DOI QR코드

DOI QR Code

Propensity score methods for estimating treatment delay effects

생존자료분석에서 성향 점수를 이용한 treatment delay effect 추정법에 대한 연구

  • Jooyi Jung (Department of Biostatistics, Korea University College of Medicine, Korea University) ;
  • Hyunjin Song (Department of Biostatistics, Korea University College of Medicine, Korea University) ;
  • Seungbong Han (Department of Biostatistics, Korea University College of Medicine, Korea University)
  • 정주이 (고려대학교 의학통계학협동과정) ;
  • 송현진 (고려대학교 의학통계학협동과정) ;
  • 한승봉 (고려대학교 의학통계학협동과정)
  • Received : 2023.04.04
  • Accepted : 2023.05.08
  • Published : 2023.10.31

Abstract

Oftentimes, the time dependent treatment covariate and the time dependent confounders exist in observation studies. It is an important problem to correctly adjust for the time dependent confounders in the propensity score analysis. Recently, In the survival data, Hade et al. (2020) used a propensity score matching method to correctly estimate the treatment delay effect when the time dependent confounder affects time to the treatment time, where the treatment delay effects is defined to the delay in treatment reception. In this paper, we proposed the Cox model based marginal structural model (Cox-MSM) framework to estimate the treatment delay effect and conducted extensive simulation studies to compare our proposed Cox-MSM with the propensity score matching method proposed by Hade et al. (2020). Our simulation results showed that the Cox-MSM leads to more exact estimate for the treatment delay effect compared with two sequential matching schemes based on propensity scores. Example from study in treatment discontinuation in conjunction with simulated data illustrates the practical advantages of the proposed Cox-MSM.

생존 자료에서 Hade 등 (2020) 은 시간-의존 교란 변수가 환자의 처치 시점에 영향을 미칠 때, 해당 효과를 보정하여 treatment delay effect를 올바르게 추정하기 위해 성향 점수 매칭 방법을 이용하였다. 이 때, treatment delay effect란 환자가 관심 있는 지연 시점만큼 늦게 처치를 받는 경우 제 때 받는 경우에 비해 사건 발생 위험에 미치는 영향을 의미한다. 본 연구에서는 또 다른 성향 점수 기반 모형인 Cox-MSM 모형 또한 해당 효과를 올바르게 추정할 수 있는지 모의 실험을 통해 확인 및 기존 매칭 모형과 비교하였다. 모의실험 결과, 세 가지 모형 모두 다양한 시나리오 내에서 treatment delay effect를 올바르게 추정함을 확인하였다. 특히 모든 시나리오 내에서 Cox-MSM의 제곱근평균제곱오차의 값이 가장 낮았으며, restricted Cox matching 모형에서 가장 큰 값을 가지는 것으로 나타났다. 결론적으로, 성향 점수에 기반하나 매칭이 아닌 방법 또한 treatment delay effect 적용이 가능하다는 결과를 제공한다. 추후 G-formula과 같이 성향 점수 기반이 아닌 모형에서도 적용이 가능한지에 대한 상세 연구가 필요하다고 사료된다.

Keywords

Acknowledgement

본 연구는 과학기술정보통신부 재원 한국연구재단 (No. 2022R1F1A1063027)과 보건복지부 재원 한국보건산업진흥원의 보건의료기술연구개발사업 (No. HI22C045400)의 지원을 받아 수행된 연구임.

References

  1. Austin PC (2012). The performance of different propensity score methods for estimating marginal hazard ratios, Statistics in Medicine, 32, 2837-2849. https://doi.org/10.1002/sim.5705
  2. Austin PC (2014). The use of propensity score methods with survival or time-to-event outcomes: Reporting measures of effect similar to those used in randomized experiments, Statistics in Medicine, 33, 1242-1258. https://doi.org/10.1002/sim.5984
  3. Bender R, Augustin T, and Blettner M (2005). Generating survival times to simulate Cox proportional hazards models, Statistics in Medicine, 24, 1713-1723. https://doi.org/10.1002/sim.2059
  4. Hade EM, Nattino G, Frey HA, and Lu B (2020). Propensity score matching for treatment delay effects with observational survival data, Statistical Methods in Medical Research, 29, 695-708. https://doi.org/10.1177/0962280219877908
  5. Hernan M ' A, Brumback B, and Robins JM (2000). Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men, Epidemiology, 11, 561-570. https://doi.org/10.1097/00001648-200009000-00012
  6. Hernan MA and Robins JM (2010). Causal Inference: What If, CRC Boca Raton, FL.
  7. Lu B (2005). Propensity score matching with time-dependent covariates, Biometrics, 61, 721-728. https://doi.org/10.1111/j.1541-0420.2005.00356.x
  8. Lu B, Cai D, and Tong X (2018). Testing causal effects in observational survival data using propensity score matching design, Statistics in Medicine, 37, 1846-1858. https://doi.org/10.1002/sim.7599
  9. Robins JM, Hernan MA, and Brumback B (2000). Marginal structural models and causal inference in epidemiology, Epidemiology, 11, 550-560. https://doi.org/10.1097/00001648-200009000-00011
  10. Stuart EA (2010). Matching methods for causal inference: A review and a look forward, Statistical Science, 25, 1-21. https://doi.org/10.1214/09-STS313
  11. van der Wal WM and Geskus RB (2011). ipw: An R-package for inverse probability weighting, Journal of Statistical Software, 43, 1-23. https://doi.org/10.18637/jss.v043.i13
  12. Xiao Y, Abrahamowicz M, and Moodie EE (2010). Accuracy of conventional and marginal structural Cox model estimators: A simulation study, The International Journal of Biostatistics, 6, 1-28. https://doi.org/10.2202/1557-4679.1208
  13. Yang S and Holloway ST (2021). contTimeCausal-Continuous Time Causal Models, Available from: https://cran.rproject.org/web/packages/contTimeCausal/index.html
  14. Young JG, Hernan MA, Picciotto S, and Robins JM (2010). Relation between three classes of structural models for the effect of a time-varying exposure on survival, Lifetime Data Analysis, 16, 71-84. https://doi.org/10.1007/s10985-009-9135-3
  15. Zhao Q-Y, Luo J-C, Su Y, Zhang Y-J, Tu G-W, and Luo Z (2021). Propensity score matching with R: Conventional methods and new features, Annals of Translational Medicine, 9, 1-39. https://doi.org/10.21037/atm-2020-132