Objective. To evaluate the comparative effectiveness of nonbiologic disease-modifying antirheumatic drugs (DMARD) versus biologic DMARD (bDMARD) for treatment of rheumatoid arthritis (RA), using 2 common analytic approaches. Methods. We analyzed change in Clinical Disease Activity Index (CDAI) scores in patients with RA enrolled in a US-based observational registry from 2001 to 2008 using multivariable (MV) regression and propensity score (PS) matching. Among patients who initiated treatment with a nonbiologic DMARD (n = 1729), we compared patients who switched to, or added, another nonbiologic (n = 182) or a bDMARD (n = 342) at 5, 9, and 24 months after treatment change. Results. Both analytic approaches showed that patients switching to or adding another nonbiologic DMARD demonstrated improvement across 9 and 24 months (both p < 0.001). Both approaches also demonstrated greater improvement in CDAI among recipients of bDMARD relative to a second nonbiologic DMARD at 5 months (p < 0.02). The MV regression approach upheld these results at 9 and 24 months (p < 0.03). In contrast, the PS-matching approach did not show a sustained advantage with bDMARD at these later timepoints, possibly because of lower statistical power and/or lower baseline disease activity in the PS-matched cohort. Conclusion. Patients in both treatment groups generally experienced lower CDAI scores across time. Patients switching to bDMARD demonstrated greater improvement than patients switching to nonbiologic DMARD with both analytic approaches at 5 months. Relative advantages with bDMARD were observed at 9 and 24 months only with MV regression. These analyses provide a practical example of how findings in comparative effectiveness research can diverge with different methodological approaches. Copyright © 2013 The Journal of Rheumatology.