Levels of preoperative cerebrospinal fluid pro‐inflammatory mediators and chronic pain after total knee arthroplasty surgery

Patients undergoing total knee arthroplasty (TKA) surgery are at high risk of chronic postsurgical pain (CPSP). Accumulating evidence suggests an active role of neuroinflammation in chronic pain. However, its role in the progression to CPSP following TKA surgery remains unanswered. Here, we examined the associations between preoperative neuroinflammatory states and pre‐ and postsurgical chronic pain in TKA surgery.

Conclusions: We identified the CSF fractalkine level as a potential predictor for CPSP severity following TKA surgery. In addition, our study provided novel insights into the potential role of neuroinflammatory mediators in the pathogenesis of CPSP.

K E Y W O R D S
cerebrospinal fluid, chronic postsurgical pain, fractalkine, neuroinflammation, total knee arthroplasty

Editorial Comment
Persistent pain can be an undesirable result after total knee arthroplasty and the mechanism for development of this are not completely understood. In this explorative study, cerebrospinal fluid fractalkine level was found to be associated with severity of chronic postsurgical pain in this cohort. Further study of postoperative neuroinflammation and persistent pain mechanisms is warranted.

| INTRODUCTION
Despite significant improvements in perioperative pain management, chronic postsurgical pain (CPSP), which persists even after the surgical wound has healed, remains a huge medico-social problem. 1 Although numerous risk factors, such as younger age, female sex, genetic factors, psycho-emotional factors, and preoperative chronic pain, have been reported, 2 the lack of understanding of the mechanisms underlying progression to CPSP has precluded the implementation of effective preventive and therapeutic measures.
Total knee arthroplasty (TKA), a joint replacement surgery for knee joint destruction due to osteoarthritis (OA), rheumatoid arthritis (RA), or trauma, has been recognized as a high-risk procedure for CPSP, with a reported incidence of 13%-44%. 2 Although the mechanisms by which preoperative pain leads to CPSP are unclear, preoperative chronic knee arthralgia is a possible risk factor. 3 Neuroinflammation is a local inflammation of tissues within the peripheral and central nervous systems (CNS), which is characterized by the activation of immune cells and increased production of inflammatory mediators at these sites. 4 Accumulating evidence suggests a pivotal role of CNS neuroinflammation in the development and maintenance of chronic pain. 5,6 Notably, the microglia, which are bone marrow-derived immune cells in the CNS, may be key players in neuroinflammation via the release of pro-inflammatory mediators in response to peripheral tissue injury. 7,8 These centrally released mediators purportedly induce central sensitization, leading to the augmentation and prolongation of pain. [9][10][11] Such preoperatively established neuroinflammation in painful knee diseases may contribute to the progression of CPSP.
Moreover, emerging evidence suggests that preoperative painrelated psycho-emotional distress, such as anxiety, pain catastrophizing, and sleep disturbances, may contribute to the transition to CPSP. 2,12 However, little is known about the association between neuroinflammation and a maladaptive psycho-emotional status.
In this prospective cohort study with patients undergoing TKA surgery, we first examined the correlations between preoperative multidimensional pain-related parameters and pro-inflammatory mediator levels in cerebrospinal fluid (CSF) collected immediately before surgery. Second, we analyzed the associations between preoperative multidimensional pain-related parameters, CSF proinflammatory mediator levels, and CPSP intensity 6 months postsurgery to explore the involvement of preoperative neuroinflammation in the transition to CPSP. Finally, we sought to establish a prediction model for the intensity of CPSP using perioperative variables, including preoperative CSF pro-inflammatory mediator levels.

| Study setting and participants
This prospective cohort study was conducted with patients who underwent TKA between September 2017 and October 2019 at Keio University Hospital, Tokyo, Japan. The inclusion criteria were: age of 40-80 years, OA or RA as the primary reason for elective TKA surgery, and chronic knee arthralgia (average numerical rating score >4/10) with a pain duration >3 months. In addition, patients were excluded if they had any of the following conditions: contraindications for spinal anesthesia, severe cognitive or mental disorders, active CNS disease, or pregnancy.

| Data collection
Data on baseline demographics, comorbidities, preoperative painrelated parameters, and surgical and anesthetic parameters were prospectively collected from electronic medical records.

| Multidimensional preoperative pain assessment
All the patients were asked to complete questionnaires, after admission to the hospital, before TKA. All the questionnaires were validated in Japanese versions. Preoperative pain severity and interference associated with knee arthralgia were measured using the brief pain inventory (BPI), which includes 11-point numeric rating scales. 13 Pain intensity was calculated as the mean of the four items that assessed pain at its worst, its least, on average, and currently, while pain interference was the mean of the items that evaluated pain interference related to seven domains of life (i.e., general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life). Preoperative anxiety and depression were assessed using the hospital anxiety and depression scale (HADS), a 14-item questionnaire with scores ranging from 0 to 3. 14,15 Half of the questions assessed anxiety, and the rest assessed depression. Each subscale score ranges from 0 to 21. For each item, a score of ≥11 indicates a clinically significant disorder, whereas a score of 8-10 suggests a mild disorder. 15 The involvement of neuropathic pain-like components in knee arthralgia was evaluated using the painDETECT questionnaire, which consists of nine items with possible scores ranging from 0 to 38 and evaluates pain quality, pattern, and radiation. 16 PainDETECT scores ≧19 indicate the presence of neuropathic pain-like symptoms. 17 Preoperative pain catastrophizing thoughts were quantified using the pain catastrophizing scale (PCS), which reflects the frequency of 13 types of pain-related catastrophizing thoughts or feelings in patients (each scored from 0 to 4). The total PCS scores range from 0 to 52, with a cut-off value ≧30 indicating clinical significance. 18,19 2.4 | Collection of CSF and serum samples  ical pathway for TKA was to prepare for patients to discharge from hospital 2-3 weeks after surgery. Patients were allowed to discharge from hospital when they became able to walk with a cane, independently with all transfers.

| Acute postoperative pain assessment
Patients were asked to rate their pain intensity at rest on the visual analog scale (VAS, ranging from 0 = no pain to 100 = worst pain imaginable) 6, 12, 24, 36, 48, 60, and 72 h postsurgery. In addition, the cumulative amount of fentanyl used postoperatively via iv-PCA and the number of PCA bolus requests by postoperative Day 3 were also recorded.

| CPSP assessment
The chronic pain after surgery was characterized by the questionnaires sent to the patients 6 months postsurgery. Based on the answers to the questionnaires and information from the medical records, the diagnosis of CPSP was ascertained by the following criteria, as defined by the IASP classification for ICD-11: postsurgical pain developed or increased in intensity, which had persisted more than 3 months after surgery; pain localized to the surgical field, projected to the innervation territory of a nerve situated in the area or referred to a dermatome; other causes of pain (e.g., infections, malignancy, etc.) were excluded. 20 The intensity of CPSP was quantified in the same manner as BPI pain intensity, as previously described. A BPI pain intensity less than or equal to the median value (= 1.25) was defined as no-to-mild CPSP, while the others were defined as moderate-to-severe CPSP.

| Statistical analyses
In descriptive statistical analyses, categorical data were presented as numbers and percentages, and continuous data as medians and interquartile ranges (IQR). The associations between variables were assessed using Spearman's correlation coefficient (rho) using the cor-

| Associations between CSF/serum analytes and preoperative parameters
The score distributions in the preoperative multidimensional pain assessments and the concentrations of the neuroinflammation markers in the CSF and serum samples are shown in Figure 2. Correlation analyses between the CSF analytes ( Figure 3A, Table S1)  However, no significant correlation was observed between the CSF and serum concentrations of either analyte ( Figure 3B, Table S2). The correlations between the concentrations of CSF analytes and the preoperative demographic and pain-related data are shown in Figure 3C and

| Association of perioperative parameters and persistent postoperative pain 6 months after TKA surgery
The correlations between the preoperative demographics, painrelated data, levels of CSF analytes, and the severity of CPSP 6 months after surgery are shown in Figure 4 and Table S5. Among shown in Table 2. Patients with moderate-to-severe CPSP had significantly higher scores for preoperative BPI pain severity (median Furthermore, the participants with moderate-to-severe CPSP had significantly lower levels of fractalkine (median 3.84 ng mL À1 , IQR: 3.59-4.40 ng mL À1 ; versus no-to-mild CPSP: median 5.15 ng mL À1 , IQR: 4.14-6.24 ng mL À1 ; p < .001) and CSF-1 (median 8.51 pg mL À1 , IQR: 7.48-9.79 pg mL À1 ; versus no-to-mild CPSP: median 10.9 pg mL À1 , IQR: 9.23-12.3 pg mL À1 ; p = .008) in the preoperative CSF than patients with no-to-mild CPSP. After Bonferroni adjustment (0.05/33 = 0.0015), the differences between the groups in the painDETECT and PCS scores, and fractalkine levels, remained significant. However, no significant differences were observed in other parameters between the participants with no-to-mild and moderate-to-severe CPSP 6 months postsurgery. p < .01) and CSF fractalkine levels (β: À.62; 95% CI: À1.10 to À0.15; p = .012) were significant predictors of the severity of CPSP 6 months after TKA (Table 4).

| DISCUSSION
In this study, the preoperative CSF fractalkine level and the PCS score were independent predictors of CPSP intensity 6 months after TKA surgery. To the best of our knowledge, this is the first study to longi- of these spinally released mediators on the initiation and maintenance of pathological pain via activation of microglia, reported in rodent models of chronic pain. 35,36 In addition to elucidating chronic pain models, emerging clinical evidence has also revealed the anti-inflammatory effects of fractalkine and CSF-1 in neurodegenerative disorders, such as Alzheimer's disease and Parkinson's disease. [37][38][39][40][41][42][43] These two factors promote microglial polarization toward an M2-like anti-inflammatory phenotype, suppressing the ability to release proinflammatory factors in rodent models. 44,45 Recently, Palada et al.
reported the protective effects of neuroinflammation in OA patients, which included elevated levels of fractalkine and CSF-1 in the CSF. 32 Considering the pleiotropic effects of these factors, higher preopera-  how the neuroinflammatory status changed before and after surgery; therefore, the mechanism by which the preoperative neuroinflammatory status influenced the development of CPSP remains speculative.
Although ethically more challenging, repeated postoperative CSF sampling may provide more mechanistic insights. A recent study showed the CSF levels of most pro-inflammatory mediators, including IL-6 and IL-8, were comparable before and after surgery in hip OA patients. Interestingly, in this longitudinal study, increases in the levels of CSF IP-10 after surgery were found to be associated with increases in pressure pain thresholds. 47 Additional studies with wider repertoire of inflammatory mediators are needed to establish the causative link between neuroinflammation and CPSP. In addition, the levels of pro-inflammatory mediators in the CSF can potentially be affected by preoperative medications.
The previous report showed no major effects of NSAIDs use on the levels of pro-inflammatory mediators in the CSF. 48  The association between the perioperative pain management and transition to CPSP warrants further research. Furthermore, the quantification of pain and psychological/emotional status in this study were mostly derived from self-evaluated patient responses to the questionnaires.
The incorporation of other modalities, such as functional MRI studies targeting the aforementioned brain area activities, and quantitative sensory testing, may provide additional objective and quantitative information on the associations between preoperative neuroinflammation, preoperative pain status, and CPSP. Moreover, our study lacked data on pro-inflammatory mediator levels in the CSF of healthy controls. Therefore, it was impossible to determine whether the mediator levels in our study patients were relatively high or low compared to those in the general population. Although a Swedish study reported higher fractalkine levels in the CSF of chronic knee OA patients compared to those in CSF controls from younger headache patients with no signs of inflammation in the CNS, 32 this aspect may be worth investigating in larger scales with optimal healthy control groups.
Cumulatively, our study results corroborate the utility of incorpo-

DATA AVAILABILITY STATEMENT
The full-de-identified datasets are available from the corresponding author upon reasonable request.