The main finding in the present study is that non-responders had more joint destruction and higher levels of VEGF in synovial fluid than responders. Neither the autoimmune profile, the ongoing cartilage turnover nor local or systemic inflammation seems to influence the relapse rate.
In this large prospective cohort study we have analysed several possible outcome predictors of IAGC for knee synovitis. Many previous outcome investigations include patients with different diagnoses. As different joints and different rheumatic diseases may have different predictors, in this study we have focused on knee injections in RA patients.
Smoking has been identified as an exogenous factor of importance in the pathogenesis of RA. Furthermore, treatment with methotrexate and biological agents is less successful in smokers , but results in the present study indicate that smoking has no influence on the outcome of IAGC in RA.
Postinjection rest of weight-bearing joints has been shown to improve the response to IAGC in arthritis (4), but our findings show that the body weight itself did not differ between responders and non-responders.
In a randomised controlled trial Jahangier et al.  compared knee injections with the combination of yttrium-90 + 20 mg THA with the combination of placebo + 20 mg THA on 66 patients with different arthritides. The overall response rate was 48% and no significant difference between the groups was found at six months. Pretreatment synovial biopsies showed significantly more synovial macrophages in patients responding on IAGC. Furthermore, Luukiainen et al.  studied 30 RA patients with knee synovitis after IAGC with 30 mg THA and found a significant correlation between polymorphonuclear leucocytes in synovial fluid and reduction of joint circumference after six months. This parameter had no influence on the relapse rate in our study.
The CIMESTRA study  was designed as a randomised controlled trial on recent onset RA with intense IAGC (14 mg betamethasone for each knee injection) combined with either cyclosporine A and methotrexate or sulfasalazine and methotrexate. The tight control regimen used made early detection of arthritis relapse possible. A 48% relapse rate after 200 days was found among the 89 included patients with knee synovitis. In the present investigation we used 20 mg THA on patients with longer disease duration (median 10 years) and found a 40% relapse rate after 180 days and we defined relapse as a request for another injection confirmed with synovitis on clinical examination. Despite differences in study design and relapse definition the difference in general relapse rate between the studies is small.
The CIMESTRA trial showed that the effect of the first joint injection had a longer duration than a subsequent injection in the same joint and the presence of anti-CCP was weakly associated with longer injection survival. This is not confirmed in our study, which, in contrast, showed a non-significant trend towards more relapse among the anti-CCP positive patients.
In a study by af Klint et al., 31 patients were examined with synovial biopsies before and two weeks after IAGC using 40 mg THA for knee synovitis . The results showed marked reduction of synovial T-lymphocytes, TNF, IL1 and VEGF, but no significant change in synovial vascularity or the expression of VEGF on endothelial cells. The authors discussed whether this finding was important for the transient effect of IAGC. This hypothesis is supported by our finding of significantly higher levels of VEGF in synovial fluid of non-responders in the present study.
VEGF is normally produced by synovial cells in response to hypoxia and proinflammatory cytokines, such as TNF and IL6 . In accordance, levels of IL6 and VEGF in synovial fluid correlate strongly in our study. VEGF stimulates the formation of new blood vessels and increases vascular permeability. This is important for production of the synovial fluid that represents an ultrafiltrate of plasma as the synovial membrane lacks basement membranes towards the synovial cavity. VEGF may also protect synoviocytes from apoptosis , thereby further contributing to synovial hyperplasia. The high levels of VEGF in the synovial fluid of IAGC non-responders in the present study may have stimulated synovial vascularity and rapid recurrence of blood perfusion and probably the larger joint effusion as well. This facilitates recruitment of inflammatory cells and may explain the increased risk for relapse. The fact that this difference in VEGF levels between responders and non-responders was found locally in the treatment target organ, but not systemically, further argues for a pathogenetic impact of VEGF in the joints in steroid-resistant RA patients. The role of sf-VEGF in this study is in agreement with animal studies on the role of VEGF in arthritis. VEGF knockout mice showed reduced pathology and less synovial angiogenesis in both antigen-induced and collagen-induced models of arthritis . An anti-VEGF antibody has also been shown to prevent collagen-induced arthritis and reduced established disease activity in mice . Antibodies to VEGF receptor 1 have also been shown to attenuate disease activity in mice . These findings suggest the possibility that local anti-VEGF administration might be an adjunct therapy to THA in RA patients with high synovial fluid levels of VEGF.
Compared to synovial fluid, the VEGF levels in serum are significantly lower and show no difference between the responder groups, indicating that VEGF levels or VEGF-associated pathology in the local knee environment determines treatment outcome. None of the systemic variables analysed seems to be important, suggesting a limited value of analysing serum samples before IAGC to predict clinical outcome.
However, in juvenile chronic arthritis a high erythrocyte sedimentation rate was correlated with a better response to IAGC for knee synovitis . Remission of synovitis of the knee after IAGC lasted longer, not only if concomitantly treated with methotrexate, but also with a procedure including general anaesthesia . The latter finding was unexpected and the authors discussed whether the injection placement in a child is easier during general anaesthesia.
Not surprisingly, the radiographic changes in our patients were correlated both with disease duration and age. For elderly RA patients comorbidity with osteoarthritis (OA) cannot be excluded. How these joint diseases interact is not known, and comparing the outcome of IAGC between OA and RA is difficult because of differences in pathogenesis, but in the late stages of both diseases there are more mechanical than immunological causes for inflammation. RA patients with severe joint destruction may, therefore, respond more like patients with severe OA. A systematic review of outcome predictors in OA  concluded that results are not consistent, but some of the reviewed studies suggest that the presence of effusion and the severity of disease have an impact on treatment outcome. As in several OA studies, we found an association between degree of joint damage on radiographs and outcome of IAGC. This has never been shown in RA before. COMP levels, however, did not differ between responders and non-responders, which suggest that the ongoing cartilage turnover has only minor effects on the effect of IAGC in RA.
A weakness of the study may be the relapse definition, based on the complaints of the patient, and a confirming clinical examination. The patient delay may vary due to individual thresholds for discomfort, pain and need for help. On the other hand, this method has been used before [3, 24] and is relevant because it reflects the real life in clinical practice.
In this investigation we have chosen to use 20 mg THA, as this is the recommended dose for knee injections in Sweden. Higher doses are often used, but there are no published dose finding studies comparing the efficacy of different THA dosages for knee injections, and consequently, there is a large variation in the tradition of using THA.