Evaluating a surrogate endpoint at three levels, with application to vaccine development

Identification of an immune response to vaccination that reliably predicts protection from clinically significant infection, i.e. an immunological surrogate endpoint, is a primary goal of vaccine research. Using this problem of evaluating an immunological surrogate as an illustration, we describe a hierarchy of three criteria for a valid surrogate endpoint and statistical analysis frameworks for evaluating them. Based on a placebo-controlled vaccine efficacy trial, the first level entails assessing the correlation of an immune response with a study endpoint in the study groups, and the second level entails evaluating an immune response as a surrogate for the study endpoint that can be used for predicting vaccine efficacy for a setting similar to that of the vaccine trial. We show that baseline covariates, innovative study design, and a potential outcomes formulation can be helpful for this assessment. The third level entails validation of a surrogate endpoint via meta-analysis, where the goal is to evaluate how well the immune response can be used to predict vaccine efficacy for new settings (building bridges). A simulated vaccine trial and two example vaccine trials are presented, one supporting that certain anti-influenza antibody levels are an excellent surrogate for influenza illness and another supporting that certain anti-HIV antibody levels are not useful as a surrogate for HIV infection.

HIV-1 vaccine-induced immunity in the test-of-concept Step Study: a case-cohort analysis

BACKGROUND

In the Step Study, the MRKAd5 HIV-1 gag/pol/nef vaccine did not reduce plasma viraemia after infection, and HIV-1 incidence was higher in vaccine-treated than in placebo-treated men with pre-existing adenovirus serotype 5 (Ad5) immunity. We assessed vaccine-induced immunity and its potential contributions to infection risk.

METHODS

To assess immunogenicity, we characterised HIV-specific T cells ex vivo with validated interferon-gamma ELISPOT and intracellular cytokine staining assays, using a case-cohort design. To establish effects of vaccine and pre-existing Ad5 immunity on infection risk, we undertook flow cytometric studies to measure Ad5-specific T cells and circulating activated (Ki-67+/BcL-2(lo)) CD4+ T cells expressing CCR5.

FINDINGS

We detected interferon-gamma-secreting HIV-specific T cells (range 163/10(6) to 686/10(6) peripheral blood mononuclear cells) ex vivo by ELISPOT in 77% (258/354) of people receiving vaccine; 218 of 354 (62%) recognised two to three HIV proteins. We identified HIV-specific CD4+ T cells by intracellular cytokine staining in 58 of 142 (41%) people. In those with reactive CD4+ T cells, the median percentage of CD4+ T cells expressing interleukin 2 was 88%, and the median co-expression of interferon gamma or tumor necrosis factor alpha (TNFalpha), or both, was 72%. We noted HIV-specific CD8+ T cells (range 0.4-1.0%) in 117 of 160 (73%) participants, expressing predominantly either interferon gamma alone or with TNFalpha. Vaccine-induced HIV-specific immunity, including response rate, magnitude, and cytokine profile, did not differ between vaccinated male cases (before infection) and non-cases. Ad5-specific T cells were lower in cases than in non-cases in several subgroup analyses. The percentage of circulating Ki-67+BcL-2(lo)/CCR5+CD4+ T cells did not differ between cases and non-cases.

INTERPRETATION

Consistent with previous trials, the MRKAd5 HIV-1 gag/pol/nef vaccine was highly immunogenic for inducing HIV-specific CD8+ T cells. Our findings suggest that future candidate vaccines have to elicit responses that either exceed in magnitude or differ in breadth or function from those recorded in this trial.

Development of an automated analysis system for data from flow cytometric intracellular cytokine staining assays from clinical vaccine trials

Intracellular cytokine staining (ICS) by multiparameter flow cytometry is one of the primary methods for determining T-cell immunogenicity in HIV-1 clinical vaccine trials. Data analysis requires considerable expertise and time. The amount of data is quickly increasing as more and larger trials are performed, and thus there is a critical need for high-throughput methods of data analysis. A web-based flow cytometric analysis system, LabKey Flow, was developed for the analyses of data from standardized ICS assays. Using a gating template created manually in commercially available flow cytometric analysis software, the system automatically compensates and analyzes all data sets. Quality control queries were designed to identify potentially incorrect sample collections. Comparison of the semiautomated analysis performed by LabKey Flow and the manual analysis performed using FlowJo software demonstrated excellent concordance (concordance correlation coefficient > 0.990). Manual inspection of the analyses performed by LabKey Flow for eight-color ICS data files from several clinical vaccine trials indicated that template gates can appropriately be used for most data sets. Thus, the semiautomated LabKey Flow analysis system can accurately analyze large ICS data files. Routine use of the system does not require specialized expertise. This high-throughput analysis will provide great utility for rapid evaluation of complex multiparameter flow cytometric measurements collected from large clinical trials.

Genital herpes has played a more important role than any other sexually transmitted infection in driving HIV prevalence in Africa

Background

Extensive evidence from observational studies suggests a role for genital herpes in the HIV epidemic. A number of herpes vaccines are under development and several trials of the efficacy of HSV-2 treatment with acyclovir in reducing HIV acquisition, transmission, and disease progression have just reported their results or will report their results in the next year. The potential impact of these interventions requires a quantitative assessment of the magnitude of the synergy between HIV and HSV-2 at the population level.

Methods and Findings

A deterministic compartmental model of HIV and HSV-2 dynamics and interactions was constructed. The nature of the epidemiologic synergy was explored qualitatively and quantitatively and compared to other sexually transmitted infections (STIs). The results suggest a more substantial role for HSV-2 in fueling HIV spread in sub-Saharan Africa than other STIs. We estimate that in settings of high HSV-2 prevalence, such as Kisumu, Kenya, more than a quarter of incident HIV infections may have been attributed directly to HSV-2. HSV-2 has also contributed considerably to the onward transmission of HIV by increasing the pool of HIV positive persons in the population and may explain one-third of the differential HIV prevalence among the cities of the Four City study. Conversely, we estimate that HIV had only a small net impact on HSV-2 prevalence.

Conclusions

HSV-2 role as a biological cofactor in HIV acquisition and transmission may have contributed substantially to HIV particularly by facilitating HIV spread among the low-risk population with stable long-term sexual partnerships. This finding suggests that prevention of HSV-2 infection through a prophylactic vaccine may be an effective intervention both in nascent epidemics with high HIV incidence in the high risk groups, and in established epidemics where a large portion of HIV transmission occurs in stable partnerships.

Identifying trait clusters by linkage profiles: application in genetical genomics

MOTIVATION

Genes often regulate multiple traits. Identifying clusters of traits influenced by a common group of genes helps elucidate regulatory networks and can improve linkage mapping.

METHODS

We show that the Pearson correlation coefficient, rho L, between two LOD score profiles can, with high specificity and sensitivity, identify pairs of genes that have their transcription regulated by shared quantitative trait loci (QTL). Furthermore, using theoretical and/or empirical methods, we can approximate the distribution of rho L under the null hypothesis of no common QTL. Therefore, it is possible to calculate P-values and false discovery rates for testing whether two traits share common QTL. We then examine the properties of rho L through simulation and use rho L to cluster genes in a genetical genomics experiment examining Saccharomyces cerevisiae.

RESULTS

Simulations show that rho L can have more power than the clustering methods currently used in genetical genomics. Combining experimental results with Gene Ontology (GO) annotations show that genes within a purported cluster often share similar function.

SOFTWARE

R-code included in online Supplementary Material.

Direct inference of SNP heterozygosity rates and resolution of LOH detection

Single nucleotide polymorphisms (SNPs) have been increasingly utilized to investigate somatic genetic abnormalities in premalignancy and cancer. LOH is a common alteration observed during cancer development, and SNP assays have been used to identify LOH at specific chromosomal regions. The design of such studies requires consideration of the resolution for detecting LOH throughout the genome and identification of the number and location of SNPs required to detect genetic alterations in specific genomic regions. Our study evaluated SNP distribution patterns and used probability models, Monte Carlo simulation, and real human subject genotype data to investigate the relationships between the number of SNPs, SNP HET rates, and the sensitivity (resolution) for detecting LOH. We report that variances of SNP heterozygosity rate in dbSNP are high for a large proportion of SNPs. Two statistical methods proposed for directly inferring SNP heterozygosity rates require much smaller sample sizes (intermediate sizes) and are feasible for practical use in SNP selection or verification. Using HapMap data, we showed that a region of LOH greater than 200 kb can be reliably detected, with losses smaller than 50 kb having a substantially lower detection probability when using all SNPs currently in the HapMap database. Higher densities of SNPs may exist in certain local chromosomal regions that provide some opportunities for reliably detecting LOH of segment sizes smaller than 50 kb. These results suggest that the interpretation of the results from genome-wide scans for LOH using commercial arrays need to consider the relationships among inter-SNP distance, detection probability, and sample size for a specific study. New experimental designs for LOH studies would also benefit from considering the power of detection and sample sizes required to accomplish the proposed aims.

A framework for assessing immunological correlates of protection in vaccine trials

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.

Chromosomal instability in Barrett's esophagus is related to telomere shortening

Barrett's esophagus is a useful model for the study of carcinogenesis, as the metaplastic columnar epithelium that replaces squamous esophageal epithelium is at elevated risk for development of adenocarcinoma. We examined telomere length and chromosomal instability (CIN) in Barrett's esophagus biopsies using fluorescence in situ hybridization. To study CIN, we selected centromere and locus-specific arm probes to chromosomes 17/17p (p53), 11/11q (cyclin D1), and 9/9p (p16 INK4A), loci reported to be involved in early stages of Barrett's esophagus neoplasia. Telomere shortening was observed in Barrett's esophagus epithelium at all histologic grades, whereas CIN was highest in biopsies with dysplastic changes; there was, however, considerable heterogeneity between patients in each variable. Alterations on chromosome 17 were strongly correlated with telomere length (r = 0.55; P < 0.0001) and loss of the 17p arm signal was the most common event. CIN on chromosome 11 was also associated with telomere shortening (r =0.3; P = 0.05), although 11q arm gains were most common. On chromosome 9p, arm losses were the most common finding, but chromosome 9 CIN was not strongly correlated with telomere length. We conclude that CIN is related to telomere shortening in Barrett's esophagus but varies by chromosome. Whether instability is manifested as loss or gain seems to be influenced by the chromosomal loci involved. Because telomere shortening and CIN are early events in Barrett's esophagus neoplastic progression and are highly variable among patients, it will be important to determine whether they identify a subset of patients that is at risk for more rapid neoplastic evolution.

The clustering of regression models method with applications in gene expression data

Identification of differentially expressed genes and clustering of genes are two important and complementary objectives addressed with gene expression data. For the differential expression question, many "per-gene" analytic methods have been proposed. These methods can generally be characterized as using a regression function to independently model the observations for each gene; various adjustments for multiplicity are then used to interpret the statistical significance of these per-gene regression models over the collection of genes analyzed. Motivated by this common structure of per-gene models, we proposed a new model-based clustering method--the clustering of regression models method, which groups genes that share a similar relationship to the covariate(s). This method provides a unified approach for a family of clustering procedures and can be applied for data collected with various experimental designs. In addition, when combined with per-gene methods for assessing differential expression that employ the same regression modeling structure, an integrated framework for the analysis of microarray data is obtained. The proposed methodology was applied to two microarray data sets, one from a breast cancer study and the other from a yeast cell cycle study.

Joint Inferences on Vaccine Efficacy Against Infection and Disease with Application to the First HIV Vaccine Efficacy Trial

In many clinical trials, subjects are followed for two stages of outcomes, and it is of interest to compare the incidence of each outcome between two randomized groups. The outcome of the first stage may influence the outcome of the second stage. Moreover, the relative risks of the two outcomes may be linked, with the time-dependent profile of relative risk for the second outcome functionally dependent on that of the first. For example, during exposure to HIV, virologic and host factors simultaneously impact the probability of infection and the subsequent viral trajectories, and the efficacy of a tested vaccine to prevent infection and to prevent viral failure may work in concert. We address this problem by modeling the relationship between the stage two hazard function and covariates via Cox's proportional hazards model (Cox, 1972), with the stage one log-hazard ratio theta(*) at the first event time Tl, included as a covariate. With theta(*) estimated using three methods, 1) nonparametric kernel smoothing; 2) locally parametric penalized splines; and 3) fully parametric cubic linear splines, we subsequently develop inference procedures for the regression parameter in the stage two Cox model based on each of the estimator of theta(*). The inferential procedures are studied in simulations and are illustrated with application to data from the world's first preventive HIV vaccine efficacy trial.

Statistical positivity criteria for the analysis of ELISpot assay data in HIV-1 vaccine trials

The enzyme-linked immunospot (ELISpot) assay is one of the leading technologies used to measure cellular responses in many settings including HIV vaccine clinical trials. HIV-1-specific effector T lymphocyte responses are considered necessary in the control of infection. Accurate measurement and summary of cellular immune responses are critical to HIV research. ELISpot assay readout is often dichotomized into positive/negative responses according to some pre-specified criteria. Given the increase in the number of replicates used for each stimulation with current assay configurations in the HIV Vaccine Trials Network (HVTN) laboratories, a new approach is now possible. We propose an objective criteria based on a hypothesis-driven method that controls the false positive rate while maintaining sensitivity to each of the antigen-specific responses under study. The new approach is compared to other commonly employed criteria using real and simulated data.

Peptide selection for human immunodeficiency virus type 1 CTL-based vaccine evaluation

Dozens of human immunodeficiency virus-type 1 (HIV-1) vaccine candidates specifically designed to elicit cytotoxic T-lymphocyte (CTL) responses have entered the pipeline of clinical trials. Evaluating the immunogenicity and potential efficacy of these HIV-1 vaccine candidates is challenging in the face of the extensive viral genetic diversity of circulating strains. Standardized peptide reagents to define the magnitude and potential breadth of the T-cell response, especially to circulating strains of HIV-1, are needed. For this purpose we developed a biometric approach based on T-cell recognition pattern for defining standardized reagents. Circulating strains in the Los Alamos database were evaluated and standardized algorithms to define all potential T-cell epitopes (PTEs) were generated. While many unique PTEs could be identified, a finite number based upon prevalence of circulating strains in the database, which we define as vaccine-important PTEs (VIPs), were used to select a common standardized panel of HIV-1 peptides for CTL-based vaccine evaluation. The usability of PTE peptide set was manifested by detection of Nef-specific CTL responses in HIV-1 subtype B infections.

On modeling HIV and T cells in vivo: assessing causal estimators in vaccine trials

The first efficacy trials--named STEP--of a T cell vaccine against HIV/AIDS began in 2004. The unprecedented structure of these trials raised new modeling and statistical challenges. Is it plausible that memory T cells, as opposed to antibodies, can actually prevent infection? If they fail at prevention, to what extent can they ameliorate disease? And how do we estimate efficacy in a vaccine trial with two primary endpoints, one traditional, one entirely novel (viral load after infection), and where the latter may be influenced by selection bias due to the former? In preparation for the STEP trials, biostatisticians developed novel techniques for estimating a causal effect of a vaccine on viral load, while accounting for post-randomization selection bias. But these techniques have not been tested in biologically plausible scenarios. We introduce new stochastic models of T cell and HIV kinetics, making use of new estimates of the rate that cytotoxic T lymphocytes--CTLs; the so-called killer T cells--can kill HIV-infected cells. Based on these models, we make the surprising discovery that it is not entirely implausible that HIV-specific CTLs might prevent infection--as the designers explicitly acknowledged when they chose the endpoints of the STEP trials. By simulating thousands of trials, we demonstrate that the new statistical methods can correctly identify an efficacious vaccine, while protecting against a false conclusion that the vaccine exacerbates disease. In addition to uncovering a surprising immunological scenario, our results illustrate the utility of mechanistic modeling in biostatistics.

Statistical evaluation of HIV vaccines in early clinical trials

The HIV pandemic is a pressing threat to global public health; HIV vaccine development is critical. Clinical evaluation of HIV vaccine candidates differs from the standard therapeutics trial framework primarily due to the fact that healthy individuals are studied. We present an early stage evaluation program developed for the HIV Vaccine Trials Network (HVTN) motivated by characteristics unique to the vaccine setting. The program consists of 3 prototypical stages (Phase I, Ib, II) that provide a unified yet flexible approach to the safety and immunogenicity evaluation of diverse vaccine regimens. The goal of these early trials is to narrow the number of candidate vaccines to the most promising candidates worthy of further study in efficacy trials.

How many human immunodeficiency virus type 1-infected target cells can a cytotoxic T-lymphocyte kill?

The antiviral role of CD8+ cytotoxic T lymphocytes (CTLs) in human immunodeficiency virus type 1 (HIV-1) infection is poorly understood. Specifically, the degree to which CTLs reduce viral replication by killing HIV-1-infected cells in vivo is not known. Here we employ mathematical models of the infection process and CTL action to estimate the rate that CTLs can kill HIV-1-infected cells from in vitro and in vivo data. Our estimates, which are surprisingly consistent considering the disparities between the two experimental systems, demonstrate that on average CTLs can kill from 0.7 to 3 infected target cells per day, with the variability in this figure due to epitope specificity or other factors. These results are compatible with the observed decline in viremia after primary infection being primarily a consequence of CTL activity and have interesting implications for vaccine design.

Denoising array-based comparative genomic hybridization data using wavelets

Array-based comparative genomic hybridization (array-CGH) provides a high-throughput, high-resolution method to measure relative changes in DNA copy number simultaneously at thousands of genomic loci. Typically, these measurements are reported and displayed linearly on chromosome maps, and gains and losses are detected as deviations from normal diploid cells. We propose that one may consider denoising the data to uncover the true copy number changes before drawing inferences on the patterns of aberrations in the samples. Nonparametric techniques are particularly suitable for data denoising as they do not impose a parametric model in finding structures in the data. In this paper, we employ wavelets to denoise the data as wavelets have sound theoretical properties and a fast computational algorithm, and are particularly well suited for handling the abrupt changes seen in array-CGH data. A simulation study shows that denoising data prior to testing can achieve greater power in detecting the aberrant spot than using the raw data without denoising. Finally, we illustrate the method on two array-CGH data sets.

[Randomized trial of breast self-examination in 266,064 women in Shanghai]

OBJECTIVE

A randomized trial of breast self-examination (BSE) program was carried out to evaluate whether the intensive BSE can reduce the death number of women from breast cancer.

METHODS

A total of 266,064 women (age of 30 to 64 years) associated with 519 textile factories in Shanghai had been randomly assigned to a BSE instruction group (132,979 women) or a control group (133,085 women) since 1989. Initial instruction in BSE group included demonstration of proper palpation techniques. It was followed by 2 reinforcement sessions during the subsequent 4 years including video shows, BSE instruction sessions and BSE practice under medical supervision. These activities were continued for 5 years. Attendance at all events was recorded. The cohort was followed through July 2000 for development of breast diseases, and the breast cancer cases were followed up through 2001 for vital status. The data analysis methods used included Kaplan-Meier plots, Log-rank test and Cox modeling.

RESULTS

Among women under instruction, 864 breast cancers were detected and 133 breast cancer deaths occurred, and 896 breast cancers were detected and 130 deaths recorded in the control group. The tumor size (P = 0.07), TNM stage (P = 0.39) and cumulative breast cancer mortality rate (P = 0.72) were not significantly different between the 2 groups. However, more and smaller fibroadenomas were detected in the instruction group than in the control group (P < 0.01).

CONCLUSION

Intensive instruction in BSE can not reduce mortality rate of breast cancer, but more and smaller benign breast lumps can be detected.

Correlation between immunologic responses to a recombinant glycoprotein 120 vaccine and incidence of HIV-1 infection in a phase 3 HIV-1 preventive vaccine trial

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 log(10) higher neutralization titer against HIV-1(MN), 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.

On simulating strongly interacting, stochastic population models. II. Multiple compartments

In Wick and Stelf [Math. Biosci. 187 (2004) 1], we showed how to simulate a pair of strongly interacting biological populations evolving stochastically over many orders-of-magnitude. Here we generalize the method to any (finite) number of compartments; transitions including births, deaths, progression through life-stages, and mitoses; and arbitrary rate functions. We illustrate the technique for a seven-compartment model of the cellular immune response to a viral infection.

Recognition of the peripheral self by naturally arising CD25+ CD4+ T cell receptors

Naturally arising CD25+ CD4+ regulatory T cells (TR) play an important role in the prevention of autoimmunity. TCR specificity is thought to play a critical role in TR development and function, but the repertoire and specificity of TR TCRs remain largely unknown. We find by sequencing of TRAV14 (Valpha2) TCRalpha chains associated with a transgenic TCRbeta chain that the TRand CD25- CD4+ TCR repertoires are similarly diverse, yet only partially overlapping. Retroviral expression of TCRalpha genes in TCR transgenic RAG-deficient T cells revealed that a high frequency of TCRs derived from CD25+ but not CD25- CD4+ T cells confers the ability to rapidly expand upon transfer into a lymphopenic host. Thus, these data show that a large proportion of naturally arising TR have substantially more efficient interactions with MHC class II bound peptides from the peripheral self than CD25- T cells.

Statistical considerations for the design and analysis of the ELISpot assay in HIV-1 vaccine trials

Effector T lymphocyte responses are considered critical for controlling human immunodeficiency virus type-1 (HIV-1) infection. The enzyme-linked immunospot (ELISpot) assay has emerged as a primary means of assessing HIV-specific T cell responses, and the development of objective methods that distinguish positive and negative ELISpot responses while properly controlling the rate of false positives is critical. In this paper, we consider several statistical methods that are helpful in defining such a positive criterion. Simulation results under a variety of scenarios suggest that a permutation-based criterion using a resampling adjustment for multiple comparisons yields the desired false positive rate while remaining competitive with other potential criteria in terms of sensitivity. These results also provide guidance on the effect of the number of experimental and negative control replicate wells on assay sensitivity. Application of different potential positive criteria using ELISpot assay results from IFN-gamma-secreting T cells of HIV-1 seropositive and seronegative donors confirmed several of the results obtained under simulation. Our findings support the application of statistically-based positive criteria such as the permutation-based resampling approach in assessing HIV vaccine-induced T cell responses. Moreover, the proposed methods have potential utility in related HIV immunopathogenesis studies and in non-HIV clinical vaccine trials.

Endpoints in vaccine trials

In this paper we discuss statistical considerations regarding endpoints in preventive vaccine trials. Brief discussion is given to preclinical, Phase I, and Phase II trials, with the bulk of attention paid to endpoint choice and analysis in Phase III efficacy trials. In addition to traditional efficacy measures of vaccine effects for immunized individuals, consideration is given to waning, strain specific efficacy, correlates of protective immunity, postinfection endpoints, and cluster randomized trials.

On simulating strongly-interacting, stochastic population models

Simulating strongly-interacting biological populations over many orders-of-magnitude can be challenging even on fast modern computers. Here we describe a regime-switching technique that can speed up simulations from stochastic models by large factors, while retaining third-order accuracy at each time-step. We apply it to an elementary model of the immune response to an infectious organism.

On the analysis of viral load endpoints in HIV vaccine trials

First generation HIV vaccines are not likely to provide complete protection from HIV-1 infection. Therefore, it is important to assess a vaccine's effect on disease progression and infectiousness of infected vaccinees in an efficacy trial; however, direct assessment of such vaccine effects is not feasible within current trial designs. Viral load in HIV-infected individuals correlates with infectiousness and disease progression in a natural history setting, and thus is a reasonable candidate for a surrogate outcome in vaccine efficacy trials. We consider comparisons of viral load of infected vaccinees to that of infected trial participants in the control group. Dramatic differences in viral loads between these groups would suggest a vaccine effect on disease progression. However, modest differences, even if statistically significant, could be consistent with an imperfect vaccine effect on susceptibility to infection and not an effect on disease progression, that is, a selection effect of the vaccine. Thus, the usual statistical tests for no difference between groups do not test the biologically and clinically relevant hypothesis. We propose a model for the possible selective effects of a vaccine and develop several test statistics for assessing a direct effect of the vaccine on viral load given this selection model. Finite sample properties of these tests are evaluated using computer simulations.