Methods for Feature Selection in Down-Selection of Vaccine Regimens Based on Multivariate Immune Response Endpoints
- PDF / 1,930,177 Bytes
- 23 Pages / 439.37 x 666.142 pts Page_size
- 85 Downloads / 200 Views
Methods for Feature Selection in Down‑Selection of Vaccine Regimens Based on Multivariate Immune Response Endpoints Ying Huang1 · Aliasghar Tarkhan2 Received: 5 May 2019 / Revised: 3 December 2019 / Accepted: 19 March 2020 © International Chinese Statistical Association 2020
Abstract In clinical trials, it is often of interest to compare and order several candidate regimens based on multiple endpoints. For example, in HIV vaccine development, immune response profiles induced by vaccination are key for selecting vaccine regimens to advance to efficacy evaluation. Motivated by the need to rank and choose a few vaccine regimens based on their immunogenicity in phase I trials, Huang et al. (Biostatistics 18(2):230–243, 2017) proposed a ranking/filtering/selection algorithm that down-selects vaccine regimens to satisfy the superiority and non-redundancy criteria, based on multiple immune response endpoints. In practice, many candidate immune response endpoints can be correlated with each other. An important question that remains to be addressed is how to choose a parsimonious set of the available immune response endpoints to effectively compare regimens. In this paper, we propose novel algorithms for selecting immune response endpoints to be used in regimen down-selection, based on importance weights assigned to individual endpoints and their correlation structure. We show through extensive simulation studies that pre-selection of endpoints can substantially improve performance of the subsequent regimen down-selection process. The application of the proposed method is demonstrated using a real example in HIV vaccine research, although the methods are also applicable in general to clinical research for dimension reduction when comparing regimens based on multiple candidate endpoints. Keywords Correlation · Down-selection · Feature selection · Importance weight · Measurement error · Vaccine trial
Ying Huang and Aliasghar Tarkhan have contributed equally. Electronic supplementary material The online version of this article (https://doi.org/10.1007/s1256 1-020-09275-2) contains supplementary material, which is available to authorized users. * Ying Huang [email protected] Extended author information available on the last page of the article
13
Vol.:(0123456789)
Statistics in Biosciences
1 Introduction When developing a vaccine against a rare disease such as HIV/AIDS, efficacy trials are typically large and operationally challenging to conduct, making it critical to select and rank candidate vaccine regimens based on their immunogenicity in phase I studies before the regimens can be advanced to efficacy evaluation. We and others in the HIV Vaccine Trials Network (HVTN) have been developing statistical approaches and frameworks for this process. For example, out of 15 vaccine regimens studied in Phase I trials by the HVTN, combining 5 unique prime-boost types and 3 Env dose × adjuvant types, Huang et al. [13] described statistical approaches for selecting up to 3 regimens to advance for concurrent testing in
Data Loading...
