Infliximab (IFX), a chimeric anti-tumor necrosis factor monoclonal antibody, is effective in the treatment of rheumatoid arthritis (RA) and other inflammatory diseases [1-4]. However, treatment fails in around 40% of patients, in many cases due to immunogenicity. An important contributor to treatment failure is the presence of antibodies towards IFX (ATI). The formation of immune complexes between antibodies and biological drugs may increase drug clearance, reduce serum levels and result in a loss of efficacy [5-8].

Several publications have reported an association between serum IFX levels and clinical response [5, 6, 9, 10]. A good clinical response is usually associated with high circulating IFX levels [11, 12]. On the other hand, low or undetectable circulating IFX levels may indicate ATI development with the consequent loss of drug efficacy.

Most of the methods available to detect ATI are hindered by the simultaneous presence of drug and antibodies in serum, being the bridging enzyme-linked immunosorbent assay (ELISA) the most susceptible one to this effect [13]. Drug interference accounts for important discrepancies between studies, so current efforts are focused on the development of assays lacking drug interference that can reveal the true prevalence of immunogenicity.

The association between early-stage IFX serum concentrations, clinical outcomes and immunogenicity has recently been studied for inflammatory diseases such as RA and ulcerative colitis (UC) [12, 14]. It is not clear why some patients show faster drug clearance from treatment commencement, being the hypothesis that circulating drug levels are affected by patient characteristics associated with baseline disease activity and an early anti-drug antibody production [15, 16].

The aim of this study is to analyze whether serum IFX trough levels (ITL) at the induction phase are associated with IFX clearance and clinical outcomes in the first year of treatment.

The implication of different factors potentially accounting for the presence of low levels at early stages of treatment is also studied.

This observational retrospective study included 66 patients with RA recruited from a prospective observational cohort of patients treated with biological drugs at the rheumatology department of La Paz University Hospital (Madrid). Patients fulfilled the 1987 revised criteria of the American College of Rheumatology [17], had moderate or high disease activity [18] (despite previous treatment with disease-modifying anti-rheumatic drugs), and were naïve to treatment with biologics; they were followed-up for at least one year, except when lost to follow-up for medical reasons and sera from most of the visits (at 0-2-6-14-22-54 weeks) had to be available. Patients were given intravenous infusions of 3 mg/kg of IFX at week (W) 0, 2 and 6 and every 8 weeks thereafter. Serum samples were collected at baseline and immediately before each infusion. Disease activity was assessed at baseline, W22 and W54 using the Disease Activity Score for 28 joints (DAS28) measured according to the erythrocyte sedimentation rate. Therapeutic response was evaluated using the criteria of the European League Against Rheumatism (EULAR) [19]. All patients were treated with standard dose (3mg/kg) every 8 weeks.

The study was approved by La Paz University Hospital Ethics Committee and all patients signed an informed consent form.

Serum IFX levels were measured by a capture ELISA as previously described [5]. Coating antibody (MoAb7 anti-TNF) and buffer (High Performance ELISA buffer) were supplied by Sanquin (Amsterdam, The Netherlands). Recombinant TNF was supplied by Preprotech and biotinylated anti-idiotype monoclonal antibody for detection was supplied by Progenika Biopharma (Derio, Spain). Peroxidase-Streptavidin and TMB were used for reaction developing, and optical density was read at 450 nm.

The cut-off for positivity was 10ng/ml, established with a control group of 250 sera (150 from healthy donors and 100 from untreated RA patients). IFX serum levels were evaluated at baseline (W0) and at W2, W6, W14, W22 and W54 after initiation of IFX treatment. Serum samples from the 6 time points were available from 43 patients, missing a maximum of 2 samples /patient in the rest.

To determine whether the quantitative or qualitative data obtained by our capture ELISA could be extrapolated, we compared our results from University Hospital La Paz (UHLP) with two different commercial ELISAs; Pomonitor-IFX (Progenika Biopharma, Derio, Spain) and IFX ELISA-Compact produced by Sanquin (Amsterdam, The Netherlands). The Pearson´s correlation coefficients were 0.97 and 0.88, respectively (p<0.001 in each comparison), and no discrepancies in positive or negative results were found among all three assays (Kappa correlation index ULPH vs Sanquin =1; Kappa UHLP vs Promonitor=1) Fig. (S1).

The agreement between drug levels of the three different assays was analyzed using an intra-class correlation coefficient (ICC); UHLP vs Sanquin (0.907; 95%IC 0.87-0.93; p<0.001) and UHLP vs Promonitor (0.64; 95%IC 0.39-0.78; p<0.001). The Bland-Altman plot Fig. (S2) gave an average of differences with a 95% limit of agreement, indicating that the two assay methods produce similar results.

Fig. (S1). Correlation between UHLP capture ELISA and two different ELISAs (PROMONITOR and SANQUIN). A) Infliximab (IFX) trough levels (ITL) in patients with rheumatoid arthritis (n=124) measured by UHLP capture ELISA and SANQUIN ELISA. Results are shown correlations between both assays. B) Infliximab (IFX) trough levels (ITL) in patients with rheumatoid arthritis (n=77) measured by UHLP capture ELISA and PROMONITOR ELISA. Results are shown correlations between both assays.

Fig. (S2). Comparison of IFX methods by Bland-Altman analysis. A) Bland-Altman analysis between UHLP and SANQUIN, the differences between the two measurements (Y-axis in µg/ml) is plotted against the average of each pair of measurements (X-axis in µg/ml), the 95% limits of agreement are represented by dot line. B) Bland-Altman analysis between UHLP and Promonitor, the differences between the two measurements (Y-axis in µg/ml) is plotted against the average of each pair of measurements (X-axis in µg/ml), the 95% limits of agreement are represented by dot line.

ATI levels were assessed in sera from W0, W2, W6, W14 and W22 of treatment in 2 different ways:

An in-house 2-site (bridging) ELISA was used to detect uncomplexed (free) ATI as previously described [5]. The cut-off value for the presence of ATI was 50 arbitrary units (AU)/mL, established from the same serum control group as for IFX levels.

A commercial kit called IDKmonitor (Immundiagnostik, Bensheim, Germany) (A), and an acid-dissociation radioimmunoassay (ARIA, Sanquin) (B), were used to measure total (free and complexed) ATI. The (A) assay performed acid dissociation of the serum to acquire free ATI. Peroxidase conjugate-IFX and biotinylated-IFX were added to replace the unmarked therapeutic antibody and the marked IFX formed a complex with ATI that bound via biotin to a streptavidin-coated microtiter plate. Qualitative results higher than 10 AU/mL were considered positive. In the (B) assay, an acid dissociation was also performed to acquire free ATI. Samples were neutralized in a pH 7.6 buffer [20] with an excess of IFX F(ab´)₂-biotin (to replace the unmarked therapeutic antibody). Antibodies were subsequently captured by Protein A Sepharose and radiolabelled streptavidin was used to detect specific ATI, analogue to the adalimumab ARIA [20].

Because of the retrospective nature of the study not all serum samples were available to be tested in all assays for ATI. Available samples were as follows: W0 (60 by brigding ELISA, 37 by ARIA and 59 by IDKmonitor); W2 (60 by brigding ELISA, 58 by ARIA and 60 by IDKmonitor); W6 (64 by brigding ELISA, 63 by ARIA and 63 by IDKmonitor); W14 (61 by brigding ELISA, 51 by ARIA and 58 by IDKmonitor); W22 (60 by brigding ELISA, 51 by ARIA and 57 by IDKmonitor).

Descriptive statistics were reported as mean and standard deviation (SD) or median and interquartile range (IQR) depending on normality. Patients were classified into two groups depending on the presence or absence of circulating serum IFX at W54, due to the ATI presence is detected in the most of patients during the first year of the treatment. In addition, we extended the follow-up time to W54 to include not only patients who develop primary inefficacy but also patients who develop secondary inefficacy associated with the development of ATI.

Differences in clinical and biological characteristics between both groups were assessed using Pearson’s chi-square test or Fisher’s exact test for categorical variables and the Mann-Whitney U test or the Student t-test for continuous variables. Last observation carried forward (LOCF) was employed to impute data for patients who discontinued treatment before the end of the first year.

Serum-dependent receiver operating characteristic (ROC) curves at W2, W6, W14 and W22 were used to determine the ITL that best predicted the absence of free IFX at W54. The predictive cut-off was determined as the value for IFX levels that showed maximum sensitivity, maximum specificity and higher positive likelihood ratio at each studied time point. IFX treatment survival was studied using Kaplan-Meier curves and groups were compared using the log-rank test.

Univariable and multivariable logistic regression models were employed to investigate the association between early-stage ITL cut-off determined by ROC curves analysis and IFX at W54. Demographic, clinical and serological characteristics (including ITL cut-off at W2 and W6 as categorical variables) were included in the models as independent variables and the IFX levels as the dependent variable.

For all the analysis, GraphPad Prism 6 (San Diego, CA, USA) and SPSS 21.0 software were employed and p-values <0.05 were considered statistically significant.

Baseline characteristics for the 66 patients with RA starting IFX are summarized in Table 1. Most patients were female (86%) and were positive for both rheumatoid factor (82%) and anti–cyclic citrullinated peptide antibody (82%). At inclusion time, all patients had active disease (DAS28: 5.5±1.3).

Median disease duration prior to IFX treatment was 14 years (IQR: 9-18). Out of the 66 patients, 42 (64%) received concomitant treatment with methotrexate (MTX, mean dose 15 mg/week). Eight out of 66 (12%) patients discontinued the IFX treatment before the first year (5 due to lack of efficacy and 3 due to adverse events).

Patients were divided into 2 groups depending on the presence (ITLpos) or absence (ITLneg) of circulating serum IFX at W54 Table 1. At W54, ITL were undetectable in 25 patients (38%), 23 of whom were ITLneg already since W22. ITLpos patients had lower baseline DAS28 scores compared to ITLneg patients (5.3±1.1 vs 5.8±1.2; p=0.06) and their mean (SD) IFX levels at W54 were 2.1 (1.8) µg/mL.

Less ITLneg patients were treated with methotrexate (MTX) (11 (26%)) than ITLpos patients (32 (74%)); p≤0.01. ITLpos patients were older than ITLneg patients (59 years (51-69) vs 51 years (42-65)); p≤0.05. And the baseline C-reactive protein (CRP) was higher in ITLneg patients than in ITLpos patients (23±2 vs 15±17; p≤0.05).

The ITLneg patients at W54 had lower early-stage ITL than the ITLpos patients (W2: 20.0 ±12.7 µg/mL vs 29.7±14.5 µg/mL (p=0.015); W6: 4.2±5.9 µg/mL vs 15.7±11.1 µg/mL (p<0.0001); W14: 0.1±0.2 µg/mL vs 4.1±5.3 µg/mL (p<0.0001); and W22: 0.01±0.04 µg/mL vs 2.8±3.3 µg/mL (p<0.0001)) Fig. (1A).

At all studied time points, the areas under the curve (AUC) of the ROC curves were statistically different from 0.5 Fig. (1B), enabling cut-off levels to distinguish between ITLpos and ITLneg patients at W54 (Table 2).

Out of the 25 ITLneg patients at W54, 12 (48%), 18 (72%) and 25 (100%) had ITL below the predictive cut-off at W2, W6 and W14, respectively. At W14, in 19 out of 25 patients (76%) ITL were undetectable. The W6 predictive cut-off (4.4 μg/mL) showed the best association with IFX absence at W54, with a sensitivity of 70% (95% confidence interval (CI): 45.7-88.1), a specificity of 95% (95% CI: 83.1-99.4) and a positive likelihood ratio of 14. Therefore, the W6 predictive cut-off was considered the reference value to predict clinical and serological outcomes for further analysis.

Fig. (1). A) Infliximab (Ifx) trough levels (ITL) in patients with rheumatoid arthritis at early treatment stages, according to week (W) 54 Ifx status. Results are shown as medians (solid lines within boxes), interquartile ranges (upper and lower box boundaries) and maximums and minimums. Black dots indicate outliers. B) Receiver operating characteristic curves for ITL at W2, W6, W14 and W22 for prediction of Ifx absence at W54. AUC, Area Under the Curve.

In the univariable analysis, three factors were significantly associated with IFX absence at W54 Table 3A: i) having an ITL below the cut-off at W2 (odds ratio (OR): 12.40; 95% CI: 3.48-44.15) or at W6 (OR: 44.33; 95% CI: 7.99-246.03), ii) non-use of MTX (OR: 4.20; 95% CI: 1.33-13.32) and iii) being older (OR: 1.04; 95% CI: 1.03-1.07).

In the multivariable logistic regression analysis, even after adjusting for possible confounders (age, gender and baseline DAS28), the ITL below the cut-off at W2 (OR: 15.85; 95% CI: 2.95-85.03; p=0.01) or at W6 (OR: 86.64; 95% CI: 6.58-1139.99) and also the non-use of MTX (OR: 12.26; 95% CI: 1.83-82.22) remained significantly associated with IFX absence at W54 Table 3B.