Efficacy and safety of a mosaic HIV-1 vaccine regimen in men who have sex with men and transgender individuals (HVTN 706/HPX3002/Mosaico): a global, randomised, double-blind, placebo-controlled, phase 3 trial

Summary

Background

There is a high unmet need for effective HIV prevention options, including vaccines, for individuals at high risk of HIV acquisition who choose not to use pre-exposure prophylaxis or other prevention strategies. This study evaluated the efficacy and safety of an HIV-1 vaccine regimen consisting of tetravalent mosaic adenovirus serotype 26-based vaccine (Ad26.Mos4.HIV) and bivalent clade C glycoprotein (gp) 140–mosaic gp140 vaccine, in a population with high seroincidence.

Methods

This randomised, double-blind, phase 3 trial enrolled adult cisgender men and transgender individuals without HIV from 52 academic medical centres, health departments, and community-based clinics in Latin America, Europe, and the USA. Participants were randomly assigned (1:1) by use of a centrally prepared, computer-generated randomisation schedule to receive intramuscular injections of vaccine or saline placebo in stratified permuted blocks. Study participants, study site personnel (except for those with primary responsibility for study vaccine preparation and dispensing), and investigators were masked from vaccine allocation. The vaccine regimen consisted of Ad26.Mos4.HIV administered at months 0 and 3 followed by Ad26.Mos4.HIV administered concurrently with aluminium phosphate-adjuvanted clade C gp140–mosaic gp140 at months 6 and 12. The primary endpoint was vaccine efficacy in preventing HIV-1 acquisition between months 7 and 24 or between months 7 and 30 in the per-protocol population, which included all randomly assigned participants who received at least one study vaccination and who had not been diagnosed with HIV-1 4 weeks after the third vaccination, had received all planned vaccinations at the first three vaccination visits within the respective visit windows, and had no major protocol deviations linked to incorrect product administration. Safety outcomes were assessed in all randomised participants who received at least one study vaccination. The trial is registered with ClinicalTrials.gov (NCT03964415) and is complete.

Findings

Between Nov 4, 2019, and Aug 13, 2021, 3900 participants were enrolled and randomly assigned; 3887 received at least one study vaccination (1942 assigned to vaccine, 1945 to placebo). 3870 (99·6%) of 3887 participants were assigned male at birth, seven (0·2%) were assigned female, and one participant’s sex at birth was undifferentiated; 3557 (91·5%) participants identified as male gender, 48 (1·2%) as female gender, and 278 (7·2%) as transgender or non-binary. The per-protocol population included 1525 participants in the vaccine group and 1494 in the placebo group. From month 7 to month 24 in the per-protocol population, HIV-1 incidence per 100 person-years was 3·63 (95% CI 2·77 to 4·67) in the vaccine group and 3·35 (2·53 to 4·36) in the placebo group, with an estimated vaccine efficacy (months 7–24) of −0·7% (95% CI −50·9 to 32·8; p=0·97). Vaccine efficacy (months 7–30) was −149·1% (−737·7 to 26·0; p=0·14). Most solicited local and systemic adverse events were mild or moderate and short-lived. Medically attended adverse events occurred in 999 (51·4%) of 1942 participants given vaccine and 1002 (51·5%) of 1945 participants given placebo; serious adverse events occurred in 82 (4·2%) of 1942 participants given vaccine and 77 (4·0%) of 1945 participants given placebo. No fatal adverse events considered related to the study vaccine occurred. Adverse events of special interest (thrombotic events or thrombocytopenia) occurred in four participants with vaccine and two with placebo; none had thrombosis with thrombocytopenia syndrome.

Interpretation

The lack of efficacy in this and other HIV vaccine trials points to the importance of current efforts to develop vaccines that generate broadly neutralising antibodies.

Introduction

A key challenge in developing an effective HIV-1 vaccine is the high genetic diversity of HIV-1 across populations due to the high mutation rate and recombination potential during viral replication.¹ Only one HIV-1 vaccine trial, RV144, has shown statistically significant, but modest, vaccine efficacy (31·2% at 3·5 years) in reducing the risk of acquiring HIV;² this was insufficient for licensure. Subsequent trials have evaluated other vaccine regimens without demonstrating efficacy.³

Computationally designed mosaic antigens were developed to provide immunological coverage against various global HIV-1 strains.^4,5 The vaccine regimen evaluated in this study was derived and refined through a series of preclinical and early phase clinical studies.^6-12 The phase 1/2a ASCENT trial investigated a heterologous vaccine regimen consisting of a tetravalent adenovirus serotype 26 (Ad26)-based vaccine encoding two mosaic Env antigens and two mosaic Gag–Pol antigens (Ad26.Mos4.HIV) and a bivalent combination of an aluminium phosphate-adjuvanted clade C glycoprotein (gp) 140 and an aluminium phosphate-adjuvanted mosaic gp140.¹² The regimen increased the breadth, magnitude, and durability of binding antibodies to cross-clade panels of Env antigens compared with a single-valent clade C gp140 regimen, without compromising responses against clade C.^12,13 The magnitude and response rate of CD4 cell responses to Env were also statistically significantly higher with the bivalent clade C–mosaic gp140 regimen versus the single-valent clade C gp140 regimen.¹² The clinical safety and immunogenicity data from ASCENT provided the basis to advance the Ad26.Mos4.HIV and bivalent clade C–mosaic gp140 regimen to this phase 3 study (HVTN 706/HPX3002/Mosaico), which enrolled cisgender men who have sex with men and transgender individuals in Europe, the USA, and Latin America. Clade B is the predominant HIV-1 subtype in these regions.

A complementary study to Mosaico, the phase 2b HVTN 705/HPX2008/Imbokodo trial, evaluated a similar vaccine regimen (Ad26.Mos4.HIV/clade C gp140) in young women with increased likelihood of acquiring HIV in southern Africa, where clade C is most prevalent.^14,15 The vaccine regimen was safe and well tolerated, but did not show efficacy in preventing HIV-1. Imbokodo was ongoing at the time Mosaico completed recruitment and, although Imbokodo showed no benefit, the independent data and safety monitoring board (DSMB) for Mosaico advised the funders to continue the study. This recommendation was based on key differences in vaccine composition, study population, geographical region, circulating HIV-1 clades, use of pre-exposure prophylaxis (PrEP), and mode of transmission (mainly intravaginal vs intrarectal) between the two trials. We report the primary efficacy, safety, and immunogenicity results of the Mosaico study.

Methods

Study design and participants

This randomised, double-blind, placebo-controlled, phase 3 study was done at 52 sites, consisting of academic medical centres, health departments, and community-based clinics, in eight countries (Argentina, Brazil, Italy, Mexico, Peru, Poland, Spain, and the USA, including Puerto Rico). All sites complied with Johnson & Johnson’s policies for site selection and were required to have an acceptable volume of potentially eligible patients. Investigators were treating clinicians with appropriate expertise and experience. Investigators and study staff underwent study-specific training before trial initiation. The study enrolled cisgender men and transgender individuals aged 18–60 years without HIV-1 and at high risk for exposure to HIV who were having sex with cisgender men or transgender individuals, or both. Potential participants were first counselled about PrEP. Those who opted for PrEP were linked to low-cost or no-cost PrEP services and were not screened for the trial. Participants who did not want to take PrEP were offered study screening; those who met all eligibility criteria and agreed to join were enrolled. Full inclusion and exclusion criteria are provided in the study protocol (appendix pp 102-06). Gender and ethnicity data were self-reported by participants; options for gender are listed in the appendix (p 19). All participants received access to the highest standard of prevention, including, but not restricted to, HIV testing, HIV prevention counselling, provision of condoms and lubricants, and access to management of sexually transmitted infections. Each site had a protocol-approved PrEP plan that included access to low-cost or no-cost PrEP services, usually at the site itself. After enrolment access to PrEP was discussed at each clinic visit, and all participants who requested PrEP reported having been placed on it.

The study consisted of a 12-month vaccination period and a follow-up period of at least 18 months after the fourth vaccination (until month 30) in participants who had not acquired HIV-1 or up to 6 months after diagnosis in participants who acquired HIV-1.

This study adhered to the principles of the Declaration of Helsinki, Good Clinical Practice guidelines, and applicable regulatory and country-specific requirements. The protocol, protocol amendments, recruitment materials, and other relevant documents were approved by institutional review boards and ethics committees at each participating site (appendix p 12). No study participants or members of the public were involved in the design, conduct, and reporting of the trial. All participants provided written informed consent in a language that the potential participant could read and/or understand via consent forms approved by the sponsor and the reviewing institutional review board or ethics committee. The study was monitored regularly for safety and non-efficacy by an independent DSMB. Safety was also monitored by the HVTN 706/HPX3002/Mosaico Safety Review Team. This study is registered with ClinicalTrials.gov (NCT03964415) and is complete.

Randomisation and masking

Participants were randomly assigned 1:1 to receive either the vaccine regimen or placebo based on a computer-generated randomisation schedule prepared centrally by or under supervision of the sponsor. The randomisation was balanced using permuted blocks (size 4) and stratified by study site. A pharmacist at each site prepared all doses for administration and placed an overlay on all syringes to preserve masking. A masked, qualified health-care provider from the study site administered the vaccines. Study participants, study site personnel (except for those with primary responsibility for study vaccine preparation and dispensing), and investigators were masked to vaccine allocation until the end of the study. The sponsor and its partners were masked to study vaccine allocation until the primary analysis, when the last participant reached their month x visit. The timing of x was based on a minimum follow-up of 24 months but could be extended up to 30 months until the target number of HIV-1 acquisitions was reached to sufficiently power the primary analysis (appendix p 16). Site staff performed HIV testing within the study to avoid unmasking due to vaccine-induced seropositivity (VISP) elicited by the vaccine. A prespecified diagnostic algorithm was used to distinguish HIV-1 acquisition from VISP.

Procedures

Participants received Ad26.Mos4.HIV or placebo administered intramuscularly at months 0 and 3, followed by either Ad26.Mos4.HIV administered concurrently with aluminium phosphate-adjuvanted clade C gp140 and mosaic gp140 or saline placebo at months 6 and 12. The study vaccine was given to participants by a qualified health-care provider from the study site. Additional details on study vaccinations are provided in the appendix (p 14).

We observed participants at the study site for at least 30 min after each vaccination and requested that participants record solicited signs and symptoms in a diary for 7 days post-vaccination. Unsolicited adverse events were recorded up to and including 28 days after each vaccination. Adverse events judged to be of special interest (ie, because of clinical importance, known or suspected class effects, or based on non-clinical signals; thrombotic events and thrombocytopenia, or both) were collected up to and including 6 months after the last vaccination. Serious adverse events, medically attended adverse events (including sexually transmitted infections), and adverse events leading to discontinuation were collected until the end of the study.

We tested participants for HIV (using a test that distinguished from VISP) at months 0, 3, 6, 7, 9, 12, 13, 15, and every 3 months thereafter in a central laboratory. Participants were advised to refrain from receiving any HIV testing outside of the study protocol due to VISP concerns. Participants diagnosed with HIV via a confirmed HIV test remained in the study and were followed up for 6 months, but no further scheduled vaccinations were administered. Additional details on HIV testing are provided in the appendix (p 14).

To assess the effect of PrEP or post-exposure prophylaxis (PEP) use, we collected data from multiple sources, including questionnaires completed by study participants, concomitant medications, and dried blood spot (DBS) samples for quantitative assessment of drug concentrations. DBS samples were drawn from participants present on site on the first or 15th day of each month to measure PrEP or PEP uptake, which provided a random subsample of participants and timepoints. DBS analysis methods are described in the appendix (p 14).

Blood samples collected to assess pre-existing neutralising antibody (NAb) responses to the Ad26 vector at baseline were quantified in a virus neutralisation assay. Vaccine-induced binding antibody responses were evaluated using an HIV-1 Env strain-specific ELISA for the mosaic 1 (Mos1) gp140 antigen (additional details provided in the appendix, pp 14-15).

Because of the COVID-19 pandemic, screening activities, enrolment of new participants, and further vaccinations were temporarily paused between March 17, and June 1, 2020; safety follow-up visits were conducted remotely during the temporary pause. Study participants and staff adhered to all non-pharmaceutical regulations for medicine clinic visits during the COVID-19 pandemic (appendix p 14).

Outcomes

Analyses were done in different study populations (panel and appendix p 14). The primary endpoint was vaccine efficacy, defined as 1 minus the cumulative incidence ratio (vaccine vs placebo), in preventing HIV-1 acquisition between month 7 and month x (with month 24≤ × ≤month 30) after the first vaccination in the per-protocol population.

Panel: Study populations.

Full analysis set (FAS)

All randomly assigned participants who received at least one study vaccination.

Modified intention-to-treat (mITT) cohort

Participants in the FAS who had not been diagnosed with HIV-1 at the date of the first vaccination.

Modified intention-to-treat-2 (mITT-2) cohort

Participants in the FAS who had not been diagnosed with HIV-1 4 weeks after the third vaccination visit.

Modified intention-to-treat-3 (mITT-3) cohort

Participants in the FAS who had not been diagnosed with HIV-1 4 weeks after the third vaccination visit, who received all planned vaccinations at the first three vaccination visits regardless of whether the vaccinations were within the protocol-specified visit windows, and who had no major protocol deviations linked to incorrect product administration.

Immunogenicity analysis set (IAS)

A subset of participants from the four Latin American countries that included cases (participants diagnosed with HIV-1) and controls (participants not diagnosed with HIV-1, excluding participants with protocol deviations judged to affect immunological responses). Participants were from the mITT-3 cohort. Participants in this cohort were selected from an interim database (data cutoff March 29, 2023), while the study was still running. As a result, in the final database, some participants who were initially selected as controls became cases.

Per-protocol (PP) cohort (primary efficacy analysis)

Participants in the FAS who had not been diagnosed with HIV-1 4 weeks after the third vaccination visit, who received all planned vaccinations at the first three vaccination visits within the respective visit windows, and who had no major protocol deviations linked to incorrect product administration.

Full immunisation set (FIS)

Participants in the FAS who had not been diagnosed with HIV-1 4 weeks after the fourth vaccination visit and who received all planned vaccinations within the respective visit windows.

Secondary efficacy endpoints were vaccine efficacy in preventing HIV-1 over different time intervals (eg, 0 – x months, 13 – x months) and in different populations (appendix pp 195-96), and vaccine efficacy by country. The different populations for secondary efficacy analyses are modified intention-to-treat (mITT; participants in the full analysis set who were HIV-1 negative at the date of first vaccination), mITT-2 (participants in the full analysis set who had a negative HIV test 4 weeks after the third vaccination visit), mITT-3 (participants in mITT-2 who received all planned vaccinations at the first three vaccination visits regardless of whether the vaccinations were within the visit windows; participants with major protocol deviations were excluded), and the full immunisation set (participants who were HIV-1 negative 4 weeks after the fourth vaccination visit who received all vaccinations within the respective windows).

Secondary safety endpoints were the incidence of solicited local and systemic adverse events, unsolicited adverse events, adverse events of special interest, medically attended adverse events, serious adverse events, and discontinuations from the study or vaccination due to adverse events. Safety analyses were done in the full analysis set. Exploratory endpoints included assessment of perceived treatment allocation collected through questionnaires completed by study participants (full analysis set), assessment of pre-existing NAb responses to the Ad26 vector at baseline, vaccine-induced binding antibody responses (immunogenicity analysis set), and frequency of confirmed VISP (VISP analysis set).

Statistical analysis

We selected a sample size of 3800 (1900 each in the vaccine and placebo groups) to provide 90% power to reject a null hypothesis of vaccine efficacy (7–x months) of ≤20% (with month 24≤ x ≤30), under an alternative vaccine efficacy of 65% and with a one-sided error rate of 2·5%. The annual HIV-1 incidence rate in the placebo group was assumed to be 2·0%. Additional details are provided in the appendix (pp 15-16).

For the primary analysis, first the treatment group-specific cumulative incidence between months 7 and x (vaccine efficacy [7–x]) was estimated using the transformed Nelson–Aalen method. Next, vaccine efficacy was calculated as 1 minus the cumulative incidence ratio (vaccine vs placebo), and the Wald CIs were constructed for the log-cumulative incidence ratio and then mapped into a CI for vaccine efficacy. The primary endpoint would be achieved if the lower bound of the 95% CI for vaccine efficacy (7–x) was greater than 20% at the primary analysis. To assess potential time effects of vaccine efficacy, the Nelson–Aalen method was used to plot the estimated cumulative incidence rates over time for the vaccine and placebo groups. Secondary vaccine efficacy analyses were assessed with the same methods as those used for the primary endpoint.

Unsolicited adverse events, adverse events of special interest, medically attended adverse events, and serious adverse events were summarised using Medical Dictionary for Regulatory Activities version 26.0 system organ class and preferred terms. Solicited adverse events were summarised by class (administration site, systemic) and preferred term. Additional details are provided in the appendix (p 17). Formal statistical testing comparing adverse events in the vaccine and placebo groups was not done.

Descriptive statistics, including geometric means with their corresponding 95% CIs, were calculated for the immunogenicity assessments (immunogenicity analysis set). A binding antibody response to Mos1 gp140 was defined as at least a three-fold increase from baseline or a post-baseline value greater than or equal to the limit of detection if the sample was below the limit of detection or missing at baseline. Both Pearson correlation and ordinary least squares regression between baseline Ad26 NAb titres and binding antibody response at month 7 were assessed. Regression assumptions were checked using a quantile–quantile plot and residuals scatter plot.

All statistical analyses were done with SAS version 9.4. Data are presented from the database lock on Sept 22, 2023, and March 20, 2024, for the DBS analysis. The statistical analysis plan is published at the end of the appendix.

Role of the funding sources

Johnson & Johnson and the HIV Vaccine Trials Network contributed to study design; collection, analysis, and interpretation of data; and writing of the report. All funders contributed to the decision to submit the paper for publication. Johnson & Johnson provided the vaccine.

Results

The study enrolled participants from Nov 4, 2019, to Aug 13, 2021. Of 4976 participants screened, 3900 were randomly assigned to treatment and 3887 received one or more doses of study vaccine (figure 1). The full analysis set included 1942 participants assigned to receive vaccine and 1945 to receive placebo. Due to out-of-window vaccinations related to COVID-19, recruitment was increased by 5%. Overall, 1715 (88·3%) of 1942 participants in the vaccine group and 1708 (87·8%) of 1945 participants in the placebo group received all four study vaccinations, including those who received vaccinations both within and outside of the study windows (appendix p 18). Two (0·1%) of 1942 participants in the vaccine group discontinued study treatment for reasons related to COVID-19; the effect of COVID-19 on the study was minimal. The per-protocol population included 1525 participants in the vaccine group and 1494 in the placebo group.

Figure 1: Trial profile.

MedDRA=Medical Dictionary for Regulatory Activities. mITT=modified intention-to-treat. *Six additional participants were randomly assigned but did not receive any vaccine doses. †Seven additional participants were randomly assigned but did not receive any placebo doses. ‡Adverse events that were considered COVID-19 related (associated) were based on events that coded to a COVID-19 MedDRA term. §Participants can be in more than one category. ¶Includes nine participants in the vaccine group and seven in the placebo group who completed the fourth vaccination but might have missed an earlier vaccination.

At baseline, mean age was 29·8 years (SD 8·4), and most participants were assigned male at birth, were either White or of multiple race, and were of Hispanic or Latino ethnicity (table 1; appendix p 19). In the USA (excluding Puerto Rico), 51 (27·0%) of 189 participants were of Hispanic or Latino ethnicity. Most participants identified as male gender (3557 [91·5%] of 3887); 278 (7·2%) of 3887 participants were transgender or non-binary. Demographic characteristics were similar between the vaccine and placebo groups.

Table 1:

Baseline participant demographic and clinical characteristics (full analysis set)

	Vaccine (n=1942)	Placebo (n=1945)
Age, years	29·7 (8·3)	29·9 (8·6)
Region
Europe	231 (11·9%)	234 (12·0%)
Latin America	1614 (83·1%)	1609 (82·7%)
USA	97 (5·0%)	102 (5·2%)
Sex at birth
Male	1933 (99·5%)	1937 (99·6%)
Female	9 (0·5%)	7 (0·4%)
Undifferentiated	0	1 (0·1%)
Participant-identified gender at baseline
Male	1768 (91·0%)	1789 (92·0%)
Transgender female	102 (5·3%)	94 (4·8%)
Female	26 (1·3%)	22 (1·1%)
Transgender male	8 (0·4%)	5 (0·3%)
Other*	38 (2·0%)	35 (1·8%)
Sexual orientation
Gay, lesbian, or homosexual	1539 (79·2%)	1546 (79·5%)
Bisexual	281 (14·5%)	272 (14·0%)
Straight or heterosexual	66 (3·4%)	75 (3·9%)
Additional category†	25 (1·3%)	26 (1·3%)
Queer	19 (1·0%)	12 (0·6%)
Prefer not to answer	7 (0·4%)	7 (0·4%)
Not sure	5 (0·3%)	7 (0·4%)
Race‡
White	868 (44·7%)	862 (44·3%)
Multiple§	814 (41·9%)	809 (41·6%)
Black or African American	140 (7·2%)	161 (8·3%)
American Indian or Alaska Native	63 (3·2%)	59 (3·0%)
Unknown	24 (1·2%)	21 (1·1%)
Not reported	17 (0·9%)	13 (0·7%)
Asian	13 (0·7%)	12 (0·6%)
Native Hawaiian or other Pacific Islander	3 (0·2%)	7 (0·4%)
Ethnicity
Hispanic or Latino	1690 (87·0%)	1675 (86·1%)
Not Hispanic or Latino	238 (12·3%)	246 (12·6%)
Not reported	14 (0·7%)	24 (1·2%)
BMI, kg/m2	26·0 (4·9)	25·8 (4·7)
Positive for sexually transmitted infection
Chlamydia¶	214 (11·0%)	206 (10·6%)
Gonorrhoea‖	158 (8·1%)	180 (9·3%)
Syphilis**††	177 (9·1%)	177 (9·1%)
Sexual activity, number of partners within the last month
Insertive anal or vaginal sex without a condom§§	1·0 (1·0–3·0)	1·0 (1·0–3·0)
Receptive anal sex without a condom¶¶	1·0 (0·0–2·0)	1·0 (0·0–2·0)

Data are mean (SD), n (%), or median (IQR). Percentages may not total 100% due to rounding. For participants who switched site, site and country at baseline were used for analysis.

^*Includes gender variant or gender non-conforming, gender queer, self-identify (including, but not limited to, crossdresser, gender fluid, male or non-binary, non-binary, non-binary transgender, non-binary transvestite, non-compliance, sissy, and X), and prefer not to answer.

^†Includes androsexual, asexual, cis-men oriented, homoflexible, modern, non-binary, pan, pansexual, trans male, and transsexual.

^‡n=1944 for the placebo group.

^§If multiple race categories were indicated, the race was recorded as multiple.

^¶n=1941 for the vaccine group and n=1942 for the placebo group.

^‖n=1941 for the vaccine and placebo groups.

^**Participants were categorised as having syphilis if both their treponemal and non-treponemal tests were positive.

^††n=1939 for the vaccine and placebo groups.

^§§n=1522 for the vaccine group and n=1515 for the placebo group.

^¶¶n=1397 for the vaccine group and n=1353 for the placebo group.

The decision to unblind the study occurred on Jan 18, 2023, after the recommendation by the DSMB to the Governance Committee when study stopping rules due to a lack of efficacy were met (appendix p 20). Efficacy analyses were censored for HIV-1 acquisitions that occurred after study unblinding. Reasons for censoring are summarised in the appendix (p 21). As the study was terminated early, 2890 (74·3%) of 3887 participants were at risk of HIV-1 acquisition at month 12, whereas 578 (14·9%) of 3887 participants were at risk at month 24 (appendix p 26). Between months 7 and 24 in the per-protocol cohort, the estimated incidence rate of HIV-1 acquisition was 3·63 (95% CI 2·77–4·67) per 100 person-years in the vaccine group and 3·35 (2·53–4·36) per 100 person-years in the placebo group (figure 2, table 2). In the per-protocol cohort, the estimated vaccine efficacy (months 7–24) was −0·7% (95% CI −50·9 to 32·8; p=0·97) and the estimated vaccine efficacy (months 7–30) was −149·1% (−737·7 to 26·0; p=0·14; table 2). The estimated vaccine efficacy (months 7–30) was primarily driven by a single acquisition of HIV-1 that occurred late in the follow-up period among the four participants remaining at risk in the vaccine group (appendix p 26). The protocol-defined criteria to show efficacy were not met, as the lower boundary of the 95% CI for vaccine efficacy between months 7 and 24 or 30 in the per-protocol cohort was not more than 20%. Secondary analyses of vaccine efficacy across different time intervals and analysis populations (mITT, mITT-2, mITT-3, and full immunisation set) also showed that the lower bound of the 95% CI for vaccine efficacy was not more than 20% (table 2).

Figure 2: Cumulative incidence of HIV-1 over 7–24 months* (per-protocol cohort).

When primary analysis was done at month x (between months 24 and 30), the times at risk of participants with a last HIV-1 negative test at or after the month x visit was right-censored at the right edge of the month x visit window. For the month 24 visit, the upper visit window is 24·84. *Censored at study unblinding date.

Table 2:

Incidence of HIV-1 and point estimates of vaccine efficacy, excluding acquisitions that occurred after study unblinding^*

	Vaccine				Placebo				Vaccine efficacy point estimate, % (95% CI)	p value†
	Number of HIV-1 acquisitions, n/N	Total person-years of follow-up	Median (IQR) total person- years of follow-up	Incidence rate per 100 person- years, % (95% CI)	Number of HIV-1 acquisitions, n/N	Total person-years of follow-up	Median (IQR) total person- years of follow-up	Incidence rate per 100 person- years, % (95% CI)
Primary efficacy endpoint
Per-protocol cohort, months 7–24	60/1525	1653·7	1·7 (1·5–1·9)	3·6% (2·8–4·7)	55/1494	1640·5	1·7 (1·5–1·9)	3·4% (2·5–4·4)	−0·7% (−50·9 to 32·8)	0·97
Per-protocol cohort, months 7–30	62/1525	1688·8	1·7 (1·5–1·9)	3·7% (2·8–4·7)	56/1494	1673·5	1·7 (1·5–1·9)	3·3% (2·5–4·3)	−149·1% (−737·7 to 26·0)	0·14
Secondary efficacy endpoints
mITT cohort, months 0–24	119/1940	3044·0	1·7 (1·4–1·9)	3·9% (3·2–4·7)	119/1938	3037·5	1·7 (1·4–1·9)	3·9% (3·2–4·7)	4·4% (−25·6 to 27·2)	0·75
mITT cohort, months 0–30	121/1940	3084·4	1·7 (1·4–1·9)	3·9% (3·3–4·7)	120/1938	3077·9	1·7 (1·4–1·9)	3·9% (3·2–4·7)	−59·8% (−265·6 to 30·1)	0·27
mITT-2 cohort, months 7–24	70/1812	1953·9	1·7 (1·5–1·9)	3·6% (2·8–4·5)	67/1796	1953·3	1·7 (1·5–1·9)	3·4% (2·7–4·4)	3·4% (−40·2 to 33·4)	0·86
mITT-2 cohort, months 7–30	72/1812	1994·3	1·7 (1·5–1·9)	3·6% (2·8–4·6)	68/1796	1993·6	1·7 (1·5–1·9)	3·4% (2·6–4·3)	−91·8% (−440·6 to 31·9)	0·22
mITT-3 cohort, months 7–24	70/1793	1939·6	1·7 (1·5–1·9)	3·6% (2·8–4·6)	67/1775	1934·9	1·7 (1·5–1·9)	3·5% (2·7–4·4)	3·6% (−39·8 to 33·6)	0·85
mITT-3 cohort, months 7–30	72/1793	1979·7	1·7 (1·5–1·9)	3·6% (2·8–4·6)	68/1775	1974·0	1·7 (1·5–1·9)	3·4% (2·7–4·4)	−90·0% (−434·5 to 32·5)	0·22
Full immunisation set, months 13–24	17/1383	891·5	1·7 (1·5–2·0)	1·9% (1·1–3·1)	19/1376	881·7	1·7 (1·5–2·0)	2·2% (1·3–3·4)	17·7% (−70·3 to 60·3)	0·60
Full immunisation set, months 13–30	19/1383	926·3	1·7 (1·5–2·0)	2·1% (1·2–3·2)	20/1376	914·2	1·7 (1·5–2·0)	2·2% (1·3–3·4)	−238·2% (−1509·2 to 28·9)	0·13

mITT=modified intention-to-treat.

^*Censored at unblinding date.

^†p values represent two-sided Wald test (α=0·025 level) of the equality of log cumulative hazard functions for the vaccine group and the placebo group.

The highest incidence of HIV-1 was in Latin American countries (appendix p 27), which corresponded to the countries with the highest enrolment. In Brazil, the incidence rate was higher in the vaccine group than in the placebo group: 14 of 332 HIV-1 acquisitions in the vaccine group (incidence rate 3·99 per 100 person-years during months 7–24, 95% CI 2·18–6·69) versus three of 342 HIV-1 acquisitions in the placebo group (0·82, 0·17–2·39). In Peru, the rate was slightly lower in the vaccine group (31 of 606 HIV-1 acquisitions, incidence rate 4·74 [3·22–6·72] per 100 person-years during months 7–24) than in the placebo group (40 of 573 HIV-1 acquisitions, 6·44 [4·60–8·76]).

PrEP adoption in the per-protocol cohort increased throughout the study from 1·3% (17 of 1279) in the vaccine group and 2·0% (25 of 1229) in the placebo group at months 0–3 to 10·3% (72 of 698) and 12·7% (89 of 699) at months 19–21 according to questionnaire responses (appendix p 28) and from 1·0% (15 of 1525) and 1·5% (22 of 1494) to 10·7% (154 of 1444) and 12·8% (182 of 1425) on the basis of medication review (appendix p 29). PEP use based on concomitant medications increased through months 4–6, remained relatively consistent through months 16–18, and decreased thereafter (appendix p 29). The percentage of participants with any PrEP or PEP use (according to questionnaire responses, concomitant medications, or DBS analysis) from months 7 to 24 in the per-protocol cohort was 19·6% (299 of 1525) in the vaccine group and 22·4% (334 of 1494) in the placebo group (appendix p 22). Among Latin American countries, participants in Peru had the lowest PrEP or PEP use from months 7 to 24 (38 [6%] of 606 in the vaccine group vs 53 [9%] of 573 in the placebo group), whereas participants in Brazil had the highest PrEP or PEP use (111 [33%] of 332 vs 124 [36%] of 342).

Of the 3282 DBS samples analysed in the per-protocol cohort, 200 (12·1%) of 1649 samples in the vaccine group and 217 (13·3%) of 1633 samples in the placebo group had detectable tenofovir diphosphate; 106 (6·4%) of 1649 samples in the vaccine group and 115 (7·0%) of 1633 samples in the placebo group had effective tenofovir diphosphate concentrations (appendix p 23). The percentage of participants with detectable tenofovir diphosphate generally increased over time in the vaccine and placebo groups (months 0–3: 8% [35 of 417] in the vaccine group vs 8% [33 of 427] in the placebo group; months 16–18: 14% [24 of 176] vs 16% [27 of 174]; appendix p 30). A similar trend was observed for effective tenofovir diphosphate concentrations (months 0–3: 3% [13 of 417] in the vaccine group vs 4% [15 of 427] in the placebo group; months 16–18: 9% [16 of 176] vs 8% [14 of 174]; appendix p 30). From month 19 onwards, there was an inadequate number of samples available for evaluation of detectable or effective tenofovir diphosphate concentrations, making estimates less precise.

When questioned about perceived treatment allocation, 1736 (51·6%) of 3364 participants in the full analysis set responded “active” or “placebo” at month 12. Of these respondents, 1265 (72·9%) correctly identified their assigned treatment. Other responses were “don’t know”, chosen by 1575 (46·8%) of 3364 participants, and “prefer not to answer”, chosen by 53 (1·6%) participants.

In the full analysis set, solicited local adverse events were reported after any vaccination by 1520 (78·3%) of 1942 participants in the vaccine group and 700 (36·0%) of 1945 participants in the placebo group (table 3). The most frequent solicited local adverse event in both groups was vaccination-site pain (appendix pp 31-32). Solicited systemic adverse events were reported by 1541 (79·4%) participants in the vaccine group and 1196 (61·5%) participants in the placebo group after any vaccination (table 3). The most frequently reported solicited systemic adverse events observed in more than 40% of participants in either group were fatigue, headache, myalgia, chills, and arthralgia (appendix pp 33-34). Most solicited local and systemic adverse events were of grade 1 or 2 in severity and short-lived.

Table 3:

Solicited and unsolicited adverse events (full analysis set)

	Vaccine (n=1942)	Placebo (n=1945)
Solicited local adverse events following any study injection	1520 (78·3%)	700 (36·0%)
Vaccination-site pain	1519 (78·2%)	695 (35·7%)
Vaccination-site swelling	89 (4·6%)	9 (0·5%)
Vaccination-site erythema	47 (2·4%)	7 (0·4%)
Solicited systemic adverse events after any study injection	1541 (79·4%)	1196 (61·5%)
Fatigue	1323 (68·1%)	928 (47·7%)
Headache	1181 (60·8%)	864 (44·4%)
Myalgia	1124 (57·9%)	563 (28·9%)
Chills	858 (44·2%)	343 (17·6%)
Arthralgia	806 (41·5%)	381 (19·6%)
Nausea	521 (26·8%)	362 (18·6%)
Pyrexia	400 (20·6%)	138 (7·1%)
Vomiting	128 (6·6%)	86 (4·4%)
Unsolicited adverse events	708 (36·5%)	716 (36·8%)
Grade 1	298 (15·3%)	313 (16·1%)
Grade 2	366 (18·8%)	352 (18·1%)
Grade 3	23 (1·2%)	20 (1·0%)
Grade 4	2 (0·1%)	1 (0·1%)
Grade 5 (fatal)	4 (0·2%)*	6 (0·3%)†
Adverse events related to vaccine or placebo	84 (4·3%)	71 (3·7%)
Adverse events leading to vaccination discontinuation	6 (0·3%)	3 (0·2%)
Serious adverse events	82 (4·2%)	77 (4·0%)
Related serious adverse events	2 (0·1%)	1 (0·1%)
Adverse events of special interest	4 (0·2%)‡	2 (0·1%)§
Medically attended adverse events	999 (51·4%)	1002 (51·5%)

Data are participants with events (%).

^*Fatal adverse events in the vaccine group included cardiac arrest, cerebral haemorrhage, completed suicide, and COVID-19.

^†Fatal adverse events in the placebo group included multiple injuries, traumatic lung injury, asphyxia, septic shock, death (unknown cause), and head and neck cancer.

^‡Adverse events of special interest in the vaccine group included pulmonary embolism (two participants), thrombocytopenia, and venous thrombosis.

^§Adverse events of special interest in the placebo group included thrombocytopenia and deep vein thrombosis.

Unsolicited adverse events after any vaccination were reported in 708 (36·5%) participants in the vaccine group and 716 (36·8%) participants in the placebo group (table 3). Most unsolicited adverse events were grade 1 or 2 in severity. Serious adverse events were reported in 82 (4·2%) participants in the vaccine group and 77 (4·0%) participants in the placebo group. Two serious adverse events in the vaccine group (rhabdomyolysis and pulmonary embolism) and one in the placebo group (pericarditis) were considered related to study vaccination by the investigator. Four participants in the vaccine group and six in the placebo group died; no death was considered related to study vaccination. Adverse events of special interest were reported in four participants in the vaccine group (pulmonary embolism [two participants], thrombocytopenia, venous thrombosis) and two in the placebo group (thrombocytopenia, deep vein thrombosis); none of these events qualified for further assessment for thrombosis with thrombocytopenia syndrome. No cases of thrombosis with thrombocytopenia syndrome were observed. The frequencies of chlamydia and gonorrhoea infections were similar in vaccine and placebo groups and were consistent over the course of the study (appendix p 24).

A summary of participants in the immunogenicity analysis set is provided in the appendix (p 25; sampling design for the immunogenicity analysis set population is described in the appendix on p 15). A total of 204 (56%) of 364 participants in the vaccine group and ten (44%) of 23 participants in the placebo group were seropositive for Ad26 NAbs at baseline (appendix p 35). Within the vaccine group, the percentage of seropositive participants was similar in the case (participants who acquired HIV-1) and control (participants without HIV-1) groups: 44 (63%, 95% CI 51–74) of 70 cases versus 60 (54%, 49–60) of 294 controls.

Vaccine-matched gp140 Mos1 binding antibody geometric mean concentrations among participants in the immunogenicity analysis set increased from 57·0 ELISA units (EU)/mL (95% CI 54.7–59.4) at baseline to 22 574·3 EU/mL (20 235.9–25 183.1) at month 7 in the vaccine group (figure 3). The magnitude of response in the placebo group was similar at baseline and month 7. Among vaccine recipients in the case–control analysis with available ELISA data, the percentage of responders at month 7 was greater than 99% (323 of 324) overall, 98% (62 of 63) for the case group, and 100% (261 of 261) for the control group. The geometric mean concentrations at month 7 in the vaccine group were similar for cases and controls.

Figure 3: Binding antibody concentrations against gp140 Mos1 antigen (immunogenicity analysis set).

Values of n are number of participants with available data. Solid horizontal lines show the GMC, the error bars show the 95% CI, and the dots and triangles represent individual participants. The LLOQ is 100 EU/mL, LOD is 116 EU/mL, and the ULOQ is 1 869 824 EU/mL. Values below the LLOQ were imputed as LLOQ/2 when calculating geometric mean. Samples below the LLOQ are shown as imputed values, samples above or equal to the LLOQ are shown as actual values. A responder was defined if the sample was below the LOD or missing at baseline and post-baseline values were above or equal to the LOD, or if the sample was above or equal to the LOD at baseline and the post-baseline value increased at least three-fold from baseline. EU=ELISA units. GMC=geometric mean concentration. gp140=glycoprotein 140. LLOQ=lower limit of quantification. LOD=limit of detection. ULOQ=upper limit of quantification.

Participants who were Ad26 seronegative at baseline had higher binding antibody response magnitude at month 7 versus those who were Ad26 seropositive (32 001 EU/mL vs 17 713 EU/mL; appendix p 36). Among participants who were Ad26 seropositive at baseline, there was weak negative correlation between binding antibody responses at month 7 and baseline Ad26 NAb titres (slope −0·16; Pearson correlation coefficient −0·224, p=0·0028; appendix p 36).

In the VISP analysis set, the percentage of participants with VISP, as shown by the ARCHITECT HIV Ag/Ab Combo assay (Abbott Laboratories, Abbott Park, IL, USA), was 15·8% (277 of 1757) after the first vaccination (month 3), 82·6% (1424 of 1723) following the second vaccination (month 6), and reached 99·6% (1687 of 1693) by the third vaccination (month 7; appendix p 38). VISP remained near 100% for the duration of the study among participants.

Discussion

The heterologous vaccine regimen of Ad26.Mos4.HIV complemented with a bivalent clade C and mosaic gp140 did not show efficacy in preventing HIV-1 in a diverse population of cisgender men and transgender individuals who have sex with cisgender men and transgender individuals, enrolled in the Americas and Europe. The lack of efficacy observed in the Mosaico study follows the Imbokodo study, which found that a similar HIV-1 vaccine regimen (Ad26.Mos4.HIV plus clade C gp140) did not protect against HIV-1 acquisition in a population of young women in southern and eastern Africa where clade C is more prevalent (vs clade B in Mosaico) and transmission is primarily intravaginal (vs mainly intrarectal in Mosaico). The regimen used in Mosaico included a bivalent gp140 boost that provided greater binding antibody and T-cell responses to clade B envelope than that used in Imbokodo, forming the basis for initiating the trial. These enhanced immune responses were insufficient for the prevention of HIV acquisition.

The high HIV incidence observed in Mosaico highlights a major need for a vaccine and other biomedical interventions to reduce HIV acquisition in cisgender men and transgender individuals, as there is for African heterosexual women. We offered to link participants to PrEP available in their jurisdiction before and during the trial; only those declining oral PrEP before entering were enrolled. Approximately 25% of participants used oral PrEP during the study after extensive counselling. This approach of enrolling participants who opted not to use a highly effective prevention method and linking these participants to PrEP or PEP services illustrates the dilemma and realities of HIV prevention modalities—that many cisgender men and transgender individuals with a high likelihood of HIV acquisition desire alternative choices for HIV prevention.

Despite somewhat higher point estimates of PrEP use over time in the placebo group, subgroup analyses of vaccine efficacy by PrEP or PEP use showed no consistent differences from the main efficacy populations or between vaccine and placebo groups (data not shown). Additionally, analysis of vaccine efficacy by country showed high variability, which might in part be explained by differences in PrEP use by country.

The Mosaico vaccine regimen was generally safe, well tolerated, and had an acceptable reactogenicity profile. The solicited local and systemic adverse event profiles were consistent with those reported in the phase 1/2a ASCENT trial.¹² Although rare events of thrombosis with thrombocytopenia syndrome were observed with the Ad26.COV2.S vaccine against SARS-CoV-2,¹⁶ no cases of thrombosis with thrombocytopenia syndrome were observed in this study, and no cases have been reported in the overall programme for this HIV vaccine regimen.

Binding antibody responses were lower in Mosaico than those observed with the same regimen in ASCENT,¹² which might be due to geographical differences. Peak binding antibody responses in Imbokodo were also lower than those observed with the Ad26.Mos4.HIV plus clade C gp140 regimen in ASCENT.¹⁷ In Mosaico, participants who were Ad26 seropositive at baseline had lower binding antibody responses; however, neither APPROACH nor ASCENT showed any effect of Ad26 NAb titres.^10,12 There could be a geographical factor involved; previously for the Ad26.COV2.S vaccine, a lower spike protein binding antibody response was observed in Ad26-seropositive participants versus Ad26-seronegative participants from Brazil but not in those from South Africa.¹⁸ In addition, there were few observations of Ad26 NAb titres greater than 1000 90% inhibitory concentration at baseline in APPROACH and ASCENT;^10,12 however, these high baseline titres did not affect the Ad26.ZEBOV, MVA-BN-Filo Ebola vaccine regimen in adults and children in African countries.^19-22 Overall, no consistent trend has been observed between baseline Ad26 NAb titres and vaccine-induced, insert-specific immune responses post-vaccination for Ad26-based vaccines targeting HIV, COVID-19, Ebola virus, and respiratory syncytial virus.^10,12,18-25

Although the study vaccine regimen met the safety profile requirements of a candidate HIV vaccine, the prediction we made based on immunological markers that correlate with protection (ie, per exposure risk reduction) in preclinical non-human primate studies¹⁰ did not translate into the same prediction of protection against HIV acquisition in our human clinical trial. The lack of efficacy of this four-dose Ad26 gp140 regimen contrasts markedly with high vaccine efficacy observed with single-dose and two-dose regimens of Ad26 vaccines incorporating the SARS-CoV-2 spike protein,^26,27 illustrating the scientific complexity of developing effective HIV vaccines. The antigenic diversity of HIV is many times greater than that of SARS-CoV-2, and antigenic alterations in envelope variants markedly reduce vaccine efficacy with SARS-CoV-2 variants,^28-30 indicating the importance of matching vaccine antigens with circulating strains of HIV-1. To this end, neutralising antibodies to the divergent circulating strains of HIV-1 in regions where the Imbokodo or Mosaico trials were done were not seen.¹² These data suggest that vaccine strategies eliciting neutralising antibodies against a wide spectrum of circulating strains are likely to be required for any successful HIV vaccine, perhaps reflecting the necessity of preventing HIV-1 acquisition rather than a reduction in symptomatic disease.

One limitation of the study was the high rate of VISP (near 100% by month 7). VISP, in cases of HIV testing outside of the study, along with reactogenicity, are potential causes of unmasking and might have resulted in a higher percentage of participants (73% of questionnaire respondents) being able to correctly identify their treatment group than would be expected by random chance; this could have subsequently affected safe sex practices or PrEP use. Additionally, the study enrolled only cisgender men and transgender individuals, limiting the generalisability of the results.

In conclusion, the Ad26.Mos4.HIV and clade C gp140–mosaic gp140 vaccine regimen was not effective in preventing HIV-1 acquisition in a population of cisgender men and transgender individuals who have sex with cisgender men or transgender individuals enrolled in Latin America, Europe, and the USA. Additional immunological analyses are planned to better understand why the vaccine regimen did not provide protection in this population.

Research in context.

Evidence before this study

Efforts to develop a safe and effective prophylactic HIV-1 vaccine have been unsuccessful. A PubMed search for HIV-1 vaccine clinical trials (published from Nov 1, 2004, with no language restrictions), conducted on Nov 1, 2024, using the MeSH term “AIDS Vaccines[mh]” and the “Randomized Controlled Trial” filter yielded 324 articles published within the past 20 years. Of these, eight reported results from HIV-1 vaccine efficacy trials. Only one trial (RV144) demonstrated modest efficacy.

HVTN 705/HPX2008/Imbokodo was the first trial to evaluate the efficacy of a mosaic-based HIV-1 vaccine regimen. Young women in southern Africa were administered a tetravalent mosaic adenovirus type 26-based vaccine (Ad26.Mos4.HIV) and an aluminium phosphate-adjuvanted clade C glycoprotein (gp) 140. The vaccine regimen did not show efficacy in preventing HIV-1, and the trial was stopped early with no safety concerns.

Added value of this study

HVTN 706/HPX3002/Mosaico was complementary to Imbokodo and evaluated Ad26.Mos4.HIV and a clade C gp140 plus mosaic gp140 boost in men and transgender individuals in Latin America, Europe, and the USA. This vaccine regimen was designed to induce immune responses against a wide array of global HIV-1 strains. In the preceding phase 1/2a ASCENT study, the regimen was shown to improve clade B responses compared with the regimen used in Imbokodo. Because Mosaico was conducted in regions where clade B is more common (vs higher clade C prevalence in Imbokodo) and the route of transmission was expected to be predominantly intrarectal (vs intravaginal in Imbokodo), Mosaico continued despite the lack of efficacy in Imbokodo.

Mosaico enrolled individuals considered to be particularly susceptible to HIV-1 acquisition (eg, transgender individuals, people from Latin America) and was also unique in that the study offered pre-exposure prophylaxis (PrEP) to potential participants in the community, but only enrolled those who chose not to use PrEP. Participants who chose to use PrEP post-enrolment were linked to PrEP services and allowed to remain in study.

Implications of all the available evidence

Despite the lack of efficacy results, Mosaico was an innovative, inclusive, and ethically conducted study that will help inform the design of future HIV-1 prevention trials. The study highlights the need for prevention options beyond PrEP. The available evidence from Mosaico together with numerous other efficacy trials of failed vaccines that elicited non-neutralising antibodies, indicate that novel approaches that induce broadly neutralising antibodies might be required to protect against HIV-1.

Data sharing

The data sharing policy of Janssen Pharmaceutical Companies of Johnson & Johnson is available at https://www.janssen.com/clinical-trials/transparency. As noted on this site, requests for access to the study data can be submitted through Yale Open Data Access Project site at http://yoda.yale.edu.