The relationship between the metabolic activity assessed with 18F–FDG PET (Beckers et al., 2004; Beckers et al., 2006; Polisson et al., 1995; Palmer et al., 1995) or PET/CT (Kubota et al., 2011; Kubota et al., 2009; Okamura et al., 2012) and the inflammatory activity in RA has been repeatedly documented. We have previously shown that changes in SUV were significantly correlated with changes in CRP and MMP-3 serum levels, as well as with various MRI parameters, in patients who were receiving a 4-week anti-TNF-α therapy (Beckers et al., 2006). Okamura et al. showed a significant correlation between changes in SUV and changes in DAS28 in patients receiving, for 6 months, infliximab and etanercept, two anti-TNF-α therapies (Okamura et al., 2012). These two studies suggest that 18F–FDG PET/CT could be used to evaluate the response of RA patients to biologic treatments. In addition, using 18F–FDG PET, Elzinga et al. also showed that changes in SUV 2 weeks after initiating a treatment with infliximab correlated with the clinical evolution at 14 and 24 weeks (Elzinga et al., 2011). We applied 18F–FDG PET/CT to assess, and possibly predict, the response to treatment with rituximab in a highly selected population of RA patients previously resistant to anti-TNF-α therapies. First, we were able to demonstrate a high interobserver reproducibility, as both the visual analysis and SUV measurements showed high agreement, whether at baseline or during treatment. It should be noted however that the overall agreement was superior at baseline than at week 16 in particular in the small articulations such as the MCPs and PIPs. The decreased level of joint inflammation on rituximab combined with partial volume effect may explain this observation. Even though image interpretation was greatly facilitated by the CT part of the PET/CT study, the interobserver agreement was not improved compared to a series obtained with standalone PET (Beckers et al., 2004). US is generally considered a reliable tool for synovitis assessment but the interobserver agreement may be quite variable, with kappa values ranging from 0.22 to 0.868 for the B-mode, according to a recent systematic review (Cheung et al., n.d.). Further, the reproducibility of US in the therapeutic evaluation of RA is unknown.
In the present series, the metabolic results at baseline are correlated with the biological data, the DAS28 and the US results, which is consistent with the literature (Beckers et al., 2004; Beckers et al., 2006; Polisson et al., 1995; Palmer et al., 1995; Kubota et al., 2011; Kubota et al., 2009; Okamura et al., 2012; Elzinga et al., 2011). The principal endpoint of the study was the clinical response at week 24, according to the clinical criteria DAS28. Nevertheless, we added a subcategory of patients according to the clinical response at week 16: the “transient” responders, which corresponds to a response at week 16 followed by an increase in DAS28 at week 24. Indeed, non-responders at week 16 were not eligible for further treatment with rituximab due to Belgian regulations. On the other hand, a transient and a sustained response to a single injection of rituximab may well each represent two different patterns of evolution for the disease. According to the clinical response at week 16, 9 out of the 14 assessable patients were responders and the remaining five were non-responders. At week 24, two months later, the five non-responders remained non-responding, and 6 out of the 9 responders had flares. It must be first be noted that in our limited series, none of the parameters measured at baseline, whether clinical, biological, ultrasonographic or metabolic, were able to anticipate the response to treatment. During follow-up, all 3 patients with a sustained clinical response showed decrease in CRP levels, number of US-positive joints and cumulative synovial thickness. The clinical relevance of these parameters is however greatly limited by the very poor specificity and positive predictive value of both tests. PET/CT was less sensitive as it failed to identify a clinical responder. This patient had only 5 PET-positive joints at baseline, which increased to 6 at week 16. Both US and CRP showed a favorable evolution. A possible explanation may be the inflammatory rebound known under rituximab which may lead to overestimating the disease activity with PET, a classical observation in lymphoma treatment (Moskowitz et al., 2011), although we would expect to observe the phenomenon in a larger proportion of patients. On the contrary, 6/11 patients that eventually did not respond at week 24 showed initial clinical improvement at week 16, but 10/11 did not experience any improvement in at least one of the metabolic parameters analyzed. The most clinically relevant finding of this study is the capacity of PET to identify early on those patients who will eventually fail to respond to the treatment. Indeed, the absence of decrease in the number of hypermetabolic joints predicted the subsequent clinical failure in 91% of the cases. This may lead to a possible integration of PET/CT in the management algorithm, that would allow to move on more rapidly to another line of treatment, in hope of preventing structural damages to the joints.
Elzinga et al. found a strong correlation between the glucose metabolic rate (Mrglu) obtained by Patlak graphical analysis and the SUVs (Elzinga et al., 2011). We compared various methods for assessing the metabolic burden and found that the visual analysis was the most accurate for predicting the outcome. While the visual analysis was strongly correlated with the clinical and biological parameters at baseline, such relationship was lost at week 16 and it turned out to present the highest accuracy for predicting the clinical status at week 24. The cumulative SUVs performed poorly and a composite index taking into account the number of diseased joints fared better, but did not match the results of the visual evaluation. It seems that extinction of inflammation in target joints as evidenced by a decrease in PET-positive joints is a better predictor than a lowering of the overall inflammation, as reflected by the cumulative SUV and the composite SUV index. Contrary to Elzinga et al., we did not cluster the metacarpophalangeal joints into one and we evaluated individual joints, including the knees. This, along with differences in the patients population and treatment schemes, may contribute explaining the differences between our results and theirs. In any event, both series are quite limited and larger trials will be needed to go beyond the feasibility study. The group at Gunma University has performed extensive clinical research in RA patients treated with biologics (Okamura et al., 2012; Okamura et al., 2014; Suto et al., 2016; Suto et al., 2016; Yonemoto et al., 2016). Using the total SUVmax, i.e. the sum of the SUV in the considered joints, they showed a correlation between the evolution of the metabolic activity and the response to tocilizumab (Okamura et al., 2014). Similarly the baseline cumulative SUVmax was related to the subsequent joint destruction as assessed by X-rays (Suto et al., 2016; Yonemoto et al., 2016). In these studies however, the metabolic activity was evaluated in 8 to 12 of larger joints only, i.e. shoulders, elbows, wrists, knees, and possibly hips and ankles. Smaller joints such as the MCPs and PIPs are clinically relevant however, and we show in the present study that the metabolic activity can be recored with a high reproducibility, both using visual and semiquantitative measurements. Dedicated PET/CT devices could further refine the quality of the metabolic assessment (Chaudhari et al., 2016).
The second major observation in our series is that correlations between PET/CT and US parameters remained at week 16, and that changes observed in both technical approaches remained significantly correlated. We therefore confirm in this new set of patients the relationship between the anatomical observation of synovitis through US and the metabolic identification of these synovitis through 18F–FDG-PET/CT assessment, as shown earlier (Beckers et al., 2004; Beckers et al., 2006), as well as between the changes induced by rituximab, as already shown with anti-TNF-α treatments (Beckers et al., 2006). Various other methods are being tested for objectively assessing the response to rituximab treatment. High resolution US provided evidence of significant reduction in synovial hyperplasia after rituximab treatment, identifying response or resistance to rituximab and showing a link between B-cell-directed immune modulation and clinical symptoms (Ziswiler et al., 2009). Nevertheless, the evaluation of response to rituximab in this study was realized 6 months after the first administration, which seems rather late considering that the optimum rituximab re-treatment intervals is not defined and appears highly variable when based on individual’s disease progression (Smolen et al., 2007). A recent study estimated the response to treatment by performing MRI of the metacarpophalangeal joints in 10 patients, at baseline and at week 26 (Fritz et al., 2009). Decrease in volume of synovial enhancement and early rapid enhancement was associated with clinical response at 52 weeks in a subgroup of patients. Along with US and MRI, 18F–FDG PET/CT may provide insights into the biological activity of the disease and help guiding the selection of patients more likely to benefit from a sustained response.
Among the limitations of the present study, we acknowledge in particular that both the number of patients and the duration of the clinical follow up need to be extended. These preliminary results support the proof of concept but larger series are obviously needed to confirm the clinical relevance of this technology in rheumatoid arthritis,