Validation of Epidemiological Methods for Studying the Infection Risk in Rheumatoid Arthritis with Methotrexate Combined Biologicals using Propensity Scores
- VernacularTitle:Methotrexate 投与関節リウマチ患者における生物学的製剤併用時の感染リスクに関する Propensity Score を用いた複数の疫学的方法の妥当性の検討
- Author:
Takuma KOINUMA
1
;
Manabu AKAZAWA
2
Author Information
- Keywords: rheumatoid arthritis; biological; immortal time bias; propensity score; database
- From:Japanese Journal of Pharmacoepidemiology 2021;26(1):27-40
- CountryJapan
- Language:Japanese
- Abstract: Objective：In epidemiological studies, changes in patient conditions caused by treatment would be chronologically repeated. Thus, the manner of representing this change can create time-dependent bias which researchers should address. In this study, we aimed to validate the estimators obtained using various epidemiological methods based on the infection risk between the administration of methotrexate (MTX) alone and MTX combined biologicals.Design：Cohort studyMethods：We extracted data regarding 3769 rheumatoid arthritis (RA) patients, consisting of 2805 patients with MTX alone and 964 patients with MTX combined biologicals from the claims data from JMDC Inc.. We represented each time course using the time axis of the elapsed time, the prescription number, and the administration time to make the corresponding data set. Subsequently, we performed time-conditional propensity score (PS) matching for matched points in each time axis. We also performed Inverse Probability Weighting Estimator (IPW) and Augmented Inverse Probability Weighting Estimator (AIPW) analyses.Results：The Odds Ratios (OR) estimated by each method using the time axis of the elapsed time, the prescription number, and the administration time were 1.48 (95％CI 0.71-3.11), 1.60 (95％CI 0.72-3.55), and 1.04 (95％CI 0.58-1.86), respectively. We performed PS weighting, of each Average Treatment Effect obtained from IPW, and the AIPW were estimated to be 0.31％ (95％CI −0.91-1.53) and 0.29％ (95％CI −0.91-1.49), respectively, and the average treatment effect on the treated was estimated to be 0.10％ (95％CI −1.11-1.32). We support the findings of a previous study which showed that the combination of biologicals was not statistically associated with increased infection risk.Conclusion：This study suggests that estimators from different perspectives might be obtained by using some epidemiological methods. Therefore, our results could contribute to the establishment of an improved methodology.
