Vaginal microbiome and cervical antiretroviral exposure: pharmacometric and machine learning analysis
Specific vaginal taxa — not overall microbiome diversity — are associated with variability in cervical exposure to tenofovir, lamivudine, and emtricitabine, with bacterial vaginosis-associated genera showing an unfavorable direction.
| Population | Women receiving tenofovir-, lamivudine-, or emtricitabine-containing regimens |
|---|---|
| Exposure | Vaginal microbiome characterization by sequencing + pharmacometric modeling of cervical drug exposure |
| Comparator | Absence vs. presence of specific taxa (internal comparisons via nonlinear mixed-effects and machine learning models) |
| Outcome | Cervical tenofovir (TFV) exposure; Cervical lamivudine (3TC) exposure; Cervical emtricitabine (FTC) exposure; Overall microbiome diversity and cervical drug exposure |
Summary of findings
| Outcome | Effect | 95% CI | Certainty | Clinical relevance | Notes |
|---|---|---|---|---|---|
| Cervical tenofovir (TFV) exposure | Association with Gemella, Megasphaera, Falsiporphyromonas; effect size not reported | — | Low | — | 1 studies |
| Cervical lamivudine (3TC) exposure | Association with Dialister, Prevotella, Atopobium, Streptobacillus, Gardnerella; effect size not reported | — | Low | — | 1 studies |
| Cervical emtricitabine (FTC) exposure | Association with Bifidobacteriaceae, Lachnospiraceae, Prevotellaceae; effect size not reported | — | Low | — | 1 studies |
| Overall microbiome diversity and cervical drug exposure | No significant association; alpha/beta diversity not associated with TFV, 3TC, or FTC cervical exposure | — | Low | — | 1 studies |
Context
Oral PrEP with TDF/FTC is less effective in women than in men who have sex with men, a gap not fully explained by adherence alone. The vaginal microbiome may modulate local antiretroviral pharmacokinetics, potentially compromising HIV protection in women — particularly in sub-Saharan Africa, where incidence in young women is more than three times that of young men.
What the study showed
Overall alpha/beta microbiome diversity was not associated with cervical exposure to any of the three drugs. Specific taxa showed significant associations: Gemella, Megasphaera, and Falsiporphyromonas with TFV; Dialister, Prevotella, Atopobium, Streptobacillus, and Gardnerella with 3TC; Bifidobacteriaceae, Lachnospiraceae, and Prevotellaceae with FTC. Absolute effect sizes (e.g., concentration differences in ng/mL) and individual 95% CIs per taxon were not reported in the available text, limiting precise quantification.
How it was done
Observational study combining nonlinear mixed-effects pharmacometric (PK) modeling with machine learning algorithms to analyze high-dimensional microbiome data alongside cervical antiretroviral concentrations. Sample size (n) and follow-up duration were not specified in the available excerpt. Population: women using TFV, 3TC, or FTC.
Effect magnitude
Effect sizes (RR, OR, SMD, or concentration differences with 95% CI) were not provided in the available text; statistical significance of identified taxa is reported as 'significant' by the authors without accessible numeric values in this excerpt.
Risk of bias
Observational design precludes causal inference between taxa and reduced drug exposure. Sample size, risk-of-bias tool (ROBINS-I would be most applicable), and adjustment for confounders (adherence, menstrual cycle, co-infections) are not detailed in the excerpt. High dimensionality of microbiome data increases overfitting risk in machine learning models; external validation is not mentioned.
What this study does NOT prove
This study does not prove that identified taxa cause reduced cervical drug exposure or that microbiome modification improves PrEP efficacy. Results cannot be generalized without validation in populations with different microbiological and epidemiological profiles.
In clinical practice
Evidence is insufficient to change clinical practice based on this single study. Clinicians should note that bacterial vaginosis and taxa such as Gardnerella and Prevotella may be associated with lower cervical antiretroviral exposure, supporting screening and treatment of bacterial vaginosis in PrEP users. Await prospective validation before any microbiome-guided dosing adjustment.
Limitations
Observational design precludes causal inference between taxa and reduced drug exposure. Sample size, risk-of-bias tool (ROBINS-I would be most applicable), and adjustment for confounders (adherence, menstrual cycle, co-infections) are not detailed in the excerpt. High dimensionality of microbiome data increases overfitting risk in machine learning models; external validation is not mentioned.
What is still missing
Prospective studies with direct cervical concentration measurements, quantified effect sizes, and independent validation of machine learning models in distinct cohorts are needed to establish causality and clinical relevance.
Technical appendix
Version history
- 1.0 · 2026-11-15 — Auto-generated under Evidence Standard v1.0
