Adverse Event Risk Factors in Rheumatoid Arthritis

Next, we selected patients with RA who were ≤100 years old as judged from the INDI and DEMO tables and conducted a multiple logistic regression analysis to calculate aRORs for the adverse events. After excluding cases with unknown sex, we set each adverse event as the objective variable and age, sex, and treatment patterns of MTX as explanatory variables. We defined four treatment patterns of MTX: i) MTX group that did not use FA or TNFi, ii) MTX + FA group that did not use TNFi, iii) MTX + TNFi group that did not use FA, and iv) MTX + FA + TNFi group. TNFi was used if at least one TNFis (infliximab, adalimumab, etanercept, golimumab, or certolizumab) was employed. In our preliminary analysis to establish a logistic model, we confirmed that higher variance inflation factor (VIF) values were obtained with a logistic model incorporating the use of MTX, FA, and TNFi as covariables and factors of drug combination expressed as products (e.g., MTX*FA or MTX*FA*TNFi). Thus, we used an alternative model for logistic analysis as follows: $\mathit{Log}\left(RORs\right)={\beta }_0+{\beta }_{1}A+{\beta }_{2}S+{\beta }_3{M}_1+{\beta }_4{M}_2+{\beta }_5{M}_3+{\beta }_6{M}_4$
$(\mathrm{A}=\mathrm{a}\mathrm{g}\mathrm{e},\mathrm{S}=\mathrm{s}\mathrm{e}\mathrm{x},{\mathrm{M}}_1=\mathrm{M}\mathrm{T}\mathrm{X}\left(\mathrm{n}\mathrm{o}\mathrm{F}\mathrm{A},\mathrm{T}\mathrm{N}\mathrm{F}\mathrm{i}\right),{\mathrm{M}}_2=\mathrm{M}\mathrm{T}\mathrm{X}+\mathrm{F}\mathrm{A}\left(\mathrm{n}\mathrm{o}\mathrm{T}\mathrm{N}\mathrm{F}\mathrm{i}\right),{\mathrm{M}}_3=\mathrm{M}\mathrm{T}\mathrm{X}+\mathrm{T}\mathrm{N}\mathrm{F}\mathrm{i}\left(\mathrm{n}\mathrm{o}\mathrm{F}\mathrm{A}\right),{\mathrm{M}}_4=\mathrm{M}\mathrm{T}\mathrm{X}+\mathrm{F}\mathrm{A}+\mathrm{T}\mathrm{N}\mathrm{F}\mathrm{i})$
Using this logistic model, we confirmed that all VIF values were ˂ 1.4, and the deviance value was statistically significant, supporting the model’s suitability.
Statistical significance was determined if the upper 95% CI of the ROR was ˂ 1.0 or the lower 95% CI of the ROR was ˃1.0. Fisher’s exact test was used to calculate the p-values of cRORs. Data mining and all statistical analyses were performed using Microsoft Access 2016 (Microsoft Inc. Tokyo, Japan), R version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria), EZR version 1.36 (Kanda, 2013 (link)), and GraphPad Prism ver. 9.2 (GraphPad Software, San Diego, CA).