Search

Background The ALVAC/gp120 + MF59 vaccines in the HIV Vaccine Trials Network (HVTN) 702 efficacy trial did not prevent human immunodeficiency virus-1 (HIV-1) acquisition. Vaccine-matched immunological endpoints that were correlates of HIV-1 acquisition risk in RV144 were measured in HVTN 702 and evaluated as correlates of HIV-1 acquisition. Methods Among 1893 HVTN 702 female vaccinees, 60 HIV-1–seropositive cases and 60 matched seronegative noncases were sampled. HIV-specific CD4+ T-cell and binding antibody responses were measured 2 weeks after fourth and fifth immunizations. Cox proportional hazards models assessed prespecified responses as predictors of HIV-1 acquisition. Results The HVTN 702 Env-specific CD4+ T-cell response rate was significantly higher than in RV144 (63% vs 40%, P = .03) with significantly lower IgG binding antibody response rate and magnitude to 1086.C V1V2 (67% vs 100%, P Conclusions HVTN 702 and RV144 had distinct immunogenicity profiles. However, both identified significant correlations (univariate or interaction) for IgG V1V2 and polyfunctional CD4+ T cells with HIV-1 acquisition. Clinical Trials Registration . NCT02968849.

Background Respiratory syncytial virus (RSV) can cause serious lung infections in young children and there is currently no available vaccine. Methods We used complementary statistical frameworks to analyze 4 RSV serology measurements in mothers and their infants in South Africa who participated in a phase 3 maternal immunization trial of an RSV F protein nanoparticle vaccine as correlates of risk and of protection against different RSV disease endpoints. Results We found evidence to support each antibody measurement—encompassing RSV-neutralizing antibodies and F surface glycoprotein-binding antibodies—as an inverse correlate of risk of RSV-associated acute lower respiratory tract infection with severe hypoxia in at least 1 framework, with vaccine-induced fold-rise from the maternal enrollment to day 14 samples of anti-F immunoglobulin G (IgG) binding antibodies having the most consistent evidence. This evidence includes a significant association of fold-rise anti-F IgG with vaccine efficacy (VE); achieving a baseline covariate-adjusted VE of 75% requires a vaccine-induced maternal anti-F IgG fold-rise of around 16. Neither multivariable logistic regression nor superlearning analyses showed benefit to including multiple time points or assays in the same model, suggesting a parsimonious correlate. Post hoc exploratory analyses supported adherence of vaccine-induced maternal anti-F IgG fold-rise to the Prentice criteria for a valid surrogate endpoint. Conclusions Our results suggest that the vaccine induced protective anti-F antibody responses. If this finding is confirmed, VE could potentially be augmented by increasing these responses.

Although interim results from several large, placebo-controlled, phase 3 trials demonstrated high vaccine efficacy (VE) against symptomatic coronavirus disease 2019 (COVID-19), it is unknown how effective the vaccines are in preventing people from becoming asymptomatically infected and potentially spreading the virus unwittingly. It is more difficult to evaluate VE against severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection than against symptomatic COVID-19 because infection is not observed directly but rather is known to occur between 2 antibody or reverse-transcription polymerase chain reaction (RT-PCR) tests. Additional challenges arise as community transmission changes over time and as participants are vaccinated on different dates because of staggered enrollment of participants or crossover of placebo recipients to the vaccine arm before the end of the study. Here, we provide valid and efficient statistical methods for estimating potentially waning VE against SARS-CoV-2 infection with blood or nasal samples under time-varying community transmission, staggered enrollment, and blinded or unblinded crossover. We demonstrate the usefulness of the proposed methods through numerical studies that mimic the BNT162b2 phase 3 trial and the Prevent COVID U study. In addition, we assess how crossover and the frequency of diagnostic tests affect the precision of VE estimates.

Motivation A single monoclonal broadly neutralizing antibody (bnAb) regimen was recently evaluated in two randomized trials for prevention efficacy against HIV-1 infection. Subsequent trials will evaluate combination bnAb regimens (e.g. cocktails, multi-specific antibodies), which demonstrate higher potency and breadth in vitro compared to single bnAbs. Given the large number of potential regimens, methods for down-selecting these regimens into efficacy trials are of great interest. Results We developed Super LeArner Prediction of NAb Panels (SLAPNAP), a software tool for training and evaluating machine learning models that predict in vitro neutralization sensitivity of HIV Envelope (Env) pseudoviruses to a given single or combination bnAb regimen, based on Env amino acid sequence features. SLAPNAP also provides measures of variable importance of sequence features. By predicting bnAb coverage of circulating sequences, SLAPNAP can improve ranking of bnAb regimens by their potential prevention efficacy. In addition, SLAPNAP can improve sieve analysis by defining sequence features that impact bnAb prevention efficacy. Availabilityand implementation SLAPNAP is a freely available docker image that can be downloaded from DockerHub (https://hub.docker.com/r/slapnap/slapnap). Source code and documentation are available at GitHub (https://github.com/benkeser/slapnap and https://benkeser.github.io/slapnap/).

BackgroundThe recombinant tetravalent live-attenuated dengue vaccine based on the YF 17D vaccine virus backbone (CYD-TDV) demonstrated vaccine efficacy (VE) against symptomatic, virologically confirmed dengue of any serotype from month 13 to month 25 (VCD-DENV-AnyM13→M25) in the CYD14 (2–14-y-olds) and CYD15 (9–16-y-olds) phase 3 trials. Fifty percent plaque reduction neutralization test (PRNT50) titers are a potential surrogate for immunobridging VE to adults.MethodsUsing PRNT50 calibration datasets, we applied immunobridging approaches using baseline and/or M13 PRNT50 titers to estimate VE against VCD-DENV-AnyM0→M25 and against hospitalized VCD (HVCD)-DENV-AnyM0→M72 in hypothetical 18–45-y-old and 46–50-y-old CYD14 and CYD15 cohorts.ResultsBaseline and M13 geometric mean PRNT50 titers were greater in 18–45-y-olds and in 46–50-y-olds vs 9–16-y-olds for most comparisons. Estimated VE (95% CIs against VCD-DENV-AnyM0→M25 ranged from 75.3% to 90.9% (52.5% to 100%) for 18–45-y-olds and 74.8% to 92.0% (53.4% to 100%) for 46–50-y-olds. Estimated VE (95% CIs) against HVCD-DENV-AnyM0→M72 ranged from 58.8% to 78.1% (40.9 to 98.9%) for 18–45-y-olds and 57.2% to 78.4% (40.5 to 97.6%) for 46–50-y-olds. Corresponding predictions among baseline-seropositive individuals yielded comparable or higher VE estimates.ConclusionsVE M0→M25 against DENV-Any and VE against HVCD-DENV-AnyM0→M72 are both expected to be higher in 18–45 and 46–50-y-olds vs CYD14 and CYD15 9–16-y-olds.

Summary Left-truncation poses extra challenges for the analysis of complex time-to-event data. We propose a general semiparametric regression model for left-truncated and right-censored competing risks data that is based on a novel weighted conditional likelihood function. Targeting the subdistribution hazard, our parameter estimates are directly interpretable with regard to the cumulative incidence function. We compare different weights from recent literature and develop a heuristic interpretation from a cure model perspective that is based on pseudo risk sets. Our approach accommodates external time-dependent covariate effects on the subdistribution hazard. We establish consistency and asymptotic normality of the estimators and propose a sandwich estimator of the variance. In comprehensive simulation studies we demonstrate solid performance of the proposed method. Comparing the sandwich estimator with the inverse Fisher information matrix, we observe a bias for the inverse Fisher information matrix and diminished coverage probabilities in settings with a higher percentage of left-truncation. To illustrate the practical utility of the proposed method, we study its application to a large HIV vaccine efficacy trial dataset.

The recombinant vesicular stomatitis virus (rVSV) Ebola vaccine was shown to be very efficacious in a novel ring vaccination trial in Guinea. However, no correlates of vaccine protection have been established for Ebola vaccines. Several Ebola vaccine candidates are available, but conducting randomized trials of additional candidates in outbreaks is difficult. Establishing correlates of vaccine protection is essential. Here we explore power and sample-size calculations to evaluate potential correlates of risk during an Ebola vaccination campaign in an outbreak. The method requires that a blood draw be made at a predetermined time after vaccination. The statistical analysis estimates the relative risk of the Ebola endpoint occurring from after the blood draw through to the end of follow-up, contrasting vaccine recipients with different values of the immune response marker. The analysis can be done assuming a trichotomous or continuous marker. Under certain assumptions, at an overall vaccine efficacy of 75%, 50 Ebola endpoints in the vaccinees provided good power. At an overall vaccine efficacy of 90%, 20 Ebola endpoints gave good power. Power was highest when more vaccinees were in the high- and low-responder groups versus the middle group and when vaccine efficacy differed the most between the high- and low-responder groups.

The scale of the HIV-1 epidemic underscores the need for a vaccine. The multitude of circulating HIV-1 strains together with HIV-1’s high evolvability hints that HIV-1 could adapt to a future vaccine. Here, we wanted to investigate the effect of vaccination on the evolution of the virus post-breakthrough infection. We analyzed 2,635 HIV-1 env sequences sampled up to a year post-diagnosis from 110 vaccine and placebo participants who became infected in the RV144 vaccine efficacy trial. We showed that the Env signature sites that were previously identified to distinguish vaccine and placebo participants were maintained over time. In addition, fewer sites were under diversifying selection in the vaccine group than in the placebo group. These results indicate that HIV-1 would possibly adapt to a vaccine upon its roll-out.

A large number of studies are being conducted to evaluate the efficacy and safety of candidate vaccines against coronavirus disease 2019 (COVID-19). Most phase 3 trials have adopted virologically confirmed symptomatic COVID-19 as the primary efficacy end point, although laboratory-confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is also of interest. In addition, it is important to evaluate the effect of vaccination on disease severity. To provide a full picture of vaccine efficacy and make efficient use of available data, we propose using SARS-CoV-2 infection, symptomatic COVID-19, and severe COVID-19 as dual or triple primary end points. We demonstrate the advantages of this strategy through realistic simulation studies. Finally, we show how this approach can provide rigorous interim monitoring of the trials and efficient assessment of the durability of vaccine efficacy.

Summary Doubly robust estimators are widely used to draw inference about the average effect of a treatment. Such estimators are consistent for the effect of interest if either one of two nuisance parameters is consistently estimated. However, if flexible, data-adaptive estimators of these nuisance parameters are used, double robustness does not readily extend to inference. We present a general theoretical study of the behaviour of doubly robust estimators of an average treatment effect when one of the nuisance parameters is inconsistently estimated. We contrast different methods for constructing such estimators and investigate the extent to which they may be modified to also allow doubly robust inference. We find that while targeted minimum loss-based estimation can be used to solve this problem very naturally, common alternative frameworks appear to be inappropriate for this purpose. We provide a theoretical study and a numerical evaluation of the alternatives considered. Our simulations highlight the need for and usefulness of these approaches in practice, while our theoretical developments have broad implications for the construction of estimators that permit doubly robust inference in other problems.

Background It is important to identify vaccine-induced immune responses that predict the preventative efficacy of a human immunodeficiency virus (HIV)-1 vaccine. We assessed T-cell response markers as correlates of risk in the HIV Vaccine Trials Network (HVTN) 505 HIV-1 vaccine efficacy trial. Methods 2504 participants were randomized to DNA/rAd5 vaccine or placebo, administered at weeks 0, 4, 8, and 24. Peripheral blood mononuclear cells were obtained at week 26 from all 25 primary endpoint vaccine cases and 125 matched vaccine controls, and stimulated with vaccine-insert-matched peptides. Primary variables were total HIV-1-specific CD4+ T-cell magnitude and Env-specific CD4+ polyfunctionality. Four secondary variables were also assessed. Immune responses were evaluated as predictors of HIV-1 infection among vaccinees using Cox proportional hazards models. Machine learning analyses identified immune response combinations best predicting HIV-1 infection. Results We observed an unexpectedly strong inverse correlation between Env-specific CD8+ immune response magnitude and HIV-1 infection risk (hazard ratio [HR] = 0.18 per SD increment; P = .04) and between Env-specific CD8+ polyfunctionality and infection risk (HR = 0.34 per SD increment; P < .01). Conclusions Further research is needed to determine if these immune responses are predictors of vaccine efficacy or markers of natural resistance to HIV-1 infection.

Using Super Learner, a machine learning statistical method, we assessed varicella zoster virus-specific glycoprotein-based enzyme-linked immunosorbent assay (gpELISA) antibody titer as an individual-level signature of herpes zoster (HZ) risk in the Zostavax Efficacy and Safety Trial. Gender and pre- and postvaccination gpELISA titers had moderate ability to predict whether a 50-59 year old experienced HZ over 1-2 years of follow-up, with equal classification accuracy (cross-validated area under the receiver operator curve = 0.65) for vaccine and placebo recipients. Previous analyses suggested that fold-rise gpELISA titer is a statistical correlate of protection and supported the hypothesis that it is not a mechanistic correlate of protection. Our results also support this hypothesis.

Biomarker endpoints measuring vaccine-induced immune responses are essential to HIV vaccine development because of their potential to predict the effect of a vaccine in preventing HIV infection. A vaccine’s immune response profile observed in phase I immunogenicity studies is a key factor in determining whether it is advanced for further study in phase II and III efficacy trials. The multiplicity of immune variables and scientific uncertainty in their relative importance, however, pose great challenges to the development of formal algorithms for selecting vaccines to study further. Motivated by the practical need to identify a set of promising vaccines from a pool of candidate regimens for inclusion in an upcoming HIV vaccine efficacy trial, we propose a new statistical framework for the selection of vaccine regimens based on their immune response profile. In particular, we propose superiority and non-redundancy criteria to be achieved in down-selection, and develop novel statistical algorithms that integrate hypothesis testing and ranking for selecting vaccine regimens satisfying these criteria. Performance of the proposed selection algorithms are evaluated through extensive numerical studies. We demonstrate the application of the proposed methods through the comparison of immune responses between several HIV vaccine regimens. The methods are applicable to general down-selection applications in clinical trials.

Summary: Analysis of HIV-1 virions from participants infected in a randomized controlled preventive HIV-1 vaccine efficacy trial can help elucidate mechanisms of partial protection. By comparing the genetic sequence of viruses from vaccine and placebo recipients to the sequence of the vaccine itself, a technique called ‘sieve analysis’, one can identify functional specificities of vaccine-induced immune responses. We have created an interactive web-based visualization and data access tool for exploring the results of sieve analyses performed on four major preventive HIV-1 vaccine efficacy trials: (i) the HIV Vaccine Trial Network (HVTN) 502/Step trial, (ii) the RV144/Thai trial, (iii) the HVTN 503/Phambili trial and (iv) the HVTN 505 trial. The tool acts simultaneously as a platform for rapid reinterpretation of sieve effects and as a portal for organizing and sharing the viral sequence data. Access to these valuable datasets also enables the development of novel methodology for future sieve analyses. Availability and Implementation: Visualization: http://sieve.fredhutch.org/viz. Source code: https://github.com/nkullman/SIEVE. Data API: http://sieve.fredhutch.org/data. Contact: agartlan@fredhutch.org

Summary The paper develops hypothesis testing procedures for the stratified mark-specific proportional hazards model in the presence of missing marks. The motivating application is preventive human immunodeficiency virus (HIV) vaccine efficacy trials, where the mark is the genetic distance of an infecting HIV sequence to an HIV sequence represented inside the vaccine. The test statistics are constructed on the basis of two-stage efficient estimators, which utilize auxiliary predictors of the missing marks. The asymptotic properties and finite sample performances of the testing procedures are investigated, demonstrating double robustness and effectiveness of the predictive auxiliaries to recover efficiency. The methods are applied to the RV144 vaccine trial.

(See the editorial commentary by Wittes on pages 1523–5.) Herpes zoster (HZ), caused by reactivation of latent varicella zoster virus (VZV) [1], entails rash and pain that decreases daily functioning and health-related quality of life, more so with increasing age [1–10]. The phase III Shingles Prevention Study (SPS) showed that the live, attenuated zoster vaccine (ZV; Zostavax, Merck, Whitehouse Station, NJ) had an estimated 51% vaccine efficacy (VE; 95% confidence interval [CI], 44%–58%) to reduce the incidence of HZ in persons aged ≥60 years [11]. Subsequently, another phase III Zostavax Efficacy and Safety Trial (ZEST) was conducted to assess the VE of ZV in persons aged 50–59 years and showed an estimated 70% VE (95% CI, 54%–81%) [12]. Secondary analyses of the SPS suggested that 3 different measures of VZV-specific immunity were correlated with protection from HZ: responder cell frequency–determined cell-mediated immunity (CMI), interferon γ (IFN-γ) enzyme-linked immunosorbent spot–determined CMI, and glycoprotein-based enzyme-linked immunosorbent assay (gpELISA)–determined antibody level [13]. Because the gpELISA is easier to use and more validated as a reproducible quantitative assay and because the combination of titers and fold rise appeared to best correlate with protection in the SPS [14], the ZEST included an objective to assess VZV antibody titers by gpELISA but not by the CMI measures as a correlate of protection (CoP). Both week 6 and fold rise in VZV antibody titers determined by gpELISA were strong inverse correlates of risk (CoRs) of HZ in vaccine recipients [15], where a CoR is defined as an immune biomarker that is associated with the rate of a study end point used to measure VE in a defined population [16]. Here we directly assess whether the identified CoRs are also good CoPs [16]. A CoP may be defined as an immune biomarker measuring a response to vaccination that reliably predicts the level of VE to prevent a clinically meaningful end point. The literature on approaches for assessing CoPs has used different meanings for the term “CoP,” with no consensus [16–20]. Plotkin and Gilbert [21] advocated for the common use of the definition specified above, which is solely about reliable statistical prediction and is silent about whether the biomarker is also a mechanism of protection. These authors also advocated for annotating CoPs as mCoPs (mechanistic CoPs) or nCoPs (nonmechanistic CoPs) for scenarios in which experiments have determined whether the biomarker is a true mechanism of protection. Validated CoPs benefit vaccine development by serving as reliable surrogate end points for true clinical end points of interest, thereby allowing using surrogate end points in phase I and II trials to provide reliable answers about clinical VE without requiring phase III field trials. Validated CoPs confer many other benefits, as extensively discussed elsewhere [16–20, 22]. Given that reliable prediction of VE is very useful in itself, even as a so-called black box predictor with no associated mechanism, nCoPs and mCoPs are highly beneficial for vaccine development. The concept of a CoP is the same as a valid surrogate end point in the clinical trials statistical literature. Two major statistical frameworks have been developed for assessing the validity of a surrogate end point based on an efficacy trial, the Prentice framework [16, 23–25] and the principal stratification or VE curve framework [26, 27]. While both are useful for assessing the validity of the candidate gpELISA CoPs, this article focuses on the VE curve framework. This approach directly estimates the VE curve, defined as the VE against the clinical end point for each of a spectrum of participant subgroups defined by their biomarker readout measuring vaccine-induced immune response. The major challenge posed to using this approach is that the vaccine-induced immune response is only measured in subjects assigned to receive the vaccine (it is unknown in subjects assigned to receive placebo). Consequently, its effective application requires availability of a subject characteristic before immunization that predicts the candidate CoP [28]; such a baseline immunogenicity predictor (BIP) allows inclusion of an estimate of the immune response to the vaccine that placebo recipients would have had had they been vaccinated, thereby enabling estimation of the VE curve. This approach can be used for estimating the VE curve for subgroups defined either by levels of the vaccine-induced immune response or by both the levels of this response and the response under assignment to placebo. Several statistical methods have been developed for estimating the VE curve based on a BIP [27–32]. However, because adequately predictive BIPs have not been available, the methods have not yet been effectively applied. Fortunately, a high-quality BIP is available in the ZEST, and here we report, for the first time, an application of this framework to produce clear results that validate one biomarker as an excellent CoP (fold rise in gpELISA titers) and another biomarker as a poor CoP (week 6 gpELISA titers). By directly assessing how VE varies with participant subgroups defined by gpELISA readouts, this framework provides clearly interpretable results that lend themselves to central applications, such as comparing the utility of different immune response biomarkers as CoPs and estimating seroprotection levels.

The contribution of host T-cell immunity and HLA class I alleles to the control of human immunodeficiency virus (HIV-1) replication in natural infection is widely recognized. We assessed whether vaccine-induced T-cell immunity, or expression of certain HLA alleles, impacted HIV-1 control after infection in the Step MRKAd5/HIV-1 gag/pol/nef study. Vaccine-induced T cells were associated with reduced plasma viremia, with subjects targeting ≥3 gag peptides presenting with half-log lower mean viral loads than subjects without Gag responses. This effect was stronger in participants infected proximal to vaccination and was independent of our observed association of HLA-B*27, –B*57 and –B*58:01 alleles with lower HIV-1 viremia. These findings support the ability of vaccine-induced T-cell responses to influence postinfection outcome and provide a rationale for the generation of T-cell responses by vaccination to reduce viremia if protection from acquisition is not achieved. Clinical trials identifier: NCT00095576.

It is widely believed that a neutralizing antibody (NAb) response of sufficient magnitude, breadth, and duration would be highly beneficial for human immunodeficiency virus type 1 (HIV-1) vaccines [1–3]. Indeed, Nabs protect against experimental challenge with simian human immunodeficiency virus (SHIV) in nonhuman primates [4–6], and they exert strong selective pressure on HIV-1 after infection in humans [7, 8]. Neutralization occurs when antibodies bind to functional envelope glycoprotein (Env) spikes on the virus surface to prevent entry into host cells [9–11]. Each Env spike consists of 3 surface gp120 molecules bound noncovalently to 3 transmembrane gp41 molecules [12, 13]. These glycoproteins exhibit an extraordinary degree of genetic and antigenic variability that poses major challenges for vaccine development [14, 15]. Moreover, the virus uses a number of mechanisms to evade NAbs [7, 12, 16]. As a result, a minor subset of circulating variants exhibit a highly neutralization-sensitive tier 1 phenotype and are often susceptible to vaccine-elicited NAbs, whereas most circulating strains exhibit a less sensitive tier 2 phenotype and have proven difficult to target with vaccines[1, 2, 15]. A recently completed HIV-1 vaccine efficacy trial in Thailand (RV144) showed that priming with a recombinant canarypox vector (vCP1521) and boosting with this vector plus bivalent gp120 protein (AIDSVAX B/E) can provide partial protection against the acquisition of HIV-1 infection in a community-based heterosexual population [17]. The same bivalent gp120 immunogen, when used alone and with an increased number of inoculations (Vax003 trial), showed no protection in a cohort of Thai injection drug users [18]. In addition, no overall protection was seen when a similar regimen of bivalent gp120 (AIDSVAX B/B) was used alone in a cohort of mostly men who have sex with men (MSM) in North America and the Netherlands (Vax004 trial) [19]. The gp120 protein component in all 3 clinical trials was designed to elicit NAbs [20, 21]. In Vax004, strong NAb responses were seen against a subset of tier 1 viruses, and sporadic weak responses were seen against tier 2 viruses [22]. Here we assessed the magnitude and breadth of NAb responses in RV144 and Vax003.

Cell-mediated immunity in human immunodeficiency virus (HIV)–infected patients may control viral replication and slow HIV disease progression [1–8]. In preclinical primate models, most vaccines that elicit cell-mediated immunity do not prevent infection, but they do lower post-infection viral loads and slow retroviral disease progression [9–11]. Therefore, a number of HIV-1 vaccine candidates have been developed that aim to elicit cell-mediated immunity. The goal of these vaccines is to improve immune control of viral replication after infection, to decrease HIV-1 RNA levels, and to thereby slow HIV-1 disease progression and reduce sexual transmission to others. The Step study was a randomized trial of the Merck adenovirus type 5 (Ad5) trivalent HIV-1 vaccine that was designed to elicit cell-mediated immunity. Three thousand adults in North America, the Caribbean, South America, and Australia were randomized to receive placebo or the vaccine. Vaccinations were halted and the study was unblinded in 2007, when the first interim analysis indicated that the vaccine would not achieve efficacy for the two primary study end points, HIV-1 acquisition and plasma HIV-1 RNA level 3 months after the diagnosis of HIV infection [12]. In the current analysis, we evaluate the effect of the vaccine on HIV/AIDS disease progression in subjects who became HIV-1–infected during the Step trial. Measures of disease progression include HIV-1 RNA levels, CD4+ T cell counts, time to antiretroviral therapy (ART), and AIDS-free survival during post-infection follow-up [13–15].

Recently, the RV144 randomized, double-blind, efficacy trial in Thailand reported that a prime-boost human immunodeficiency virus (HIV) vaccine regimen conferred ∼30% protection against HIV acquisition. However, different analyses seemed to give conflicting results, and a heated debate ensued as scientists and the broader public struggled with their interpretation. The lack of accounting for statistical principles helped flame the debate, and we leverage these principles to provide a more scientific interpretation. We first address interpretation of frequentist results, including interpretation of P values, synthesis of results from multiple analyses (ie, intention-to-treat versus per-protocol/fully immunized), and accounting for external efficacy trials. Second, we address how Bayesian statistics, which provide clearly interpretable statements about probabilities that the vaccine efficacy takes certain values, provide more information for weighing the evidence about efficacy than do frequentist statistics alone. Third, we evaluate RV144 for completeness of end point ascertainment and integrity of blinding, necessary tasks for establishing robustly interpretable results.

In the past decade, several principal stratification–based statistical methods have been developed for testing and estimation of a treatment effect on an outcome measured after a postrandomization event. Two examples are the evaluation of the effect of a cancer treatment on quality of life in subjects who remain alive and the evaluation of the effect of an HIV vaccine on viral load in subjects who acquire HIV infection. However, in general the developed methods have not addressed the issue of missing outcome data, and hence their validity relies on a missing completely at random (MCAR) assumption. Because in many applications the MCAR assumption is untenable, while a missing at random (MAR) assumption is defensible, we extend the semiparametric likelihood sensitivity analysis approach of Gilbert and others (2003) and Jemiai and Rotnitzky (2005) to allow the outcome to be MAR. We combine these methods with the robust likelihood–based method of Little and An (2004) for handling MAR data to provide semiparametric estimation of the average causal effect of treatment on the outcome. The new method, which does not require a monotonicity assumption, is evaluated in a simulation study and is applied to data from the first HIV vaccine efficacy trial.

Simulation studies were conducted to estimate the statistical power of repeated low-dose challenge experiments performed in nonhuman primates to detect the effect of a candidate human immunodeficiency virus vaccine. The effect of various design parameters on power was explored. Results of simulation studies indicate that repeated low-dose challenge studies with a total sample of size 50 (25 animals/arm) typically provide adequate power to detect a 50% reduction in the per-exposure probability of infection resulting from vaccination. Power generally increases with the maximum number of allowable challenges per animal, the per-exposure risk of infection in control animals, and the proportion of animals susceptible to infection.

A central goal of vaccine research is to identify a vaccine-induced immune response that predicts protection from infection or disease. The term "correlate of protection" has been used to refer to at least 3 distinct concepts, which has resulted in confusion surrounding this topic. We propose precise definitions of these different concepts of immune correlates, using the nomenclature "correlate of risk," "level 1 surrogate of protection," and "level 2 surrogate of protection." We suggest a general framework for assessing these 3 levels of immune correlates in vaccine efficacy trials. To demonstrate the proposed principles, we analyze data from a 1943 influenza vaccine field trial, supporting Weiss strain A-specific antibody titers as a level 1 surrogate of protection. Other real and simulated examples are also discussed.

The world's first efficacy trial of a preventive HIV vaccine was completed in 2003. Study participants who became HIV infected were followed for 2 years and monitored for HIV viral load and initiation of antiretroviral therapy (ART). In order to determine if vaccination may have altered HIV progression in persons who acquired HIV, a pre-specified objective was to compare the time until a composite endpoint between the vaccine and placebo arms, where the composite endpoint is the first event of ART initiation or viral failure (HIV viral load exceeds a threshold x(vl) copies/ml). Specifically, with vaccine efficacy, VE(tau, x(vl)), defined as one minus the ratio (vaccine/placebo) of the cumulative probability of the composite endpoint (with failure threshold x(vl)) occurring by tau months, the aim was to estimate the four parameters {VE(tau, x(vl)): x(vl) is an element of {1500, 10,000, 20,000, 55,000} copies/ml} with simultaneous 95% confidence bands. A Gaussian multipliers simulation method is devised for constructing confidence bands for VE(tau, x(vl)) with x(vl) spanning multiple discrete values or a continuous range. The new method is evaluated in simulations and is applied to the vaccine trial data set.

Background. An objective of the first efficacy trial of a candidate vaccine containing recombinant human immunodeficiency virus (HIV) type 1 envelope glycoprotein 120 (rgp120) antigens was to assess correlations between antibody responses to rgp120 and the incidence of HIV-1 infection. Methods. Within the randomized trial (for vaccinees, n = 3598; for placebo recipients, n = 1805), binding and neutralizing antibody responses to rgp120 were quantitated. A case-cohort design was used to study correlations between antibody levels and HIV-1 incidence. Results. Peak antibody levels were significantly inversely correlated with HIV-1 incidence. The relative risk (RR) of infection was 0.63 (95% confidence interval, 0.45-0.89) per log10 higher neutralization titer against HIV-1MN, and the RRs of infection for second-, third-, and fourth-quartile responses of antibody blocking of gp120 binding to soluble CD4 versus first-quartile responses (the lowest responses) were 0.35, 0.28, and 0.22, respectively. Conclusions. Despite inducing a complex, robust immune response, the vaccine was unable to reduce the incidence of HIV-1. Two interpretations of the correlative results are that the levels of antibodies (i) caused both an increased (low responders) and decreased (high responders) risk of HIV-1 acquisition or (ii) represented a correlate of susceptibility to HIV-1 but had no causal effect on susceptibility. Although the data cannot definitively discriminate between these 2 explanations, (ii) appears to be more likely.

Summary To help to design vaccines for acquired immune deficiency syndrome that protect broadly against many genetic variants of the human immunodeficiency virus, the mutation rates at 118 positions in HIV amino-acid sequences of subtype C versus those of subtype B were compared. The false discovery rate (FDR) multiple-comparisons procedure can be used to determine statistical significance. When the test statistics have discrete distributions, the FDR procedure can be made more powerful by a simple modification. The paper develops a modified FDR procedure for discrete data and applies it to the human immunodeficiency virus data. The new procedure detects 15 positions with significantly different mutation rates compared with 11 that are detected by the original FDR method. Simulations delineate conditions under which the modified FDR procedure confers large gains in power over the original technique. In general FDR adjustment methods can be improved for discrete data by incorporating the modification proposed.

To the Editor—Richie and Villasante [1] pose thoughtful questions about the implications of our recently published analysis of the Step Study data with extended follow-up time [2]. Although some of the points are well taken, we would like to clarify several questions raised about the study methods and conclusions.

SUMMARY The following problem is treated: given s possibly selection biased samples from an unknown distribution function, and assuming that the sampling rule weight functions for each of the samples are mathematically specified up to a common unknown finite-dimensional parameter, how can we use the data to estimate the unknown parameters? We propose a simple maximum partial likelihood method for deriving the semiparametric maximum likelihood estimator. A discussion of assumptions under which the selection bias model is identifiable and uniquely estimable is presented. We motivate the need for the methodology by discussing the generalised logistic regression model (Gilbert, Self & Ashby, 1998), a semiparametric selection bias model which is useful for assessing from vaccine trial data how the efficacy of an HIV vaccine varies with characteristics of the exposing virus. We show through simulations and an example that the maximum likelihood estimator in the generalised logistic regression model has satisfactory finite-sample properties.

Selecting antiretroviral therapies for human immunodeficiency virus type 1-infected persons is complicated by the availability of a vast number of potentially useful drug combinations and by extensive variation among patients in their resistance to various drugs. AIDS clinical trials have used designs in which a handful of drug regimens in a few patient classes can be compared. Here is proposed implementation of innovative designs with factorial structure that permit assessment of many treatment arms and patient classes in a single trial; when and how they can be appropriately used are discussed. These designs are efficient, permit systematic investigation of correlations between genetic mutations and in vivo drug resistance, and provide insight into important drug interactions in people that conventional designs are unable to provide. Through creative application of these designs, identification of superior drug combinations and the science of understanding in vivo joint drug dynamics and genotypic resistance will progress at an optimum pace.

To the Editor—The commentary by Dunning helpfully raises the question about the sensitivity of estimation of the vaccine efficacy (VE) curve to the choice of the baseline immunogenicity predictor (BIP) that is used. In the Zostavax Efficacy and Safety Trial (ZEST) example in our article the VE curve is estimated as a function of 2 binding antibody markers: fold rise and week 6 titer, in each case using baseline titer as the BIP [1]. We apply several statistical methods for estimating the VE curve, each of which separately estimates 2 terms: the disease risk in vaccine recipients conditional on the marker and the disease risk in placebo recipients conditional on the marker, which for a placebo recipient is the marker that s/he would have had if, contrary to fact, s/he had been assigned to receive the vaccine. We call these risks Vrisk(s) and Prisk(s), respectively, for subgroups with marker level s, where the VE curve is VE(s) = 1 – Vrisk(s)/Prisk(s), which measures VE ranging over subgroups defined by the marker level s. Vrisk(s) can be straightforwardly estimated using regression statistical methods without the need for any BIP, because the marker is observed in vaccinated individuals, such that the key issue is estimation of Prisk(s). In our Discussion, we noted that if Vrisk(s) and Prisk(s) are correctly modeled such that valid estimates are obtained, then the VE curve is accurately (unbiasedly) estimated, and this statement holds for any BIP that is reasonably well correlated with the marker, as it is for the ZEST example [1]. Therefore, we recast the question raised by Dunning as “How does a scientist jointly select the BIP and a model for Prisk(s) to obtain a valid estimate of Prisk(s), and how can one evaluate the validity?”