Serum IL-33, a new marker predicting response to rituximab in rheumatoid arthritis

A total of 224 patients with RA for at least 6 months and fulfilling the American College of Rheumatology (ACR) 1987 criteria were included in the SMART study (NCT01126541). This study is a 2-year, national, multicenter, randomized open-label study evaluating the efficacy and tolerability of two doses of RTX for re-treatment after one initial course of RTX at a usual dose (1000 mg on days 1 and 15) described previously [21]. All patients had active disease, defined by a Disease Activity Score in 28 joints (DAS28) using C-reactive protein (CRP) (DAS28-CRP) >3.2, with ≥6/66 swollen and ≥6/68 tender joints, or a CRP ≥10 mg/L, or an erythrocyte sedimentation rate (ESR) ≥28 mm/h. Erosive status was based on the reading of hand and feet X-rays by the investigators in each center. Each patient received a stable dose of methotrexate (MTX) (≥10 mg/week for at least 4 weeks) and had experienced an inadequate response or intolerance to TNF inhibitors, or had contraindications to TNF inhibitors. All patients in the SMART study received one course of RTX (two 1000 mg infusions, given on days 1 and 15) with usual pre-medication (methylprednisolone, acetaminophen, and antihistamine). In this ancillary study, 111 patients from the 224 were included due to the availability of the serum samples for IL-33 assessment and were designated as cohort 1.

We also studied 32 RA patients from Leeds (UK) and 42 from Clermont-Ferrand (France) treated in real life with a first course of RTX using the same schedule. In order to match the numbers of patients in cohort 1, these two groups of patients from Leeds and Clermont-Ferrand were merged and called cohort 2. There were no overlaps between the two cohorts.

For all patients, treatment efficacy was evaluated 24 weeks after the first RTX infusion according to European League Against Rheumatism (EULAR) response [22]. Patients were then classified as responders (good or moderate) or non-responders after this first course of RTX.

We first used an ELISA IL-33 kit (DuoSet, R&D Systems) for protein assessment, and preliminary presented as an abstract an association between serum IL-33 level and RTX response in a single cohort [23]. However, this kit was not validated for human sera and additional experiments have confirmed the need for caution about the accuracy of this kit for sera measurements [24]. Recommendations from the manufacturer advise us to use another assay for human sera, named Quantikine and also provided by R&D Systems. We aimed to use this more accurate ELISA in order to investigate the possible association between a detectable serum level of IL-33 and a response to RTX in RA patients in the different cohorts.

Using a sample obtained prior to the initiation of RTX in the two cohorts, we measured the serum IL-33 level using the Quantikine kit, which is a solid-phase ELISA with a 6.25 pg/mL lower limit threshold according to the manufacturer’s instruction [24]. The assessment of rheumatoid factor (RF), anti-cyclic citrullinated peptide (anti-CCP) antibodies, and serum IgG level were also performed. The upper limit of normal (ULN) for serum IgG was 12.7 g/L and high serum IgG level was a value above this cut-off, as described previously [25].

Continuous data are described as means and standard deviations (mean ± SD) or as medians and ranges (median (range)).

For all analyses, since serum IL-33 is not systematically detectable, we presented qualitative results as detectable serum IL-33 or undetectable serum IL-33. Among patients of the SMART cohort, we compared patients who were assessed for serum IL-33 level and those who were not due to the availability of the serum samples. Student and Wilcoxon tests were used to compare quantitative values and the Chi-squared test was used for qualitative values.

We analyzed by logistic regression the relationship between EULAR response at 24 weeks and the following four explanatory variables: 1) high disease activity defined as DAS28-CRP >5.1; 2) high serum IgG level; 3) auto-antibody status (i.e., the presence or absence of RF and/or anti-CCP antibodies), since all of them have been found previously as associated with subsequent RTX response [25, 26]; and 4) detectable serum IL-33. All these explanatory variables were entered in a stepwise multivariate model systematically adjusted on RF or anti-CCP positivy (entry level, p = 0.15; level for staying in the model, p = 0.10). Results are expressed as odds ratios (ORs) with 95% confidence intervals (CIs). A p value <0.05 was considered as significant. Statistical analysis was performed with SAS 9.4 software.

The characteristics of patients from cohort 1 are presented in Table 1. There was no significant difference between the 111 patients that underwent IL-33 assessment and the others that participated in the main SMART trial (n = 113) who did not (Table 1). Available characteristics of patients from cohort 2 are presented in Additional file 1 (Table S1).

Using the validated Quantikine assay, most of the patients had undetectable serum IL-33: 84/111 (76%) and 39/74 (53%) in cohorts 1 and 2, respectively. In patients with detectable IL-33, mean serum levels were 63.39 ± 87.31 pg/mL and 45.35 ± 87.19 pg/mL in cohorts 1 and 2, respectively (Fig. 1).

Among the 111 patients, a total of 84 patients (76%) were classified as EULAR responders to RTX at week 24; 27 patients (24%) were non-responders. Detectable serum IL-33 was found in 27 (24.3%) patients. In univariate analysis, the three explanatory variables including DAS28-CRP >5.1, high serum IgG level, and auto-antibody status were associated with RTX response and were included in the multivariate model since they had a p value <0.15 for the association with EULAR response (Table 2).

The multivariate analysis indicated that high disease activity (OR 5.24, 95% CI 1.85–14.85) and auto-antibody status (OR 3.48, 95% CI 1.04–11.60) were independently associated with RTX response. A non-significant association between EULAR response and high serum IgG level was found (OR 2.82, 95% CI 0.97–8.14) (Table 2). There was no association with serum IL-33 detection in multivariate analysis in cohort 1.

In order to replicate these results in RA patients from routine care, we tested whether the DAS28-CRP >5.1, high serum IgG level, auto-antibody status, and detectable serum IL-33 were associated with response to RTX in cohort 2 in which there were 73% EULAR responders among 74 patients.

In univariate analysis, high disease activity and detectable serum IL-33 were associated with RTX response, but not auto-antibody status or high serum IgG (Table 2). In multivariate analysis, detectable serum IL-33 was the sole factor associated with RTX response (OR 3.73, 95% CI 1.12–12.38), with a non-significant association for auto-antibody status (OR 32.76 (0.7– 999). The presence of RF and anti-CCP auto-antibodies in 97% of the patients in cohort 2 might limit analysis on these markers.

Multivariate analyses indicated that four factors were independently and significantly associated with response to RTX: high disease activity (OR 4.10, 95% CI 1.90–8.85), auto-antibody status (OR 3.27, 95% CI 1.13–9.46), high IgG level (OR 2.32, 95% CI 1.01–5.33), and detectable serum IL-33 (OR 2.40, 95% CI 1.01–5.72) (Table 3 and Fig. 2). Since the presence of RF and/or anti-CCP antibodies and high serum IgG level have already been identified as predictive markers of subsequent EULAR response, we aimed to investigate whether the addition of detectable serum IL-33 increased the likelihood of response. While 55% of patients without these three factors responded to RTX, the presence of RF and/or anti-CCP antibodies and elevated serum IgG level further increased the likelihood of this response (77% responders; OR = 2.72, 95% CI 0.67–10.98; p = 0.16 versus patients without any of these three factors). When we added detectable serum IL-33 to these two factors, 100% of the patients with the three risk factors (corresponding to 9% of the combined population) responded to RTX (OR 29.61, 95% CI 1.30–674.79; p = 0.034 versus patients with none of the three risk factors) (Fig. 3).