Background The introduction of tumour necrosis factor inhibitors represents a major advance in the treatment of rheumatoid arthritis (RA). Despite this, up to 40% of patients fail to respond- either due to primary inefficacy or loss of response. One explanation is immunogenicity leading to the development of anti-drug antibodies (ADAb) and low drug levels. Radioimmunoassay (RIA) is a sensitive method to detect ADAb and is less prone to drug interference vs. ELISA. The clinical utility of pharmacological monitoring in routine rheumatology practice is still debated with conflicting opinions.
Objectives To evaluate whether ADAb formation and serum drug levels may predict future treatment response of anti-TNF initiated RA patients.
Methods 331 patients were selected from the Biologics in RA Genetics and Genomics Study Syndicate prospective cohort [n=160 adalimumab (AD); n=171 etanercept (ETA)]. Serum samples were collected at 3, 6 and 12 months following initiation of therapy. ADAb were measured using RIA and drug levels using ELISA at 3, 6 and 12 months; the timing of blood samples was not consistently pre-dose. Disease activity (DAS28) scores were measured at each visit. Descriptive statistics, multiple linear regression and generalised estimating equation were used as appropriate.
Results 835 serial samples were suitable for pharmacological testing (n=414 AD; n=421 ETA). Mean age: 56±13 years; 75% female; baseline DAS28 score 5.9±0.8; median BMI 27.5 (IQR 23.6-32.3). 85% were on a DMARD (56% MTX). ADAbs to AD were detected in 24.8% (31/125 patients at ≥1 time points by 12 months) and in none of the ETA patients. The presence of ADAbs were significantly associated with lower AD drug levels (p <0.0001; rs -0.51; if ADAb titres>100AU p=0.0041; rs-0.66). At 3 months, ADAb formation and low drug levels were a significant predictor of poor ΔDAS28 at 6 and 12 months (p<0.0001, RC -0.0048 95% confidence intervals (CI) -0.0071 to -0.0025). By 12 months, patients who developed ADAbs showed less improvement in disease activity (mean ΔDAS28 2.35; CI 1.67-3.03) than patients without (mean ΔDAS28 3.15; CI 2.86-3.35; p=0.022). Patients who did not develop ADAbs were more likely to be co-treated with MTX (61.4% vs. 43.7% p=0.01). AD drug level was the most significant independent predictor of ΔDAS28 at all time points after adjusting for confounders (p <0.0001, RC 0.12, CI: 0.06-0.18). At 12 months, patients on ETA with higher drug levels (>15μg/mL) were more likely to achieve a good EULAR response than patients with sub-therapeutic levels (<0.0035 μg/mL) (p=0.01). High BMI was the strongest predictor of low drug levels (ETA, p<0.0001, RC -5.97; CI -8.75 to -3.19; AD, p<0.0001, RC-3.86; CI -5.72 to -2.00).
Conclusions At 3 months ADAb formation and low AD drug levels are a significant predictor for poor treatment response at 6 and 12 months. In addition low AD drug levels were the strongest independent predictor of poor treatment response at all time points.
Acknowledgements Supported by Medical Research Council [grant number G1000417/94909], Arthritis Research UK  and NIHR Manchester Musculoskeletal BRU.
Disclosure of Interest None declared