Statistics from Altmetric.com
Tapering of TNF inhibitors (TNFi) is feasible in many patients with rheumatoid arthritis (RA), but sometimes leads to flaring. TNFi trough serum levels and antidrug antibodies (ADAs) have been proposed as predictors for successful dose reduction or discontinuation,1–3 suggesting that: (1) a patient with low or undetectable serum levels (with or without ADAs) should be able to successfully stop the TNFi and (2) a patient with high serum levels should be able to reduce the dose. As obtaining trough levels is often not practical, our aim was to investigate whether random timed serum drug levels and ADAs of adalimumab and etanercept are predictive for successful dose reduction or discontinuation of these TNFi in patients with RA doing well.
For these analyses 118 patients with RA from an open randomised clinical trial investigating a dose reduction strategy of adalimumab or etanercept with 18 months follow-up were included.4 Serum samples were collected at baseline (before start of dose reduction) at a regular visit, unrelated to time of injection. Serum levels were measured after completion of the study, with clinical outcome assessed blinded for assay results. Serum levels and ADAs were assessed,5 ,6 and receiver operating characteristic (ROC) curves were created for TNFi levels versus the outcomes successful discontinuation and successful dose reduction separately, and compared with the ‘no dose reduction possible’ group. Sensitivity analyses were done based on timing (tertile) of serum sampling in relation to last injection.
At 18 months, TNFi could be stopped in 19% (95% CI 12% to 27%) of patients, the interval increased in 44% (95% CI 35% to 53%). In 37% (95% CI 29% to 47%) of patients no dose reduction was possible. All patients had detectable drug levels. Mean drug levels and ADAs were not significantly different between outcome subgroups (table 1). Antiadalimumab ADAs (low titres, 15–46 U/mL) were detected in four patients (10%, 95% CI 4% to 23%) and were not predictive for successful discontinuation. Drug levels in patients with antiadalimumab ADAs were low (1.8–4.9 mg/L). No antietanercept ADAs were detected.
ROC analyses showed no significant predictive value of TNFi serum levels for successful dose reduction or discontinuation, except for a significant but small inverse association between etanercept levels and chance for successful dose reduction (figure 1). Sensitivity analyses (ROCs not shown) showed that for intermediately timed (middle tertile) serum etanercept level, low levels were associated with a higher chance of successful dose reduction (area under the curve (AUC) 0.28 95% CI 0.08 to 0.47, optimal cut-off 2.5 mg/L). For adalimumab, high trough levels were associated with successful dose reduction (AUC 0.86, 95% CI 0.58 to 1.00, optimal cut-off 7.8 mg/L).
One explanation for the absent predictive value—which is, of note, not conflicting with other data, as thus far these hypotheses were not supported by adequate evidence—is that, although on a group level serum level response curves have been demonstrated, the same curves show high interindividual variation in effective serum drug levels.6 A certain serum drug level is therefore too low for some patients, but supratherapeutic for others, complicating use of one threshold for different patients. Indeed, recently data supporting this view have been published.7 The use of random timed sampling of serum levels, instead of trough level sampling might have led to underestimation of the association between (anti)drug level and outcome. However, it should be noted that the differences between peak and trough levels in subcutaneous TNFi are not very high.8 Also, subanalyses using timing tertiles did not result in relevant changes. Finally, the requirement for trough level sampling might hamper the use in daily clinical practice. We welcome validation of our findings, and of other suggested test algorithms using serum (anti)drug level measurement, using the appropriate study designs.9
The authors thank Leo Schiebergen for help with handling the serum samples. The study was designed, executed and reported according to the STARD (Standard for Reporting of Diagnostic accuracy studies) guideline.
Funding This study received no external funding.
Contributors NvH, AAdB, AvdM, BJFvdB, FHJvdH, CAMB, RFvV and JWB were involved in the study design. NvH, CAMB, AAdB, AvdM and FHJvdH were involved in the data collection. NvH, AAdB, CAMB, BJFvdB and AvdM performed the data analyses. All authors were involved in writing, revision and final approval of the manuscript.
Competing interests JWB reports grants and personal fees from AbbVie, BMS, Roche, UCB, Pfizer, MSD, during the conduct of the study. RFvV reports grants from AbbVie, BMS, GSK, Pfizer, Roche, UCB, personal fees from AbbVie, Biotest, BMS, Crescendo, GSK, Janssen, Lilly, Merck, Pfizer, Roche, UCB, Vertex, outside the submitted work.
Ethics approval Commissie Mensgebonden Onderzoek region Arnhem-Nijmegen.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The authors commit to making the relevant anonymised patient level data available on reasonable request.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.