Comparison of efficacy of arthroscopic lavage plus administration of corticosteroids, arthroscopic lavage plus administration of placebo, and joint aspiration plus administration of corticosteroids in arthritis of the knee: A randomized controlled trial

Patients.

Patients were recruited from the outpatient clinic of the Department of Rheumatology of the Leiden University Medical Center (LUMC) between July 2002 and December 2003. Eligibility criteria included a minimum age of 18 years and arthritis of the knee not due to gout, osteoarthritis, or infection. Exclusion criteria were hemorrhagic disease, oral prednisone >10 mg/day, recent (<6 weeks) change of disease-modifying antirheumatic drugs (DMARDs), recent joint injection (any joint), and hypersensitivity to methylprednisolone or local anesthetics. Patients were allowed to use acetaminophen, nonsteroidal antiinflammatory drugs, other analgesics, and DMARDs, and any changes in dose regimen were recorded. The protocol was approved by the LUMC Ethics Committee and written informed consent was obtained from all patients.

Interventions.

Patients were randomized to 1 of 3 interventions based on a computer-generated list prepared by the statistician (JKS). Allocation of the interventions was based on the chronological order of registration and was performed by an independent physician unaware of patients' identities. Patients and the research nurses and treating rheumatologists who served as assessors (see below) were not informed about the results of randomization. Each procedure was performed in an outpatient setting under aseptic conditions in a designated room with a setup aimed at maximal blinding of the patient (dimmed lights, supine position). The 3 interventions were arthroscopic lavage with administration of corticosteroids (ALC), arthroscopic lavage with administration of placebo (ALP), and joint aspiration with administration of corticosteroids (JAC). All interventions were performed by one person (MO) and included the following consecutive steps: 1) maximal aspiration of synovial fluid from the suprapatellar pouch, 2) intraarticular injection of 30 ml of lidocaine 0.5% through the aspiration needle, and 3) additional infiltration of the skin and joint capsule with 10 ml of lidocaine 1%. The subsequent steps of each intervention are described below.

Outcome measures and followup.

The primary outcome measure was event-free survival defined as the time in months after treatment until local re-treatment (joint aspiration or injection, arthroscopy, or [radio]synovectomy) because of recurrence or persistence of arthritis of the knee as determined by the treating rheumatologist, or nonimprovement of knee joint score assessed by trained research nurses. The knee score (range 0–7) encompasses knee tenderness as scored from 0 to 3 (0 = no tenderness, 1 = tenderness when asked, 2 = tenderness on pressure, and 3 = tenderness and wincing) (10), knee swelling as scored from 0 to 3 (0 = no swelling, 1 = little swelling, 2 = moderate swelling, and 3 = abundant swelling) (11), and a patient visual analog scale (VAS; range 0–100 where 0 = no pain and 100 = maximal pain) for knee pain as scored from 0 to 1 (VAS score divided by 100). Secondary outcomes were adverse events, serologic, and immunohistochemical parameters. Furthermore, tender and swollen 28-joint scores were obtained in all patients. In patients with rheumatoid arthritis (RA) (12), a Disease Activity Score in 28 joints was calculated, which is a composite score for joint pain and tenderness in 28 joints, erythrocyte sedimentation rate (ESR), and a VAS for general well-being (range 0–100) (13). Laboratory investigations included IgM rheumatoid factor, full blood count, ESR, C-reactive protein (CRP) level, and renal and liver function tests. At baseline, standard weight-bearing radiographs of both knees were made. Radiographs were scored for severity of osteoarthritis on the Kellgren and Lawrence (K/L) scale (14) for joint space narrowing, osteophytes, and sclerosis. This score was divided into 2 categories depending on the absence (K/L score 0–1) or presence (K/L score 2–4) of osteoarthritis.

Synovial tissue analysis.

Synovial biopsy samples were obtained from all patients in a standardized manner (15). Synovial specimens were embedded in paraffin, sectioned at 3 μm, and stained with hematoxylin and eosin. Coded sections were scored for the following histologic features: mean synovial lining layer thickness from 6 random places; mean number of infiltrating lymphocytes, plasma cells, and neutrophils from 3 random high-power fields (× 400); vascularity expressed as the number of blood vessels with detectable endothelial cells in 3 high-power fields (× 200); and presence of fibrosis and fibrin (range 0–4, where 0 = absent and 4 = abundantly present).

Statistical analysis.

The determination of the sample size was based on detectable differences in the primary end point, event-free survival. A difference in event-free survival of 25% after 9 months (9) between any 2 of the 3 interventions was calculated to have a 90% power with a sample size of 26 patients per treatment arm (α = 0.05) and a dropout rate of 20%. The cumulative incidence of events for treatment groups was determined by constructing Kaplan-Meier survival curves. A possible difference in recurrence rate between the treatments was examined through a variant of the Cox proportional hazards regression model that allows for the participation of patients in all treatments by robust variance estimates (16). Dichotomous variables were created for clinical and synovial characteristics. The Cox model was used to evaluate a possible confounding effect of dichotomized characteristics on the recurrence rate. Furthermore, we assessed whether the recurrence rate after treatments differed between categories of clinical parameters. The Cox model produces a regression coefficient for each covariate that can be interpreted as a relative risk (RR) when taken exponentially. Other statistical analyses were performed using Student's t-test, Wilcoxon's signed rank test, analysis of variance, and chi-square test where appropriate. All statistical analyses were performed with the statistical software package STATA 6.0 (Stata, College Station, TX). P values less than 0.05 were considered statistically significant.

Patient characteristics.

A total of 78 patients with arthritis of the knee were enrolled in the study. Of these patients, 75 were analyzed: 51 with RA, 15 with spondylarthropathy, and 9 with undifferentiated arthritis. Three patients allocated to ALP were excluded from analysis because of a treatment violation (the patients insisted on treatment with ALC and were treated accordingly) and 3 patients (1 in each group) were lost to followup. A flow diagram of the study is shown in Figure 1. The treatment groups were comparable at baseline regarding demographic, clinical, and synovial characteristics except for synovial lining layer thickness and neutrophils, which were statistically lower in the JAC group (P = 0.005 and P = 0.04, respectively) (Table 1).

Clinical efficacy.

ALC was the most effective therapy, followed by JAC and ALP (Figure 2). The median event-free survival time was 9.6 months after ALC, 3.0 months after JAC, and 1.0 month after ALP. The estimated RR of an event during the 9 months following intervention was 2.2 (95% confidence interval [95% CI] 1.2–4.2, P = 0.02) for JAC and 4.7 (95% CI 2.3–9.4, P < 0.0001) for ALP compared with ALC. These results were independent of the number of previous joint injections (data not shown). The RR between ALP and JAC was 2.0 (95% CI 1.1–3.8, P = 0.01) in favor of joint aspiration. Twenty-two patients (13 after ALC, 8 after JAC, and 1 after ALP; P < 0.01) completed the 9 months without recurrence of arthritis. Thirty-seven patients (8 after ALC, 13 after JAC, and 16 after ALP; P = 0.26) reached the primary end point because of re-treatment and 15 patients (5 in each group) reached the end point because of nonimprovement of the knee score. Thirteen of 22 patients who did not experience an event had no DMARD change during followup (9 after ALC, 3 after JAC, and 1 after ALP; P = 0.02). The number of patients with a change of DMARD therapy (regardless of recurrence of arthritis) per month followup was skewed between interventions: 0.04 for ALC versus 0.06 for ALP and JAC (P = 0.36). There was no difference in disease activity or ESR between the groups during the followup period (P = 0.73 and P = 0.12, respectively). Eighteen patients had >1 local reintervention of the knee within the year following treatment.

The knee score showed a greater decrease after ALC compared with both of the other interventions (Figure 3). The difference in decrease was significant for ALC compared with ALP after 1 month (1.93 versus 0.08; P < 0.01) and 3 months (1.63 versus 0.86; P < 0.04), and for JAC compared with ALP at 1 month (1.77 versus 0.08; P < 0.01). After 6 months, too many patients had had a reintervention to make meaningful comparisons (6 in ALC, 13 in JAC, and 13 in ALP).

Clinical and synovial factors and treatment effect.

Using the Cox proportional hazards regression analysis, no clinical or synovial factors were found to have a confounding effect on the differences in responses between treatment groups. For example, although the number of patients with a diagnosis of spondylarthropathy was lower in the ALP group, this was not associated with a significant difference in outcome (P = 0.35). Similarly, the difference in synovial lining thickness in biopsy samples obtained at baseline did not have a significant impact on outcome either (P = 0.22). Clinical and synovial factors were also studied as predictors for treatment outcome and were dichotomized to calculate an RR within one treatment arm between both subgroups of a variable. In continuous variables, the median served as the arbitrary cutoff point for dichotomization. A greater degree of fibrosis compared with a lesser degree in the synovial tissue (≥2 versus <2, 2 representing the median) conferred an RR of 5.7 (95% CI 1.6–20.5, P < 0.01) after ALC (Figure 4). The effect of fibrosis on outcome was confirmed when the knee score was taken as a parameter. ALC resulted in a more pronounced decrease in knee score in patients with a low versus high extent of fibrosis: 2.5 versus 1.5 at 1 month (P = 0.02), 2.5 versus 1.0 at 3 months (P < 0.01), and 1.8 versus 1.5 at 6 months (P = 0.27). No differences were detectable in the other treatment arms. None of the other histologic features had a statistically significant impact on outcome (data not shown). Patients with >1 previous injection had an RR of 2.4 (95% CI 1.5–6.3, P = 0.006), and patients with 0–1 injections had an RR of 2.0 (95% CI 1.2–3.2, P = 0.04) for recurrence within 9 months after JAC when compared with ALC.

Combination of variables.

Correlation analysis demonstrated an incomplete overlap of absence of fibrosis and lymphocyte infiltration in a proportion of the patients. The presence of fibrosis showed an inverse relationship with the number of infiltrating lymphocytes in all 3 groups (Spearman's ρ = −0.43, P = 0.01). ESR at baseline correlated inversely with synovial fibrosis in the lavage with corticosteroid group (ρ = 0.55, P = 0.016). A correlation was found between the leukocyte count (granulocytes) in synovial fluid and CRP levels (ρ = 0.63, P < 0.001) and between synovial fluid leukocyte count and knee swelling (ρ = 0.25, P = 0.045).

Adverse events.

No treatment-related infections occurred during the study. One patient (ALC) died of sudden cardiac arrest 6 months after the intervention. One patient (ALP) was diagnosed with lymphoma shortly after treatment. One patient (JAC) developed hemarthrosis, 1 patient (ALP) experienced prolonged skin bleeding, and 3 patients (1 in each group) experienced transient local symptoms around the superior skin incision. All intervention-related side effects resolved spontaneously.