Technical Appendix: MS RTX versus alternative aTNF
Methods:
Study population: Within the Swiss Clinical Quality Management for RA (SCQM-RA) system, patients are assessed for disease activity, joint damage, RA symptoms and drug AEs by the treating rheumatologist.[1] Clinical information is collected systematically by the patient’s physician every 6 – 12 months and further updated at every significant change in antirheumatic therapy. To collect more complete data, we contacted the rheumatology units of all academic hospitals and asked for data on RA patients not included in the SCQM-RA cohort. All therapeutic decisions were made by the treating rheumatologist as he felt best and without reference to any binding treating guidelines. The SCQM registry does not have any predefined study objectives, which implies that the rheumatologists performing clinical assessments are de facto blinded to the hypotheses of subsequent analyses, thus preventing any systematic assessment bias. Regulatory authorities in Switzerland recommend continuous monitoring of all patients receiving biologic agents and selected the SCQM system. Approximately half of the patients come from private rheumatology practices, 30 % from non-academic hospital centres and 20% from academic centres. Based on a comparison with sales data from the industry, approximately 70-80% of all Swiss RA patients receiving aTNF were included in SCQM-RA in 2003.[2,]
Study design. Patients may interrupt aTNF therapy for various reasons (i.e. ineffectiveness, AE, personal preferences, pregnancy wish, or prolonged international travel). Observational studies suggested that the effectiveness of subsequent aTNF agents varies according to the reason for switching.[3-5] ‘Loss of efficacy’, ‘resistance’ or ‘inadequate response’ to aTNF are not univocally defined.[6] We created a dummy variable corresponding to a history of aTNF failure due to ineffectiveness based on the reasons provided by the treating rheumatologist for switching therapies. Some physicians motivated aTNF interruption by a combination of reasons and some others discontinued several aTNF agents for different reasons; therefore a given patient could have more than one cause of prior aTNF discontinuation (N = 35). Patients who indicated insufficient response as one of the reasons for interrupting therapy were analysed within the aTNF inefficacy group.
The literature also suggests that the type of aTNF switch may influence the effectiveness of the subsequent aTNF therapy and most studies highlight the benefits of switching from a soluble receptor aTNF (etanercept) to a monoclonal antibody aTNF (infliximab or adalimumab) or vice versa to avoid a class effect.[7] We created a second dummy variable corresponding to an aTNF switch with a change in aTNF type (receptor antagonist to monoclonal antibody or vice versa). Furthermore, it is well established that the likelihood of a favourable response to aTNFs declines with the number of previous aTNF failures.[2, 3, 7] We created a third dummy variable corresponding to a single prior aTNF failure versus several prior aTNF failures. Finally, while some aTNF agents may be given alone, it is well established that the effectiveness of these agents increases when given concomitantly with conventional DMARDs.[8] We created a fourth dummy variable corresponding to the presence of co-therapy with conventional DMARDs (MTX, Leflunomide, or other).
Propensity Score: The propensity score was computed using a logistic-regression model that ‘predicted’ which treatment patients received as a function of baseline variables potentially associated with disease progression :baseline disease activity (DAS28), disease characteristics (RF+, disease duration), demographic characteristics (age, sex), socio-economic characteristics at enrolment (education level) and treatment characteristics (co-therapy with low dose glucocorticoids). Potential effect modifiers were not included into the propensity score to allow an analysis of their effects, but final analyses were corrected for differences in potential effect modifiers using multivariate regression adjustments. In various sensitivity analyses, we used alternative techniques to adjust for confounding factors, such as multivariate adjustments or propensity score matching, to explore the robustness of our findings.
Statistical Analysis: The longitudinal improvement in DAS28 was analyzed using linear mixed model analysis for longitudinal data. In longitudinal study designs, outcomes are assessed repeatedly over time and can thus no longer be assumed to be independent observations. The advantage of the linear mixed model over traditional analytic approaches to longitudinal data is that it models explicitly the dependency in a covariance matrix. Therefore, the fixed parameter estimates become more efficient and the model turns out to be more powerful in terms of testing the effects associated with the repeated measures. This approach is also more robust than traditional univariate and multivariate models. We chose the best fitting and parsimonious covariance structure using series of likelihood ratio tests based on restricted maximum likelihood estimates. The final model was based on an ‘unstructured’ covariance structure, and the specified random effects were a ‘random intercept’ for the subject and ‘random slope’ for time. We then examined whether time as a linear trend or as a quadratic function best fitted the data and if the multivariate normal assumption for longitudinal models was satisfied. Point estimates of the regression model were used to produce the result graphs. Pair-wise comparisons within effect modifier subgroups were planed a priori, but were considered only if the overall comparison indicated a significant difference (p for effect modification < 0.05).
Table 2: Detailed results of the regression model examining effect modification by prior anti-TNF ineffectiveness on longitudinal DAS28 improvement
Rituximab [Yes/No] / -0.196 / 0.24
Ineffectiveness to prior anti-TNF [Yes/No] / 0.170 / 0.31
Absence of concomitant conventional DMARD [Yes/No] / -0.057 / 0.64
Inadequate response to several anti-TNF [Yes/No] / 0.009 / 0.91
Time [Mo] º / - 0.021 / 0.14
Time² [Mo²] º / 0.018 / < 0.001
Time * RTX [Mo] ⁿ / 0.030 / 0.30
Time² * RTX [Mo²] ⁿ / 0.009 / 0.15
Time * Ineffectiveness [Mo] † / - 0.070 / 0.04
Time² * Ineffectiveness [Mo²] † / -0.005 / 0.49
Time * RTX * Ineffectiveness [Mo] ‡ / - 0.070 / 0.04
Time² * RTX * Ineffectiveness [Mo²] ‡ / -0.005 / 0.49
Propensity score, 1st strata [Yes/No] Ñ / -1.573 / < 0.001
Propensity score, 2nd strata [Yes/No] Ñ / -0.891 / < 0.001
Propensity score, 4th strata [Yes/No] Ñ / 0.917 / < 0.001
Propensity score, 5th strata [Yes/No] Ñ / 1.524 / < 0.001
Constant / 3.485 / < 0.001
Legend Table 2: The table provides the detailed results of the mixed linear regression model used to examine effect modification by prior anti-TNF ineffectiveness. º Time was centerred at 6 months and modelled as a linear trend and as a quadratic function. ⁿ Interaction terms used to examine whether the evolution of patients without prior aTNF ineffectiveness on RTX is different from patients on aTNF. † Interaction terms used to examine whether the evolution of patients on aTNF is different with prior aTNF ineffectiveness. ‡ Interaction terms used to examine whether the evolution of patients on RTX is different with prior aTNF ineffectiveness. Significance of the combined effect of the two interaction terms was examined using a log-likelihood ratio test (on maximum likelihood estimators): p = 0.005. Ñ Analyses were adjusted for baseline disease activity (DAS28), disease differences characteristics (RF+, disease duration), demographic characteristics (age, sex), socio-economic characteristics at enrolment (education level) and treatment characteristics (co-therapy with low dose glucocorticoids) via propensity score stratification.
References
1 Finckh A, Simard JF, Duryea J, Liang MH, Huang J, Daneel S, et al. The effectiveness of anti-tumor necrosis factor therapy in preventing progressive radiographic joint damage in rheumatoid arthritis: a population-based study. Arthritis Rheum, 2006, 54(1), 54-59.
2 Finckh A, Simard JF, Gabay C and Guerne PA. Evidence for differential acquired drug resistance to anti-tumour necrosis factor agents in rheumatoid arthritis. Ann Rheum Dis, 2006, 65(6), 746-752.
3 Hyrich KL, Lunt M, Watson KD, Symmons DP and Silman AJ. Outcomes after switching from one anti-tumor necrosis factor alpha agent to a second anti-tumor necrosis factor alpha agent in patients with rheumatoid arthritis: results from a large UK national cohort study. Arthritis Rheum, 2007, 56(1), 13-20.
4 Gullick N, Da Silva C and Kirkham B. Failure of Adalimumab in Patients with Rheumatoid Arthritis [abstract]. Arthritis Rheum, 2007; 56(9; suppl): S183.
5 Bombardieri S, Ruiz AA, Fardellone P, Geusens P, McKenna F, Unnebrink K, et al. Effectiveness of adalimumab for rheumatoid arthritis in patients with a history of TNF-antagonist therapy in clinical practice. Rheumatology (Oxford), 2007, 46(7), 1191-1199.
6 Haraoui B. Is there a rationale for switching from one anti-tumor necrosis factor agent to another? J Rheumatol, 2004, 31(6), 1021-1022.
7 Gomez-Reino JJ and Carmona L. Switching TNF antagonists in patients with chronic arthritis: an observational study of 488 patients over a four-year period. Arthritis Res Ther, 2006, 8(1), R29.
8 Klareskog L, van der Heijde D, de Jager JP, Gough A, Kalden J, Malaise M, et al. Therapeutic effect of the combination of etanercept and methotrexate compared with each treatment alone in patients with rheumatoid arthritis: double-blind randomised controlled trial. Lancet, 2004, 363(9410), 675-681.
- 6 -