Article Text

other Versions

Identification of biological risk factors for persistent postoperative pain after total knee arthroplasty
  1. Alexandra Sideris1,
  2. Michael-Alexander Malahias2,
  3. George Birch1,
  4. Haoyan Zhong1,
  5. Valeria Rotundo1,
  6. Brian J Like1,
  7. Miguel Otero3,
  8. Peter K Sculco2 and
  9. Meghan Kirksey1,4
  1. 1Department of Anesthesiology, Critical Care, and Pain Management, Hospital for Special Surgery, New York, New York, USA
  2. 2Stavros Niarchos Foundation Complex Joint Reconstruction Center, Hospital for Special Surgery, New York, New York, USA
  3. 3Orthopedic Soft Tissue Research Program, Weill Cornell Medical College, New York, New York, USA
  4. 4Department of Anesthesiology, Weill Cornell Medicine, New York, New York, USA
  1. Correspondence to Dr Meghan Kirksey, Department of Anesthesiology, Critical Care, and Pain Management, Hospital for Special Surgery, New York, NY 10021, USA; KirkseyM{at}HSS.EDU


Background There is growing evidence that cytokines and adipokines are associated with osteoarthritis (OA) severity, progression, and severity of associated pain. However, the cytokine response to total knee arthroplasty (TKA) and its association with persistent postoperative pain is not well understood. This study aims to describe the perioperative systemic (plasma) and local (synovial fluid) cytokine profiles of patients who do and do not develop persistent pain after TKA.

Methods Patients undergoing primary unilateral TKA for end-stage OA were prospectively enrolled. Demographic and clinical data were gathered preoperatively and postoperatively. Synovial fluid was collected pre arthrotomy and plasma was collected at multiple time points before and after surgery. Persistent postoperative pain (PPP) was defined as Numerical Rating Score≥4 at 6 months. Cytokine levels were measured using the V-Plex Human Cytokine 30-Plex Panel (Mesoscale—Rockville, Maryland, USA). Cytokine levels were compared between PPP and minimal pain groups. Given that the study outcomes are exploratory, no adjustment was performed for multiple testing.

Results Incidence of persistent pain at 6 months post TKA was 15/162 (9.3%). Postoperative plasma levels of four cytokines were significantly different in patients who developed persistent postoperative pain: interleukin (IL)-10, IL-1β, vascular endothelial growth factor, and IL12/IL23p40. Significantly lower IL-10 levels in the prearthrotomy synovial fluid were associated with development of postoperative persistent pain.

Conclusions This prospective cohort study described a distinct acute perioperative inflammatory response profile in patients who developed persistent post-TKA pain, characterized by significant differences in four cytokines over the first 2 postoperative days. These results support the growing evidence that the patient-specific biologic response to surgery may influence longer-term clinical outcomes after TKA.

Trial registration number NCT02626533.

  • pain
  • postoperative
  • chronic pain
  • outcomes

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information. All relevant data are included in the article or as uploaded supplementary information.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The number of annual total knee arthroplasties (TKAs) performed in the USA is estimated to reach 1.28 million by 2030.1 Despite this high number of TKA procedures, the clinical and biological contributors to persistent postoperative pain following TKA are not well understood. Approximately 20% of TKA patients experience persistent pain at 6 months postoperatively.2 3 In addition, persistent postoperative pain is the highest predictor of patient dissatisfaction following TKA and negatively impacts overall quality of life.4 5 The high volume of TKA procedures performed each year and the relatively high rates of long-term postoperative pain need to be better understood, particularly the patient-specific biologic risk factors that may influence persistent postoperative pain.

There is increasing evidence that cytokines and adipokines are associated with osteoarthritis (OA)-associated pain severity. Rodent studies have associated a number of knee soft tissue or serum cytokines, such as interleukin (IL)-1β, IL-6, IL-10, and tumor necrosis factor (TNF)-α, to OA-associated pain syndromes, including central sensitization and inflammatory pain.6 7 There are many surgeon, patient, and implant-related factors that may contribute to persistent postoperative pain after TKA for OA, including retained osteophytes,8 incorrect sizing, nerve injury, and malalignment.9 Studies are just beginning to explore how patient-specific biologic response to surgery may influence postoperative TKA outcomes.10 Low levels of adiponectin, high levels of leptin, and low adiponectin/leptin ratios in synovial fluid have been associated with increased preoperative pain severity in patients undergoing TKA.11 In addition, leptin and IL-6 have also been associated with increased pain levels among preoperative patients.12 Finally, a recent pilot study has suggested an association between chemokine pathways and persistent pain after TKA.13 While these limited studies indicate associations between preoperative biologic markers and pain, significant gaps remain in our understanding of how the immune and inflammatory systems may contribute to pain in the postoperative period following TKA.

The primary aim of this study was to analyze the relationship, if any, between persistent postoperative pain scores and perioperative cytokine levels in a large, well-characterized, prospectively enrolled cohort of patients undergoing unilateral TKA. Both systemic (plasma) and local (synovial fluid) perioperative cytokine profiles were analyzed to test our hypothesis that distinct perioperative plasma and synovial fluid cytokine profiles would be associated with persistent postoperative pain 6 months following TKA.


One hundred seventy-nine patients with end-stage OA were enrolled in this single-institution, prospective cohort study as previously described.14 Data were hosted and collected electronically using Clinical and Translational Science Center’s Research Electronic Data Capture (REDCap). A total of 17 patients were excluded from analysis following enrolment for reasons including discovery of exclusion criteria following enrolment, protocol violations, or withdrawal/loss to follow-up, leaving 162 patients for analysis (figure 1).

Figure 1

Patient flow diagram. NSAIDs, non-steroidal anti-inflammatory drugs; OA, osteoarthritis.

Enrolment criteria included adult patients undergoing unilateral TKA for severe end-stage OA from a single academic institution between May 2016 and February 2018. OA was deemed severe by radiologist or surgeon using descriptors of ‘severe narrowing’ and/or ‘bone on bone’.

Exclusion criteria included contraindications to non-steroidal anti-inflammatory drugs, acetaminophen, dexamethasone, or regional anesthesia. Patients with a history of daily opioid use of 6 weeks or greater or any usage of non-prescribed opioids were excluded. Patients receiving a periarticular injection for postoperative pain were also excluded. A history or diagnosis of any rheumatic or autoimmune disease, post-traumatic OA, crystalline arthropathy, American Society of Anesthesiologists physical status score>3, current pregnancy, and any active infections or current antibiotic use were also excluded.

Intraoperatively, all patients received neuraxial anesthesia and sedation with midazolam, propofol, and up to 100 mcg of fentanyl at the discretion of the anesthesiologist. Four milligrams of ondansetron and four milligrams of dexamethasone were given to all patients intraoperatively for nausea prophylaxis. For multimodal postoperative pain control, patients received an ultrasound guided adductor canal/subsartorial saphenous nerve block, patient-controlled epidural analgesia until the afternoon of postoperative day 1 (POD1), non-steroidal anti-inflammatories (ketorolac, meloxicam, and/or celecoxib), acetaminophen, and a sliding scale oral opioid as needed. Following anesthesia timeout but prior to surgery, the anesthesiologist drew approximately 8 mL of blood. Synovial fluid was drawn by the surgeon from the knee capsule prior to arthrotomy. Blood was drawn postoperatively on arrival to the postanesthesia care unit (PACU) and again at 24 and 48 hours postoperatively. Blood samples were centrifuged for plasma extraction and plasma and synovial fluid were stored in a −20 freezer overnight before being moved to a −80°C freezer the following day. Numerical Rating Scores (NRS) (0–10) at surgical site (to identify pain at rest, with movement, worst pain, and least pain) were collected immediately preoperatively on day of surgery, POD1, POD2, 4 weeks postoperatively and 3 months and 6 months postoperatively. Responses were collected as a mix of in-person and by phone call follow-up. Pain Catastrophizing Scale (PCS) scores were collected by questionnaire from all patients on POD2. Finally, a painDETECT Score, used to determine neuropathic pain, was collected immediately preoperatively on day of surgery, 4 weeks postoperatively, 3 months postoperatively, and 6 months postoperatively. Patients who reported an NRS pain score of 4 or greater with activity at 6 months postoperatively were designated as the persistent postoperative pain (PPP) group. Patients with NRS pain scores of 3 or less with activity at 6 months were defined as the minimal pain (MP) group.

Cytokine levels were measured using the V-Plex Human Cytokine 30-Plex Panel and two adipokines (adiponectin, leptin) were measured using individual assay plates (Mesoscale—Rockville, Maryland, USA).

Statistical analysis

Categorical demographics and baseline variables were summarized as counts and percentages (%) and were compared between PPP and MP groups using χ2 or Fisher’s exact test. Continuous demographic and baseline variables are presented as means and SD and were compared between groups using t-test.

NRS pains scores measured at multiple time points before 6 months (preoperatively, POD1, POD2, 4 weeks, 3 months) were compared between groups using generalized estimating equations (GEEs) with a random intercept and unstructured covariance structure.

All continuous cytokine values are presented as medians and interquartile ranges. All cytokine values were log transformed for analysis due to skewed distributions. Undetectable cytokine levels with a minimum detection limit are imputed as 1/2 of the detection limit value. Undetectable cytokine levels with a maximum detection limit are imputed as the detection limit value.15 Batch effect was identified in synovial fluid samples. A linear model for microarray data package in R was used after normalizing the cytokine values to remove the batch effect.16

Log-transformed plasma cytokine levels were compared between the PPP and MP groups via generalized linear modeling using the GEE approach. GEE was used to account for the correlation between repeated cytokine measurements at preoperatively baseline, early intraoperatively, POD1, and POD2, for the same patient, adjusting for log-transformed baseline plasma cytokine levels. Effect sizes are presented as exponentiated geometric means with 95% CIs for each group and ratios of geometric means between PPP and MP groups with 95% CIs. Log-transformed joint fluid cytokine levels were compared at baseline between PPP and MP groups using Wilcoxon rank test.

Given the study outcomes are exploratory, no adjustment was performed for multiple testing. All statistical hypothesis tests were two sided. Statistical analyses were performed with SAS V.9.4 (SAS Institute) and R V.3.6.0.


Overall, 15 out of 162 (9.3%) patients met criteria for persistent postoperative pain (PPP) at 6 months following TKA (NRS≥4). The remaining patients (147/162) had NRS pain of 3 or below (MP group). PPP and MP patients were comparable in their demographic and baseline clinical characteristics (table 1). In addition, tobacco use, statin use, exposure to intraoperative tranexamic acid, and postoperative deep vein thrombosis prophylaxis did not differ between groups (table 1). PPP patients had higher baseline worst NRS pain and NRS with movement and immediately postoperatively had higher worst NRS and NRS at rest and with movement (table 2). Patients who had a high PCS scores were more likely to develop PPP compared with those who had lower scores (online supplemental table 1).

Supplemental material

Table 1

Patient and baseline characteristics

Table 2

NRS pain scores over time in patients who develop PPP versus MP patients at 6 months

Of the 32 cytokines and adiponectins assayed, perioperative levels of 4 cytokines differed significantly between PPP and MP patients (figure 2, online supplemental table 2). When comparing blood cytokines levels at each time point, IL-1β is significantly higher on POD1 in the PPP patients (p<0.001; figure 2A), IL-10 is significantly lower in the PACU (early postop) of PPP patients (p=0.03, figure 2B), IL12/IL23p40 is significantly lower on POD1 (p=0.02) and POD2 (p=0.04) in PPP patients (figure 2C), and vascular endothelial growth factor (VEGF) is significantly lower on POD1 in PPP patients (p=0.02, figure 2D). With regard to prearthrotomy synovial fluid cytokine levels, only IL-10 significantly differed between the two groups and is higher in patients with MP at 6 months (table 3, online supplemental table 3).

Supplemental material

Supplemental material

Table 3

Synovial fluid cytokine levels between persistent postoperative pain and minimal pain (MP) groups at baseline

Figure 2

Log-transformed cytokine levels from baseline to POD2. Plots represents geometric means with 95% CIs from calculated data for (A) interleukin-1 (IL-1) β, (B) IL-10, (C) IL-12–23_p40, and (D) vascular endothelial growth factor (VEGF). Bold lines represent patients with MP and dotted lines represent patients with PPP at 6 months. *p<0.05; ***p<0.001. MP, minimal pain; PACU, postanesthesia care unit; POD, postoperative day.


In this study of 162 patients undergoing primary unilateral TKA at a single institution, the incidence of persistent postoperative pain (PPP) at 6 months was 15/162 (9.3%). PPP was defined as an NRS≥4 with activity, which is considered as moderate-to-severe pain and is well established as a trigger for a more comprehensive assessment and treatment for pain.17 18 While the underlying cause of PPP is likely multifactorial, we found significant associations between specific perioperative cytokine profiles and the risk of PPP at 6 months following TKA. Specifically, patients suffering from PPP were found to have had significantly different levels of four cytokines (IL-10, IL-1β, VEGF, IL-12/23p40) in the perioperative period. We have recently described correlations between early postoperative cytokine levels and development of stiffness 6 weeks after TKA.14 Based on the findings presented here, we suggest that the acute inflammatory biologic response to TKA surgery in the first 2 days postoperatively may also correlate with longer-term clinical/functional outcomes, including risk for chronic pain after TKA. Various extrinsic risk factors for persistent pain after TKA have been described, including preoperative knee pain, pain in other sites, pain catastrophizing, anxiety, depression, younger age, and comorbidities.19 This study adds to the limited literature exploring potential biological risk factors for developing PPP.

Pre-existing pain sensitization in the central nervous system is a preoperative factor that likely influences the risk of chronic postoperative pain following TKA.20 Indeed, our findings suggest that higher pain catastrophizing scores are associated with PPP after TKA. It is possible that the perioperative cytokine differences described in this paper reflect elements of an underlying predisposition to pain that is reflected in the acute immune/inflammatory response to surgery. However, the statistically significant differences in levels of specific cytokines between PPP and MP patients should be interpreted with caution, as our findings indicate an association and do not prove a clinically significant causal relationship. The mechanisms through which the immune/inflammatory system may act in the pathogenesis of chronic postoperative pain are not known. Buvanendran et al recently described differences in chemokine gene expression profiles in four patients who developed chronic pain in a pilot study of 16 TKA patients.13 Our results describing differing cytokine levels in 15 PPP patients from a cohort of 162 reinforce the need for further clinical and preclinical studies to elucidate the potential role of the immune system in the development of persistent pain after TKA.

Previous studies suggest an association of VEGF with OA severity and pain in patients with knee OA;21–23 however, little is known about the impact of perioperative VEGF on outcomes following TKA. Here, we present evidence that VEGF plasma levels rise markedly on POD1; however, this rise is less pronounced in patients who developed PPP (p=0.02). The potential association between early postoperative VEGF plasma levels, VEGF periarticular levels, and development of persistent postoperative pain is an area worthy of further study. It may be that lower VEGF levels immediately after surgery contribute to development of a relatively hypoxemic environment in the joint that contributes to or is reflective of pain sensitization.

IL-1β is a proinflammatory cytokine that has been implicated in both acute and chronic pain states.24 In this study, we describe a significantly higher spike in IL-1β plasma levels on POD1 in patients who went on to develop PPP (p<0.001). Similarly, in a recent study, IL1_beta levels were significantly lower (p < 0.03) in 10 patients with stable, painless, well-functioning, cemented TKAs or total hip arthroplasties compared with patients with aseptically loosened, painful arthroplasties.25 Our results support the hypothesis that IL-1β may play a significant role in the development of persistent pain following TKA.

Clinical observations have established that IL-12/23p40 is integral to numerous immune-mediated disorders, such as psoriasis, psoriatic arthritis, rheumatoid arthritis, systemic lupus erythematosus, spondyloarthritis, and Crohn’s disease. However, to our knowledge, this is the first study to report a significant association between early postoperative plasma levels of IL-12/23p40 (p=0.02 at POD1; p=0.04 at POD2) and PPP after TKA. Further evidence is required to clarify the impact of IL-12/23p40 on the immune/inflammatory response to surgery and investigate pathways mediated by IL-12/23p40 that may impact persistent postoperative pain following TKA.

The only cytokine for which we detected significantly different baseline synovial fluid levels in PPP patients compared with MP patients was IL-10 (lower in PPP; p=0.05). IL-10 was also the only cytokine found significantly different between groups when measured on PACU arrival (lower in PPP; p=0.03). IL-10 is an anti-inflammatory cytokine that has been shown to have immunosuppressive effects in inflammatory arthritis26 as well as specific chondroprotective effects.27 Lower baseline synovial fluid levels of IL-10 may be associated with a distinct pattern of cartilage damage and/or inflammatory end-stage OA phenotype that predisposes patients to develop PPP. The low PACU plasma levels of IL-10 are presumably due to acute spillover into the blood during surgery and may serve as confirmation of our findings in the synovial fluid. In a secondary analysis of a randomized controlled trial, Langkilde et al reported that surgery-induced IL-10 early response was associated with decreased functional performance (6 min walk test) 26 weeks after TKA.28 In another randomized trial, Singh et al reported that after injection of botulinum toxin A for chronic painful TKA, increased serum IL-10 levels correlated with improved response.29 In the context of the established literature, our findings support the hypothesis that both preoperative and postoperative IL-10 play a role in the pathogenesis of persistent TKA pain. Moreover, findings from experimental models of arthritis suggest that IL-10 may be a pathway amenable to intervention.27

Several cytokines and chemokines which have previously been associated with pain were not found to be differentially represented in our outcome groups. Gandhi et al identified greater synovial fluid concentrations of TNF-α, matrix metalloproteinase 13 (MMP-13), and IL-6 as predictors of a lesser pain improvement 2 years after TKA.30 In another recent study, serum IL-6 and leptin levels were associated with preoperative pain levels in TKA patients,12 but not acute postoperative pain. In our study, perioperative TNF-α, IL-6, and leptin levels were not found to be significantly different between patients determined to have PPP and MP at 6 months postoperatively. These distinct findings may reflect differences in assay sensitivity, sample size, patient demographics, and metrics and timing of pain assessment. MMP-13 was not measured in our study. Clinically relevant associations of immune/inflammatory markers with pain chronicity after TKA must be validated through more detailed exploration in large well-characterized and diverse cohorts.

Identification of factors associated with outcomes serves numerous purposes including characterization of biological mechanisms associated with the development of and protection against PPP, subsequent stratification of patients based on risk, and potential interventions with available or emerging therapeutics that may modify the course of recovery. For example, promising gene therapy immune modulators, such as XT-150, are currently under investigation in clinical trials for pain via their ability to increase IL-10 levels.

This study was not without limitations. Implants were not standardized and there were variations in surgical technique, rehabilitation protocols, and scheduled length of stay (LOS). However, it can be argued that these limitations strengthen our findings, since they might apply across different types of implants, incisions, rehabilitation protocols, and LOS. In addition, while postoperative persistent pain was determined at 6 months, we did not measure cytokine levels at time points after discharge. We were thus unable to determine the later kinetics of the cytokines that differed on POD1 and POD2. Knowing the temporal pattern of cytokine expression after POD2 may better inform our mechanistic understanding of development of PPP and would be necessary prior to consideration of potential therapeutic interventions and dosing regimens.


This prospective cohort study described a distinct perioperative cytokine profiles in patients who develop persistent post-TKA pain at 6 months, characterized by significant differences in four plasma cytokines during the first 2 PODs after TKA. These results support the growing evidence that patient-specific biologic responses to surgery may influence longer-term clinical outcomes after TKA. Future research directed toward early control of inflammatory cytokines may identify interventions to reduce post-TKA persistent pain.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information. All relevant data are included in the article or as uploaded supplementary information.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and ethical approval for this work was obtained from Institutional Review Board of Hospital for Special Surgery (IRB#2015-361). Participants gave informed consent to participate in the study before taking part.


Supplementary materials


  • Contributors AS: helped in data analysis, interpretation of results, manuscript preparation, and review. M-AM: helped in interpretation of results, manuscript preparation, and review. GB: helped in study data collection, interpretation of results, and manuscript review. HZ: conducted data analysis, interpretation of results, manuscript preparation, and review. VR: helped in study planning, data collection, manuscript preparation, and review. MO: helped in study design/planning, interpretation of results, manuscript preparation, and review. BJL: helped with data interpretation, manuscript preparation, and review. PKS: helped in study design/planning, interpretation of results, manuscript preparation, and review. MK: study design/planning, interpretation of results, manuscript preparation and review, and guarantor.

  • Funding Research reported in this publication was supported by the National Center for Advancing Translational Science of the National Institute of Health Under Award Number UL1TR002384, the Department of Anesthesiology, Critical Care & Pain Management Research and Education Fund, and the Adult Reconstruction and Joint Replacement Marmor Award at Hospital for Special Surgery (HSS). MK was also supported by the HSS Department of Anesthesiology Young Investigator Award.

  • Competing interests AS is supported by the C.V. Starr Foundation and is an unpaid consultant on a cannabis research study grant funded by Colombia Cientifica/Colciencias awarded to Pontifical Xavierian University.

  • Provenance and peer review Not commissioned; externally peer reviewed.