Inclusion and Exclusion Criteria
Inclusion criteria were an age of eighteen years or older, knee pain of recent onset (less than six months), and physical examination findings and magnetic resonance imaging results consistent with meniscal pathology.
Exclusion criteria were an age of less than eighteen years, a recent history (within three months) of an intra-articular injection of a corticosteroid, and a past or current history of autoimmune disease (such as rheumatoid arthritis). In addition, no patient involved in a Workers' Compensation claim or personal injury litigation was enrolled in the study.
Subjects and Knee Lavage
This prospective cohort study was approved by the institutional review boards of Parkway Regional Medical Center and the University of Miami, Jackson Memorial Medical Center. The trial registration number is NCT00836966 (). From June 2006 through June 2007, fifty subjects who met the above inclusion criteria were offered enrollment in the study and all fifty provided informed consent to participate. These subjects were selected from a cohort of approximately 200 consecutive patients referred for the evaluation of knee pain to two board-certified orthopaedic surgeons who were fellowship-trained in sports medicine. For all patients, a history was obtained, a physical examination was performed, and magnetic resonance imaging scans of the knee were acquired. The magnetic resonance imaging scans were read and interpreted by a single, independent, and experimentally blinded, board-certified radiologist. The study cohort consisted of adult patients who had knee pain for less than six months with either an acute or insidious onset and who had failed nonoperative treatment and elected to have arthroscopic management. Indications for surgery included the presence of mechanical symptoms; a physical examination with a positive McMurray sign or joint-line tenderness, or both1; an absence of severe joint-space narrowing on plain radiographs10,11; and grade-III signal changes on magnetic resonance imaging12,13 in an anatomic location consistent with the history and physical examination.
Eighteen subjects were later excluded from the study because of the presence of an anterior cruciate ligament tear and/or a large joint effusion (eleven patients), sample loss during shipment (five patients), or surgical exploration that was negative for meniscal injury despite positive findings on magnetic resonance imaging (two patients). Therefore, thirty-two subjects in the surgical group contributed data to the study.
Prior to arthroscopy and/or lavage, study patients were asked to rate the pain in both knees on a previously validated verbal pain-rating scale14 of 0 to 10 points, with a score of 0 indicating no pain and a score of 10, the worst pain imaginable. A knee with any score other than 0 was considered to have pain for the sake of grouping. These scores were later correlated to the intra-articular cytokine results. Patients deemed appropriate to undergo arthroscopic surgery as part of their treatment were invited to participate in the study and consented to intraoperative lavage of the involved knee or both knees. With the patient under regional anesthesia, a synovial fluid aspirate was obtained by means of lavage in the operatively treated knee or both knees, with use of approximately 15 mL of sterile physiologic saline solution, allowing the withdrawal of 3 to 5 mL of knee lavage aspirate. The aspirated fluid was then placed into 2-mL tubes containing 130 µL of protease inhibitor cocktail tablets (Roche Diagnostics, Indianapolis, Indiana) dissolved in phosphate-buffered saline solution (0.045 tablet per milliliter of sample) with a pH of 7.4 and was frozen at -20°C temporarily until being shipped on dry ice to Stanford University, where samples were subsequently aliquoted and stored at -80°C. Intraoperative findings were catalogued, and any chondral defects were graded with use of the Outerbridge classification for subjects in the arthroscopy group15.
In addition to the surgical group, fifteen volunteers were enrolled in the study after consent was obtained for lavage of a knee without a recent injury or current pain and with no history of knee surgery. These individuals underwent office-based knee lavage. The skin was sterilized with 10% povidone-iodine solution and was anesthetized with a topical anesthetic. The knee lavage was performed as described above for intraoperative lavage.
The two orthopaedic surgeons who performed the knee lavage procedures and operations, as well as the radiologist, completed all aspects of the study in a manner blinded to the assay procedure and results.
Cytokine Analysis
The concentrations of seventeen inflammatory cytokines and chemokines (IFN-? [interferon gamma], IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12, IL-13, IL-17, G-CSF [granulocyte-colony-stimulating factor], GM-CSF [granulocyte-macrophage colony-stimulating factor], TNF-a, IL-1ß, MCP-1, and MIP-1ß [macrophage inflammatory protein-1 beta]) were quantified in knee lavage samples with use of a human multiplex inflammatory cytokine panel and the Bio-Plex 200 System (Bio-Rad, Hercules, California), following the manufacturer's protocol in a ninety-six-well plate format. The choice of cytokines reflects the standard, prechosen panel available through Bio-Rad. This cohort was intended to represent typical inflammatory molecules in a variety of conditions that we hypothesized were also likely to be present in our study. This assay utilizes antibodies linked to polystyrene beads containing different levels of fluorophores and has been validated against standard enzyme-linked immunosorbent assays (ELISAs) of human blood samples16.
ELISA Cross-Validation
Although the Bio-Plex assay has been cross-validated against ELISA with use of human blood samples16, we performed another cross-validation experiment for two cytokines, with a majority of the knee lavage samples. Sample volume limitations prevented further cross-validation testing.
IL-6
Fifty-six samples were assayed with use of a human IL-6 Quantikine ELISA kit (D6050; R and D Systems, Minneapolis, Minnesota) according to the manufacturer's protocol.
MCP-1
Forty-six samples were assayed with use of a human MCP-1 ELISA kit (DCP00; R and D Systems) according to the manufacturer's protocol.
Statistical Analyses
Multiple analyses of variance were performed to compare mean concentrations of all seventeen cytokines between samples from asymptomatic knees and painful knees. Multiple hypotheses were tested in the analysis of data because of the multivariate nature of the dependent variables. The testing of multiple hypotheses can lead to false-positive results, a phenomenon that is encountered in many panel-type assays (for example, DNA microarrays and multiplexed bead immunoassays). This problem is mitigated by the use of multivariate analysis of variance, which partially accounts for this effect for any one test. However, p values in the range 0.005 to 0.05 should be viewed with some caution because of the multiple comparisons made. The asymptomatic knee group includes the contralateral knees of symptomatic subjects, which may violate the independence assumptions of multivariate analysis of variance. To address this important issue, we analyzed the data using a repeated-measures approach including a within-subjects effect in addition to the between-subjects effect. The within-subjects effect was not found to be significant by multivariate analysis of variance with a single binary within-subjects factor (p > 0.05).
Correlation analyses of cytokine concentration compared with pain score or one cytokine concentration compared with another cytokine concentration were performed with use of Spearman correlation coefficient analysis. A p value of <0.05 was taken as significant.
Source of Funding
No external funding was used for this project.
Cytokine analysis was completed for forty-seven subjects, consisting of thirty-two patients undergoing unilateral knee arthroscopy for acute pain and fifteen asymptomatic control subjects. Of the thirty-two patients who underwent knee arthroscopy and completed the study, twenty-three also provided a sample from the contralateral knee. Therefore, cytokine assay data were collected from a total of seventy knees. All patients undergoing arthroscopy had physical examination findings consistent with meniscal abnormality (joint-line tenderness and/or a positive McMurray sign) on the operatively treated side.
Pain Scores, Age, and Sex Profile of the Patients by Group
The mean self-reported pain score (and standard error of the mean) for the normal (volunteer) control knees, the operatively treated knees, and the nonoperatively treated, contralateral knees of the patients was 0 ± 0, 6.3 ± 0.2, and 0.7 ± 0.2, respectively. The mean age of the normal control group, the operatively treated knee group, and the nonoperatively treated knee group was 43 ± 3, 54 ± 3, and 54 ± 3 years, respectively (p > 0.05, analysis of variance). The male-to-female ratio of the normal control group, the operatively treated knee group, and the nonoperatively treated knee group was 7:8, 23:9, and 17:6, respectively. The ratio of male to female was significantly different from 1:1 in the operatively treated and nonoperatively treated knee groups (p < 0.05).
Magnetic Resonance Imaging Findings by Group
Asymptomatic Volunteers
Fifteen subjects completed magnetic resonance imaging evaluation of the knee, which identified nonpainful meniscal abnormalities in three knees (the posterior horn of the medial meniscus was involved in one knee and the medial meniscus, in two).
Symptomatic Subjects—Operatively Treated Knee
All symptomatic, operatively treated knees had a meniscal injury identified on magnetic resonance imaging.
Symptomatic Subjects—Nonoperatively Treated Knee
Nine of the twenty-three patients who agreed to have lavage of the contralateral knee also completed magnetic resonance imaging of that knee. All nine of them were found to have a degenerative meniscal abnormality (seven involved the posterior horn of the medial meniscus; one, the anterolateral portion of the medial meniscus; and one, the lateral meniscus). None of these knees were found to have an anterior cruciate ligament injury.
Intraoperative Findings
Of thirty-four patients who underwent knee arthroscopy, thirty-two had confirmation of the preoperative diagnosis of a meniscal tear on magnetic resonance imaging. However, for two knees, the operative exploration failed to confirm the meniscal injury identified by magnetic resonance imaging, representing possible false-positive findings.
Cytokine Assay Results
Cytokine data were grouped according to the presence or absence of acute knee pain. Results from all thirty-one nonpainful knees (fifteen normal controls and sixteen contralateral, nonoperatively treated knees) were grouped and compared with those from all thirty-nine painful knees (Fig. 1, Table I). Multivariate analysis of variance detected significantly greater mean concentrations of cytokines (and standard error of the mean) in lavage samples from painful knees compared with those from nonpainful knees for eight of the seventeen cytokines (in picograms per milliliter of knee lavage fluid) including IFN-? (1558 ± 556 and 188 ± 113, respectively; p = 0.03), IL-6 (214 ± 83.0 and 0.0 ± 0.0; p = 0.026), MCP-1 (58.0 ± 15.1 and 2.4 ± 1.3; p = 0.002), MIP-1ß (24.2 ± 5.0 and 4.2 ± 3.0; p = 0.002), IL-2 (5.6 ± 2.1 and 0.0 ± 0.0; p = 0.023), IL-4 (2.0 ± 0.7 and 0.2 ± 0.1; p = 0.025), IL-10 (3.9 ± 1.4 and 0.1 ± 0.0; p = 0.016), and IL-13 (3.4 ± 1.3 and 0.3 ± 0.3; p = 0.036) (Fig. 1).
A significant positive correlation was detected between patient-reported pain scores and concentrations of the following cytokines: IL-6 (Spearman ? = 0.7; p < 0.001), MCP-1 (? = 0.8; p < 0.001), MIP-1ß (? = 0.6; p < 0.001), and IFN-? (? = 0.6; p < 0.001). In addition, the concentration of these four cytokines was significantly correlated to each other (? = 0.5 to 0.7; p < 0.001) (MCP-1 compared with IL-6 and MIP-1ß). Although the p value reached significance in the covariant analysis for IL-2, IL-13, IL-7, IL-10, and G-CSF compared with reported pain scores, the correlation coefficient values were low (? = 0.16 to 0.26). With the numbers studied, there was no correlation between the duration of symptoms in the arthroscopy group and the cytokine concentration (p > 0.05). No significant differences in cytokine levels were observed with regard to sex (p > 0.05).
Within the operatively treated group, no significant difference was detected, with the numbers studied, between the mean concentration of cytokines measured in knee lavage samples from the thirteen knees with moderate effusions compared with the nineteen knees with minimal effusions (p < 0.05).
The median Outerbridge classification score in the arthroscopy group determined intraoperatively was 3 (range, 0 to 4). A weak, negative correlation was observed between the Outerbridge score and intra-articular IL-2 (p < 0.05; ? = 0.39); the correlation was not significant for any other cytokine.
Table II presents the sensitivity of IL-6, MCP-1, MIP-1ß, and IFN-? alone or in a combination compared with magnetic resonance imaging of the knee alone.
ELISA Compared with Bio-Plex Cross-Validation Assays
IL-6
The concentrations of IL-6 in knee samples, as determined by the Bio-Plex assay, were not found to differ significantly from those measured with use of standard ELISA (p = 0.23, Student paired t test), and a significant correlation was detected between the two assays (Pearson correlation coefficient, r = 0.7; p < 0.001; two tailed) (data not shown).
MCP-1
The concentration of MCP-1 determined by ELISA was, on the average, fivefold greater than that determined by the Bio-Plex assay, and the two assay means were significantly different (p < 0.001). However, relative to each other, the two assays were significantly correlated (r = 0.9; p < 0.0001; two tailed) (data not shown).
We attempted to identify an intra-articular inflammatory cytokine profile that may assist the physician in evaluating patients with knee pain. We analyzed and compared the intra-articular aspirate of thirty-nine painful knees with that of thirty-one asymptomatic knees. In addition, these results were compared with magnetic resonance imaging findings from both groups of knees and the intraoperative findings in knees undergoing arthroscopy. Although we observed significantly greater mean concentrations of eight inflammatory cytokines in the painful knees compared with the nonpainful knees, four of these cytokines, IL-6, MCP-1, MIP-1ß, and IFN-?, also had a significant positive correlation to self-reported pain scores and were present most consistently in painful knees. This cytokine profile was not present in knees with nonpainful meniscal abnormalities identified on magnetic resonance imaging.
It has been previously demonstrated in animal models that various inflammatory cytokines are elevated in response to tissue injury17-19, possibly related to leukocyte-mediated wound-healing mechanisms19,20. This process may result in enhanced nociception or pain21-23 by means of cytokine-induced sensitization of nociceptors24-26, which have been shown to innervate the outer third of the human meniscus27. An acute response diminishing over time has been suggested, with several studies that have described cytokine proliferation in the acute phase of injury9,17,19,28. It has recently been shown that IL-6 can sensitize nociceptive C-fibers innervating the rat knee joint29, possibly by inducing the cyclooxygenase-prostaglandin-E2 inflammatory pathway30,31. There is also recent evidence that IL-6 may be involved in the transition from peripheral tissue injury and inflammation to central sensitization and thus pathological pain states32. However, the pattern of cytokine release has not yet been correlated to patient-reported knee pain.
We propose that a diagnostic test panel of these four cytokines fulfills criteria for diagnostic testing and meets the qualifications for validity proposed in the Sackett and Haynes model33. The approach proposed by Sackett and Haynes centers on four fundamental questions or so-called phases that, they argue, any diagnostic test needs to address in order to achieve maximum validity. Although currently the mainstay of decision-making in the treatment of meniscal tears, magnetic resonance imaging clearly fails the first Sackett and Haynes criterion, that is, diagnostic test results in affected patients must differ from those in normal individuals. The fact that many physicians may base operative treatment on magnetic resonance imaging findings of knee pathology has been called into question by numerous studies that have identified age-related meniscal changes as pathological3,34-36.
When used as a diagnostic test to detect painful meniscal tears in the present study, the presence of MCP-1, IL-6, MIP-1ß, or IFN-? alone was highly sensitive and specific (Table II). If the presence of any one of these cytokines is taken as positive (i.e., MCP-1 or IL-6, etc.), the sensitivity and specificity was 100%, outperforming magnetic resonance imaging compared with the so-called gold standard of intraoperative findings. This suggests that the cytokine test result might distinguish patients with and without painful meniscal tears among patients in whom it is clinically reasonable to suspect a meniscal injury. A larger study is necessary to confirm this preliminary finding, but these observations are promising for the development of a diagnostic test that is highly accurate, rapid, and inexpensive.
Although substantial effort was made to adhere to a standard sample collection and assay protocol, there was unavoidable variability in some sample handling parameters. Interindividual precision of the volume of sterile saline solution injected into and withdrawn from the knee joint was difficult to maintain. In addition, some knee lavage samples were more contaminated with blood than others. Although the injected volume of saline solution varied, depending on the amount that was required to enable withdrawal of a 2 to 3-mL fluid sample, it raised the possibility for bias toward greater dilution of cytokines in patients without an effusion compared with those with some effusion. However, this variation had no apparent pattern and thus was equally distributed among subjects in the operatively treated, nonoperatively treated, and control groups, who all had lavage performed in the same manner. Therefore, we do not believe that this factor introduced substantial bias or confounding effects. Furthermore, immunoassay measurements in the comparison of biological fluids are more likely to provide relative concentrations rather than absolute values. The lavage technique whereby saline solution was injected into the joint space was necessary since fluid aspiration is difficult or impossible in most patients unless a large effusion is present.
We believe there was a potential selectivity bias in the study for the participation of subjects who consented to have lavage of the contralateral knee; younger, healthier patients might be less likely to allow the healthy knee to be lavaged, for fear of a negative outcome. Indeed, a substantial number of subjects refused to have the contralateral knee lavaged. We recognize the inherent but unavoidable selection bias as approximately 28% of the patients in our study with a painful knee declined lavage of the other, nonpainful knee. It is likely that this bias, however, would tend toward reducing the difference between symptoms in the knee undergoing arthroscopy and those in the contralateral knee since younger, less symptomatic subjects would be less likely to volunteer the healthy knee for an invasive lavage procedure.
Finally, the asymptomatic knees include the contralateral knees of symptomatic subjects, which may violate the independence assumptions of multivariate analysis of variance. To address this important issue, we analyzed the data using a repeated-measures approach including a within-subjects effect in addition to the between-subjects effect. The within-subjects effect was not found to be significant by multivariate analysis of variance with a single binary within-subjects factor (p > 0.05).
In conclusion, we observed a greater concentration of a combination of four inflammatory cytokines and/or chemokines, IL-6, MCP-1, MIP-1ß, and IFN-?, in intra-articular lavage samples from painful knees of patients with an intraoperatively confirmed meniscal injury compared with lavage samples from nonpainful, control knees. These cytokines were nearly absent in the knees of fifteen normal, age-matched, asymptomatic control subjects. Compared with the findings of preoperative magnetic resonance imaging and intraoperative diagnosis, the assay of these four cytokines performed as well as or better than magnetic resonance imaging. These findings suggest that these cytokines may serve as the basis for further development of a diagnostic test. 