by applying Fisher's r-to-z transform. The baseline vlPAG rs-FC of KOA knee pain patients and healthy controls were compared by employing a two-sample t-test. Pre-and post-treatment values for each group were compared using paired t-tests. Next, factorial design modules were utilized to assess alterations in rs-FC (posttreatment minus pre-treatment) between the verum acupuncture and sham acupuncture groups, acupuncture (verum + sham) and celecoxib groups, acupuncture and placebo medication groups, as well as acupuncture and waiting list groups. Covariables included gender, age, and body mass index (BMI) at baseline. As previous studies mentioned, variables that were percentage transformed could mitigate potential scaling problems. 73 The association between clinical outcomes and vlPAG rs-FC was examined by Spearman correlation analysis on the percentage variations (post-treatment minus pretreatment/pre-treatment) in rs-FC before and after treatment in the acupuncture group and the corresponding percentage changes in SF-MPQ. The SPM12 (http://www.fil.ion.ucl.ac.uk/spm/softw are/ spm12) was used to analyze functional connectivity in the rest status.

| RE SULTS
Knee osteoarthritis knee pain patients 180 were recruited, and 31 patients did not complete the study or had excessive head movement. This trial included 41 HCs of the matched age and gender; however, five HCs could not enroll in the MRI scan because of schedule issues. Therefore, 149 KOA knee pain patients and 36 matched HCs were involved in the final analyses ( Figure 2).

| Baseline characteristics
There was no statistically significant variation between HCs and KOA knee pain subjects following gender, age, or BMI (p > 0.05). The VA, SA, SC, PB, and WL groups showed no significant variation in terms of gender, age, BMI, disease duration, VAS, SF-MPQ, WOMAC, and SF-12 (p > 0.05) ( Table 1). Besides, there was no significant difference in knee pain intensity on fMRI scan in KOA patients among all groups (p > 0.05) (Table S1).

| Clinical outcomes
All groups demonstrated improvement in VAS, SF-MPQ, WOMAC, and SF-12 scores relative to the initial group (p < 0.05; Table 2).
Mann-Whitney U test showed the changes in VAS, SF-MPQ, and WOMAC were statistically different between the SC and PB groups (Table 3). Nevertheless, there was no statistical difference between the VA and SA groups in all outcomes (VAS, SF-MPQ, WOMAC, and SF-12), which was similar to the previous research on RCTs and meta-analyses, implying that the particular impact of perforating acupoints over perforating non-acupoints for knee pain is none or only a small to moderate. 33,[74][75][76][77][78] Besides, AES scores did not differ significantly between VA and SA groups for expectations of acupuncture (p = 0.777; Table S2). The average C-MASS ratings for all ten descriptors showed no statistical difference in acupuncture sensations between VA and SA groups (p = 0.080; Table S2). Based on the above, the VA and SA groups were combined into a single acupuncture group (AG) to study the overall acupuncture effect. Mann-Whitney U test with Bonferroni correction was applied to compare AG and SC, AG and PB, and AG and WL groups, respectively. Results showed no statistically significant differences between the AG and SC groups, which was similar to the previous RCT, suggesting that acupuncture and NSAIDs had a similar analgesic impact on controlling pain. 19 There were statistical differences between the AG and PB groups, and the AG and WL groups, in the changes of VAS and SF-MPQ but not in WOMAC and SF-12 ( Table 3).

| KOA vs. HCs
Compared with HCs, KOA knee pain patients showed increased vlPAG rs-FC with bilateral thalamus ( Figure 3C, Table 4). Spearman correlation analyses demonstrated that rs-FC levels of the bilateral thalamus had a small but significant positive correlation with SF-MPQ (r = 0.187, p = 0.022) ( Figure 3C).

| Modulation effects of different treatments
After longitudinal VA treatment, KOA knee pain subjects had increased vlPAG rs-FC with the right dorsolateral prefrontal cortex (DLPFC), and bilateral angular ( Figure 4A, Table 4). Nevertheless, there were no significant alterations in vlPAG rs-FC in KOA knee pain subjects following longitudinal medication with celecoxib, placebo, F I G U R E 2 Study flow chart. HCs, healthy controls; PB, placebo group; SA, sham acupuncture group; SC, celecoxib group; VA, verum acupuncture group; WL, waiting list group. or waiting list. Moreover, VA and SA groups in terms of vlPAG rs-FC alterations showed no significant variation ( Table 4).
According to the absence of significant variation in clinical outcomes and vlPAG rs-FC alterations between VA and SA groups (Table 4), the two groups were combined into a single acupuncture group (AG). After the longitudinal AG treatment, KOA knee pain subjects had increased rs-FC vlPAG with the bilateral dorsolateral prefrontal cortex (DLPFC), bilateral angular, and bilateral precuneus ( Figure 4B, Table 4). More comparisons were made to investigate the potential variation in vlPAG rs-FC regulation impact between acupuncture and celecoxib, placebo, and waiting list. In contrast with the celecoxib group, the AG exhibited increased vlPAG rs-FC with the right DLPFC and right angular ( Figure 4C, Table 4). Compared with the placebo group, the AG showed greater vlPAG rs-FC with the right DLPFC and right angular ( Figure 4D, Table 4). Contrary to the waiting list group, the AG exhibited significantly increased vlPAG rs-FC with the right DLPFC and right precuneus ( Figure 4E, Table 4). As additional controls, the previously mentioned analysis was performed using the seeds from the ventricular aqueduct near the vlPAG and from the fourth ventricle. No outcome was detected, which additionally validated the outcomes of vlPAG rs-FC in this investigation.

| Medication uses
No patient reported pain medication usage during the 4-week observation period in this study. However, three patients used plaster on the local knee joint for pain relief (two in the placebo group and one in the celecoxib group).

| Patients safety
All 180 patients were monitored for safety and tolerability. One patient in the VA group discontinued the study due to a fractured upper limb unrelated to the treatment. Four patients in SA and four patients in VA groups developed mild bruising around the acupoints/ non-acupoints, all of which were back to normal after 2 weeks of follow-up. One patient discontinued celecoxib during the intervention period due to intolerable gastric discomfort, and no further gastric discomfort occurred after 2 weeks of follow-up.

| DISCUSS ION
This study explored the vlPAG rs-FC alternations and the modulation effects of verum, sham acupuncture, celecoxib, placebo, and waiting TA B L E 1 Baseline characteristics of KOA knee pain patients in different treatment groups and healthy controls.

| Verum acupuncture versus sham acupuncture
This study found that verum acupuncture (puncturing acupoints) and sham acupuncture (puncturing non-acupoints) were all equally effective for KOA knee pain relief and had similar effects on vlPAG rs-FC, which is similar to the findings from previously published literature on RCTs and meta-analyses. Some research reported no advantage of puncturing acupoints over non-acupoints for treating knee pain, [74][75][76] while other studies showed that puncturing acupoints was more effective than puncturing non-acupoints, but the impact size was only a small to moderate. 33,77,78 The specificity effect of acupoints for pain relief remains controversial, which needs further investigation.

| Acupuncture versus celecoxib
This study found no clinical therapeutic efficacy difference in knee pain relief between acupuncture treatment and celecoxib, which further validated the analgesic effect of acupuncture. Celecoxib is the most widely used NSAIDs for the medication of KOA. 21 Several researchers also declared that acupuncture was at least as effective as celecoxib in relieving knee pain for KOA patients. 19,79 Although both celecoxib and acupuncture could relieve knee pain for KOA patients in clinical practice, they have different analgesia mechanisms.
The neuroimaging results showed that KOA knee pain patients had increased vlPAG rs-FC with bilateral thalamus compared with HCs. Neuroanatomical studies have found that the thalamus has neural fiber connections with PAG through the thalamus-PAG pathway, [80][81][82][83] which is part of the pathway in DPMS. This finding indicated that KOA knee pain might have impaired DPMS function. This study also found that the acupuncture (VA + SA) group exhibited a significantly attenuated rs-FC between the vlPAG and the right DLPFC and right angular, compared to the celecoxib group which were also correlated with suppression in KOA pain severity. These results indicate that celecoxib and acupuncture treatment might have different modulation effects on DPMS in KOA knee pain patients. Previous studies have reported that celecoxib could exert an analgesic effect through peripheral and central mechanisms for many chronic pain conditions. [26][27][28][29][30] Nevertheless, this study found that celecoxib did not affect vlPAG rs-FC in KOA patients. It is believed that celecoxib might    (7) the fMRI data with cardiac-gating was not obtained 104 and the respiratory or heart rate records were not noted to account for probable physiological and movement artifacts. In data pre-processing, however, the motion, CSF, and white matter signals were all regressed away. Notably, the possible physiological/movement impact required no region-specificity. Additionally, the absence of significant control seeds in the ventricle confirmed the conclusion. Further investigations may be required to compare fMRI data in the absence or presence of cardiac gating and physiological adjustments.

| CON CLUS ION
Acupuncture treatment, celecoxib, and placebo medication have different modulation effects on vlPAG DPMS in KOA knee pain patients. Acupuncture could modulate vlPAG rs-FC with brain regions associated with cognitive control, attention, and reappraisal for knee pain relief in KOA patients, compared with celecoxib and placebo medication. No.PC2019012).

CO N FLI C T O F I NTER E S T S TATEM ENT
None of the authors have any conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.