Cervicovaginal microbiome diversity was not associated with mucosal pharmacokinetics of systemically delivered HIV broadly neutralizing antibodies

Abstract

Broadly neutralizing antibodies (bNAbs) are a promising HIV prevention strategy due to their potent antiviral activity and potential for long-acting protection. While the vaginal microbiome can influence mucosal immunity and the efficacy of topical interventions, its effect on the pharmacokinetics of systemically administered bNAbs remains unclear. Forty-two women were included in a retrospective analysis of the CAPRISA 012B clinical trial evaluating passive immunization for HIV prevention. Vaginal microbiota were profiled using 16 S rRNA gene sequencing and classified into three community state types (CSTs): CST I (Lactobacillus crispatus-dominated), CST III (Lactobacillus iners-dominated), and CST IV (diverse, non-Lactobacillus-dominated). Mucosal concentrations of CAP256V2LS were measured from Soft-cup^® cervicovaginal fluid using the Meso Scale Discovery (MSD) platform with electrochemiluminescence (ECL) detection. Longitudinal analyses assessed CST stability, transitions, and associations with mucosal antibody pharmacokinetics. Lactobacillus-dominated CSTs were most frequent (CST I, 5.3%; CST III, 57.9%), whereas BV-associated CST IV subtypes were less common (CST IV-A, 2.6%; CST IV-B, 34.2%). Lactobacillus-dominated communities were generally stable, while high-diversity CST IV communities were more dynamic, with transitions toward Lactobacillus-dominated states observed over time. Despite these microbial shifts, mucosal bNAb kinetic patterns appeared broadly similar across CSTs. Within the limits of this exploratory analysis, we did not observe clear evidence of an association between CST composition and the timing or magnitude of mucosal bNAb accumulation. These observations were descriptive and not derived from inferential statistical or pharmacokinetic modelling analyses. In this exploratory sub-analysis, cervicovaginal microbiome composition was not clearly associated with differences in mucosal concentrations of systemically administered bNAbs. These findings suggest that systemic bNAb delivery may achieve measurable genital tract exposure across diverse vaginal microbial communities; however, larger studies incorporating inferential pharmacokinetic and immunological analyses are needed to confirm these observations and exclude subtle microbiome-associated effects.

Introduction

Although effective behavioral and pharmacological strategies exist to prevent HIV transmission, an estimated 1.3 million people worldwide were newly diagnosed with HIV in 2024¹. Existing strategies are insufficient to meet the 2030 global targets for reducing HIV incidence, largely owing to persistent barriers to access and suboptimal adherence to pre-exposure prophylaxis (PrEP), including both oral agents and long-acting injectable formulations². Long-acting monoclonal antibodies (mAbs), including broadly neutralizing antibodies (bNAbs), constitute a promising HIV prevention strategy, characterized by an extended half-life permitting infrequent dosing, potential for subtype-specific targeting, and the feasibility of subcutaneous administration^3,4. Passive infusion studies in non-human primates have shown that bNAbs can protect against mucosal challenge with simian human immunodeficiency virus (SHIV), demonstrating a proof of concept for passive immunization^2,5–7. In humans, bNAb investigational products, such as VRC01, have been shown in antibody-mediated prevention (AMP) trials to reduce the acquisition of HIV strains susceptible to the antibody, and high serum concentrations correlate with reduced HIV risk⁸ Recent studies, such as the CAPRISA 012 A and 012B trials, have evaluated the safety, pharmacokinetics, and neutralization activity of bNAbs, including VRC07-523LS, PGT121, and CAP256V2LS, in young women in South Africa^9,10. These trials demonstrated that subcutaneous administration is safe, well-tolerated, and maintains detectable bNAb concentrations for up to six months, supporting their potential as long-acting HIV prevention agents and highlighting their ability to address the adherence challenges associated with conventional PrEP.

While these studies provide critical evidence for the safety and half-life profiles of passively administered bNAbs, the factors that influence antibody distribution and retention in mucosal tissues remain poorly understood. The cervicovaginal environment, the primary site of heterosexual HIV transmission, may shape bNAb pharmacokinetics and effectiveness, yet the role of host factors, especially the cervicovaginal microbiome, remains uncharacterized. A healthy vaginal microbiome, often dominated by beneficial Lactobacillus species, plays a crucial role in maintaining a balanced local immune environment and is associated with a lower risk of HIV infections^11,12. In contrast, the cervicovaginal microbiome with low Lactobacillus abundance and high diversity is significantly associated with bacterial vaginosis (BV), which is highly prevalent in sub-Saharan Africa^13,14. Such highly diverse bacterial communities have shown significant correlations with elevated pro-inflammatory cytokines, defining genital inflammation^13–15. Genital inflammation has been implicated in the reduced efficacy of topical tenofovir-based PrEP^16,17 and adverse reproductive health outcomes, including preterm birth, cervical dysplasia, and increased susceptibility to sexually transmitted infections, including HIV^13,18,19. Despite its well-established role in mucosal immunity, the vaginal microbiome’s influence on systemically administered bNAb pharmacokinetics in mucosal tissues remains unknown.

This study aims to investigate the role of the cervicovaginal microbiome composition on the pharmacokinetics of systemically administered bNAbs (CAP256V2LS and VRC07-523LS) in the female genital tract. Addressing this question is crucial to determine whether variations in the vaginal microbiome affect the mucosal bioavailability of bNAbs, with implications for the consistency and generalizability of antibody-based prevention strategies across heterogeneous populations.

Methodology

General study design

This is a retrospective analysis of the stored specimens from the first-in-human CAPRISA 012 B Phase I trial conducted at the CAPRISA eThekwini Clinical Research Site in Durban, South Africa. The CAPRISA 012 B Phase I trial evaluated the safety, tolerability, and pharmacokinetics of bNAb CAP256V2LS, administered either intravenously to HIV-negative and HIV-positive women or subcutaneously alone and in combination with VRC07-523LS in HIV-negative women. Eligible participants were enrolled in three groups: Groups 1 and 2 were evaluated in an open-label design, while Group 3 was conducted as a double-blind, placebo-controlled study. Group 1 participants received CAP256V2LS intravenously (5 or 10 mg/kg), whereas Group 2 participants received subcutaneous dosing at 5–20 mg/kg with extended dosing periods, including repeat doses at 16 or 24 weeks in Groups 2d and 2 f. Groups 3a and 3b each enrolled five participants (4:1 active-to-placebo) who received CAP256V2LS and VRC07-523LS subcutaneously at 10 or 20 mg/kg with extended dosing¹⁰. For this study, we included only cervicovaginal swab samples from participants who received active bNAb products before and after passive immunization; placebo recipients were excluded. All experiments were conducted in accordance with relevant ethical guidelines and regulations, and approved by the University of KwaZulu-Natal Biomedical Research Ethics Committee (BREC00000857/2019) and the South African Health Products Regulatory Authority (SAHPRA 20200123). Informed consent was obtained from all subjects prior to participation. The trial was registered with the Pan African Clinical Trial Registry (PACTR202003767867253) on 20 March 2020.

16S rRNA gene amplicon sequencing and taxonomy assignment

Bacterial DNA was extracted from the vaginal swabs using a PowerSoil DNA kit (MoBio, CA, USA). The V3-V4 region of the 16S rRNA gene was PCR-amplified from bacterial DNA using the universal primers 319 F/806R, and purified libraries of approximately 120 pooled samples were sequenced on the Illumina MiSeq platform using paired-end v3 chemistry. Raw sequence quality was assessed using FastQC to guide the downstream processing. Amplicon sequence variants (ASVs) were inferred using the Divisive Amplicon Denoising Algorithm (DADA2), which includes quality filtering and trimming. Taxonomic classification of ASVs to the genus or higher level was performed using the SILVA rRNA database (version 14)²⁰. The ASV and taxonomic tables, combined with relevant metadata, were consolidated into a phyloseq object using the phyloseq R package²¹. CSTs were inferred using VALENCIA (VAginaL community state typE Nearest CentroId clAssifier)²², a nearest centroid classification method for vaginal microbial communities based on composition, implemented in Python (version 3.6), and has the pandas module as a dependency²³. VALENCIA uses distance matrices to classify the similarity of bacterial community structures between individual samples based on species proportions to each reference centroid, and was calculated using the Yue-Clayton θ²⁴.

bNAb concentrations measurements

The Meso Scale Discovery (MSD) platform (Rahway, New Jersey, USA) was used to determine CAP256V2LS and VRC07-523LS concentrations in mucosal soft-cup supernatants. This platform uses electrochemiluminescence (ECL) labels conjugated to a detection antibody via SULFO-TAG to detect antibody concentrations. Briefly, to detect CAP256V2LS, an anti-idiotype (ID) antibody was used to coat MSD 96-well assay plates, while VRC07-523LS plates were coated with a 5C9 ID antibody (Rahway, New Jersey, USA) and incubated at 4 °C overnight or until use. The plates were then blocked for 1 h at room temperature with MSD blocking solution. Thereafter, the reference standards and diluted soft-cup supernatants (containing either CAP256V2LS or VRC07-523LS) were added and incubated for 1 h at room temperature. An anti-human IgG detection antibody conjugated to SULFO-TAG was added. For CAP256V2LS, this secondary antibody bound to the anti-ID antibody CAP256V2LS complex, while for VRC07-523LS, it bound to the 5C9 ID antibody VRC07-523LS complex. An MSD reading solution containing an ECL substrate was added to each well, and the plates were read using the MSD Sector 600 instrument (Rahway, New Jersey, USA). In the instrument, an electric current was applied under the plate, exciting the SULFO-TAG and emitting a light that was measured and used to determine the concentration of CAP256V2LS or VRC07-523LS in each sample. All calculations were performed using Microsoft Excel and GraphPad 7.0 Prism software (Boston, Massachusetts, USA). Assay development, method evaluation, assay specificity, precision, and reproducibility were conducted according to the NIH, Vaccine Immunology Programme (VIP) guidelines.

Statistical analysis

All downstream statistical analyses were performed using the R software version 4.3.1 (http://www.r-project.org). Descriptive statistics were used to compare the demographic, behavioral, and clinical characteristics of the women stratified by mucosal bNAb detection status. The R package gt_summary was used to compute the demographic table outlining the median and IQR values of all continuous variables, whereas the total and percentage of qualitative variables were reported for qualitative variables. The clinical characteristics assessed included the STI and Bacterial vaginosis status. To assess the microbial shift, the probability of transitioning from one CST to another across visits was computed. Alpha diversity measures were computed using the R vegan library. Chao 1, Inverse Simpson Observed and Shannon index values were compared by Lactobacillus dominance within the group with bNAB concentrations detected, and within the group that did not have detectable concentrations. Multidimensional scaling (NMDS) was performed to assess differences between taxonomy profiles of samples with detected concentrations compared to those with undetected concentrations based on the Bray-Curtis distance of ASV relative abundance using the metaMDS function. Given the exploratory design and limited sample size of this sub-analysis, analyses were conducted descriptively, and pharmacokinetic modelling analyses were not performed.

Results

Study participants

This sub-analysis included 42 women, of whom 38/42 (90.5%) had detectable bNAB levels, 4/42 (9.5%) had undetectable levels. Women with detectable bNAb levels were younger (median 24 years, IQR 21–26) than those with undetected levels (median 28 years, IQR 24–36). Women with undetectable bNAb levels had a higher median number of pregnancies [2, IQR 1–2] compared with women with detectable levels [1, IQR 1–1], and most women reported that their partner was circumcised (94% of women with detectable bNAb levels and 100% of those with undetectable levels). Age at sexual debut was similar between women, with a median of 18 years (IQR 17–19) for those with detectable bNAb levels and 19 years (IQR 17–22) for those with undetectable levels. Condom use in the past 30 days was reported by 82% (31/38) of women with detectable bNAb levels and 75% (3/4) of those with undetectable levels, with Depo-Provera being the most commonly used contraceptive (66% (25/38) and 75% (3/4), respectively). BV prevalence by Nugent score did not differ markedly between groups. Among women with detectable bNAb, 37% (14/38) had no BV, 42% (16/38) had an intermediate microbiota, and 21% (8/38) were BV-positive. Similarly, in women without detectable bNAbs, 25% (1/4) had no BV, 50% (2/4) had an intermediate microbiota, and 25% (1/4) were BV-positive. No STIs were detected in any of the participants, and yeast was present in 13% (5/38) of women with detectable bNAb levels, compared with none in women with undetectable levels.

Table 1.

Behavioural and clinical characteristics of women by detection of mucosal bNAbs.

Variable	Characteristic	bNAbs detected N = 38 N (%)	bNAbs undetected N = 4 N (%)
Age	Median (IQR)	24 (21–26)	28 (24–36)
Race	Black	38 (100)	4 (100)
Education	Secondary school	30 (79)	4 (100)
	Tertiary	6 (16)	0 (0)
	Primary school	2 (5.3)	0 (0)
Number of sex partners	Median (IQR)	1 (1–1)	1 (1–1)
Partner circumcision	Yes	34 (94)	4 (100)
	No	1 (2.8)	0 (0)
	Don’t know	1 (2.8)	0 (0)
	Unknown	2	0
Number of pregnancies	Median (IQR)	1 (1–1)	2 (1–2)
Age at first sex	Median (IQR)	18 (17–19)	19 (17–22)
Sexual practice	Vaginal sex	25 (66)	2 (50)
	Oral sex	3 (7.9)	1 (25)
	Condom use	31 (82)	3 (75)
Contraception use	Depo provera	25 (66)	3 (75)
	Nuristerate	10 (26)	1 (25)
	Other	0 (0)	0 (0)
	Iucd	2 (5.3)	0 (0)
	Tubal ligation	0 (0)	0 (0)
	Implant use	4 (11)	1 (25)
STIs	Not detected	38 (100)	4 (100)
Yeast present	Negative	33 (87)	4 (100)
	Positive	5 (13)	0 (0)
BV category	No BV	14 (37%)	1 (25%)
	Intermediate	16 (42%)	2 (50%)
	BV by nugent	8 (21%)	1 (25%)

Microbial shift patterns and the probability of transitioning between CSTs during bNAb exposure

Based on the composition and relative abundance of the bacterial species, we assigned bacterial communities into three main CSTs (Fig. 1A). Lactobacillus-dominated communities, particularly L. crispatus and L. iners, were highly abundant in participants classified as CST I (5.3%, 2/38) and III (57.9%, 22/38), respectively, while diverse, non-Lactobacillus-dominated communities, including BVAB, Gardnerella vaginalis, Fannyhessea vaginae, Prevotella spp., and Sneathia sanguinegens, were prevalent in CST IV subtypes. BVAB-dominant was classified as CSTIV-A (2.6%, 1/38), whereas the Gardnerella vaginalis dominant state was classified as CSTIV-B (34.2%, 13/38). CSTIV-C was characterized by a mixed community dominated by Streptococcus.

Fig. 1.

Longitudinal vaginal microbiome dynamics (A) and transition probabilities of vaginal CSTs (B). CSTs were defined based on the dominant bacterial taxa: CST I (green, Lactobacillus crispatus-dominated), CST III (yellow, L. iners-dominated), and three BV-associated CSTs; CST IV-A (pink), CST IV-B (red), and CST IV-C (blue). Gray denotes missing samples. CST assignments for each participant (n = 38) across seven time points from Day 0 (pre-dose) to Day 168 (A). The heatmap summarizes the likelihood of transitioning from one CST to the other. Diagonal values indicate the probability of CST stability (B).

Over time, several participants exhibited transitions between community state types (CSTs). Lactobacillus-dominated communities tended to persist, whereas BV-associated, high-diversity communities were more dynamic, with some shifting toward Lactobacillus-dominant states (Fig. 1B). Participants initially classified as CST I showed a strong tendency to transition to CST III [80%, 8/10; probability = 0.8], and a smaller proportion transitioned to CST IV-B [1%, 1/10; probability = 0.1]. Similarly, CST III participants were most likely to remain in the same state [79%, 59/74; probability = 0.79%], although 12% (9/74) transitioned to CST IV-B (probability = 0.12) and some transitioned to CST IV-C [2%, 2/74; probability = 0.02], a mixed community characterized by the presence of Streptococcus. CST IV-B exhibited a high likelihood of transitioning to CST III [27%, 11/41, probability = 0.27] or to CST IV-C [2%, 1/41); probability = 0.02]. Overall, Lactobacillus-dominant communities demonstrated greater temporal stability, whereas high-diversity CST IV profiles exhibited greater transition frequencies; however, these observations were descriptive and not derived from inferential statistical modeling.

Baseline alpha diversity and spatial segregation of the vaginal microbiome by bNAb detection

We next examined whether baseline vaginal microbial diversity differed by Lactobacillus dominance and mucosal bNAb detection status. Alpha diversity was assessed using four indices: Observed species, Chao1, Shannon, and inverse Simpson. Among women with detectable mucosal bNAbs, alpha diversity was significantly higher in those with non-Lactobacillus-dominant communities compared to those with Lactobacillus-dominant communities across all indices (Observed: mean 35 vs. 12, p < 0.001; Chao1: 35 vs. 12, p < 0.001; Shannon: 2.17 vs. 0.90, p < 0.001; InvSimpson: 6.29 vs. 2.19, p < 0.001; Fig. 2A). In contrast, among women without detectable mucosal bNAbs, alpha diversity did not differ between the groups (all p > 0.9). Median values for richness and diversity indices were comparable (Observed: 29 vs. 28; Shannon: 1.65 vs. 1.91; InvSimpson: 3.41 vs. 5.02; Fig. 2A). To evaluate the possible impact of vaginal microbial communities on bNAb detection, we performed a non-metric multidimensional scaling (NMDS) on Bray-Curtis distance at baseline. The presence of bNAb did not influence spatial segregation. No clustering was detected based on detectable bNAb (Fig. 2B).

Fig. 2.

Alpha diversity indices at baseline by community state type and antibody detection (A). Bar graphs show Observed species, Chao1, Shannon, and inverse Simpson diversity indices in women with detectable (left panel) and undetectable (right panel) mucosal antibodies, stratified by Lactobacillus-dominant and non-Lactobacillus-dominant communities. Non-metric multidimensional scaling (NMDS) of Bray-Curtis showing no clustering dissimilarity of species-level relative abundances in bNAb detected and undetected groups at baseline (B). Colour indicates CST classification, while shape represents mucosal bNAB detection status (circles for detected and triangles for undetected).

Mucosal bNAb levels are consistent across community state types following systemic administration

Next, we evaluated whether the cervicovaginal microbiome environment has a measurable effect on how long and how much of the bNAbs are present in the female genital tract post administration and whether this effect is consistent across different CSTs. Consistent with known pharmacokinetic principles, IV administration resulted in the earlier detection of bNAbs at the mucosal surface. In contrast, SC dosing resulted in delayed peak mucosal concentrations, reflecting slower systemic absorption and subsequent bNAb transport to mucosal tissues^25,26. These route-dependent kinetic patterns were observed across women harboring diverse CST profiles, including Lactobacillus-dominant and non-Lactobacillus-dominant microbiomes (Fig. 3). No clear CST-specific differences in the timing or magnitude of mucosal bNAb accumulation were observed. Within the limits of this cohort, we did not observe evidence that cervicovaginal microbiome composition substantially altered the transudation or measured mucosal concentrations of systemically administered bNAbs.

Fig. 3.

Longitudinal mucosal concentrations of broadly neutralizing antibodies CAP256V2LS (blue circles) and VRC07-523LS (red circles) measured in cervicovaginal secretions of women following IV or SC administration at varying doses (5–20 mg/kg). Each panel represents an individual subject, identified by unique study ID. Microbiome composition was classified into CSTs, indicated by color-coded boxes below each plot: yellow (CST I), green (CST III), pink (CST IV-A), red (CST IV-B), and blue (CST IV-C). Filled circles indicate time points where cervicovaginal microbiome profiling was performed, while open circles indicate time points without microbiome data. Data are presented on a logarithmic scale over time (days post-administration).

Discussion

BNAbs are a novel approach in HIV prevention and cure strategies²⁷. However, it remains unclear whether the composition of the cervicovaginal microbiota influences bNAb concentrations and persistence within the female genital tract. In this study, we investigated longitudinally whether the cervicovaginal microbiome modulates the magnitude or duration of mucosal antibody concentrations following intravenous or subcutaneous administration of CAP256V2LS and VRC07-523LS. In this sub-analysis of women who received bNAbs via systemic administration, we did not observe clear evidence that cervicovaginal microbial composition, defined by CSTs, was associated with differences in the timing or magnitude of mucosal bNAb accumulation.

Previous studies have demonstrated that cervicovaginal microbiome with low Lactobacillus abundance and high diversity is significantly associated with elevated genital inflammation, altered barrier function, as well as reduced efficacy of topical (and not oral) antiretroviral products through microbial metabolism of active drug or enhanced mucosal inflammation^{13,15,16,28,29}. Given this precedent, it was critical to determine whether vaginal microbiome could similarly affect the pharmacokinetics or persistence of systemically delivered bNAbs that transudate to the genital tract. Our findings suggest that bNAb pharmacokinetics in the systemic compartment and route of administration remain the dominant determinants of mucosal bNAb levels, with minimal influence from microbiome composition. In addition, consistent with this, L. iners-dominated CST III exhibited relative stability, whereas high-diversity CST IV profiles were more variable and frequently transitioned toward Lactobacillus-dominated states. Notably, these microbial shifts did not measurably affect systemic bNAb pharmacokinetics. In this sub-analysis, we did not observe evidence of the association between cervicovaginal microbial community state type and the timing or magnitude of mucosal bNAb accumulation following systemic administration. While prior studies have demonstrated that high-diversity vaginal microbiota may reduce efficacy of some topical antiretroviral products, our findings suggest that, in this cohort, systemic pharmacokinetics and route of administration appeared to be the primary determinants of mucosal bNAb levels^16,29,30. While topical microbicides are directly exposed to vaginal microbial enzymes¹⁶ and inflammatory mediators³¹ that can degrade or inactivate the drug, systemically administered antibodies reach the mucosa primarily through plasma exposure and Fc receptor-mediated transport³⁰. These mechanisms did not appear to be measurably altered by differences in cervicovaginal microbial community structure, although subtle effects cannot be excluded.

The strength of this study includes the integration of microbiome profiling with mucosal pharmacokinetic measurements in the context of a first-in-human bNAb trial testing CAP256 V2 LS in combination with VRC07 523-LS. Nonetheless, certain limitations should be acknowledged. Being a Phase 1 study, the modest sample size, particularly the small number of women without detectable mucosal bNAbs, limited statistical power for between-group comparisons, constrained our ability to detect more subtle microbiome-associated effects, and less common CST subtypes were underrepresented. In addition, given the exploratory nature of this sub-analysis, participants were not stratified or normalized according to administration route (IV versus SC), dose level, HIV status, or receipt of combination versus single-bNAb therapy. These groups were therefore analyzed descriptively to assess overall trends in mucosal bNAb detection across CSTs. Consequently, clinical and pharmacokinetic heterogeneity within the cohort may have confounded subtle differences in mucosal bNAb distribution associated with cervicovaginal microbiome composition. Larger studies incorporating formal pharmacokinetic modelling and stratified analyses will be necessary to determine whether these variables independently influence mucosal bNAb pharmacokinetics within the female genital tract. The Phase 1 design and inclusion of low-risk women may limit generalizability to broader populations, particularly those with higher STI prevalence. Additionally, we did not assess additional mucosal immune factors, such as cytokines, to assess the state of genital inflammation or immunoglobulin subclasses or isotypes, that may account for bNAb distribution. In the absence of data regarding epithelial barrier integrity and ensuing proteolytic activity, our understanding of whether the bNAbs remain functional is limited. Furthermore, functional assays evaluating neutralization capacity or Fc-mediated bNAb effector function within cervicovaginal secretions were not performed.

In conclusion, we did not observe evidence that cervicovaginal microbial composition was associated with differences in mucosal concentrations of systemically administered bNAbs. These findings suggest that systemic delivery achieves measurable genital tract exposure across diverse vaginal microbiome states; however, larger studies with integrated assessments of mucosal inflammation, epithelial integrity, and antibody functionality are required to confirm these observations.

Author contributions

AM and AHT conducted the laboratory assays. NZ, DA, and AS conducted the PK analysis. SM and NG contributed to clinical investigations and sample collection. AM, SN, ML, and AK contributed to data analysis and interpretation. AM and SN wrote the first draft of the manuscript. All authors reviewed the final draft of the manuscript and approved the final version.

Funding

This study and the CAP012B clinical trial was supported principally by the European and Developing Countries Clinical Trials Partnership (EDCTP Grant number: RIA2017S). AM received funding from NRF Thuthuka (REF: TTK240315209308).

Data availability

The datasets generated and/or analysed during the current study are available in the ENA repository, [Accession: PRJEB108584].

Declarations

Ethics approval

The study protocol was approved by the University of KwaZulu-Natal Biomedical Research Ethics Committee (BREC00000857/2019 and the South African Health Products Regulatory Authority (SAHPRA 20200123).

Competing interests

The authors declare no competing interests.