서브메뉴
검색
Statistical Methods for Recurrent Events and First Non-Fatal Events in Clinical Trials
Statistical Methods for Recurrent Events and First Non-Fatal Events in Clinical Trials
상세정보
- 자료유형
- 학위논문 서양
- 최종처리일시
- 20260202103024
- ISBN
- 9798315701705
- DDC
- 574
- 저자명
- Sun, Jiren.
- 서명/저자
- Statistical Methods for Recurrent Events and First Non-Fatal Events in Clinical Trials
- 발행사항
- [Sl] : The University of Wisconsin - Madison, 2025
- 발행사항
- Ann Arbor : ProQuest Dissertations & Theses, 2025
- 형태사항
- 177 p
- 주기사항
- Source: Dissertations Abstracts International, Volume: 86-11, Section: B.
- 주기사항
- Advisor: Mao, Lu;Cook, Thomas D.
- 학위논문주기
- Thesis (Ph.D.)--The University of Wisconsin - Madison, 2025.
- 초록/해제
- 요약This dissertation addresses four statistical questions related to the analysis of first non-fatal events and recurrent events-non-fatal events that occur repeatedly within the same subject-in clinical trials, and develops four corresponding statistical methods.In the first project, we investigate the question: How can we quantify the difference between the commonly reported cause-specific hazard ratio (CSHR) and the direct effect of treatment on the underlying first non-fatal event process in the presence of death? To answer this, we introduce the Proportional Principal Stratum Hazards (PPSH) model within the principal stratification framework. The PPSH model estimates the principal stratum hazard ratio (PSHR), which reflects the direct effect on the underlying first non-fatal event process, assuming correct model specification. By reporting the PSHR alongside the CSHR, researchers can gain a more comprehensive understanding of the direct effect on the underlying first non-fatal event process.In the second project, we address the question: How can we improve the precision of area under the curve estimation for the mean cumulative function while preserving its unconditional interpretability? To this end, we propose a nonparametric covariate adjustment approach that ensures efficiency gains over unadjusted analyses and applies universally to various randomization schemes, including both simple and covariate-adaptive designs.In the third project, we explore the question: How can we estimate treatment effects under a hypothetical scenario where the intercurrent event-post-randomization events affecting outcome interpretation or existence-does not occur? We apply inverse probability weighting to the widely used Lin-Wei-Yang-Ying and negative binomial models, appropriately adjusting for baseline and internal time-varying covariates, to obtain unbiased estimates of hypothetical treatment effects. Simulation studies demonstrate that our approach outperforms alternative analytical methods in terms of bias and power.In the final project, we examine the question: How can we estimate treatment effects on recurrent events and recover the underlying trajectory of recurrent events in the presence of death? We propose a parametric shared frailty model that enables formal testing of recurrent event trends and offers greater power than traditional time-to-first-event analyses in heterogeneous clinical trial populations.
- 일반주제명
- Biostatistics
- 일반주제명
- Statistics
- 일반주제명
- Mathematics
- 일반주제명
- Applied mathematics
- 키워드
- Causal inference
- 키워드
- Clinical trials
- 키워드
- Competing risks
- 키워드
- Recurrent events
- 기타저자
- The University of Wisconsin - Madison Biomedical Data Science
- 기본자료저록
- Dissertations Abstracts International. 86-11B.
- 전자적 위치 및 접속
- 로그인 후 원문을 볼 수 있습니다.
MARC
008260126s2025 us c eng d■001000017356726
■00520260202103024
■006m o d
■007cr#unu||||||||
■020 ▼a9798315701705
■035 ▼a(MiAaPQ)AAI31844988
■040 ▼aMiAaPQ▼cMiAaPQ
■0820 ▼a574
■1001 ▼aSun, Jiren.
■24510▼aStatistical Methods for Recurrent Events and First Non-Fatal Events in Clinical Trials
■260 ▼a[Sl]▼bThe University of Wisconsin - Madison▼c2025
■260 1▼aAnn Arbor▼bProQuest Dissertations & Theses▼c2025
■300 ▼a177 p
■500 ▼aSource: Dissertations Abstracts International, Volume: 86-11, Section: B.
■500 ▼aAdvisor: Mao, Lu;Cook, Thomas D.
■5021 ▼aThesis (Ph.D.)--The University of Wisconsin - Madison, 2025.
■520 ▼aThis dissertation addresses four statistical questions related to the analysis of first non-fatal events and recurrent events-non-fatal events that occur repeatedly within the same subject-in clinical trials, and develops four corresponding statistical methods.In the first project, we investigate the question: How can we quantify the difference between the commonly reported cause-specific hazard ratio (CSHR) and the direct effect of treatment on the underlying first non-fatal event process in the presence of death? To answer this, we introduce the Proportional Principal Stratum Hazards (PPSH) model within the principal stratification framework. The PPSH model estimates the principal stratum hazard ratio (PSHR), which reflects the direct effect on the underlying first non-fatal event process, assuming correct model specification. By reporting the PSHR alongside the CSHR, researchers can gain a more comprehensive understanding of the direct effect on the underlying first non-fatal event process.In the second project, we address the question: How can we improve the precision of area under the curve estimation for the mean cumulative function while preserving its unconditional interpretability? To this end, we propose a nonparametric covariate adjustment approach that ensures efficiency gains over unadjusted analyses and applies universally to various randomization schemes, including both simple and covariate-adaptive designs.In the third project, we explore the question: How can we estimate treatment effects under a hypothetical scenario where the intercurrent event-post-randomization events affecting outcome interpretation or existence-does not occur? We apply inverse probability weighting to the widely used Lin-Wei-Yang-Ying and negative binomial models, appropriately adjusting for baseline and internal time-varying covariates, to obtain unbiased estimates of hypothetical treatment effects. Simulation studies demonstrate that our approach outperforms alternative analytical methods in terms of bias and power.In the final project, we examine the question: How can we estimate treatment effects on recurrent events and recover the underlying trajectory of recurrent events in the presence of death? We propose a parametric shared frailty model that enables formal testing of recurrent event trends and offers greater power than traditional time-to-first-event analyses in heterogeneous clinical trial populations.
■590 ▼aSchool code: 0262.
■650 4▼aBiostatistics
■650 4▼aStatistics
■650 4▼aMathematics
■650 4▼aApplied mathematics
■653 ▼aCausal inference
■653 ▼aClinical trials
■653 ▼aCompeting risks
■653 ▼aIntercurrent events
■653 ▼aRecurrent events
■653 ▼aShared frailty model
■690 ▼a0308
■690 ▼a0463
■690 ▼a0405
■690 ▼a0364
■71020▼aThe University of Wisconsin - Madison▼bBiomedical Data Science.
■7730 ▼tDissertations Abstracts International▼g86-11B.
■790 ▼a0262
■791 ▼aPh.D.
■792 ▼a2025
■793 ▼aEnglish
■85640▼uhttp://www.riss.kr/pdu/ddodLink.do?id=T17356726▼nKERIS▼z이 자료의 원문은 한국교육학술정보원에서 제공합니다.


