Safety and efficacy assessment of the HVTN 503/Phambili Study: A double-blind randomized placebo-controlled test-of-concept study of a Clade B-based HIV-1 vaccine in South Africa

Introduction

South Africa (SA), the country with the largest number of people infected with HIV-1 (5.7 million) ¹ and an HIV-1 prevalence of 11% in the general population ² is in dire need of a biomedical intervention to prevent HIV-1 infection. The HVTN 503 or Phambili (Zulu for “forward!”) study was the second phase IIb HIV vaccine test-of-concept study to evaluate the efficacy of the MRKAd5 HIV-1 gag/pol/nef sub-type B vaccine. This vaccine was designed to elicit T-cell mediated immune responses capable of providing complete or partial protection from HIV-1 infection or a decrease in viral load post-acquisition^3–9. The Ad5 vaccine was shown to elicit immune responses in participants regardless of Ad5 serostatus¹⁰. With promising safety and immunogenicity data from earlier clinical studies ^{10, 11}, the Step test-of-concept of study¹² was designed to assess the efficacy of this vaccine in regions of the world where the predominant circulating sub-type is clade B. The Phambili study was designed soon after to test the vaccine in a clade C region of the world^{13, 14}, in populations with high levels of pre-existing immunity to Ad5 ^{15, 16}. Cross-clade T cell immunity to HIV-1 gag, pol and nef genes had been demonstrated in both HIV-1 infected and Clade B vaccinated HIV seronegatives, providing an immunization rationale for the trial in SA ^{17, 18}. Enrollment and vaccinations in the Phambili study, were stopped in September 2007 following the Step study’s interim analysis, which found that the vaccine did not protect against HIV-1 infection or reduce early viral load in those who acquired infection. In an exploratory analysis of Step data, the risk of HIV-1 infection also appeared to be higher in a sub-group of male vaccine recipients, those who were Ad5 sero-positive or uncircumcised ¹². We report our findings from the Phambili study on HIV-1 acquisition and disease progression as well as describe and investigate factors associated with HIV-1 acquisition as a means of guiding future biomedical interventions directed at reducing HIV-1 acquisition.

Methods

Results

We enrolled 801 participants between 24 January 2007 and 19 September 2007, before enrolment and vaccinations were discontinued, in five sites in SA (Figure 1). Of those enrolled, 360 (44.9%) were women (Table 1), and two thirds (522) were under the age of 25. Nearly 20% (154) of participants had baseline Ad5 antibody titres ≤18 and 55.6% (445) had Ad5 antibody titres of >200. At enrollment, 129 men (29.3%) were circumcised (61 on the vaccine arm and 68 on the placebo arm). An additional 109 men were circumcised post-enrollment (52 vaccine, 57 placebo); half of these occurred pre-unblinding (26 vaccine, 29 placebo). HSV-2 prevalence at baseline was significantly higher in women than in men, (177/360=49.2% vs 72/441=16.3%, respectively, p<0.001). Baseline demographic and HIV risk behaviors were similar between study arms except for women reporting drinking or taking drugs during sex in the 6 months prior to screening (26/178=14.6% of vaccinees and 14/182=7.7% of placebo recipients, p = 0.04). Our cohort was predominantly heterosexual with 457/801=57.1% of participants reporting unprotected vaginal or anal sex as a risk factor. Men were more likely than women to report more than one sexual partner (247/441=56.0% vs 56/360=15.6%, respectively, p<0.001) and to report having a casual or anonymous partner (209/441=47.4% vs 40/360=11.1%, p<0.001). Most participants (304/441=68.9% of men and 302/360=83.9% of women) had a main partner, with 349/606=57.6% of those with a main partner regularly living apart. Less than 10% of both men and women reported having an HIV positive partner, exchanging sex for money or gifts, being forced to have sex, or being diagnosed with an STI and few women reported heavy drinking (Table 1). In total, 216 (27.0%) received one study injection, 529 (66.0%) two injections, and 56 (7.0%) three injections. There were no differences between the treatment arms in the number of injections received or in reasons for discontinuation of vaccination. Overall, 63 (7.9%) participants withdrew from the study by 31 August 2009 (Figure 1), the proportion being similar between treatment arms. Post unblinding, more participants in the placebo vs vaccine arm (n=13 vs 6) refused further participation.

Vaccines were well tolerated. Adverse events, local and systemic reactogenicity and in Table 2 and pregnancies are reported in web appendix 3. Self reported STI rates were similar between vaccine and placebo arms over the entire study period (7.5 vs 9.8 per 100 person-years) as well as over the pre- and post-unblinding periods (9.9 vs 9.6 and 7.7 vs 9.9 per 100 person-years). Women were more likely than men to report an STI (11.1 vs 6.7 per 100 person-years, p=0.007).

At four weeks following the 2^nd vaccination, 83/93=89.2% of vaccinees had developed an IFN-γ secreting T cell response to Clade B peptides and 72/93=77.4% to Clade C peptides (Table 3). For Clade B, Gag-specific responses were highest (74/93=79.6%); for Clade C, Pol-specific responses were highest (65/93=69.9%). All Clade C responders also had a Clade B response. Among responders to both Clades B and C, the overall magnitude of response to the Clade B vaccine-matched panel was significantly higher than to Clade C PTE panel; the same pattern held for responses to individual antigens (Table 3). Although not statistically different, vaccinees seronegative for Ad5 (titre ≤18) at baseline tended to demonstrate consistently higher response rates for both clades, overall and by gene, than those Ad5 seropositive (Table 3). Among responders, the Ad5 seronegative vaccinees also had a greater magnitude of response than those Ad5 seropositive for both clades overall (p = 0.004 for B and 0.007 for C), for Clade B Nef and Pol (p = 0.049 and = 0.002), and Clade C Pol (p = 0.01).

There was no evidence of VE (Figure 2): the treatment HR adjusted for gender was 1.25 (95% CI: 0.76, 2.05). Accounting for the time at which participants were notified of their treatment assignment did not alter this finding: the HR adjusted for unblinding time was 1.15 (95% CI: 0.69, 1.90). The treatment HR was higher in the 6 months (26 weeks) after first vaccination, but this finding was not significant as can be seen in Web appendix Figure 1 which illustrates the log instantaneous HR for treatment over time since first vaccination. HIV incidence was similar between vaccine and placebo arms (4.54 vs 3.70/100 person-years, n=34 vs 28, Web appendix 1). The majority of these infections were among women (n=42, Figure 2) with high HIV-1 incidence rates both in the vaccine arm (6.79) and placebo arm (5.86).

In multivariate Cox proportional hazard models, gender, age, HSV-2 and the interaction of gender and HSV-2 were significant predictors of HIV-1 infection (Table 4). Baseline Ad5 titre, categorized using either 18 or 200 as a cut-point, was not a significant predictor of HIV-1 infection and adjusting for Ad5 had little effect on the treatment HR; (HR= 1.24 – 1.31, depending on model). The number of vaccinations and behavioural risk factors were also not significant predictors of infection. Ad5 titre, gender, age or HSV-2 did not modify the treatment HR, nor did the effect of age and Ad5 titre differ by gender.

HSV-2 increased the risk of HIV-1 in men but not in women (Table 4). For men, HSV-2 increased the risk of infection more than 5-fold (5.23, 95% CI: 2.09, 13.10). For those HSV-2 negative at baseline, women had an increased risk of HIV-1 compared to men (HR 4.60; 2.16, 9.79). In an analysis restricted to women, there were no significant predictors of HIV-1 infection. In multivariable analysis for men (Web appendix Table 2), HSV-2 (HR 4.90, 95% CI 2.03, 11.80) and having a main partner but living apart regularly (HR 3.61, 95% CI 1.31, 9.99) were associated with HIV-1 infection; neither Ad5 titre nor circumcision were significant. HSV-2 status post-enrollment was not assessed; among 31 participants who acquired HIV-1 who were HSV-2 seronegative at enrollment, 22 either did not change HSV-2 status (19) or acquired HSV-2 prior to HIV-1 infection (3), and all nine remaining participants were identified as HSV-2 and HIV-1 positive at the same visit.

Viral-load setpoint was calculated for 61 of the HIV-1 infected participants. One woman on the vaccine arm was excluded from the viral-load setpoint analysis because of a six month delay in confirming infection. There was no significant difference in the distribution of viral load setpoint between the vaccine arm (geometric mean titre 20,483 copies/mL, N=33) and the placebo arm (34,032 copies/mL, N=28) (Figure 3, p=0.39, stratified by gender). Women tended to have lower viral-load setpoints than men (geometric mean 19,642 vs 45,438 copies/mL), which was not statistically significant (p = 0.15). Women who received vaccine (n=21) tended to have lower viral-load setpoints than women on the placebo arm (n=20) although this was not statistically significant (11,401 vs 34,773; p=0.16). Participants with baseline Ad5 titre > 18 (n=50) tended to have higher viral-load setpoint than those ≤ 18 (n=11), (31,924 vs 9,924), but this too, was not statistically significant (p =0.12).

Neither of the p-values for the co-primary endpoints of HIV-1 infection and viral-load setpoint were statistically significant, adjusted p’s=0.39. Seven infected participants (3 vaccine, 4 placebo) started ART, all after the 3^rd month visit post HIV-1 diagnosis. Two women with CD4 counts above 350 initiated ART to prevent mother-to-infant transmission and data from these women were censored at 3 months and 1 year, respectively. Treatment was not a significant predictor of CD4 decline to less than 350 overall or among men, but among women, vaccinees were at significantly lower risk of this event than placebos: HR=0.33 (95% CI: 0.12, 0.91) (Web appendix Figure 2). No other significant predictors of time to CD4 count less than 350 were found among baseline covariates considered: site, Ad5 titre, age quartiles, HSV-2 and their interactions with treatment.

Discussion

This study agrees with the Step findings that the MRKAd5 HIV-1 vaccine did not prevent HIV-1 infection or lower early viral load in either Ad5 sero-positive or -negative vaccinees. However, Phambili had limited ability to assess the vaccine due to the discontinuation of enrollment and vaccination, possible changes in risk behavior after unblinding, and the limited number of infections particularly those occurring close to vaccination. This lack of vaccine effect is despite the relatively high frequency and magnitude of HIV-1 specific T cell responses measured by INF-γ ELISPOT to both clade B and C antigens. Most vaccinees developed an immune response to both Clades B and C, although the responses were greater to the vaccine-matched Clade B peptides. Although the vaccine was immunogenic in this and previous studies ²⁶, the responses elicited did not translate to vaccine efficacy.

The high incidence of HIV-1 in women permitted the first evaluation of VE among heterosexual women and highlights the pressing need to find effective preventive interventions for women. There was no evidence of increased risk of HIV-1 infection among vaccinated participants compared to placebo. There was also no such evidence among Ad5 seropositive women or men or among uncircumcised men.

Although our results differ from Step, the premature interruption of the study, discontinuation of vaccinations, and unblinding may have affected risk behaviour, impacted on our statistical power to show these associations; or may have altered susceptibility to infection. In addition, the men in our study were predominantly heterosexual, therefore risk factors would have differed from the predominantly homosexual population studied in Step.

Prior HSV-2 infection has been associated with increased HIV-1 infection ^27–30 and was associated with increased HIV-1 acquisition among men who have sex with men in the Step trial. In Phambili, prior HSV-2 infection was associated with HIV-1 infection among heterosexual men but not women. The finding in women that HIV-1 risk was not associated with prevalent HSV-2 has been documented previously³¹. The lack of association between oral or injectable hormonal contraception use at baseline and HIV-1 acquisition is consistent with other studies. ³² In men, HSV-2 status and living apart from the main partner were associated with HIV infection, and interventions addressing these factors need to be studied.

The lower viral-load setpoint and slower CD4 decline observed in vaccinated women may indicate a vaccine effect on disease progression, although, because enrolment in the study was halted prematurely, this study was not powered to detect this effect. Pre-existing immunity to the vaccine vector may alter VE. Further studies are required to understand the interplay between vaccination and these factors, including gender. A study of sexual partners of Phambili participants to provide new insights on the natural history of HIV infection and transmission dynamics in this population has been initiated.

Stratification by sex should be considered in future vaccine efficacy trials, to examine the effects of HIV-1 vaccines on gender. Longer follow-up of incident infections is warranted to assess the effects of vaccination on disease progression.

Supplementary Material