The optimal long-term approach to controlling the HIV type 1 (HIV-1) pandemic is an effective HIV-1 preventive vaccine that induces protective immunity against HIV-1 infection. At present, only 2 phase III HIV-1 vaccine trials have been completed; 1 in North America and Europe and 1 in Thailand. Both candidate vaccines, AIDSVAX B/B and AIDSVAX B/E, are synthetic recombinant glycoproteins (gps) representing the specific HIV-1 subtypes circulating in the geographic regions of each trial. Unfortunately, neither AIDSVAX B/B nor AIDSVAX B/E showed protective efficacy against primary HIV-1 infection.1,2
The lack of AIDSVAX B/B or AIDSVAX B/E efficacy and the absence of other candidate HIV-1 vaccines for phase III trials make it important to explore whether recently tested HIV-1 vaccines have other potential effects.1,2 For instance, as some candidate vaccines have reduced blood plasma HIV-1 RNA viral load level and delayed disease progression in animal models, viral load is used as a putative surrogate for secondary transmission and AIDS progression.3-7 We previously reported no significant differences in disease progression between AIDSVAX B/E vaccine recipients (VRs) and placebo recipients (PRs) who became HIV-1 infected, which was consistent with findings from the AIDSVAX B/B vaccine trial.1,2 In this article, we report results of our investigation of effects of vaccination with AIDSVAX B/E on blood HIV-1 RNA viral load (BVL) levels and seminal plasma HIV-1 RNA viral load (SVL) detectability among HIV-1-infected VRs compared with HIV-1-infected PRs, controlling for HIV-1-infecting subtype and time since the estimated date of seroconversion (EDS).8
A randomized, double-blind, placebo-controlled efficacy trial of phase III AIDSVAX B/E was initiated in Bangkok, Thailand, in March 1999. Details of the trial design and conduct are presented elsewhere.1,9 Briefly, after giving voluntary informed consent, 2546 HIV-1-uninfected injecting drug users were randomly assigned at a 1:1 ratio to receive 7 doses of either AIDSVAX B/E or placebo. Based on positive enzyme-linked immunoassay and confirmatory Western blot with 2 new bands other than gp120 or gp160, 230 participants were identified as HIV-1 infected during the trial. Retrospective testing of sera with a highly sensitive nucleic acid-based amplification test found 19 participants HIV-1 infected at enrollment. Immunizations were discontinued from the time of detection of HIV-1 infection. HIV-1 seroconvertors were offered enrollment into a prospective follow-up study for a total duration of 36 months; 200 enrolled. Additional follow-up of 24 months (through September 2004) was offered to seroconvertors who completed the initial 36 months for extended immunologic and virologic evaluations. Study protocols were approved by the ethical review committees of collaborating institutes.
Data Collection and Laboratory Methods
Standardized interviews collected demographic characteristics at enrollment and HIV-1 risk behaviors at every study visit. Thereafter, each participant received risk reduction counseling and was clinically evaluated by a study physician. The participants received medical care including antiretroviral therapy (ART) for their HIV-1 infection.10,11 Blood samples were collected at 2 weeks and at months 1, 2, and 4 and every 4 months after first HIV-1-positive test. Samples were processed and tested according to the study protocol. Lymphocyte immunotyping was performed on fresh anticoaggulated blood samples, using a FACScan flow cytometer (Becton Dickinson Immunocytometry Systems, San Jose, CA).12 Blood plasma samples were tested to determine HIV-1 RNA levels using the Amplicor HIV-1 Monitor Test, version 1.5 (Roche Molecular Systems, Branchburg, NJ). Genetic characterization of HIV-1 subtypes was performed by molecular sequence analysis of the C2-V4 env region.1
Additionally, 95 HIV-1-infected male participants consented to provide semen samples scheduled at months 1, 2, and 4 and every 4 months after EDS (median, 3 samples). Samples were checked for quality, and cells and sperm activity were microscopically evaluated. After separation of cells, HIV-1 RNA in seminal plasma was extracted using the NucliSen kit (Organon Teknika, Boxtel, The Netherlands) to remove an inhibitor of the Amplicor internal control and quantified using a modified Amplicor HIV-1 Monitor Test, version 1.5.13,14
The EDS was assumed to be the midpoint between the date of the last HIV-1 seronegative and the date of the first HIV-1-seropositive visit. For 122 BVL measurements beyond the assay's reliable detection range (400-750,000 copies/mL), arbitrary levels of 200 and 750,000 copies per milliliter were assigned for values below and above the limit, respectively. BVL and detectable SVL levels were log transformed to produce a normal distribution for the purpose of analysis.15-17
Demographic and behavioral characteristics and follow-up time of VRs and PRs were compared using the χ2 test or the Fisher exact test for categorical variables and the Student t test for continuous variables.18-20 Only individuals infected with HIV-1 subtype B or CRF01_AE were included; 1 PR infected with a non-B strain and 3 PRs infected with a nontypable strain were excluded, resulting in 196 individuals for this analysis. To control for confounding effects of ART on BVL and SVL levels, only data collected before ART initiation were incorporated in these analyses (total samples = 1195 for BVL and 392 for SVL).
Because HIV-1 RNA viral load has been shown to vary according to stage of HIV-1 infection, serial HIV-1 RNA viral load levels were analyzed over time. Robust, locally weighted, nonparametric, smoothed regression was used to describe temporal trends of BVL and SVL.21 Linear mixed models, with random intercepts and slopes over time, were fitted for BVL data with exchangeable correlation structure to account for intraindividual correlations.22 All variables, except time since the EDS, were entered in the models as dichotomous variables. In this study, ART initiation guidelines were based on CD4+ T-lymphocyte (CD4) counts; hence, they predict data exclusion. However, given the similarity in proportions of VRs and PRs who started ART (P = 0.36) and that CD4 was thought to be an intermediate factor between vaccination and viral load level, it was justified that approximately unbiased estimates of the vaccine efficacy on pre-ART viral load level could be obtained without CD4 level included as a covariate in the analysis. Two-way and 3-way interaction terms were generated as products of variables (study arm, HIV-1-infecting subtype, and time) and entered in the models.
The SVL levels were dichotomized into detectable or undetectable HIV-1 RNA viral load groups because HIV-1 RNA was unquantifiable in the majority of seminal plasma samples. Generalized estimating equation models were fitted with HIV-1-infecting subtype, study arm, sample collection time, and interaction terms as independent variables.23
For each infected participant, the average BVL and the average SVL were computed using all available pre-ART values from samples collected between 2 and 6 months and between 16 and 19 months after EDS. These windows were chosen because relatively large amounts of data were available and they approximately reflected initial set point viral load and a later value of viral load. BVL and SVL were compared using Wilcoxon rank sum test.24
A 2-sided P value of ≤0.05 was used to indicate statistical significance. Analyses were performed using Stata version 9.0 (StataCorp LP, College Station, TX) and SAS version 9.1 (SAS Institute Inc, Cary, NC). The following abbreviations were used: HIV-1 CRF01_AE-infected VRs (VRHIV E+), HIV-1 CRF01_AE-infected PRs (PRHIV E+), HIV-1 subtype B-infected VRs (VRHIV B+), and HIV-1 subtype B-infected PRs (PRHIV B+).
Demographic and Other Characteristics of Study Participants
Of the 196 participants, 97 (49%) were VRs and 99 (51%) were PRs. All participants were Thai, and 186 (95%) were male. The median age at diagnosis of HIV-1 infection was 27 [interquartile range (IQR), 25-34] years; 189 (96%) had completed at least primary school. A lifetime history of incarceration was reported by 172 (88%), and the median number of incarceration times was 4 (IQR, 2-6). One hundred and ninety-five (99%) had received at least 3 doses of either vaccine or placebo (at 0, 1, and 6 months). Compared with VRs, a higher proportion of PRs reported currently cohabiting with a sexual partner (P = 0.01) and reported less frequent sexual intercourse with a casual sex partner (P = 0.03).
HIV-1-Infecting Subtypes and Follow-up Time
Of the 97 VRs, 83 (86%) were infected with HIV-1 CRF01_AE and 14 (14%) with HIV-1 subtype B. Of the 99 PRs, 81 (82%) were infected with HIV-1 CRF01_AE and 18 (18%) with HIV-1 subtype B. The relative hazard (vaccine versus placebo) of HIV-1 CRF01_AE infection was similar to that of HIV-1 subtype B infection (discrete time Cox regression accounting for subtypes B and CRF01_ AE as competing risks, P = 0.84).25 All HIV-1 subtype B strains clustered on phylogenetic analysis with strains known as “Thai B.”26,27
Excluding post-ART visits, the median follow-up time was 25 (IQR, 18-32) months, corresponding to a median of 9 (IQR, 6-10) study visits. Follow-up times did not differ between VRs and PRs (P = 0.79).
Blood Plasma HIV-1 RNA Viral Load Levels
Summaries of log10 BVL levels of pre-ART measurements are presented in Figure 1. The mean durations from the EDS to the first viral load determination were 2 (IQR, 1-3) months among VRs and 2 (IQR, 2-3) months among PRs (P = 0.92). Of the 196 participants, none had a BVL level below or higher than the assay's reliable detection limit at the first determination.
VRs Versus PRs Controlling for HIV-1-Infecting Subtype
The estimated difference in BVL levels between VRHIV E+ and PRHIV E+ at first month after EDS was 0.2 log10 (1.7-fold; VRHIV E+>PRHIV E+; P < 0.01). However, this difference did not persist over time (Fig. 2A). BVL levels among VRHIV E+ were 0.2 log10 lower than those of PRHIV E+ at month 24 after EDS (1.7-fold; VRHIV E+<PRHIV E+; P < 0.01). The estimated difference in BVL levels between VRHIV B+ and PRHIV B+ was 0.3 log10 throughout follow-up time (1.8-fold; VRHIV B+>PRHIV B+; P < 0.01; Fig. 2B).
HIV-1 CRF01_AE-Infected Individuals Versus HIV-1 Subtype B-Infected Individuals Controlling for Study Arm
The estimated difference in BVL levels at first month after the EDS between VRHIV E+ and VRHIV B+ was 0.1 log10 (1.4-fold; VRHIV E+>VRHIV B+; P = 0.07). This difference decreased over time, but at approximately 12 months after EDS increased toward the end of follow-up (Fig. 2C). The estimated difference in BVL levels between PRHIV E+ and PRHIV B+ was 0.1 log10 (1.4-fold; PRHIV E+>PRHIV B+; P = 0.03) throughout the follow-up period (Fig. 2D).
Seminal Plasma HIV-1 RNA Viral Load
Detectability of Seminal Plasma HIV-1 RNA Viral Load
One hundred and twenty-nine (33%) samples from 95 participants had detectable SVL. Of these, 67 (52%) were collected from VRs (47 from VRHIV E+ and 20 from VRHIV B+) and 62 (48%) were from PRs (53 from PRHIV E+ and 9 from PRHIV B+). Fifty percent of samples were collected on the same day (IQR, 0-5) as blood. Figure 3 illustrates temporal trends of pre-ART log10 SVL levels over time by study arm and HIV-1-infecting subtype. Visually, SVL level of VRHIV E+ was similar to that of PRHIV E+ (Figs. 3C, F); SVL level of VRHIV B+ was higher than that of PRHIV B+ (Figs. 3B, E).
VRs Versus PRs Controlling for HIV-1-Infecting Subtype
PRHIV E+ and VRHIV E+ had no significant difference in SVL detectability [odds ratio (OR), 1.0; 95% confidence interval (CI), 0.5 to 2.2; P = 0.98]. VRHIV B+ were 7.4 times as likely as PRHIV B+ to have detectable SVL (OR, 7.4; CI, 1.7 to 32.5; P = 0.01).
HIV-1 Subtype B-Infected Individuals Versus HIV-1 CRF01_AE-Infected Individuals Controlling for Study Arm
VRHIV E+ were significantly less likely than VRHIV B+ to have detectable SVL (OR, 0.1; CI, 0.0 to 0.4; P = 0.01). PRHIV E+ and PRHIV B+ had no significant difference in SVL detectability (OR, 0.4; CI, 0.1 to 2.3; P = 0.28).
Association Between BVL and SVL Levels
Analysis of 374 paired samples showed that among VRs, the median SVL was 1.8 log10 copies per milliliter (IQR, 1.2-2.3) lower than BVL and among PRs, the median SVL was 1.9 log10 copies per milliliter (IQR, 1.2-2.4) lower. For the 2- to 6-month and 16- to 19-month windows, the average SVL levels were significantly lower than the average BVL levels based on paired data (P < 0.01 for both VRs and PRs). Detectable BVL and undetectable SVL was present in 257 (66%) paired samples; undetectable BVL and detectable SVL was uncommon (1 paired sample, 0.3%). Among participants with detectable average SVL, no correlation was observed between average BVL and average SVL neither for the 2- to 6-month window (Spearman r, 0.11; P = 0.62) nor for the 16- to 19-month window (Spearman r, −0.35; P = 0.20). The distribution of average 2- to 6-month BVL among those with detectable average 2- to 6-month SVL (median, 5.1 log10 copies/mL) differed (borderline) significantly from that of those with undetectable average 2- to 6-month SVL (median, 4.8 log10 copies/mL; P = 0.06). The difference was larger at the 16- to 19-month window, with median average 16- to 19-month BVL 5.2 log10 copies per milliliter compared with 4.2 log10 copies per milliliter for those with detectable and undetectable SVL, respectively (P < 0.01).
Twenty-four of 40 HIV-1 CRF01_AE-infected participants (60%) had undetectable SVL. Among the 16 detectable participants, the median 2- to 6-month average SVL was 3.1 log10 copies per milliliter (SD, 0.8). Two of 7 HIV-1 subtype B-infected participants (29%) had undetectable SVL. Among the 5 detectable participants, the median 2- to 6-month average SVL was 3.4 log10 copies per milliliter (SD, 0.6). For the 16- to 19-month window, 30 of 43 HIV-1 CRF01_AE-infected participants (70%) and 4 of 6 HIV-1 subtype B-infected participants (67%) had undetectable SVL. The median 16- to 19-month average SVL for the 13 HIV-1 subtype B-infected participants and the 2 HIV-1 CRF01_AE-infected participants with detectable SVL were 3.9 (SD, 0.5) and 3.4 log10 copies per milliliter (meaningful SD could not be calculated), respectively.
In the per protocol-specified analysis of this clinical trial, no differences in BVL levels were found between VRs and PRs.1 However, when taking into account interactions between study arm, HIV-1-infecting subtype, and time, we found that vaccination with AIDSVAX B/E was associated with slightly higher BVL levels among 83 VRHIV E+ compared with 81 PRHIV E+ during the first 12 months of the study. After that time, BVL levels among VRHIV E+ were lower than those of PRHIV E+. Fourteen VRHIV B+ had higher BVL levels compared with 18 PRHIV B+ throughout the follow-up period. VRHIV B+ were significantly more likely to have detectable SVL than PRHIV B+.
The natural history of HIV-1 infection among PRs in this study is consistent with that of Hu et al8 in that higher BVL levels were observed in PRHIV E+ compared with PRHIV B+ at the onset of HIV-1 infection.28 The slightly higher BVL levels and higher odds of SVL detectability among VRHIV B+ is noteworthy, albeit the small number of HIV-1 subtype B-infected individuals. Additionally, the analyzed groups were selected by the postrandomization event of HIV-1 infection; thus, a causal interpretation of the findings may be limited by selection bias.29 The reported changes in BVL, although relatively slight, might be of concern, as these may potentially increase infectiousness. In a systematic review of studies assessing the association between changes in BVL and risk of HIV-1 transmission during heterosexual contact, Modjarrad et al30 found evidence suggesting that an increment in BVL as small as 0.3 log10 could increase the likelihood of HIV-1 transmission by 20%. However, it should be noted that none of the included studies examined SVL, and the estimated 20% increase in transmission risk was abstracted from studies that varied by a number of factors including sample size, study population demographics, viral load assay type, and study design. In our study, BVL and SVL were concurrently evaluated, and we found that despite the subtle increase in BVL, the SVL among the majority of our participants was still below the sexual transmission threshold of 3.2 log10 copies per milliliter, leading us to believe that transmission through this mode was unlikely.31 The small increase in individuals' BVL, however, might have had some impact on HIV-1 transmission through other risk behavior, in particular, through injection drug use.
Though the full understanding of the impact of vaccination on individuals' infectiousness and clinical progression is limited, our findings emphasize the importance of inclusion of BVL and SVL as vaccine trial endpoints.32 In conjunction with virological endpoints, screening test of concept trials may also be considered to accelerate the AIDS vaccine development and to better utilize limited resources available.33
Our study is the first to address the impact of gp120 candidate vaccines on HIV-1 RNA shedding in the male genital tract. Data from 2 recent animal studies suggested that gp120 may enhance viral loads when given with other vaccine components.34,35 In both studies, vaccinated macaques showed less consistent control of postchallenge viremia with pathogenic simian immunodeficiency virus compared with controls receiving vaccine components without gp120. Statistically higher BVL levels were also observed in macaques receiving vaccine components with gp120 compared with controls, which suggests that preexposure to the Env component of the vaccine enhanced viral replication. Env 120 or 160 may raise enhancing antibodies to levels that were not evident by the in vitro assay, influence the activation of a T-helper cell subset over another, or stimulate a cytokine response profile that drives overall viral expression differently than non-Env component.34 Expansion of previously primed memory CD4 without a strong CD8+ T-lymphocyte (CD8) response, may lead to preferential infection of these lymphocytes and more rapid viral replication upon viral challenge than in controls.35 However, in another trial of the gp120 candidate vaccine (AIDSVAX B/B) in a population of predominantly men who have sex with men in North America and Europe, the pre-ART BVL was similar in the 225 VRs and 122 PRs who acquired HIV-1 subtype B infection; semen samples, though, were not collected.36
The findings may also be explained by a “sieve” effect of the vaccine.37 The epidemic of HIV-1 subtype B among injecting drug users in Bangkok preceded the epidemic of HIV-1 CRF01_AE in this population.26,27 Thus, HIV-1 subtype B-infected individuals may have had a more antigenetically diverse or more virulent virus than those with HIV-1 CRF01_AE infection, resulting in increased penetration of the vaccine barrier. Higher levels of nucleotide diversity and divergence have been shown to be associated with higher BVL levels.38 A sieve analysis may help determine if potential vaccine-induced protection depends on genotypic and phenotypic variations of the HIV-1 strain and could clarify if it was the HIV-1 strain, immune response, type of cells infected, or a combination of these factors that results in HIV-1 subtype-specific BVL and SVL differences.
The findings of this study are of particular importance after the Merck adenovirus trial (STEP trial) was halted in late 2007.39 Our data and those of the STEP trial demonstrate that either humoral or cell-mediated immunity vaccine alone did not prevent HIV-1 infection or reduce the HIV-1 viral setpoint.40 Our results, yet, cannot be directly compared with the STEP trial due to some aspects. First, AIDSVAX B/E was an HIV-1 Env vaccine, whereas the Merck vaccine was an HIV-1 Gag/Pol/Nef vaccine. Second, the 2 vaccines elicited different immune responses (humoral versus cellular). Third, the AIDSVAX B/E trial was a phase III trial conducted based on promising safety and immunologic profiles derived from phase I/II studies, whereas the STEP trial was a phase IIB test-of-concept study. To understand the complexity of viral-host interaction found in our study, additional genetic characterization of virus variation and studies of host such as CD4 and CD8 responses to gp120 in relation to virus control, CD4 phenotypes in terms of infection reservoir/enhancement, and enhancing antibodies to HIV-1 CRF01_ AE and subtype B, and Fc receptor genotypes would have been necessary. However, these immunologic and viral studies were not conducted because the vaccine did not show efficacy and breakthrough infection did not occur. Note that in the ongoing prime-boost phase III efficacy trial in Thailand, AIDSVAX B/E is administered in combination with HIV-1 vCP1521 vaccine. The efficacy of this regimen in preventing HIV-1 infection remains to be seen.
In this study, we collected a large number of paired blood and semen samples to study BVL and SVL in the setting of an HIV-1 vaccine trial. This provided an unique opportunity to study interaction between HIV-1-infecting subtype, study arm, and viral load outcomes. Though our results are based on 1 study in a specific setting with limited sample size, they may help in formulating hypotheses for further HIV-1 research toward developing an effective HIV-1 vaccine.
We are greatly indebted to all participants in our study. We thank the staff from the Bangkok Metropolitan Administration, Thailand Ministry of Public Health-US Centers for Disease Control and Prevention Collaboration, VaxGen, Inc, Centers for Disease Control and Prevention, and Department of Microbiology, Faculty of Medicine, Siriraj Hospital, Mahidol University for administrative, clinical, laboratory, and data management support. Specifically, we would like to acknowledge: Thitima Cherdtrakulkiat and Supawadee Na-Pompet for study coordination; Philip Mock for statistical consultation; Nartlada Chantharojwong for data management; Wanna Leelawiwat, Punneeporn Wasinrapee, Nancy Young, and Thanyanan Chaowanachan for laboratory support; and Chonticha Kittinunvorakoon for laboratory consultation. Last, we sincerely thank Robert Chen, Sal Butera, and John Brooks for input in the discussion section and Arthur L. Reingold, Robert Linkins, and Peter Kilmarx for critical review of this article.
1. The Bangkok Vaccine Evaluation Group. Randomized, placebo-controlled efficacy trial of a bivalent rgp120 HIV-1 vaccine
among injecting drug users
in Bangkok, Thailand. J Infect Dis
2. Flynn NM, Forthal DN, Harro CD, et al. Placebo-controlled phase 3 trial of recombinant glycoprotein 120 vaccine to prevent HIV-1 infection. J Infect Dis
3. Barouch DH, Santra S, Schmitz JE, et al. Control of viremia and prevention of clinical AIDS in rhesus monkeys by cytokine-augmented DNA vaccination. Science
4. Amara RR, Villinger F, Altman JD, et al. Control of a mucosal challenge and prevention of AIDS by a multiprotein DNA/MVA vaccine. Science
5. Shiver JW, Fu TM, Chen L, et al. Replication-incompetent adenoviral vaccine vector elicits effective anti-immunodeficiency-virus immunity. Nature
6. Hu DJ, Vitek CR, Bartholow B, et al. Key issues for a potential human immunodeficiency virus vaccine. Clin Infect Dis
7. Buchacz K, Hu DJ, Vanichseni S, et al. Early markers of HIV-1 disease progression in a prospective cohort of seroconverters in Bangkok, Thailand: implications for vaccine trials. J Acquir Immune Defic Syndr
8. Hu DJ, Vanichseni S, Mastro TD, et al. Viral load differences in early infection with two HIV-1 subtypes. AIDS
9. Vanichseni S, Tappero JW, Pitisuttithum P, et al. Recruitment, screening and characteristics of injection drug users participating in the AIDSVAX B/E HIV vaccine trial, Bangkok, Thailand. AIDS
10. Ministry of Public Health. National Guidelines for the Clinical Management of HIV Infection in Children and Adults
. 6th ed. Nonthaburi, Thailand: Division of AIDS, Department of Communicable Disease Control, Ministry of Public Health; 2000.
11. Ministry of Public Health. Guidelines for the Care and Treatment of HIV/AIDS in Children and Adults in Thailand
. 7th ed. Nonthaburi, Thailand: Division of AIDS, Department of Communicable Disease Control, Ministry of Public Health; 2002.
12. Centers for Disease Control and Prevention. 1997 Revised guidelines for performing CD4+ T-cell determinations in persons infected with human immunodeficiency virus (HIV). MMWR Recomm Rep
. 1997;46:(RR-2): 1-29.
13. Coombs RW, Speck CE, Hughes JP, et al. Association between culturable human immunodeficiency virus type 1 (HIV-1) in semen and HIV-1 RNA levels in semen and blood: evidence for compartmentalization of HIV-1 between semen and blood. J Infect Dis
14. Boom R, Sol CJ, Salimans MM, et al. Rapid and simple method for purification of nucleic acids. J Clin Microbiol
15. Lyles RH, Lyles CM, Taylor DJ. Random regression models for human immunodeficiency virus ribonucleic acid data subject to left censoring and informative drop-outs. Appl Stat
16. Smith SM, Holland B, Russo C, et al. Retrospective analysis of viral load and SIV antibody responses in rhesus macaques infected with pathogenic SIV: predictive value for disease progression. AIDS Res Hum Retroviruses
17. Hudgens MG, Hoering A, Self SG. On the analysis of viral load endpoints in HIV vaccine trials. Stat Med
18. Pearson ES. The choice of statistical test illustrated on the interpretation of data classed in a 2 × 2 table. Biometrika
19. Upton GJG. Fisher's exact test. J Roy Statist Soc A
20. “Student” (Gosset WS). The probable error of a mean. Biometrika
21. Cleveland WS. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc
22. Laird NM, Ware JH. Random-effects models for longitudinal data. Biometrics
23. Liang KY, Zeger SL. Longitudinal data analysis using general linear models. Biometrika
24. Wilcoxon F. Individual comparisons by ranking methods. Biometrics Bull
25. Lunn M, McNeil D. Applying Cox regression to competing risks. Biometrics
26. Subbarao S, Limpakarnjanarat K, Mastro TD, et al. HIV type 1 in Thailand, 1994-1995: persistence of two subtypes with low genetic diversity. AIDS Res Hum Retroviruses
27. Subbarao S, Vanichseni S, Hu DJ, et al. Genetic characterization of incident HIV type 1 subtype E and B strains from a prospective cohort of injecting drug users
in Bangkok, Thailand. AIDS Res Hum Retroviruses
28. Kivela PS, Krol A, Salminen MO, et al. High plasma HIV load in the CRF01-AE outbreak among injecting drug users
in Finland. Scand J Infect Dis
29. Gilbert PB, Bosch RJ, Hudgens MG. Sensitivity analysis for the assessment of causal vaccine effects on viral load in HIV vaccine trials. Biometrics
30. Modjarrad K, Chamot E, Vermund SH. Impact of small reductions in plasma HIV RNA levels on the risk of heterosexual transmission and disease progression. AIDS
31. Chakraborty H, Sen PK, Helms RW, et al. Viral burden in genital secretions determines male-to-female sexual transmission of HIV-1: a probabilistic empiric model. AIDS
32. Gupta SB, Jacobson LP, Margolick JB, et al. Estimating the benefit of an HIV-1 vaccine
that reduces viral load set point. J Infect Dis
33. Excler JL, Rida W, Priddy F, et al. A strategy for accelerating the development of preventive AIDS vaccines. AIDS
34. Buge SL, Ma HL, Amara RR, et al. Gp120-alum boosting of a Gag-Pol-Env DNA/MVA AIDS vaccine: poorer control of a pathogenic viral challenge. AIDS Res Hum Retroviruses
35. Staprans SI, Barry AP, Silvestri G, et al. Enhanced SIV replication and accelerated progression to AIDS in macaques primed to mount a CD4 T cell response to the SIV envelope protein. Proc Natl Acad Sci USA
36. Gilbert PB, Ackers ML, Berman PW, et al. HIV-1 virologic and immunologic progression and initiation of antiretroviral therapy among HIV-1 infected subjects in a trial of the efficacy of recombinant glycoprotein 120 vaccine. J Infect Dis
37. Gilbert P, Self S, Rao M, et al. Sieve analysis: methods for assessing from vaccine trial data how vaccine efficacy varies with genotypic and phenotypic pathogen variation. J Clin Epidemiol
38. Mani I, Gilbert P, Sankale JL, et al. Intrapatient diversity and its correlation with viral setpoint in human immunodeficiency virus type 1 CRF02_A/G-IbNG infection. J Virol
40. Buchbinder R, Osborne RH, Ebeling PR, et al. Efficacy assessment of a cell-mediated immunity HIV-1 vaccine
(the Step Study): a double-blind, randomised, placebo-controlled, test-of-concept trial. Lancet