Enhanced Clearance of HIV‐1 Broadly Neutralizing Antibody VRC07‐523‐LS During Viremia: Influences on Trial Design and Analysis

ABSTRACT

Broadly neutralizing antibodies (BNAb) are capable of neutralizing multiple HIV‐1 strains through the targeting of conserved epitopes. This study's objective was to quantify the pharmacokinetic (PK) parameters describing the distribution and elimination of the BNAb VRC07‐523‐LS in people with HIV (PWH), and to identify the influence of viral load on VRC07‐523‐LS elimination. The PK of VRC07‐523‐LS was assessed in participants with and without HIV after intravenous or subcutaneous administration. A Bayesian strategy was utilized to estimate VRC07‐523‐LS PK parameters, leveraging a previously published study in adults without HIV. The effects of static viral loads on VRC07‐523‐LS exposure and time to a target trough serum concentration of 1 mg/L was then evaluated using clinical trial simulations. VRC07‐523‐LS serum concentrations were described by a two‐compartment model with first order absorption from the subcutaneous site. Typical clearance was estimated as 99.3 mL/day, which increased to a maximum of 6.46‐fold higher with viremia. Clinical trial simulations without viremia showed that a 700–2100 mg dose of VRC07‐523‐LS is expected to remain above the IC₅₀ of 0.055 mg/L beyond 52 weeks. With a viral load of 30,000 copies/mL, a 2100 mg dose of VRC07‐523‐LS is expected to result in 90% of participants having a concentration below 1 mg/L by 18.6 weeks, a decrease of 42 weeks compared to aviremic patients. Mechanistic modeling offers the ability to identify HIV viral load effects on BNAb PK and can improve BNAb dosing, especially with consideration of acute versus maintenance treatment.

Study Highlights

• What is the current knowledge on the topic?

  • ○
    Broadly neutralizing anti‐HIV‐1 monoclonal antibodies (BNAbs) are promising therapeutics for both prevention and treatment of HIV. Current generation BNAbs have an increased spectrum of activity and can effectively reduce viral loads in human studies. Further, extended half‐life variants of BNAbs could provide year‐long coverage with a single dose.
• What question did this study address?

  • ○
    This study provides answers on how HIV‐1 viral load affects BNAb dosing requirements and how to optimize the analytical treatment interruption phase of the trial's design.
• What does this study add to our knowledge?

  • ○
    Evaluation of viral load effects shows increased clearance of BNAbs at high viral loads, necessitating higher potential doses. This study also shows how Bayesian adaptive PK trials of BNAbs can more efficiently identify important analytical treatment interruption time points.
• How might this change clinical pharmacology or translational science?

  • ○
    This study utilized a novel strategy to assess dosing requirements for BNAbs based on clinical circumstance, which could be extended to arbitrarily large BNAb cocktails. Additionally, Bayesian adaptive PK trials provide a strategic opportunity to make sampling and intervention times more efficient and less costly.

1. Introduction

Between acute and chronic HIV‐1 infection, endogenous antibodies dynamically shift in immunoglobulin type and target epitope over the first 3 months [1, 2]. After 2–4 years of natural infection, 10%–30% of individuals develop potent and cross‐reactive anti‐HIV‐1 broadly neutralizing antibodies (BNAbs) that are capable of neutralizing heterologous viral subtypes [3, 4, 5, 6]. Endogenously produced BNAbs are associated with individuals exhibiting greater viral diversity and longer duration of infection (i.e., chronic antigen exposure) [7]. Furthermore, high‐level somatic mutations (SHM) in B cells are required for broad and potent BNAbs. The most potent naturally derived BNAbs usually carry 40–110 mutations. Antibodies accumulate mutations in the antigen contact area of the complementarity determining regions (CDRs) and framework regions (FWRs) which enhance broadness and potency [8]. Thus, the passive administration of BNAbs represents a highly attractive strategy to prevent and treat HIV‐1 infection.

Identifying preferential targets for BNAbs has been critical for the development of a highly active monoclonal BNAb with clinical utility. There are six key viral targets for monoclonal BNAbs: the V1/V2 apex region [9], the V3 glycan site [10], the CD4 binding site (CD4bs) [11, 12], the membrane–proximal external region (MPER) [13], the interface fusion peptide (FP) for Env glycoproteins gp120 and gp41 [14], and the silent face of gp120 [15]. VRC07‐523 is a next‐generation engineered VRC07 variant targeting CD4bs and displaying a 7.9‐fold increase in potency compared to VRC01 and with 96% neutralization breadth [16, 17]. Co‐administration with BNAbs targeting different epitopes is required for enhanced coverage and efficacy. As a key pharmacodynamic target of efficacy, neutralization breadth is often used and represents the percentage of HIV‐1 strains neutralized at a given BNAb concentration. Potency is quantified as the BNAb concentration that causes 50% or higher neutralization (i.e., IC₅₀, IC₈₀, and IC₉₀).

To further enhance monoclonal BNAb efficacy for both active treatment and pre‐exposure prophylaxis, BNAb constructs are now routinely engineered with two mutations encoding amino acid substitutions (M482L and N434S) on their constant (F_c) region [18]. These F_c modifications, so‐called “LS” mutations, increase terminal half‐life [19]. These LS‐modified BNAbs are amenable to once to twice‐yearly dosing, which would represent a significant improvement in addressing patient adherence and protection. Although extended terminal half‐life is desirable for clinical applications, it creates challenges for study design, requiring prolonged follow‐up durations which compound trial cost and complexity.

The objective of this population pharmacokinetic study was to quantify the pharmacokinetic parameters describing VRC07‐523‐LS distribution and elimination in adults. Secondarily, this study sought to identify the influence of HIV viral load on VRC07‐523‐LS elimination. Finally, using data from a small pilot study, we also assessed the ability for maximum a posteriori‐Bayesian estimation of participant‐specific pharmacokinetic profiles as part of an adaptive PK trial design to efficiently identify the analytical treatment interruption time for each individual.

2. Materials and Methods

2.1. Statement on Informed Consent and Human Trial Guidelines

The study was approved by the Institutional Review Boards at the participating sites, and all participants provided written informed consent. The HIV Vaccine Trials Network (HVTN) Safety Monitoring Board (SMB) provided safety oversight. Trials assessing effects of viremia on VRC07‐523‐LS PK were registered with ClinicalTrials.gov (NCT02840474, NCT03015181, and NCT04340596) [20]. Pilot adaptive PK trial data were obtained under ACTG protocol A5386.

2.2. Study Design in Assessing Viremia Effects on VRC07‐523‐LS PK

In cases of BNAb use in viremic individuals, there is uncertainty as to the impact of viral load on BNAb PK. To test this effect, a population PK model was developed to assess VRC07‐523‐LS in both those without HIV and those with HIV and ongoing viremia. Data came from a phase 1, open‐label, single dose trial evaluating the safety and pharmacokinetics of VRC07‐523‐LS [21]. Participants with HIV not receiving antiretroviral therapy were included if they had at least one plasma viral load ≥ 500 copies/mL within 28 days of enrollment, but were clinically stable with CD4+ cell counts ≥ 350 copies/mL on at least 2 of 3 consecutive testing occasions. A single dose of VRC07‐523‐LS 40 mg/kg IV was administered, and participants were followed to Week 48 post‐infusion; however, participants were started on antiretroviral therapy 14 days after BNAb administration.

Pharmacokinetic data in adults without HIV came from a phase 1, dose‐escalation study evaluating VRC07‐523‐LS administered as intravenous and subcutaneous (SC) formulations. Seven different dosing groups were enrolled: 1 mg/kg IV single dose (n = 3); 5 mg/kg IV single dose (n = 3); 5 mg/kg SC single dose (n = 3); 20 mg/kg IV single dose (n = 3); 40 mg/kg IV single dose (n = 3); 5 mg/kg SC doses at Day 0 and Weeks 12 and 24; and 20 mg/kg IV doses at Day 0 and Weeks 12 and 24. Additional study details have been previously reported [20].

HIV status, sex, height, weight, age, baseline serum creatinine, and baseline hematocrit were collected to test for covariate effects on PK parameters. HIV viral load was included as a time‐dependent covariate to allow for the model to incorporate viral load effects dynamically.

2.3. VRC07‐523‐LS Pharmacokinetic Assay

VRC07‐523‐LS was quantified using a qualified Meso Scale Discovery (MSD) electrochemiluminescence immunoassay (ECLIA). VRC07‐523‐LS anti‐ID (clone 5C9) antibody was added to MSD 96‐well bare plates and placed at 4°C overnight. The next day, plates were blocked in MSD blocking buffer and washed. Serial dilutions of reference standard (VRC07‐523‐LS) and serum samples from study participants were added in duplicate and shaken for 1 h at room temperature.

Plates were washed to remove unbound sample. Sulfo‐tag labeled anti‐human IgG detection antibody was applied to the wells and allowed to associate with complexed anti‐ID and drug product within the assay wells. Plates were washed to remove unbound detection antibody. A read solution containing electrochemiluminescence (ECL) substrate was applied to the wells, and the plates were read on an MSD Sector S600 instrument. During the read, a current is applied to the plates and areas of well surface which form a full anti‐ID/drug product/anti‐human IgG‐SulfoTag complex emit light in the presence of the ECL substrate. The MSD Sector instrument quantifies the amount of light emitted and reports this ECL unit response as a result for each sample and standard. The amount of VRC07‐523‐LS sandwiched by the anti‐ID and anti‐human IgG antibodies is directly proportional to the concentration of reactive drug product in the sample wells. The sample signal was interpolated relative to the reference standard and dilution‐adjusted. Calculations were performed within Excel and GraphPad software (version 8 or later). The VRC07‐523‐LS limit of detection is 0.5 μg/mL.

2.4. Pharmacokinetic Model Development

A previously published population pharmacokinetic model of VRC07‐523‐LS was utilized as Bayesian priors for fixed and random effects [19]. Correlations, covariate effects, and HIV viremia effects were assessed in a step‐wise manner based on statistically significant correlations between fixed effects and covariates. Statistical significance was similarly determined using likelihood ratio testing as above. To internally validate the findings of a HIV viral load effect, bootstrap analysis (n = 2000 runs) was performed on the final model.

2.5. Clinical Trial Simulation and Regimen Design

To qualitatively explore the effects of static viral loads on VRC07‐523‐LS exposures, clinical trial simulations (n = 1500 simulated subjects) were conducted to assess exposure over 52 weeks after IV administration of 250, 700, 1400, and 2100 mg of BNAb (i.e., 3.6–30 mg/kg for a 70‐kg person). Each regimen was simulated assuming static HIV‐1 RNA levels ranging from 0 to 30,000 copies/mL. Static viral loads were simulated to provide conservative estimates of viral load effects at a given dose. Simulations were then analyzed to determine the time required for 90% of participants to be at or below either 10 mg/L or 1 mg/L.

When studying the effects of BNAbs as part of a regimen to suppress viremia, clinical trials incorporate antiretroviral analytical treatment interruptions to test delay in viral rebound after discontinuing antiretroviral therapy [22, 23]. In our case, simulations were performed with anticipation of designing a clinical trial incorporating an analytical treatment interruption phase after administration of VRC07‐523‐LS as part of a regimen to induce sustained plasma HIV‐1 suppression. Thus, for this work we sought to identify times where most individuals would be ready to enter the analytical treatment interruption phase of the study, which was the time a specific person reached a VRC07‐523‐LS serum concentration of ≤ 1 mg/L. In addition, a 10× target relative to the ATI target (i.e., 10 mg/L) was evaluated as an empiric resistance‐prevention endpoint.

Optimal dosing to fully account for a dynamic viral load was determined using a hybrid machine learning‐population pharmacokinetic modeling strategy. A “genetic algorithm” (GA), a machine learning optimization algorithm, was implemented to identify an optimized VRC07‐523‐LS dosage regimen such that 90% of simulated participants were at or below 1 mg/L by 48 weeks post BNAb administration. Both the concentration of 1 mg/L and the intervention time of 48 weeks were targets selected to aid in planning future studies where analytical treatment interruption was being utilized, but any concentration target or intervention time could be targeted (e.g., > 1 mg/L or 52 weeks). To identify optimal dosing in HIV‐negative or virally suppressed individuals, clinical trial simulations were performed assuming no viral load (i.e., no HIV‐induced effects on VRC07‐523‐LS clearance). Two different clinical scenarios were explored to assess BNAb dosing during viremia: single dose administration with dynamic viral load and two dose administration with dynamic viral load. The dynamic viral load was determined by fixing the time course of HIV‐1 copy number to the 90th percentile observed from the viremic participants used to generate the model. For a two‐dose strategy, the second dose was pre‐selected to occur at Week 32, such that the GA only had to identify an initial dose during the active viremia phase and at a second dose once virally suppressed. In effect, this two‐dose strategy was designed to assess a “step‐down” approach for BNAb dosing when transitioning from acute to maintenance therapy.

2.6. Participant‐Specific Interim BNAb Pharmacokinetic Analysis

Because F_c‐LS‐BNAbs have significantly prolonged half‐life compared to their F_c‐wild type counterparts, there is a significant study design burden related to the longer protocol durations. Under a Bayesian paradigm, participant‐specific pharmacokinetics can be quantified by leveraging prior population pharmacokinetic data. To test whether participant‐specific BNAb PK can be quantified accurately at the individual level to support shorter‐duration study design, we conducted a pilot analysis using interim PK samples to determine if individual time‐to‐target estimates could be obtained. Interim data came from ACTG protocol A5386, an open‐label Phase 1 trial in which adults with HIV who were virologically suppressed on ART for approximately 2 years received combination treatment with the BNAbs VRC07‐523LS (20 mg/kg) and 10‐1074LS (30 mg/kg), along with the IL‐15 superagonist N‐803 (NCT: NCT04340596). An analytic treatment interruption of ART was planned approximately 52 weeks after bNAb administration, when BNAb concentrations were believed to have fallen below pre‐defined thresholds. This strategy was motivated by the overall study protocol which utilized an analytical treatment interruption period that was prospectively designed to begin once BNAb concentrations reached 1 mg/L. Additionally, this pilot analysis was only performed using data from 3 participants.

To accomplish the interim analysis, ADAPT5 was used for individual analysis via Maximum a Posteriori estimation, using a previously published pre‐post processor to facilitate reproducible and timely reporting to the study team [24]. To ensure that the PK parameter estimation strategy accounted for N number of samples obtained for each individual, the pre‐processing software automatically adjusted the number of PK parameters to be estimated (e.g., clearance, central volume, peripheral volume, and distributional clearance) to be N‐2. This strategy was implemented to increase the amount of information available to quantify the participant‐specific PK profile.

3. Results

3.1. VRC07‐523‐LS: The Influence of Viremia on BNAb Pharmacokinetics

Participant demographics are reported in Table S1. Across 9 individuals with viremia, the average viral load at initiation of BNAb was 35,264 copies/mL (Figure 1). HIV‐1 viral load reached a nadir at Week 1 and then rebounded to a peak of 4410 copies/mL at 3 weeks. Once at peak viral load, participants were then initiated on antiretroviral therapy, which resulted in reattaining full viral suppression by Weeks 7 and 8 (average viral load of 22.2 copies/mL).

FIGURE 1.

Time course of HIV viral load in patients with viremia. For participants with viremia at time of initiation of therapy, the time course of median viral load was plotted (black line) along with empiric 75th (green shaded), 90th (blue shaded), and 95th (red shaded) percentiles. Due to ethical study design constraints, participants were started on BNAbs, but antiretroviral therapy was started once viral rebound was observed (approximately Weeks 2–4).

VRC07‐523‐LS concentrations were best described with a two‐compartment model with first order absorption from the subcutaneous administration site. VRC07‐523‐LS clearance was modeled as linear in those without viremia and being non‐linearly related to target mediated drug disposition (TMDD) utilizing the following equation:

$$\mathrm{C}{\mathrm{L}}_i\left(t\right)=\mathrm{C}{\mathrm{L}}_{\mathrm{TV}}·\left(1+\frac{E_{\max ,i}·\mathrm{HIV}{\left(t\right)}^{\gamma }}{\mathrm{E}{\mathrm{C}}_{50}^{\gamma }+\mathrm{HIV}{\left(t\right)}^{\gamma }}\right)·e^{{\eta }_{\mathrm{CL},i}}$$

Here, CL_TV represents the typical population clearance without a viral load, E _max,i represents the maximal increase in clearance because of the HIV viral load, EC₅₀ is the HIV viral load that produces 50% of E _max,i , HIV(t) is the HIV viral load at time t, γ is a shape parameter, and η is the participant‐specific adjustment to clearance.

Visual predictive checks of the final model (Figure 2) showed that the model well captured both the central tendency and variability of the data. Baseline clearance was estimated as being 99.3 mL/day, which could increase to 5.46‐fold (46.2% RSE) higher (i.e., CL = 641 mL/day) when HIV(t) > EC₅₀ (Table 1). Plasma HIV‐1 RNA levels required to reach 50% of this maximum effect (i.e., the EC₅₀) were estimated to be 16,595 copies/mL (4.5% RSE). This nonparametric bootstrap evaluation indicated that the median estimated E _max and EC₅₀ were 4.22 and 9120 copies/mL, respectively. Subcutaneous absorption of VRC07‐523‐LS was well described, with the bioavailability being estimated as being 48.6% (28% RSE). The typical absorption half‐life was estimated as being 2.02 days (41.2% RSE).

FIGURE 2.

Visual predictive check for VRC07‐523‐LS population pharmacokinetic model. Variability in model simulated median (red shaded region) along with 10th/90th prediction intervals (blue shaded region) are compared to empiric median and percentiles (solid lines) show that the model captures both the central tendency and variability of the observed concentrations for both intravenously and subcutaneously administered VRC07‐523‐LS.

TABLE 1.

VRC07‐523‐LS population pharmacokinetic parameter estimates.

	Units	Estimate (%RSE)	Bootstrap a median (95% CI)
Fixed
F bio	—	0.486 (28%)	0.486 (0.485–0.487)
T abs	day	2.02 (41%)	2.13 (1.72–2.74)
CL	mL/day	99.3 (7.1%)	97.3 (84.4–111)
V1	mL	2080 (7.4%)	2080 (1800–2380)
V2	mL	2760 (10%)	2730 (2200–3280)
CLD	mL/day	518 (18%)	556 (386–823)
CLmax	—	5.46 (46%)	4.19 (1.61–11.3)
EC50	log10 (copy/mL)	4.22 (4.5%)	3.96 (2.17–4.80)
γ	—	1 (FIXED)
Random
ω F	—	1.06 (441%)	0.744 (0.284–2.47)
ω Tabs	—	0.481 (61%)	0.563 (0.202–1.34)
ω CL	—	0.296 (18%)	0.294 (0.278–0.315)
ω V1	—	0.382 (18%)	0.378 (0.363–0.399)
ω V2	—	0.390 (17%)	0.614 (0.592–0.645)
ω CLD	—	0.611 (24%)	0.387 (0.361–0.415)
ω CLMAX		0.448 (75%)	0.440 (0.174–0.858)
ρ CL ~ V2		0.819 (14%)	0.826 (0.403–0.945)
Residual
σ a	mg/L	0.523 (25%)	0.494 (0.159–1.155)
σ p	—	0.204 (3.9%)	0.202 (0.164–0.238)

Abbreviations: ρ, correlation between random effects; σ _a, additive residual error; σ _p, proportional residual error; ω, inter‐individual variability as SD; CL, clearance; CL_D, distributional clearance; CL_max, maximum increase in CL due to viremia; EC₅₀ log, viral load as copies/mL for 50% of CL_max; F _bio, SQ bioavailability; T _abs, first order absorption half‐life; V1, central volume; V2, peripheral volume.

^a Based on 1000 bootstrap runs.

3.2. Clinical Trial Simulation and Regimen Design

Clinical trial simulations of regimens under static viral conditions (Figure 3) found a pronounced effect on total BNAb exposure. Simulated trials at a viral load of 0 copies/mL found that the 700–2100 mg dose of VRC07‐523‐LS is expected to be above the IC₅₀ of 0.055 mg/L beyond 52 weeks. However, if a high viral load is present, the BNAb is cleared more quickly (Figure 3A). With a viral load of 30,000 copies/mL, a 2100 mg dose of VRC07‐523‐LS is predicted to result in the 10%–90% prediction interval of subjects to exhibit a concentration below 1 mg/L by 7.59–18.7 weeks. By comparison, this range of simulated subjects would reach 1 mg/L by 35.4–62.3 weeks under virally suppressed/HIV‐negative simulated conditions (Figure 3B).

FIGURE 3.

VRC07‐523‐LS exposure changes under static viral loads. (A) The final model was used to simulate 1500 virtual individuals. Twenty‐one treatment groups were tested exploring three dosing levels (2100, 1400, and 700 mg) of VRC07‐523‐LS in individuals with seven static viral loads (0, 100, 300, 1000, 3000, 10,000, and 30,000 copies/mL). Viral loads 1–30 were omitted due to being indistinguishable from simulations without viremia (i.e., 0 copies/mL). (B) Clinical trial simulations were evaluated at the 10th/90th prediction interval. Thus, the times listed below are indicative of the time‐to‐target where 10%–90% of simulated participants would be at or below the listed target. Additionally, simulations were performed assuming static viral loads to provide the most conservative assessment of VRC07‐523‐LS elimination. Overall, the time‐to‐target values presented are conservative assessments, and, with sufficient viral suppression, the time‐to‐target could be longer for most individuals.

To systematically identify a VRC07‐523‐LS dose, a genetic algorithm was implemented which determined that a single 924 mg dose could be feasible in virally suppressed/HIV‐negative individuals to achieve the target concentration of 1 mg/L by 52 weeks in 90% of simulated individuals (Figure 4). By comparison, a hypothetical population who are actively viremic at the 90th percentile of viral load as outlined in Figure 1, would require a single dose of 1955 mg (2.11‐fold higher) to achieve the same result. Utilizing a two‐dose strategy can decrease the overall dosing needs by using an initial 158 mg dose during the acute viremia phase then 68 mg at Week 26 to reduce the total amount of drug required and still reach ≤ 1 mg/L in 90% of simulated participants.

FIGURE 4.

Optimization of VRC07‐523‐LS regimen design. Optimal dosing for VRC07‐523‐LS was determined without viremia (top) and with viremia (middle, bottom) using a genetic algorithm to optimize dose parameters. Viremia was simulated based on the 90th percentile of observed viral load across the 10 participants in this study. Doses were identified assuming a 1h prolonged infusion and that, for a two‐dose regimen, the second dose was administered 26 weeks after the first. The genetic algorithm’s fitness function was designed such that the 90th prediction interval of the trough concentration(s) was equal to 1 mg/L.

3.3. VRC07‐523‐LS: Adaptive Pharmacokinetic Trials Efficiently Characterize BNAb Pharmacokinetics

The final sample time collected for each of the three participants was 13, 32, and 46.6 weeks, which occurred in the terminal phase of elimination for all three (Figure 5). As part of the interim PK analysis for NCT04340596, no adverse effects were reported for the three participants included in this pilot study.

FIGURE 5.

Pilot participant‐specific Bayesian fits to identify analytical treatment interruption time. To identify time‐to‐target for individual patients, Maximum a posteriori model fits were obtained for each of three participants based on PK priors from the final population PK model from this study. Each participant had different terminal sample times, which allowed for comparison of extrapolation to a target of 1 mg/L.

For all three participants, the clearance, central volume, peripheral volume, and distributional clearance were estimated utilizing priors from the previous model (Table S2). Parameters for absorption, bioavailability, and viremia effects on clearance were fixed to their prior values. The population estimate of clearance was used in all three participants as they did not have viremia at the time of the study. Clearance was accurately estimated in all three participants (1.71%–2.63% RSE). Distributional clearance was estimated less precisely (12.2%–12.7% RSE) but was still well characterized due to rich sampling during the distribution phase for all participants. Based on individual estimates of VRC07‐523‐LS PK in each participant, individual‐specific terminal half‐lives were determined to be 45.7–59.9 days. In relation to the time‐to‐target, all three participants were expected to reach 1 mg/L by 47.3–49.9 weeks (Table S2).

4. Discussion

HIV‐1 broadly neutralizing monoclonal antibodies represent a significant shift in the strategy for HIV prevention and treatment. Regarding prevention, the ability for BNAbs to be dosed once or twice per year would drastically improve patient adherence and the proportion of time protected from infection. In the context of treatment, BNAbs can neutralize freely circulating virus while simultaneously offering improved opportunity to leverage immune effector functions to facilitate viral elimination.

VRC07‐523‐LS targets CD4bs, which is typically a viral target that induces antibody production early in infection [25]. Developing optimal dosing strategies for BNAbs in general, and VRC07‐523‐LS in particular, requires careful assessment of the influences of HIV target distribution, which can significantly alter the clearance rate of the antibodies during treatment phases where viral load is elevated. Furthermore, BNAb development with prolonged half‐lives has resulted in significant improvements to the duration of time where individuals are protected or virally suppressed, but creates challenges for trial design given the need for extended follow‐up times. Given the prolonged half‐life and the non‐linear relationship between viral load and the time‐to‐target concentration, VRC07‐523‐LS dosing may require personalization in some manner via viral load‐based dosing recommendations or concentration‐driven therapeutic drug monitoring.

In the initial assessment of VRC07‐523‐LS pharmacokinetics, viral load was indeed determined to be a significant factor on clearance. During viral suppression, VRC07‐523‐LS exhibits the typically long terminal serum half‐life of LS‐modified BNAbs: median of 36.3 days in this study. This attribute is ideal for treatment strategies focused on maintaining viral suppression in HIV‐positive individuals or for prevention. However, in the subset of individuals exhibiting viremia, BNAb clearance was statistically significantly augmented, which was estimated as being 6.46‐fold higher than under aviremic conditions. Furthermore, this effect was detected as reaching 50% of maximum at a viral load of 16,600 copies/mL. Therefore, there may be important clinical differences in required BNAb dose between viremic and aviremic conditions. The influences of disease state on the TMDD of monoclonal antibody dosing are well established and have been described both for BNAbs and for monoclonal antibodies used in other therapeutic areas [26, 27]. As such, future studies of BNAbs for active treatment during acute infection may necessitate higher doses and should simultaneously assess the changes in viral load over time to better understand the effects of viral load on BNAb clearance.

In silico studies of the final VRC07‐523‐LS population pharmacokinetic model indicated that the time to reach 1 mg/L for analytical treatment interruption was nonlinearly related to the HIV viral load. Furthermore, implementation of a machine learning‐led PK optimization scheme highlighted the different dosing requirements for individuals with HIV in the acute (viremic) and maintenance (viral suppression) phases of treatment. Specifically, the population pharmacokinetic model supports a VRC07‐523‐LS dose of 158 mg at initiation followed by a dose of 68 mg at 26 weeks that is sufficient to maintain serum BNAb levels above the threshold of > 1 mg/L. Beyond clinical effects, adjusting BNAb dose for acute versus maintenance versus prevention can also result in a significant decrease in total BNAb required, which could offer significant financial savings that can enhance access to BNAbs. There are significant challenges involved with optimizing combination anti‐infectives to ensure the concentrations of all agents are simultaneously above their respective targets. The GA‐led PK optimization illustrated in this work could be readily expanded to multiple BNAbs or BNAbs in combination with traditional antiretrovirals for improved trial planning [28, 29]. However, as a key limitation, it's still unclear what period of time is needed with higher BNAb levels prior to waning versus maintaining serum throughout lower, repeated doses.

Analytical treatment interruption is critical to assessing the durability of viral suppression longitudinally, which is necessary in any HIV trial assessing long‐term control off ART. Because LS‐modified BNAbs exhibit a terminal serum half‐life of greater than 70 days, there are significant logistical challenges and expense to providing follow‐up to study participants, in particular when study endpoints require BNAb wash out for assessment of long‐term virologic control. In a follow‐up study of VRC07‐523‐LS, an adaptive pharmacokinetic trial design was implemented to determine participant‐specific timing to initiate analytical treatment interruption. By leveraging our population pharmacokinetic model of VRC07‐523‐LS as Bayesian priors, we were able to successfully project the time of analytical treatment interruption for three participants. This strategy leveraged close coordination between the investigation site, bioanalytical laboratory, and clinical pharmacologists to provide timely feedback on participant‐specific analysis.

The results of the interim analysis of all three individuals receiving VRC07‐523‐LS indicated that the terminal half‐life and time‐to‐target could be accurately identified for each participant. The quality and quantity of VRC07‐523‐LS concentration data resulted in a highly precise individual fit, which could exhibit increased noise depending on the drug, study design, patient population, and analytical assay. As a compromise between reducing the number of PK samples and accounting for error in model predicted ATI time, a confirmatory sample can be obtained prior to a model predicted ATI window for each participant.

Clinical trial simulations in individuals without viremia indicated that after a 2100 mg dose, the bottom 10% of subjects would reach 1 mg/L by 35 weeks, whereas the top 90% would reach 1 mg/L by 60.5 weeks. In the absence of an interim analysis of participant‐specific PK, multiple samples would need to be collected for all participants within this 25‐week window to ensure the trough was captured across all individuals. However, by assessing individual concentrations just after BNAb dosing, targeted sampling can be performed for each participant. For participants 1–3, the extrapolated trough was predicted as being 6.4, 240, and 125 days after the final sample, respectively. Additionally, with optimal sample design strategies [30, 31], this method could be further refined to target most informative sample times for estimating critical PK parameters, which would further improve future trial efficiencies.

Because participants were re‐started on antiretroviral therapy at the time of viral load rebound, there may be additional effects on BNAb target mediated drug disposition based on the exact antiretroviral regimen. Furthermore, co‐administration with a second BNAb may exhibit PK drug interactions if both agents exhibit LS‐modifications and compete for FcRn, but no observations of these interactions have been observed in clinical studies evaluating combination BNAb treatment to date [32]. Combination BNAb therapy will be required to prevent development of resistance; however, the pharmacokinetics may need to be assessed alone and in combination in order to ensure proper dosages of each BNAb depending on the objectives (i.e., acute treatment, maintenance therapy, or prevention).

Future studies should seek to explore the effects of viremia on BNAb pharmacokinetics given the potentially significant target‐mediated drug distribution effects. Neutralization targets are typically defined based on in vitro derived IC₅₀ or IC₈₀ values. Though in vitro IC_50/80 values are helpful for preliminary targets, future studies may benefit from co‐modeling of BNAb pharmacokinetics and pharmacodynamics for the purposes of better quantifying variabilities in viral clearance in humans.

Overall, this study showed the utility of translational pharmacokinetic/pharmacodynamic modeling. The ability to accurately identify HIV viral load effects on BNAb pharmacokinetics may play an important role in the dosing of these agents, and tailoring to the individual's treatment goals—active treatment, maintenance, or prevention. The ability of a single BNAb to exert significant neutralization effects is promising, which may be augmented when in combination with other BNAbs or immunomodulating agents. These results support the role of mechanistic models in quantifying BNAb PK/PD and in designing clinical trials for prevention, treatment, and long‐term control of ART.

Author Contributions

N.M.S., K.J.B., L.G., G.D., R.A.K., M.P., L.S., M.C., T.J.W., T.D.W., G.D.M., C.S.V., R.C., and Q.M. wrote the manuscript and designed the research; N.M.S. and B.M.H. analyzed the data; N.M.S. contributed new analytical tools.

Funding

Research reported in this publication was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under Award Numbers R01AI177997 (Smith), R01AG063659 (Ma), UM1AI069511‐16 (Morse, Ma, Venuto), 5UM1AI068636‐16 (Morse, Cha), UM1 AI068634, UM1 AI068636, and UM1 AI106701. This research was supported in part by the Intramural Research Program of the National Institutes of Health (NIH). The contributions of the NIH authors were made as part of their official duties as NIH federal employees, are in compliance with agency policy requirements, and are considered Works of the United States Government. However, the findings and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the NIH or the U.S. Department of Health and Human Services.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Appendix S1: cts70488‐sup‐0001‐AppendixS1.docx.

Smith N. M., Ho B. M., Bar K. J., et al., “Enhanced Clearance of HIV‐1 Broadly Neutralizing Antibody VRC07‐523‐LS During Viremia: Influences on Trial Design and Analysis,” Clinical and Translational Science 19, no. 2 (2026): e70488, 10.1111/cts.70488.