Advances in Causal Inference to Support Vaccine Development and Evaluation 公开

Abstract

Communicable disease outbreaks continue to present significant challenges to human society. The incidence of various infectious diseases remains alarmingly high globally. Consequently, developing preventive vaccines has become a pivotal objective in mitigating infectious disease burden. This dissertation centers on the creation of statistical methods that support the design and evaluation of potential vaccines.

In the first section, we devise methods to identify key genetic mutations in the HIV envelope protein linked to antibody resistance. This task proves complex due to the high-dimensional and strongly correlated nature of genetic sequence data. We propose a solution using an outcome-adaptive, collaborative targeted minimum loss-based estimation approach combined with random forests, which enjoys significant advantages over existing methods. We apply this approach to the Compile, Analyze and Tally Nab Panels (CATNAP) database to identify amino acid positions causally related to resistance to neutralization by various antibodies.

In the second section, we develop methods for standardized comparisons of immunogenicity across diverse vaccine trials, involving different populations and study designs. To address this, we introduce a causal framework capable of identifying suitable causal estimands and estimators to bridge the immunogenicity of one vaccine from the trial population where it was evaluated to other trial populations. We apply the proposed technique to compare vaccine effectiveness using data from four recent HIV vaccine trials.

In the third section, we create methods to generate standardized versions of causal effects that can be used to compare the impact of vaccines on various outcomes that are measured on different scales. For example, we may wish to compare the difference in immunogenicity between two vaccines in terms of two different immunologic assays. If these assay readouts have substantially different variability, then a comparison of relative vaccine performance across assays can be challenging. To rectify this, we develop a general framework for defining standardized causal effect sizes. We develop nonparametric efficient estimators of these quantities and evaluate the estimators' performance in comprehensive numerical studies.

Table of Contents

1 Introduction 1

1.1 Overview ................................... 1

1.2 Targeted Machine Learning for Understanding HIV Resistance to Neutralizing Antibodies................................ 2

1.3 Comparing HIV Vaccine Immunogenicity across Trials with Different Populations and Study Designs.......................... 7

1.4 Standardized Causal Effect Sizes in Biomedical Research . . . . . . . . . . 10

2 Identifying HIV sequences that escape antibody neutralization using random forests and collaborative targeted learning 12

2.1 Introduction.................................. 13

2.2 Methods.................................... 14

2.2.1 Counterfactual antibody resistance probability . . . . . . . . . . . 14 

2.2.2 Causal identification ......................... 16 

2.2.3 Motivation for novel method..................... 19 

2.2.4 TMLE................................. 20 

2.2.5 CTMLE................................ 23

2.3 Simulation Study ............................... 25 

2.3.1 Design ................................ 25 

2.3.2 Results ................................ 27

2.3.3 Comparison with non-causal approach. . . . . . . . . . . . . . . . 27

2.4 Data analyses ................................. 28 

2.4.1 CATNAP datasets .......................... 28 

2.4.2 Results ................................ 29 

2.4.3 Comparison with other approaches.................. 29

2.5 Discussion................................... 30

3 Comparing HIV Vaccine Immunogenicity across Trials with Different Populations and Study Designs 35

3.1 Introduction.................................. 36

3.2 Materials and Method............................. 37

3.2.1 Notation and data structure ..................... 37

3.2.2 Causal estimands........................... 41

3.2.3 Identification of standardized immunogenicity using full data . . . . 43

3.2.4 Identification of standardized immunogenicity using observed data . 46

3.2.5 Towards estimation: efficiency theory for identifying estimands . . 48

3.2.6 Targeted minimum loss estimation.................. 49

3.2.7 Hypothesis Testings and Confidence Intervals . . . . . . . . . . . . 52

3.3 SimulationStudies .............................. 53

3.3.1 Comparing within and across early and late phase trials . . . . . . . 53 

3.3.2 Simulations inspired by HVTN trials ................ 55

3.4 Application to RV144 and HVTN Trials ................... 56

4 Standardized Causal Effect Sizes in Vaccine Research 58

4.1 Introduction.................................. 59

4.2 Considerations for defining standardized causal effect sizes . . . . . . . . . 59 

4.2.1 Causal effects............................. 59 

4.2.2 Standardized causal effect sizes ................... 60

4.2.3 Choices of measure of counterfactual variability . . . . . . . . . . 61

4.3 Identification, estimation, and inference for standard causal effect sizes . . . 63 

4.3.1 Efficiency theory ........................... 65 

4.3.2 Plug-in estimation .......................... 65 

4.3.3 One-step corrected estimation .................... 66 

4.3.4 Asymptotic study of one-step estimator . . . . . . . . . . . . . . . 67

4.4 Simulation study ............................... 68

4.5 Discussion................................... 69

Appendix A 

Appendix for Chapter 3 71

A.1 Proof of Theorem1.............................. 71

Bibliography 72

About this Dissertation

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Laney Graduate School
Department	Biostatistics
Degree	Ph.D.
Submission	Dissertation
Language	English
Research Field	Biology, Biostatistics Statistics
关键词	Infectious Disease Targeted Learning Causal Inference
Committee Chair / Thesis Advisor	David Benkeser, Emory University
Committee Members	Robert Lyles, Emory University Hao Wu, Emory University Holly Janes, Fred Hutchinson Cancer Center

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Advances in Causal Inference to Support Vaccine Development and Evaluation ()	2023-07-23 17:39:44 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions