Michael Hudgens UNC Biostatistics Repeated Low Dose Challenge Studies Regoes et al 2005 PLoS Medicine Why RLD Trials in HIV Vaccine Research Single highdose challenge high probability of infection for unvaccinated ID: 1045141
Download Presentation The PPT/PDF document "Statistical Considerations for Repeated ..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Statistical Considerations for Repeated Low Dose Challenge StudiesMichael HudgensUNC Biostatistics
2. Repeated Low Dose Challenge StudiesRegoes et al. (2005, PLoS Medicine)
3. Why RLD Trials in HIV Vaccine Research?Single high-dose challenge: high probability of infection for unvaccinatedHigh infection rates do not mirror the low probability of heterosexual HIV transmission per sexual act or low per month probability of late postnatal HIV transmission via breastfeeding Vaccines may not be equally efficacious against high-dose and low-dose challenges, such that vaccines efficacious against low-dose challenges (and hence of possible utility) may be discarded because of not demonstrating efficacy in high-dose challenge studiesUnlike human vaccine studies, # of exposures until infection observed; facilitates statistical modeling despite small n
4. OutlineMG Hudgens, PB Gilbert. Assessing vaccine effects in repeated low-dose challenge experiments. Biometrics, 65:1223-1232, 2009MG Hudgens, PB Gilbert, J Mascola, CD Wu, D Barouch, SG Self. Power to detect HIV vaccine effects in repeated low-dose challenge experiments. Journal of Infectious Diseases, 200(4):609-613, 2009.T Nolen, MG Hudgens, PK Sen, GG Koch. Analysis of repeated low-dose challenge studies. Statistics in Medicine, 34:1981–1992, 2015DM Long, MG Hudgens, C Wu. Surrogates of protection in repeated low dose challenge studies. Statistics in Medicine, 34:1747–1760, 2015.
5.
6.
7.
8. Ellenberger et al DNA-MVA RLD Challenge Study (Virology 2006)
9.
10.
11.
12.
13. Kaplan-Meier and Leaky-Model Fitted CurvesHudgens and Gilbert, Biometrics, 2009
14.
15. Simulation studies verified satisfactory operating characteristics of: Estimators and profile likelihood CIs for VES (bias corrected estimators recommended for smaller n studies) Likelihood ratio and Wald tests of H0 : VES = constant AIC should not be sole basis for model selection Extensions to accommodate heterogeneity in risk (via one additional parameter) Above based on large sample approximationsHudgens and Gilbert (Biometrics 2009)
16. Hudgens and Gilbert, Biometrics, 2009
17. Letvin et al. Immune and genetic correlates of vaccine protection against mucosal infection by SIV in monkeys. (AIDS Vaccine 2009, Paris; Science Translational Medicine 2011)Rhesus monkeys immunized intramuscularly using a plasmid DNA prime/recombinant Ad5 boost regimenExperimental animals received vectors expressing genes encoding SIVmac239 env and gag/pol; control animals received empty vectors Study 1: 20:20 Vaccine:Control Study 2: 25:25 Vaccine:ControlMonkeys challenged 4 months after final immunization with 12 weekly intra-rectal exposures of SIVsmE660 in one study, and SIVmac251 in a second studyChallenge dose titered to achieve ~50% infection per-inoculation for controlsTime to acquisition monitored by weekly assessment of plasma SIV RNA levels
18. Letvin et al E660 Challenge Study (AIDS Vaccine 2009, Paris)
19. ModelVaccineEffectHetero Trans ProbEstimatedp0All-or-noneVEsLeakyVEsOverallVEsLogLikelihoodAIC1234MixtureAll-or-noneLeakyNoneNoNoNoNo0.200.200.200.100.530.53--0.00-0.74-0.530.530.74--96.02-96.02-98.19-106.21198.04196.04200.38214.415678MixtureAll-or-noneLeakyNoneYesYesYesYes0.270.260.280.190.490.50--0.19-0.78-0.580.500.78--94.46-94.54-96.23-100.81196.92195.08198.46205.62Letvin et al E660 Challenge Study (AIDS Vaccine 2009, Paris)
20. Model 6: All-or-none, heterogenous transmission probability
21.
22. Design of RLD Challenge TrialsHudgens, M.G., Gilbert, P.B., Wu, C., Barouch, D., Mascola, J., Self, S.G. (2009). Power to detect the effects of HIV vaccination in repeated low-dose challenge experiments. Journal of Infectious Diseases 200: 609--613ConclusionsNeed 25 animals/group for 80% power to detect VE = 50%Greatest power if chance of infection per-challenge for unvaccinated is near p0=0.52:1 vaccine:control sample size maintains high powerRepeated challenges more powerful than single-challengePresence of unsusceptible animals lowers powerIn this case, the standard log-rank test performs poorly; an alternative is a model-based likelihood approach
23. Power by Total Sample Size N (1:1)Need 25 animals/group for 80% power to detect VE = 50%[RR<1 indicates protection]
24. Power by Per-Challenge Probability of Infection p0Greatest power if chanceof per-challenge infection is near 50%
25. Power by Vaccine:Control Ratio (nv:nc)2:1 vaccine:control sample size maintains high power
26. Power by Maximum Number of Challenges CmaxRepeated challenges more powerful than single-challenge
27. Power w/ some NHP not Susceptible
28. Web-Based Sample Size Calculator for RLD Trialshttp://www.scharp.org/tools/RLDwebCalc/RLD.php
29.
30.
31.
32. Power w/ some NHP not Susceptible
33.
34. Nolen et al. Stat in Med 2015Prove Regoes et al. modified version of Fisher’s exact test is not valid, including an example where actual type I error is an order of magnitude higher than the nominal level (0.5 vs 0.05) Recommend exact logrank testProvide SAS and R code
35.
36. Nolen et al, Stat in Med, 2015
37. Immunological Surrogates of ProtectionHudgens, M.G., Gilbert, P.B. Assessing vaccine effects in repeated low-dose challenge experiments. Biometrics, 65:1223-1232, 2009. DM Long, MG Hudgens, C Wu. Surrogates of protection in repeated low dose challenge studies. Statistics in Medicine, 34:1747–1760, 2015.Consider methods for evaluating possible immune surrogates of protection (SoP)
38. ReferencesMG Hudgens, PB Gilbert. Assessing vaccine effects in repeated low-dose challenge experiments. Biometrics, 65:1223-1232, 2009MG Hudgens, PB Gilbert, J Mascola, CD Wu, D Barouch, SG Self. Power to detect HIV vaccine effects in repeated low-dose challenge experiments. Journal of Infectious Diseases, 200(4):609-613, 2009.T Nolen, MG Hudgens, PK Sen, GG Koch. Analysis of repeated low-dose challenge studies. Statistics in Medicine, 34:1981–1992, 2015DM Long, MG Hudgens, C Wu. Surrogates of protection in repeated low dose challenge studies. Statistics in Medicine, 34:1747–1760, 2015.
39. Immunological Surrogates of ProtectionUtilityShortening trials and reducing costsGuiding iterative development of vaccines between basic and clinical researchRegulatory decisionsVaccination policyBridging efficacy of vaccine observed in a trial to a new setting
40. Immunological Surrogates of ProtectionConceptual framework pioneered by Gilbert and colleagues over the last decade, motivated primarily by phase III HIV vaccine trialsUtilize causal inference framework using potential outcomes/counterfactualsKey distinction: correlate of risk vs surrogate of protectionCoR: post-vaccination immunology measurement X associated with per challenge risk of infection in vaccinatedSoP: post-vaccination immunology measurement X such that vaccine effect on X is predictive of (associated with) vaccine effect on clinical endpoint of interest (infection/disease/death)
41. CoR not necessarily SoPExample:Suppose heterogeneity in susceptibility infection: weak or strong immune systemIndividuals w/ strong immune system more likely to have larger values of X when vaccinatedIndividuals w/ strong immune system more likely to escape infection when vaccinatedCoRX=0 if not vaccinatedVaccine does not protect from infectionNot SoP
42. Immunological Surrogates of ProtectionLet X(1) denote immune response if an animal is vaccinatedLet X(0) denote immune response if an animal is not vaccinated; assume X(0)=0So X(1) is effect of vaccine on XStudy how VE varies over groups of animals defined by fixed values of immune response X(1) when/if vaccinatedInterpretation: Percent reduction in risk of infection for vaccinated animals with response x1 compared to if they had not been vaccinated
43. Two Possible VE(x1) Curves
44. Estimation of VE(x1) Problem: For all control animals X(1) is missingTo estimate VE(x1), predict/impute missing X(1)’sTwo possible remedies:Baseline Immunogenicity Predictors (BIPs): Exploit correlations of X(1) with subject-specific baseline characteristics measured in both vaccine and control subjectsCloseout/crossover vaccination of uninfected control subjects