|
Activity Number:
|
91
- High Dimensional Data, Causal Inference, Biostats Education, and More
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 9, 2021 : 10:00 AM to 11:50 AM
|
|
Sponsor:
|
ENAR
|
|
Abstract #318155
|
|
|
Title:
|
Increasing efficiency and reducing bias of vaccine effect estimates, by using non-targeted virus strains
|
|
Author(s):
|
Lola Etievant* and Mitchell Gail and Joshua Sampson
|
|
Companies:
|
National Cancer Institute, NIH and NCI-DCEG and National Cancer Institute, NIH
|
|
Keywords:
|
estimating equations;
augmented estimating equations;
improving estimation;
de-biasing;
secondary outcome, negative control outcome;
HPV vaccination
|
|
Abstract:
|
Studies of vaccine efficacy often obtain data on vaccine-targeted virus strains, e.g. human papilloma virus (HPV) 16/18, and on non-targeted strains. We use as secondary outcome, or negative control outcome, non-targeted strains that are assumed not to be affected by the vaccine. We show how non-targeted strains can (i) increase the precision of the estimate of vaccine effect on targeted strains in randomized trials, and (ii) reduce confounding bias in observational studies. For objective (i), we augment the primary outcome estimating equation with a function of the secondary outcome, rather than with a function of covariates as previously proposed. For objective (ii), we jointly estimate the treatment effect on primary and secondary outcomes. Under certain assumptions on the unobserved process inducing correlation between the outcomes, the non-null effect on the secondary outcome can be used to correct the bias in the effect on the primary outcome. In simulations and analyses based on HPV data, methods using non-targeted strains modestly increase precision of the vaccine effect on primary outcome in randomized trials, and appreciably reduce bias in observational studies.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2021 program
|
