Washington D.C., USA, 22-27 July 2012www.aids2012.org Statistical Design and Analysis for Immune Correlates Assessment: Basic Concepts and RV144 Illustration.

Slides:



Advertisements
Similar presentations
Pox-Protein Public-Private Partnership (P5)
Advertisements

1 Workshop on the immunological basis of vaccine efficacy Vaccine and Infectious Disease Institute December 14, 2009 Ira M. Longini, Jr. Center for Statistical.
V.: 9/7/2007 AC Submit1 Statistical Review of the Observational Studies of Aprotinin Safety Part I: Methods, Mangano and Karkouti Studies CRDAC and DSaRM.
CONCEPTS UNDERLYING STUDY DESIGN
Clinical Trial Designs for the Evaluation of Prognostic & Predictive Classifiers Richard Simon, D.Sc. Chief, Biometric Research Branch National Cancer.
1 Case-Control Study Design Two groups are selected, one of people with the disease (cases), and the other of people with the same general characteristics.
Prevalence of and Progression to Abnormal Non-Invasive Markers of Liver Disease (APRI and FIB-4) among US HIV-infected Youth Kapogiannis B, Leister E,
Chance, bias and confounding
ODAC May 3, Subgroup Analyses in Clinical Trials Stephen L George, PhD Department of Biostatistics and Bioinformatics Duke University Medical Center.
Estimation and Reporting of Heterogeneity of Treatment Effects in Observational Comparative Effectiveness Research Prepared for: Agency for Healthcare.
Elements of a clinical trial research protocol
What is a sample? Epidemiology matters: a new introduction to methodological foundations Chapter 4.
1 Using Biostatistics to Evaluate Vaccines and Medical Tests Holly Janes Fred Hutchinson Cancer Research Center.
Model and Variable Selections for Personalized Medicine Lu Tian (Northwestern University) Hajime Uno (Kitasato University) Tianxi Cai, Els Goetghebeur,
SOME ADDITIONAL POINTS ON MEASUREMENT ERROR IN EPIDEMIOLOGY Sholom May 28, 2011 Supplement to Prof. Carroll’s talk II.
Augmented designs to assess immune responses in vaccine efficacy trials Talk adapted from Dean Follmann’s slides NIAID Biostat 578A Lecture 12.
Sample Size Determination
Cohort Studies Hanna E. Bloomfield, MD, MPH Professor of Medicine Associate Chief of Staff, Research Minneapolis VA Medical Center.
Re-Examination of the Design of Early Clinical Trials for Molecularly Targeted Drugs Richard Simon, D.Sc. National Cancer Institute linus.nci.nih.gov/brb.
Chapter 14 Inferential Data Analysis
Thoughts on Biomarker Discovery and Validation Karla Ballman, Ph.D. Division of Biostatistics October 29, 2007.
CBER Perspective VRBPAC Meeting, November 16, 2010.
BC Jung A Brief Introduction to Epidemiology - XI (Epidemiologic Research Designs: Experimental/Interventional Studies) Betty C. Jung, RN, MPH, CHES.
Statistical Issues in Data Collection and Study Design For Community Programs and Research October 11, 2001 Elizabeth Garrett Division of Biostatistics.
Cohort Study.
Chapter 4 Hypothesis Testing, Power, and Control: A Review of the Basics.
Concepts of Interaction Matthew Fox Advanced Epi.
Lecture 8 Objective 20. Describe the elements of design of observational studies: case reports/series.
ISCB Vaccines Sub-Committee Web Seminar Series November 7, 2012 Assessing Immune Correlates of Protection Via Estimation of the Vaccine Efficacy Curve.
Epidemiology The Basics Only… Adapted with permission from a class presentation developed by Dr. Charles Lynch – University of Iowa, Iowa City.
Biostatistics Case Studies Peter D. Christenson Biostatistician Session 5: Analysis Issues in Large Observational Studies.
RV 144: The Thai Phase III Trial and Development of a Globally-Effective, Multi-Clade HIV Vaccine HIV Vaccine: Quo Vadis AIDS July 2010 Dr. Merlin.
EDRN Approaches to Biomarker Validation DMCC Statisticians Fred Hutchinson Cancer Research Center Margaret Pepe Ziding Feng, Mark Thornquist, Yingye Zheng,
On Surrogate Endpoints in HIV Vaccine Efficacy Trials Steven Self, Peter Gilbert, Michael Hudgens FHCRC/UW FDA/Industry Statistics Workshop, Sept 18-19,
Flow cytometry to evaluate vaccine-induced T cell responses: standardized analysis of large numbers of FCS files Stephen De Rosa, M.D. HVTN Laboratory.
Biostatistics Case Studies 2007 Peter D. Christenson Biostatistician Session 3: Incomplete Data in Longitudinal Studies.
Epidemiologic design from a sampling perspective Epidemiology II Lecture April 14, 2005 David Jacobs.
Adaptive randomization
Sample Size Considerations for Answering Quantitative Research Questions Lunch & Learn May 15, 2013 M Boyle.
The Use of Predictive Biomarkers in Clinical Trial Design Richard Simon, D.Sc. Chief, Biometric Research Branch National Cancer Institute
EXPERIMENTAL EPIDEMIOLOGY
Causal relationships, bias, and research designs Professor Anthony DiGirolamo.
Guidelines for Appropriate OAI Data Use December 6, 2007 Yuqing Zhang, Boston University.
August 20, 2003FDA Antiviral Drugs Advisory Committee Meeting 1 Statistical Considerations for Topical Microbicide Phase 2 and 3 Trial Designs: A Regulatory.
Issues concerning the interpretation of statistical significance tests.
ProteinStart position HLAEpitopePositive responses (n) Env209A*0101SFEPIPSHY1 Env310A*0101/Cw*0401GPGPGRAFY1 Gag406A*0302RAPRKKGC WK 1 Nef9A*0101/A*0302SVVGWPAVR1.
25 Years of HIV Vaccine Research: What have we accomplished? José Esparza MD, PhD Senior Advisor on HIV Vaccines Global Health Program The Search for an.
Matching. Objectives Discuss methods of matching Discuss advantages and disadvantages of matching Discuss applications of matching Confounding residual.
Design of Clinical Research Studies ASAP Session by: Robert McCarter, ScD Dir. Biostatistics and Informatics, CNMC
Session 6: Other Analysis Issues In this session, we consider various analysis issues that occur in practice: Incomplete Data: –Subjects drop-out, do not.
Types of Studies. Aim of epidemiological studies To determine distribution of disease To examine determinants of a disease To judge whether a given exposure.
Headlines Introduction General concepts
Issues in Treatment Study Design John Whyte, MD, PhD Neuro-Cognitive Rehabilitation Research Network Moss Rehabilitation Research Institute.
Approaches to quantitative data analysis Lara Traeger, PhD Methods in Supportive Oncology Research.
CATEGORY: VACCINES & THERAPEUTICS HIV-1 Vaccines Shokouh Makvandi-Nejad, University of Oxford, UK HIV-1 Vaccines © The copyright for this work resides.
1 Study Design Imre Janszky Faculty of Medicine, ISM NTNU.
MHRP  The views expressed are those of the authors and should not be construed to represent the positions of the U.S. Army or the Department of Defense.
Purpose of Epi Studies Discover factors associated with diseases, physical conditions and behaviors Identify the causal factors Show the efficacy of intervening.
1 Borgan and Henderson: Event History Methodology Lancaster, September 2006 Session 8.1: Cohort sampling for the Cox model.
HVTN 702: A pivotal phase 2b/3 multi-site, randomized, double-blind, placebo-controlled clinical trial to evaluate the safety and efficacy of ALVAC-HIV.
04/19/ Projected effectiveness of mass HIV vaccination with multi-dose regimens to be tested in South Africa Peter Gilbert Dobromir Dimitrov Christian.
Comprehensive Evaluation Concept & Design Analysis Process Evaluation Outcome Assessment.
Methods of Presenting and Interpreting Information Class 9.
RTI International RTI International is a trade name of Research Triangle Institute. Selecting Endpoints for Clinical Trials.
HIV-1 Vaccines Shokouh Makvandi-Nejad, University of Oxford, UK
Vaccine Efficacy, Effectiveness and Impact
Aiying Chen, Scott Patterson, Fabrice Bailleux and Ehab Bassily
Effect Modifiers.
Understanding Vaccine Partial Efficacy
Joint work with Holly Janes, Peter Gilbert
Presentation transcript:

Washington D.C., USA, July 2012www.aids2012.org Statistical Design and Analysis for Immune Correlates Assessment: Basic Concepts and RV144 Illustration Yunda Huang 1, Holly Janes 1, 2, Peter Gilbert 1, 2 1 Vaccine and Infectious Disease Division Fred Hutchinson Cancer Research Center 2 Department of Biostatistics, University of Washington Seattle, Washington, USA IAS 2012 Correlates Workshop, Washington DC, USA

Washington D.C., USA, July 2012www.aids2012.org Reasons to Be a Statistician … No one knows what we do so we are always right.

Washington D.C., USA, July 2012www.aids2012.org Outline Definitions Correlates Study and Sampling Design Example: RV144 Immune Correlates Study Strategies to Evaluate Immune Correlates Summary

Washington D.C., USA, July 2012www.aids2012.org In the context of preventive HIV vaccine clinical trials Rate of HIV infection: frequency of new HIV infections during a specified time frame – can be measured in vaccine (R v ) and placebo (R p ) groups separately Vaccine efficacy: proportion of infections prevented by vaccine relative to placebo (1- R v / R p ) – need to know the rate of infection in both vaccine and placebo groups Correlates of risk (CoR ): Immune markers statistically correlated with the rate of HIV infection in the vaccine group (Qin & Gilbert et al., JID, 2007) Correlates of protection (CoP): Immune markers statistically correlated with vaccine efficacy in the vaccine and placebo groups (Qin & Gilbert et al., JID, 2007; Plotkin & Gilbert, CID, 2012)

Washington D.C., USA, July 2012www.aids2012.org Correlates of Risk (CoR) and Correlates of Protection (CoP) Vital for vaccine development –Choice of antigens included in vaccines –Bridge from previously collected protection data –Surrogate for efficacy evaluation –Population and individual level immunity measure Easily being confused and used inter-changeably Three facts: 1.A CoR may not be a CoP, but could be 2.A CoP must be a CoR 3.Not all CoRs or CoPs are created equal

Washington D.C., USA, July 2012www.aids2012.org Fact #1a: A CoR may not be a CoP It is a CoR – because the levels of the immune response are correlated with the rate of infection in the vaccine group Is it a CoP? Immune Response Rate of HIV Infection

Washington D.C., USA, July 2012www.aids2012.org Fact #1a: A CoR may not be a CoP It is a CoR It is not a CoP – if, in the same way, the rate of infection is correlated with the immune response (had it been measured) in the placebo as well Rate of HIV Infection Immune Response

Washington D.C., USA, July 2012www.aids2012.org Fact #1a: A CoR may not be a CoP It is a CoR It is not a CoP The immune responses from this biomarker are not predictive of VE (≠ CoP), although overall VE=40% Vaccine Efficacy (%) Rate of HIV Infection Immune Response

Washington D.C., USA, July 2012www.aids2012.org Fact #1a: A CoR may not be a CoP It is a CoR It is not a CoP Why: –Those individuals who could mount a strong immune response are better able to ward off infection “on their own” with no impact of the vaccine-induced immune responses of this marker –“On their own”: It may mark susceptibility to infection independent of Vaccination, e.g., risk behavior or host genetics Rate of HIV Infection Vaccine Efficacy (%) Immune Response

Washington D.C., USA, July 2012www.aids2012.org Fact #1b: A CoR could be a (perfect) CoP It is a CoR Is it a CoP? Rate of HIV Infection Immune Response

Washington D.C., USA, July 2012www.aids2012.org Fact #1b: A CoR could be a (perfect) CoP It is a CoR It is a CoP: those individuals who could mount a strong immune response are better able to remain uninfected, differently in vaccine and placebo recipients if assigned vaccine And, these immune responses (=CoR) are also predictive of vaccine efficacy (=CoP) Rate of HIV Infection Immune Response Vaccine Efficacy (%)

Washington D.C., USA, July 2012www.aids2012.org Fact #2: A CoP must be a CoR It’s equivalent to show: If not a CoR, then not a CoP Immune responses from vaccinees are not predictive of rate of infection -- not a CoR Rate of HIV Infection Immune Response

Washington D.C., USA, July 2012www.aids2012.org Fact #2: A CoP must be a CoR It’s equivalent to show: If not a CoR, then not a CoP Immune responses from placebos will not be predictive of rate of infection Rate of HIV Infection Immune Response

Washington D.C., USA, July 2012www.aids2012.org Fact #2: A CoP must be a CoR It’s equivalent to show: If not a CoR, then not a CoP Immune responses will not be predictive of vaccine efficacy Rate of HIV Infection Immune Response Vaccine Efficacy (%)

Washington D.C., USA, July 2012www.aids2012.org Fact #3a: Not all CoRs are created equal Immune Response Rate of HIV Infection

Washington D.C., USA, July 2012www.aids2012.org Fact #3a: Not all CoRs are created equal Immune Response Rate of HIV Infection

Washington D.C., USA, July 2012www.aids2012.org Fact #3b: Not all CoPs are created equal Immune Response Vaccine Efficacy (%)

Washington D.C., USA, July 2012www.aids2012.org Fact #3b: Not all CoPs are created equal Immune Response Vaccine Efficacy (%)

Washington D.C., USA, July 2012www.aids2012.org Fact #3b: Not all CoPs are created equal Immune Response Vaccine Efficacy (%)

Washington D.C., USA, July 2012www.aids2012.org Outline Definitions 1.A CoR may not be a CoP, but could be 2.A CoP must be a CoR 3.Not all CoRs or CoPs are created equal Correlates Study and Sampling design Example: RV144 Immune Correlates study Strategies to Evaluate Immune Correlates Summary

Washington D.C., USA, July 2012www.aids2012.org Time-dependent and Time-independent CoR Time-independent immune correlates analysis: discover correlates at a specific time point –e.g. immune responses 2 weeks after the last vaccination –Peak immune response time point close to baseline –Informative and practical Time-dependent immune correlates analysis: discover correlates whose levels may change over time –e.g., most recent immune responses before diagnosis of infection –Immune response near the time of exposure with the acute risk of infection –Informative about the mechanism of protection

Washington D.C., USA, July 2012www.aids2012.org CoR Study Design HIV vaccine-induced immune responses are only assessed in vaccinees Statistical power of an immune correlates study is driven by the number of HIV infections among vaccinees For a given total, the number of vaccinee infections depend on the vaccine efficacy: the higher the VE, the smaller number of infections from the vaccine group The smaller # infections is, the stronger the correlation between the immune response and the rate of infections needs to be, in order to have the same statistical power for CoR detection

Washington D.C., USA, July 2012www.aids2012.org CoR Study Sampling design With unlimited resources, we could measure the post- vaccination immune responses from every vaccinees Several cohort study designs have been developed to save resources with minimal loss of power after adjusting for the sampling design Case-cohort (traditional): controls are sampled without regard to infection time as part of a subcohort Case-control: controls are sampled after ascertainments of cases

Washington D.C., USA, July 2012www.aids2012.org CoR Study Sampling Design: Case-cohort Traditionally, controls are sampled without regard to failure times as part of a subcohort –Sampling can be done a priori without regard to case status or time –All cases are included whether they occur in the subcohort or not; controls are included only if in the subcohort –Estimate population level immune responses –Could select controls for multiple outcomes Lately, some case-cohort designs are also outcome-dependent

Washington D.C., USA, July 2012www.aids2012.org CoR Study Sampling Design: Case-control Controls are sampled after ascertainments of cases Individual matching, frequency matching or stratification to sample appropriate controls for cases Matching addresses issues of confounding in the DESIGN stage of a study as opposed to the ANALYSIS phase, providing a more efficient analysis (reduction in standard errors of estimates) Matching on non-confounders may lose efficiency compared to the non-matched case-control approach

Washington D.C., USA, July 2012www.aids2012.org Analysis Method Standard Cox or logistic regression models if data on the full cohort were available Modified Cox or logistic regression models if sub- sampling is done to account for the sampling design –Breslow and Holubkov (1997, Biometrika) –Borgan et al. estimator II (2000, Lifetime Data Analysis)

Washington D.C., USA, July 2012www.aids2012.org Outline Definitions Correlates Study and sampling design –Number of infections drives the power of the study –Sampling designs with corresponding analysis methods Example: RV144 immune correlates study Strategies to evaluate different immune correlates Summary

Washington D.C., USA, July 2012www.aids2012.org RV144 Thai Trial Primary Results 28

Washington D.C., USA, July 2012www.aids2012.org 29 RV144 Thai Trial

Washington D.C., USA, July 2012www.aids2012.org Impetus for the Correlates Study: Evidence for Partial Vaccine Efficacy Objective: To carry out an immune correlates analysis to begin to identify how the vaccine might work Years Probability of HIV Infection (%) Placebo Vaccine C. Modified Intention-to-Treat Analysis* *N=16,395 assessed; 51 Vaccine, 74 Placebo HIV-1 infected Estimated VE = 31% [95% CI 1−51%], p=

Washington D.C., USA, July 2012www.aids2012.org RV144 Correlates of Risk Results 31

Washington D.C., USA, July 2012www.aids2012.org What the RV144 Correlates Study Assessed The analysis sought to discover Correlates of Risk (CoR): Immune response variables that predict whether vaccinees become HIV-1 infected Thus, the study is designed to generate hypotheses that certain immune responses are Correlates of Protection (CoP) that would need validation in future research 32

Washington D.C., USA, July 2012www.aids2012.org RV144: Two Tiers of Studies Pilot immunogenicity studies –Multiple immunology labs to perform assays on sample-sets from HIV uninfected RV144 participants –Conducted standardized comparative analyses of all candidate assays, to down-select the best performing assays and to optimize the immune variables to study as correlates Case-control study –Assessed the selected immune variables as correlates of infection risk 33

Washington D.C., USA, July 2012www.aids2012.org RV144 Example Pilot Data: gp70-V1V2 Binding Antibodies (ELISA) 34

Washington D.C., USA, July 2012www.aids2012.org RV144: Criteria for Advancing Assays to the Case-Control Study 35 Criterion 1. Represents a niche in immunological space (not highly correlated with other assays)  2. Low false positive rate (judged in placebo recipients and Week 0 responses of vaccinees)  3. Vaccine-induced responses with broad variability  4. Relatively low noise (e.g., high reproducibility on replicate samples)  5. Relatively low specimen volume requirement  6. Previously supported as a correlate of infection in the North American VaxGen trial of AIDSVAX 

Washington D.C., USA, July 2012www.aids2012.org RV144 Case-Control Analysis: Two Tiers Primary Analysis: 6 priority immune response variables Secondary Analysis: All other immune response variables that passed pilot study criteria This division maximizes statistical power for the priority immune variables while allowing a broader exploratory analysis 36

Washington D.C., USA, July 2012www.aids2012.org Down-Selected Primary Immune Variables Primary VariablePrincipal Investigator Plasma IgA Binding (14 envelope panel)Georgia Tomaras IgG avidity score to A244 gp120Munir Alam Antibody-dependent cellular cytotoxicity-AE- 92TH023. HIV infected CD4 T cells David Evans Michael Alpert Neutralization of Tier 1 viruses (6 envelope panel) David Montefiori Rungpeung Sutthent Chitraporn Karnasutra IgG binding to scaffolded gp70-V1V2Susan Zolla-Pazner CD4 T cell intracytoplasmic cytokines (IFN , IL-2, TNF , CD154) stimulated by AE-92TH023 peptides Julie McElrath Nicole Frahm 37

Washington D.C., USA, July 2012www.aids2012.org RV144 Case-Control Study Time-independent CoR: What are the immunologic measurements at a fixed time-point (wk26) in vaccinees that predict HIV-1 infection over a 3 year follow-up? Sampling design: Balanced stratified random sampling for vaccinees− 5:1 (control:case) ratio within each of the following covariate strata Gender × Number of vaccinations × Per-protocol status –41 infected vaccinees (all available) –205 uninfected vaccinees (5:1 stratified random sample) –40 placebo recipients (simple random sample) Outcome-dependent 2-phase sampling case-cohort study 38

Washington D.C., USA, July 2012www.aids2012.org Why 5:1 Sampling? Only a Small Power Loss Moving from 5:1 to 10:1 39

Washington D.C., USA, July 2012www.aids2012.org RV144 Immune Correlates Study Main Result* Variable Relative risk per sd P-valueQ-value** IgA Binding to Envelope Panel IgG Avidity A244 gp ADCC AE.HIV-1 Infected CD4 Cells Tier 1 Neutralizing Antibodies IgG Binding to gp70-V1V CD4+ T Cell Intracellular Cytokines *Multivariate logistic regression (quantitative variables) adjusted for gender, baseline behavioral risk (low, medium, high) **1-Qvalue ≈ estimated prob. that the immune variable correlates with infection rate All 6 variables together in multivariate analysis: p=0.08 The 2 correlates in multivariate analysis: p=

Washington D.C., USA, July 2012www.aids2012.org V1V2-gp70 Scaffold ELISA Medium High Low 41

Washington D.C., USA, July 2012www.aids2012.org Cumulative Infection Rates with V1V2-gp70 Scaffold Assay Estimated Relative Risk High vs. Low = 0.29 HighV1V2 Low/Medium V1V2 42

Washington D.C., USA, July 2012www.aids2012.org Plasma IgA Binding To Envelope Panel Medium High Low 43

Washington D.C., USA, July 2012www.aids2012.org Cumulative Infection Rates with IgA Env Binding Assay Estimated Relative Risk High vs. Low = 1.89 High Env IgA Low/Medium Env IgA 44

Washington D.C., USA, July 2012www.aids2012.org Sieve Analysis is an Integral Part of Immune Correlates Assessment The correlates analysis showed V1V2 Abs predicted infection in the vaccine group only Sieve analysis examines evidence for a difference in the sequences of viruses infecting vaccinees versus placebo recipients –Observed differences can be attributed to the vaccine in a randomized trial –Detection of a ‘sieve effect’ may suggest that the vaccine blocks infection with some types of exposing HIVs If certain epitope-specific Ab responses are protective, then would expect to see a relative absence of these specific epitopes in sequences of infected vaccinees compared to infected placebo recipients Found additional evidence for vaccine pressure on the V2 mid-loop region 45

Washington D.C., USA, July 2012www.aids2012.org The gp70-V1V2 antibody CoR would be most useful for vaccine development if it strongly predicted VE (i.e., was a good CoP) What it Could Mean (Most Useful for Vaccine Development) 46

Washington D.C., USA, July 2012www.aids2012.org The gp70-V1V2 antibody CoR does not predicted VE (≠CoP) But, It Could Also Mean 47

Washington D.C., USA, July 2012www.aids2012.org Outline Definitions Correlates Study and Sampling design Example: RV144 Immune Correlates Study –Case-control study –Evidence for two correlates of infection risk in vaccinees –IgG antibodies that bind to scaffolded-V1V2 recombinant protein correlated inversely with infection rate –Plasma IgA antibodies correlated directly with infection rate Strategies to evaluate different immune correlates Summary

Washington D.C., USA, July 2012www.aids2012.org Collect the requisite data for correcting the CoR analysis for potential exposure confounding (e.g., risk behavior, host genetics) Collect the requisite data for directly assessing the ability of a CoR to predict VE (more on next few slides) Conduct sieve analysis of HIV sequences to assess whether the vaccine applied pressure on the HIV Env target(s) specific to the immune correlate Collaborate with other groups (e.g, CHAVI, CAVD, VRC) conducting experiments (e.g., in non-human primates) testing hypotheses about the CoRs 49 Strategies to Assess CoRs as VE-Predictors (CoPs) and as Mechanisms of Protection

Washington D.C., USA, July 2012www.aids2012.org Once a positive CoR is discovered in vaccinees Collect the requisite data for directly assessing the ability of a CoR to predict VE To assess the relationship between VE and an immune marker (i.e., CoP), we need to know the level of the immune marker for both vaccine and placebo recipients – fill in all the blanks GroupOutcome Immune response levels 1234 Vaccine Uninfected Infected Placebo Uninfected Infected

Washington D.C., USA, July 2012www.aids2012.org Once a positive CoR is discovered in vaccinees Collect the requisite data for directly assessing the ability of a CoR to predict VE To assess the relationship between VE and an immune marker (i.e., CoP), we need to know the level of the immune marker for both vaccine and placebo recipients -- fill in all the blanks GroupOutcome Immune response levels Total 1234 Vaccine Uninfected Infected Total Placebo Uninfected ? ? ? ? Infected ? ? ? ? Total~~~~ CoR study Clinical Trial

Washington D.C., USA, July 2012www.aids2012.org Predicting the potential HIV specific immune response (X) to HIV vaccination for placebo recipients (Follmann, Biometrics, 2006) BIP approach: baseline immunogenicity predictor (W) –W is correlated with X –At baseline, measure W in both vaccine and placebo recipients, e.g., immune responses to a non-HIV vaccine –Randomization ensures W same in both vaccine and placebo groups –Build statistical models between W and X based on vaccinees’ data –Use placebo subjects’ W to impute X and pretend X is what we would have seen, had the placebo subjects received the HIV vaccine CPV approach: close-out placebo vaccination –At the end of the trial, inoculate placebo uninfecteds with HIV vaccine –Measure immune response on the same schedule as was measured for vaccinees –Pretend that is what we would have seen, had we inoculated at baseline

Washington D.C., USA, July 2012www.aids2012.org Remarks Correlates of Risk (CoR) and Correlates of Protection (CoP) are different but both important Impact of a positive CoR –Vaccine development (other speakers) –HVTN go/no go guidelines based on RV144 immune correlates finding CoR assessment in vaccine efficacy trials –Future HVTN efficacy trials will use two-stage sequential designs* with important secondary objectives to examine immune correlates BIP and/or CPV approach for the assessment of CoP * References: Gilbert P, Grove D, Gabriel E, Huang Y, Gray G, Hammer S, Buchbinder S, Kublin J, Corey L, Self S (2011, Statistical Communications in Infectious Diseases) Corey L, Nabel GJ, Dieffenbach C, Gilbert PB, Haynes BF, Johnston M, Kublin J, Lane HC, Pantaleo G, Picker LJ, Fauci AS (2011, Science Translational Medicine)

Washington D.C., USA, July 2012www.aids2012.org Acknowledgement Research supported by –NIAID, NIH –USMHRP –Bill and Melinda Gates Foundation –Henry Jackson Foundation Statistical Center for HIV/AIDS research and Prevention (SCHARP) –Youyi Fong –Allan DeCamp –Ying Huang –Paul Edlefsen –Steve Self RV144 Immune Correlates Study –Leadership: Bart Haynes, Peter Gilbert, Jerome Kim, Nelson Michael, Julie McElrath –Host of laboratory scientists from several institutions Duke University led the antibody work (Bart Haynes, Georgia Tomaras, David Montefiori) FHCRC led the T cell work (Julie McElrath)

Washington D.C., USA, July 2012www.aids2012.org Appendix

Washington D.C., USA, July 2012www.aids2012.org Statistical Approach to Sieve Analysis Local sieve analysis (high-dimensional) –Assess Env amino acid (AA) sites as ‘signature sites’ Signature = site with different distribution of residues vaccine vs. placebo relative to an insert-residue* –Assess sets of AA sites as ‘signature sets’ E.g., 9-mers potentially constituting T cell epitopes E.g., clusters of sites potentially constituting antibody epitopes Global sieve analysis (low-dimensional) –Assess if and how vaccine efficacy depends on the distance of the exposing virus to an insert-sequence –Global = summarize a breakthrough HIV by one or a few numbers quantifying distance *3 vaccine-inserts: ALVAC-AE.92TH023, rgp120-AE.CM244, rgp120-B.MN