Overlap weighting in detection of adverse drug reaction signals: a simulated study

doi:10.19803/j.1672-8629.20230113

Abstract

Abstract: Objective To explore the effect of overlap weighting in equalizing confounding factors during detection of adverse drug reaction signals via a simulated study. Methods Point estimates, standard errors, bias, mean square errors, and 95% confidence interval coverage rate were used to evaluate the accuracy and stability of effect estimation of four methods (propensity score stratification, propensity score matching, inverse probability weighting, and overlap weighting) at four different levels of overlap between groups. Results In the simulated scenario of this study, overlap weighting yielded the most accurate and stable results of effect estimation, while the accuracy and stability of the other three methods became progressively worse as the degree of overlap decreased, and the inverse probability weighting led to the most significant difference. Conclusion Overlap weighting can be an option for equalizing confounding factors during detection of signals of adverse drug reactions.

Key words: adverse drug reaction, signal detection, confounding factors, overlap weighting, propensity score stratification, propensity score matching, inverse probability weighting

CLC Number:

R911
R994

WEI Lianhui, ZHANG Pengpeng, LIU Yongmei, GUO Xiaojing, XU Jinfang, CHEN Xiao, CHEN Chenxin, YE Xiaofei. Overlap weighting in detection of adverse drug reaction signals: a simulated study[J]. Chinese Journal of Pharmacovigilance, 2023, 20(6): 646-650.

References

[1] REN JT, WANG SF, HOU YF, et al.Common signal detection methods of adverse drug reaction[J]. Chinese Journal of Pharmacovigilance(中国药物警戒), 2011, 8(5): 294-298.
[2] VETTER TR, MASCHA EJ.Bias, confounding, and interaction: lions and tigers, and bears, oh my![J]. Anesth Analg, 2017, 125(3): 1042-1048.
[3] SCHOBER P, VETTER TR.Propensity score matching in observational research[J]. Anesth Analg, 2020, 130(6): 1616-1617.
[4] ZHANG X.Modeling of bayes propensity score and application in signal detection of adverse drug reactions[D]. Shanghai: Naval Medical University, 2014.
[5] HUANG LH, WANG YJ, WANG SZ, et al.Statistical consensus on propensity scoring methods and their normative application[J]. Chinese Journal of Health Statistics(中国卫生统计), 2020, 37: 952-958.
[6] LI F, THOMAS LE, LI F.Addressing extreme propensity scores via the overlap weights[J]. Am J Epidemiol, 2019, 188(1): 250-257.
[7] QIAN W, YE XF, WANG C, et al.Methods of controlling confounding factors in adverse drug reaction signal detection[J]. Chinese Journal of Pharmacovigilance(中国药物警戒), 2010, 7(3): 142-144.
[8] ROSENBAUM PR, RUBIN DB. Reducing bias in observational studies using subclassification on the propensity score[J]. J Am Stat Assoc, 1984, 79: 5l6-524.
[9] AUSTIN PC.An introduction to propensity score methods for reducing the effects of confounding in observational studies[J]. Multivariate Behav Res, 2011, 46(3): 399-424.
[10] D’AGOSTINO RB JR. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group[J]. Stat Med, 1998, 17(19): 2265-2281.
[11] ABADIE A, IMBENS GW.Large sample properties of matching estimators for average treatment effects[J]. Econometrica, 2006, 74: 235-267.
[12] STUART EA.Matching methods for causal inference: a review and a look forward[J]. Stat Sci, 2010, 25(1): 1-21.
[13] LEE BK, LESSLER J, STUART EA.Weight trimming and propensity score weighting[J]. PLoS One, 2011, 6(3): e18174.
[14] HIRANO K, IMBENS GW.Estimation of causal effects using propensity score weighting: an application to data on right heart catheterization[J]. Health Services Outcomes Research Methodology, 2001, 2: 259-278.
[15] STüRMER T, ROTHMAN KJ, AVORN J, et al. Treatment effects in the presence of unmeasured confounding: dealing with observations in the tails of the propensity score distribution--a simulation study[J]. Am J Epidemiol, 2010, 172(7): 843-854.
[16] CRUMP RK, HOTZ VJ, IMBENS GW, et al.Dealing with limited overlap in estimation of average treatment effects[J]. Biometrika, 2009, 96(1): 187-199.
[17] HOU YF, REN JT, JIANG J, et al.Study of adverse drug reaction signal detection[J]. Chinese Journal of Pharmacoepidemiology(药物流行病学杂志), 2010, 19(07): 369-372.
[18] LUQUE-FERNANDEZ MA, SCHOMAKER M, RACHET B, et al.Targeted maximum likelihood estimation for a binary treatment: a tutorial[J]. Stat Med, 2018, 37(16): 2530-2546.