Assay optimization was performed to identify appropriate drug conjugate and solid phase coupling concentrations to balance assay sensitivity and drug tolerance (data not shown). When testing the serially diluted positive control mAb, both assays exhibited a detection limit below 25 ng/ml and could differentiate varying concentrations of positive control mAb over a 3-log range (Figure S1). Based on the SPR experiment, the positive control mAb binds to infliximab and adalimumab with similar binding affinities (dissociation constant KD approx. 9.0 nM, Fig. S2). In comparison, adalimumab-specific HCA204 exhibited a much slower dissociation rate and higher affinity binding (KD = 0.88 nM, Fig. S2). Compared to the positive control mAb, HCA233 exhibited a much higher binding affinity towards infliximab and a very low dissociation rate, and no reliable affinity determinations could be calculated from our analysis (Fig. S2).

The 40 serum samples from healthy and drug naïve donors were used to define the screening cut points. When analyzed by the IFX-ADA assay over four different days, the average response for the samples was 120.1 RU (95% confidence interval CI: 111.4 to 135.0), with a standard deviation (SD) of 29.4 (Fig. 2a). The screening cut point of the IFX-ADA assay based on 95th percentile of the healthy controls was calculated to be 171 RU.

Infliximab and adalimumab ADAs were measured in healthy volunteers and patients receiving infliximab or adalimumab treatment. Group median and quartiles were represented in the figure and the IFX-ADA and ADL-ADA assay results were plotted on a logarithmic scale. a Infliximab ADA levels measured by IFX-ADA assay in healthy subjects (N = 40, “healthy control”), patients with positive infliximab ADAs exceeding the reportable range based on the comparison clinical ECL assay (N = 7, “high titer”), patients with positive infliximab ADAs within reportable range of the comparison assay (N = 47, “intermediate titer”), patients with negative infliximab ADAs below reportable range of the comparison assay (N = 25, “negative”) and patient with detectable infliximab ADAs but deemed negative based on the assay cutoff (N = 5, “negative titer”). Screening cut point 171 RU was plotted as the dashed horizontal line. b Adalimumab ADA levels measured by ADL-ADA assay in healthy subjects (N = 40, “healthy control”), patients with positive adalimumab ADAs exceeding the reportable range of the comparison clinical ELISA assay (N = 19, “high titer”), patients with positive adalimumab ADAs within reportable range of the comparison assay (N = 47, “intermediate”), and patients with negative adalimumab ADAs based on the comparison assay (N = 22, “negative”). Screening cut point 117 RU was plotted as the dashed horizontal line

The same set of healthy samples was analyzed by the ADL-ADA assay over two days, yielding an average of 81.9 RU (95% CI: 76.6 to 87.2), SD = 16.6 (Fig. 2B). The screening cut point for the ADL-ADA assay based on the 95th percentile was 117 RU.

The intra- and inter-assay imprecision were evaluated at three different ADA levels in 20 replicates and the coefficient of variation (CV%) were all less than 10% for both the IFX-ADA assay and the ADL-ADA assay (Table 1). Close to the screening cut point of each assay, five patient samples were also evaluated in triplicate. For both assays, the CVs% near respective screening cut point were less than 20% (Table S1).

The presence of free hemoglobin, triglyceride, or free or conjugated bilirubin did not impact the measurements in either of the ImmunoCAP ADA assays (Table S2). Similarly, the clinical ECL for infliximab ADAs and ELISA for adalimumab ADAs are also tolerant for hemolysis, lipemia, and icterus in patient serum samples (24, 26). Nevertheless, both IFX-ADA and ADL-ADA assays showed significant negative biases in measured ADA levels due to the presence of biotin at 3500 ng/mL. For IFX-ADA, three positive patient pools exhibited 88% relative reduction when biotin was added. For ADL-ADA, the relative reduction in RU was between 52 and 61%. In both IFX-ADA and ADL-ADA assays, enzyme-labeled streptavidin conjugate was used to detect the formation of the ADA bridging complex. It is postulated that free biotin in the patient sample can bind to the enzyme-labeled streptavidin conjugate. The unbound biotin-conjugate complex would be washed away prior to signal detection, resulting in negative biases in the overall assay measurement. In clinical ECL for infliximab ADAs, signal reduction was observed in serum samples containing more than 12.5 ng/mL biotin. In clinical ELISA for adalimumab ADAs, it is recommended that patient should stop taking biotin/vitamin B7 supplement at least 12 h prior to sample collection.

Patient specimens (n = 84) were analyzed using the automated IFX-ADA assay and the results were compared with the clinical ECL assay. When classified by the screening cut point of the IFX-ADA assay (171 RU) and the clinical cutoff of the ECL assay (50 U/mL), positive agreement was 1.00 and negative agreement was 0.53 (Table 2). All 14 discrepant results were positive on the IFX-ADA assay and negative on the ECL assay (Table 2). In 5 patient samples with discrepant results, there were measurable levels of infliximab ADAs based on the clinical ECL assay (Fig. 2a), but these levels were lower than the assay cutoff of 50 U/mL and considered negative. These 5 patient samples were shown to contain significantly higher levels of infliximab ADAs compared to healthy controls based on the ImmunoCAP IFX-ADA assay. There were 32 patient samples containing infliximab ADAs outside the reportable range of the clinical ECL assay (Fig. 2a). Based on 52 patient samples with quantitative measurement from both the IFX-ADA and the clinical ECL assays, we observed a strong correlation between the two assays (Fig. 3a; rs = 0.673, 95% CI: 0.484–0.802). When analyzing different dilutions of the anti-infliximab mAb HCA233 with the IFX-ADA assay, we observed a strong linear correlation (R2 = 0.9965) between observed RU and the HCA233 concentrations (Fig. 3b). This indicates a linear correlation between the IFX-ADA assay and the clinical ECL assay since HCA233 is also used as calibrator antibody in the latter assay. The discrepant results between the two assays observed in patient samples are largely due to different cutoffs established for the two assays.

Quantitative analysis of ImmunoCAP IFX-ADA and ADL-ADA assays. a Quantitative correlation between IFX-ADA and a clinical ECL assay based on patient samples within the reportable range of the ECL assay (20–1250 U/ml). b Linear responses between IFX-ADA measurements and concentrations of infliximab specific mAb HCA233 (20–1250 ng/ml). c Quantitative correlation between ADL-ADA and a clinical ELISA based on patient samples within the reportable range of the clinical ELISA (10–500 AU/ml). d Responses of ADL-ADA assay at different concentrations of adalimumab specific mAb HCA204 (50–1000 ng/ml)

Eighty-eight patient samples were evaluated by both ADL-ADA and the clinical ELISA assay. For qualitative agreement, results were classified according to the screening cut point of the ADL-ADA assay (117 RU) and the clinical assay cutoff of the ELISA assay (14 AU/mL). Positive agreement between the two assays was 0.68 and negative agreement was 1.00 (Table 3). All 21 discrepant results were negative by the ADL-ADA assay and positive by the clinical ELISA assay (Table 3). There were 40 samples outside the reportable range for the clinical ELISA assay (Fig. 3b). Based on the remaining 48 patient samples with quantitative values from both the ADL-ADA and clinical ELISA assays, we observed a moderate quantitative correlation between the two assays (Fig. 3c; rs = 0.510, 95% CI = 0.198–0.728). We investigated concentrations of adalimumab in patient samples with qualitatively agreeable and discrepant ADA results between ADL-ADA and clinical ELISA (Table S3). In 21 patient samples with discrepant results (negative by ADL-ADA, positive by clinical ELISA), their ADA results measured by the clinical ELISA ranged from 18.7 AU/mL to 215.2 AU/mL; the adalimumab concentrations varied from less than 0.8 mcg/mL (lower limit of quantification of the adalimumab assay) to 6.1 mcg/mL (median 2.1 mcg/mL). The remaining patient samples contained positive adalimumab ADAs according to both assays and the overall trend of their adalimumab drug concentrations decreased with increasing levels of adalimumab ADAs. In 21 patient samples with matching ADA levels (15.2–255.7 AU/mL based on the clinical ELISA), the adalimumab drug concentrations were from less than 0.8 mcg/mL to 6.9 mcg/mL, with a median 1.7 mcg/mL. The RF IgM in all patient samples were less than 5.0 IU/mL, indicating negative or equivocal results.

We also evaluated a subset of patient samples (n = 34) using another clinical ADA assay for adalimumab, a cell-based reporter gene activity (RGA) assay sensitive to TNF activity. Generally, the three assays agreed among each other for certain “high positive” and negative samples (Table S4). All 16 negative samples on the ADL-ADA assays were also negative on the RGA assay. Nine of the 18 positive samples on the ADL-ADA assay were negative on the RGA assay, although they were all positive on the clinical ELISA assay.

When using the ADL-ADA assay to evaluate mAb HCA204 spiked at different concentrations in serum, we observed a linear response at mAb concentrations between 0.25 and 50 mcg/mL (R2 = 0.9975, Fig. 3D). The ADL-ADA assay was not able to detect HCA204 at concentrations less than 250 ng/mL. This observation contrasted the high binding affinity observed between HCA204 and adalimumab during binding kinetics characterization (Fig. S2).

Drug tolerance of the IFX-ADA assay was evaluated by studying the inhibition profile of the measured infliximab ADAs in response to added infliximab. The addition of infliximab resulted in decreased RU in all but one patient samples (Fig. 4a). The amplitude of signal reduction varied at different concentrations of infliximab. The measured ADA levels underwent more drastic reduction at 2 and 5 mcg/mL infliximab, followed by a continued albeit lessened reduction at 10 and 15 mcg/mL infliximab. The average relative signal reduction changed from 50.4% (2 mcg/mL) to 69.2% (15 mcg/mL). The inhibition profiles in patient samples were highly dependent on their existing ADA titers. Samples with ADA titers less than 700 RU in the IFX-ADA assay were more likely to drop below the screening cut point with increasing amounts of added infliximab, while samples with ADAs levels corresponding to more than 700 RU remained positive at 15 mcg/mL infliximab (Fig. 4a). The drug tolerance profile of the mAb HCA233 at 50 and 350 ng/mL was also evaluated; the mAb HCA233 displayed a different inhibition profile compared to the patient samples, reaching maximal signal reduction below the assay screening cut point at an infliximab concentration of 2 mcg/mL (Fig. 4A).

Drug tolerance profiles of patient samples containing different levels ADAs. a Inhibition of IFX-ADA results with the addition of infliximab at different concentrations. 6 patient samples were evaluated with added infliximab at 0, 2, 5, 10 mcg/ml. 10 patient samples were evaluated with spiked infliximab at 0, 2, 5, 10, 15 mcg/ml. Human anti-infliximab mAb HCA233 at 50 ng/ml and 350 ng/ml were also evaluated with spiked infliximab at 0, 2, 5, 10, 15 mcg/ml. The measurement results (RU) based on the IFX-ADA assay were plotted on a logarithmic scale with a base of 5. Screening cut point 171 RU was plotted as the dashed horizontal line. b Inhibition of ADL-ADA results with the addition of adalimumab at different concentrations. 15 patient samples were evaluated with added adalimumab at 0, 2, 4, 8, 16 mcg/ml. 5 patient samples were evaluated with spiked adalimumab at 0, 5, 10, 20 mcg/ml. Human anti-adalimumab mAb HCA204 at 10,000 ng/ml was also evaluated with spiked adalimumab at 0, 5, 10, 20 mcg/ml. The measurement results (RU) based on the ADL-ADA assay were plotted on a logarithmic scale with a base of 5. Screening cut point 117 RU was plotted as the dashed horizontal line

When evaluating the ADL-ADA assay, increasing concentrations of adalimumab were spiked into 20 patient samples containing different levels of adalimumab ADA. While the addition of free adalimumab decreased the measured RU in some patient samples, other samples did not show significant changes (Fig. 4b). In patient specimens with adalimumab ADA levels between 300 and 2000 RU by the ADL-ADA assay, greater magnitudes of signal reduction were observed. In patient samples with ADA titers less than 200 RU, the relative response reduction to added adalimumab was not statistically significant considering the assay imprecision. The average relative signal reduction due to adalimumab addition ranged from 15.8% (2 mcg/mL) to 33.5% (16 mcg/mL). A constant decline in RU was observed when increasing adalimumab concentrations were added to the human anti-adalimumab mAb HCA204 (Fig. 4b). However, the mAb HCA204 (10,000 ng/mL) remained positive by the ADL-ADA assay (i.e. above screening cut point) in the presence of up to 20 mcg/mL adalimumab spiked to the sample.

For the IFX-ADA assay, RF serum P1 appeared to negatively interfere with the ADA measurement when spiked into the mid and high ADA serum pools, whereas RF serum P2 interfered by increasing the ADA result of the low patient serum pool (Table 4). Nevertheless, there were no false positive or false negative infliximab ADA qualitative results observed for the ADA serum pools due to RF interference.

For the ADL-ADA assay, all four individual RF serum (P3-P6) interfered in the assay by increasing the ADA results when spiked into the negative, low, and mid ADA sample pools (Table 4). RF serum P5 gave the largest relative increase, and it caused a false positive result for adalimumab ADAs when spiked into the ADA-negative serum pool (final RF IgM concentration: 176 IU/mL) (Table 4). No false positive results in the ADL-ADA assay were observed for any of the other three RF serum used for spiking (RF serum P3, P4 and P6).