Background
Total knee arthroplasty (TKA) is one of the most successful procedures performed in patients with end-stage osteoarthritis (OA) and other arthritides with refractory pain [
1]. While most patients report improved pain and function outcomes after TKA, persistent pain is reported by 6.5% or more patients post-TKA [
2].
The role that cytokines and neurotransmitters play in joint pain is of great interest and has been the focus of recent research [
3]. In an animal model of OA, cytokine levels were higher in OA joints, which were reduced by treatment with coenzyme Q-10 that also led to significant pain relief [
4]. TNF-α levels were associated with WOMAC pain scores in patients with OA [
5]. Joint fluid substance P (SP) levels in knee joints were elevated in painful knee joints with OA [
6] and in preoperative joint fluid in those with greater pain relief after knee arthroplasty [
6]. Studies showed that higher levels of cytokines such as interleukin-8 (IL-8), IL-6 and tumor necrosis factor-alpha (TNF-α) were associated with implant loosening in patients with joint arthroplasty [
7,
8]. To our knowledge, there are limited or no data regarding the role of cytokines in pain severity in patients with painful TKA or relief with interventions.
We examined the data from a randomized trial of patients with persistently painful TKA [
9] to test the hypotheses whether in patients with painful TKA, change (baseline to 2-months) in serum cytokine, chemokine and substance P (neuropeptide) levels was associated with clinically meaningful pain relief at 2-months post-injection. We also assessed whether baseline to 2-month change in cytokine/chemokine/substance P level correlated with change in pain severity from baseline to 2 months.
Results
Patients in the randomized trial had a mean age of 67 years, 84% were male and 96% were Caucasian. The mean duration of TKA pain was 4.5 years (SD, 4.8) and 75% were primary TKAs.
Compared to non-responders (
n = 23) on WOMAC pain scale at 2-months, pain responders (
n = 12) had significantly greater increase in serum levels of IL-7, IL-10, IL-12 (p70), eotaxin, IFN- γ and TNF-α from baseline to 2-months post-injection (
p < 0.05 for all; Table
2). Several other cytokines showed a non-significant trend by pain-responder status (
p ≤0.32; Table
2). In sensitivity analysis in a smaller set of patients, who reported daytime NRS pain on 0–10 scale, serum substance P decreased significantly more in the pain responders (0.54 ± 0.53;
n = 10) than in the pain non-responders (0.48 ± 1.18;
n = 9;
p = 0.023) 2-months post-injection.
Table 2
Association of change in serum cytokine and neurotransmitter levels from baseline to 2-months with pain responder status on WOMAC pain at 2-month post-injection in painful TKA
Interleukin (IL)-7 | No | 23 | 0.084 | 0.91 | 0.19 |
0.01
|
Yes | 12 | 1.07 | 1.17 | 0.34 | |
IL-10 | No | 23 | 8.51 | 20.90 | 4.36 |
0.01
|
Yes | 12 | 27.72 | 21.56 | 6.22 | |
IL-12 p70 | No | 23 | 3.36 | 8.17 | 1.70 |
0.004
|
Yes | 12 | 12.91 | 9.60 | 2.77 | |
Eotaxin | No | 23 | −2.03 | 13.92 | 2.90 |
0.046
|
Yes | 12 | 7.85 | 12.28 | 3.54 | |
Interferon gamma (IFN-γ) | No | 23 | −1.24 | 23.18 | 4.83 |
0.03
|
Yes | 12 | 15.61 | 13.35 | 3.85 | |
Tumor necrosis factor-alpha (TNF-α) | No | 23 | 4.46 | 21.32 | 4.44 |
0.03
|
Yes | 12 | 22.22 | 24.65 | 7.12 | |
IL-4 | No | 23 | −0.02 | 0.16 | 0.03 | 0.12 |
Yes | 12 | 0.06 | 0.12 | 0.03 | |
IL-6 | No | 23 | −15.90 | 275.66 | 57.48 | 0.09 |
Yes | 12 | 137.35 | 182.09 | 52.56 | |
IL-13 | No | 23 | 4.90 | 13.25 | 2.76 | 0.15 |
Yes | 12 | 12.18 | 14.87 | 4.29 | |
IL-15 | No | 23 | 0.31 | 11.87 | 2.47 | 0.09 |
Yes | 12 | 7.36 | 10.52 | 3.038 | |
Macrophage Inflammatory Protein-1 beta (mip1b) | No | 23 | 5.55 | 91.43 | 19.06 | 0.16 |
Yes | 12 | 53.35 | 96.39 | 27.82 | |
Substance P | No | 14 | 0.08 | 1.11 | 0.298 | 0.32 |
Yes | 5 | −0.46 | 0.61 | 0.273 | |
Change in several cytokine and chemokine levels from pre-injection to 2-month follow-up correlated significantly with change in WOMAC pain with correlation coefficients ranging −0.37 to −0.51: IL-2, IL-7, IL-8, IL-9, IL-16, IL-12 (p70), GCSF, IFN- γ, IP-10, MCP, MIP1b, TNF-α and VEGF (
n = 35; Table
3). In sensitivity analysis, we additionally noted a change in serum substance P from pre-injection to 2-month follow-up correlated significantly with change in daytime NRS pain, correlation coefficient was 0.53 (
p = 0.021;
n = 19).
Table 3
Non-parametric correlation of change in serum cytokine and neurotransmitter levels from baseline to 2-months with change in WOMAC pain at 2-month post-injection in painful TKA, showing statistically significant associations
IL2 | −0.37 | 0.03 |
IL7 | −0.42 | 0.01 |
IL8 | −0.51 |
0.002
|
IL9 | −0.50 |
0.002
|
IL16 | −0.48 |
0.004
|
IL12p70 | −0.56 |
<0.001
|
GCSF | −0.34 | 0.04 |
IFN gamma | −0.48 |
0.003
|
IP10 | −0.38 | 0.02 |
MCP | −0.35 | 0.03 |
MIP1b | −0.49 |
0.003
|
TNF-alpha | −0.42 | 0.01 |
VEGF | −0.38 | 0.02 |
Discussion
In this ancillary pilot study, we assessed whether changes in serum levels of cytokines, chemokines and Substance P from baseline to 2-month post-injection were associated with clinically meaningful pain relief at 2-month post-injection. We performed a mechanistic study and used data from our randomized study of intra-articular injection for painful TKA [
9]. In many patients, the pain improvement lasted through the 6-month follow-up period, indicating that the joint pain relief was somewhat durable [
9]. We found that several cytokine levels, including IL-7, IL-10, IL-12 (p70), eotaxin, IFN- γ and TNF-α, changed significantly more in WOMAC pain responders compared to pain non-responders (responders defined as those with pain decrement of 20 points or more on 0-100 scale). Correlation analyses identified additional cytokines with moderate correlations with WOMAC pain scores (baseline to 2-month change in cytokine level with baseline to 2-month change in WOMAC pain) in addition to these, including IL-2, IL-8, IL-9, IL-16, GCSF, IP-10, MCP, MIP1b and VEGF. In sensitivity analysis using a smaller dataset, serum substance P level reduction was greater in pain responders using the 0–10 daytime NRS pain than pain non-responders (pain decrement of two points or more o 0–10 scale). The direction and magnitude were similar to the responder analysis by WOMAC pain, however, standard deviations were larger in WOMAC pain analysis, leading to the difference in substance P levels being non-significant (
p = 0.32) in the main analysis, but significant in sensitivity analysis (
p = 0.023). This is a hypothesis-generating study and therefore, these findings need to be replicated in future studies.
The traditional view of cytokines/chemokines being either anti- or pro-inflammatory [
15] has been challenged [
16], since they can serve either role depending on the condition and the body organ. Higher levels of pro-inflammatory cytokines have been linked to pain [
17‐
20]. On the other hand, change in pro- and anti-inflammatory cytokines with treatment in other pain conditions do not map precisely to their associations with pain condition at baseline [
21‐
24]. Studies that have investigated the potential mechanisms of joint pain relief in intervention studies are lacking. Such studies can provide insights into mechanisms of action of an intervention, and discover mediators of joint pain relief in patients with OA. Our study begins to fill this knowledge gap by providing data among patients with painful joint arthroplasty.
Other recent uncontrolled studies have documented a potential role of cytokines, chemokines and Substance P in the failure of total joint replacement. In a recent study, both IL-1 beta and IL-2 levels were significantly lower (
p < 0.025) in 10 patients with stable, painless, well-functioning, cemented total knee or hip arthroplasties (TKA/THA) than patients with aseptically loosened, painful, arthroplasties [
25]. Genetic variants of pro-inflammatory cytokines TNF-alpha and IL-6 were associated with susceptibility to severe osteolysis after THA, a condition that is associated with increasing pain and functional limitation [
7]. Compared to OA, aseptic loosening of TKA was associated with up-regulated expression of several cytokines including IL-8 and MMP9 and low levels of inflammatory cytokines [
8].
In patients undergoing revision surgery for painful primary hip arthroplasty, nerve fibers with positive immunostaining to Substance P were found in bone-prosthesis interface membranes [
26]. Joint fluid Substance P levels were elevated in painful knee joints with osteoarthritis that underwent TKA, but not in normal/asymptomatic contralateral knees [
6]. Significantly greater pain relief after knee arthroplasty was seen in patients with an elevated preoperative joint fluid Substance P level compared to patients with normal Substance P levels [
6]. In an animal model of OA, MMP-13, IL-1b, IL-6 and IL-15 were up-regulated in OA joints and the levels were reduced by treatment with coenzyme Q-10 that led to significant pain relief [
4]. In a study of 47 OA patients, TNF-alpha levels were associated with WOMAC pain and overall scores [
5]. Thus, cytokine levels have been correlated with OA and arthroplasty joint pain in observational studies. Our study provides evidence showing that a change in cytokine levels correlated with pain relief in patients with painful TKA in a clinical trial who had a clinically meaningful pain relief after an intra-articular injection.
Our study findings must be interpreted considering study limitations. The sample size for this ancillary study was small. A protocol modification was made to the main study on receipt of federal funding that allowed us to collect specimens and perform analyses, and therefore these analyses were performed on a subset of the main trial, since almost 40% of the study cohort had already been enrolled in the study. Therefore, we likely missed several important findings due to type II error. Subsequent studies should enroll a larger number of patients to explore this important aspect of treatment for patients with painful arthroplasty. We examined the patients in the pain responder vs. non-responder categories, as specified a priori in our analytic plan. Since our hypothesis was to assess the cytokine, chemokine and neuropeptide mediators of joint pain relief in painful arthroplasty, we limited our analyses by whether patient was or was not a responder by clinically meaningful pain relief at 2-months post-injection. We adopted this approach rather than comparing Botulinum toxin group vs. placebo group, since we only had samples on half of the study participants in the RCT and both groups of patients with samples consisted of a mix of pain responders and non-responders, leading to an inadequate sample size for this comparison. We believe that the associations we noted are unlikely to be drug-specific, since responders included patients from both intervention and control arms. This analysis addressed our main objective to better understand the mechanism of pain improvement in patients who have had a failed TKA, and are undergoing a medical treatment to improve pain. Study strengths include the randomized, blinded study design, use of paired samples and focus on a condition that has significant public health impact. The presence of significant associations, even under the stringent categorization (responders/non-responders), with the current sample size, supports our underlying theoretical framework that cytokine, chemokine and substance P level changes after an intra-articular injection might be associated with pain relief.