Search

Summary: Background: The identification of baseline host determinants that associate with robust HIV-1 vaccine-induced immune responses could aid HIV-1 vaccine development. We aimed to assess both the collective and relative performance of baseline characteristics in classifying individual participants in nine different Phase 1-2 HIV-1 vaccine clinical trials (26 vaccine regimens, conducted in Africa and in the Americas) as High HIV-1 vaccine responders. Methods: This was a meta-analysis of individual participant data, with studies chosen based on participant-level (vs. study-level summary) data availability within the HIV-1 Vaccine Trials Network. We assessed the performance of 25 baseline characteristics (demographics, safety haematological measurements, vital signs, assay background measurements) and estimated the relative importance of each characteristic in classifying 831 participants as High (defined as within the top 25th percentile among positive responders or above the assay upper limit of quantification) versus Non-High responders. Immune response outcomes included HIV-1-specific serum IgG binding antibodies and Env-specific CD4+ T-cell responses assessed two weeks post-last dose, all measured at central HVTN laboratories. Three variable importance approaches based on SuperLearner ensemble machine learning were considered. Findings: Overall, 30.1%, 50.5%, 36.2%, and 13.9% of participants were categorized as High responders for gp120 IgG, gp140 IgG, gp41 IgG, and Env-specific CD4+ T-cell vaccine-induced responses, respectively. When including all baseline characteristics, moderate performance was achieved for the classification of High responder status for the binding antibody responses, with cross-validated areas under the ROC curve (CV-AUC) of 0.72 (95% CI: 0.68, 0.76) for gp120 IgG, 0.73 (0.69, 0.76) for gp140 IgG, and 0.67 (95% CI: 0.63, 0.72) for gp41 IgG. In contrast, the collection of all baseline characteristics yielded little improvement over chance for predicting High Env-specific CD4+ T-cell responses [CV-AUC: 0.53 (0.48, 0.58)]. While estimated variable importance patterns differed across the three approaches, female sex assigned at birth, lower height, and higher total white blood cell count emerged as significant predictors of High responder status across multiple immune response outcomes using Approach 1. Of these three baseline variables, total white blood cell count ranked highly across all three approaches for predicting vaccine-induced gp41 and gp140 High responder status. Interpretation: The identified features should be studied further in pursuit of intervention strategies to improve vaccine responses and may be adjusted for in analyses of immune response data to enhance statistical power. Funding: National Institute of Allergy and Infectious Diseases (UM1AI068635 to YH, UM1AI068614 to GDT, UM1AI068618 to MJM, and UM1 AI069511 to MCK), the Duke CFAR P30 AI064518 to GDT, and National Institute of Dental and Craniofacial Research (R01DE027245 to JJK). This work was also supported by the Bill and Melinda Gates Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of any of the funding sources.

Background: The identification of baseline host determinants that associate with robust HIV-1 vaccine-induced immune responses could aid HIV-1 vaccine development. We aimed to assess both the collective and relative performance of baseline characteristics in classifying individual participants in nine different Phase 1-2 HIV-1 vaccine clinical trials (26 vaccine regimens, conducted in Africa and in the Americas) as High HIV-1 vaccine responders., Methods: This was a meta-analysis of individual participant data, with studies chosen based on participant-level (vs. study-level summary) data availability within the HIV-1 Vaccine Trials Network. We assessed the performance of 25 baseline characteristics (demographics, safety haematological measurements, vital signs, assay background measurements) and estimated the relative importance of each characteristic in classifying 831 participants as High (defined as within the top 25th percentile among positive responders or above the assay upper limit of quantification) versus Non-High responders. Immune response outcomes included HIV-1-specific serum IgG binding antibodies and Env-specific CD4+ T-cell responses assessed two weeks post-last dose, all measured at central HVTN laboratories. Three variable importance approaches based on SuperLearner ensemble machine learning were considered., Findings: Overall, 30.1%, 50.5%, 36.2%, and 13.9% of participants were categorized as High responders for gp120 IgG, gp140 IgG, gp41 IgG, and Env-specific CD4+ T-cell vaccine-induced responses, respectively. When including all baseline characteristics, moderate performance was achieved for the classification of High responder status for the binding antibody responses, with cross-validated areas under the ROC curve (CV-AUC) of 0.72 (95% CI: 0.68, 0.76) for gp120 IgG, 0.73 (0.69, 0.76) for gp140 IgG, and 0.67 (95% CI: 0.63, 0.72) for gp41 IgG. In contrast, the collection of all baseline characteristics yielded little improvement over chance for predicting High Env-specific CD4+ T-cell responses [CV-AUC: 0.53 (0.48, 0.58)]. While estimated variable importance patterns differed across the three approaches, female sex assigned at birth, lower height, and higher total white blood cell count emerged as significant predictors of High responder status across multiple immune response outcomes using Approach 1. Of these three baseline variables, total white blood cell count ranked highly across all three approaches for predicting vaccine-induced gp41 and gp140 High responder status., Interpretation: The identified features should be studied further in pursuit of intervention strategies to improve vaccine responses and may be adjusted for in analyses of immune response data to enhance statistical power., Funding: National Institute of Allergy and Infectious Diseases (UM1AI068635 to YH, UM1AI068614 to GDT, UM1AI068618 to MJM, and UM1 AI069511 to MCK), the Duke CFAR P30 AI064518 to GDT, and National Institute of Dental and Craniofacial Research (R01DE027245 to JJK). This work was also supported by the Bill and Melinda Gates Foundation. The content is solely the responsibility of the authors and does not necessarily represent the official views of any of the funding sources., Competing Interests: Declaration of interests Y.H. declares contract payments to her institution from the World Health Organization to conduct statistical analysis work (outside the scope of the current work; related to COVID-19 vaccines) and plan to submit the results for publication, within the past 36 months, as well as service to WHV as an SMB (payment directly to her) within the past 36 months. K.E.S. and J.H. declare coverage of travel expenses for attendance at the HVTN Full Group Meeting in 2022 (reimbursement by the HVTN); K.E.S. additionally declares payment of a registration fee for virtual seminar attendance directly to Keystone Symposium by the HVTN. S.D.R. declares contracts awarded to his institution from Battelle and from Janssen, and grants awarded to his institution from the Gates Medical Research Institute and from the Paul G. Allen Family Foundation, within the past 36 months. P.A.G. declares consulting fees received from Johnson and Johnson within the past 36 months, as well as a patent for a COVID-19 monoclonal antibody that is not yet clinically available. K.W.C. declares funding from the Bill and Melinda Gates Foundation in the form of payments to her institution within the last 36 months. B.D.W. declares grant payments made to his institution within the past 36 months from the following entities: Centers for Disease Control and Prevention, Patient-Centered Outcomes Research Institute, National Institute of Mental Health, National Institute of Allergy and Infectious Diseases, National Cancer Institute, and the National Institute of General Medical Studies; an honorarium for delivering an invited webinar to the American Statistical Association: Statistical Learning and Data Science Section; travel support through a grant from the National Institute of Mental Health; and retirement accounts invested in mutual funds operated by Vanguard and by TIAA (he does not choose individual stocks). L.-G.B. declares payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing or educational events from Gilead Sciences, Janssen, and Merck PTY LTD within the past 36 months, as well as participation on a Data Safety Monitoring Board or Advisory Board for the PrEPVAC study within the past 36 months. G.D.T. declares consulting fees received from Axon and from Janssen (not related to this work), serving as an Advisory Board member for the NIH VRC, UNC CFAR, and Johns Hopkins (not related to this work), and serving as a scientific review member for Gilead (not related to this work), all within the past 36 months, as well as travel as part of research funding (more than 3 years prior, with funding paid to her institution). J.T. declares consulting fees from Vaccitech within the past 36 months., (Copyright © 2022 The Author(s). Published by Elsevier B.V. All rights reserved.)