Question

Disease Pathophysiology Research — Bacterial Vaginosis (MONDO:0005316)

Focus: content-completeness sweep for the dismech entry created in PR #9073. The entry was curated directly from primary literature without a deep-research provider report, and the PR review (2026-08-20) blocked on the absence of one, specifically because a completeness sweep is the kind of pass that catches missing licensed therapies — which it did.

Method: PubMed E-utilities searches (esearch + esummary + efetch) across eight dimensions, with every PMID, title, journal, year and publication type read back from the NCBI record rather than recalled. Numeric results quoted below are from the fetched abstracts.

Standing caveat. Everything here is a lead. No statement below may be copied into a YAML snippet:. Any PMID used for curation must first be fetched with just fetch-reference and the exact quote verified with just count-verified-snippets.

1. Treatment landscape — the confirmed gap

The entry as first written curated metronidazole, clindamycin, LACTIN-V and partner treatment. The sweep found three further agents with randomised data, two of them FDA-approved. All three have been added to the entry in this revision.

Notes that shaped how these were curated:

• Secnidazole and tinidazole are nitroimidazole class-mates of metronidazole, so they attach to the same Polymicrobial Anaerobic Overgrowth node and not to the ribosome node that carries clindamycin. Neither was tested against metronidazole for superiority; both were placebo-controlled, so no comparative efficacy claim is warranted.
• The tinidazole trial's cure rates (36.8% and 27.4% vs 5.1% placebo) look far worse than the 70–85% quoted elsewhere in the entry. This is an endpoint artefact, not a drug difference — Livengood used the strict five-criterion FDA definition, and the same paper states efficacy was greater under traditional criteria. Curating the number without the caveat would have produced a false comparison inside one entry. Both halves are now curated.
• Dequalinium is mechanistically distinct (membrane-active antiseptic, no resistance-selection pressure of the nitroimidazole kind), which is the trial's stated rationale.

Still uncurated treatment leads

• Astodrimer sodium 1% gel (dendrimer, approved in Europe/Australia, not US). Mini-review PMID:35246717 reports BV cure rates of 50–57% and a recurrence-prevention result against placebo. A dendrimer that acts by physically blocking adhesion would attach to the biofilm/adhesion node rather than the overgrowth node — i.e. it would be the only agent in the entry targeting the arm that metronidazole is explicitly curated as not reaching. Worth a follow-up pass; the primary trials, not the mini-review, should be the cited source.
• Vaginal microbiome transplantation (VMT). PMID:31591599 (Nat Med 2019, NCT02236429) is an exploratory case series of five women with intractable BV; four achieved long-term remission with reconstitution of a Lactobacillus-dominated microbiome, some requiring repeat transplant and one a donor change. This is a five-patient uncontrolled series and should be curated, if at all, as an EMERGING hypothesis-linked intervention with the sample size in the description — not as an established treatment. A 2025 engraftment preprint exists (PMID:40909844) and is explicitly a preprint.
• Boric acid and TOL-463 returned no BV-specific randomised evidence in this sweep. Do not curate from the searches above.

2. Adverse outcomes — where the entry is thin, and one genuine controversy

The entry curates Premature birth as OCCASIONAL from a single cohort (9/115). The sweep found the meta-analytic literature, and it contains a live dispute that is more interesting than the point estimate:

• PMID:36251068 (2023, 20 articles, 290,397 observations): OR 1.79 (1.32–2.43), RR 1.44 (1.19–1.73). Concludes BV is "undoubtedly associated" with preterm birth and argues for routine screening.
• PMID:39442804 (2025, 28 studies, 50,466 patients): OR 1.60 (1.36–1.89), I²=67%. Concludes the association is weaker than previously documented, and explicitly offers that as an explanation for why treating BV in pregnancy has not reduced preterm birth.
• PMID:36651636 (2023): individual-participant-data meta-analysis of antibiotic treatment of BV to prevent preterm delivery.

This is a candidate mechanistic_hypotheses / discussions item rather than a frequency band: a robust association whose causal interpretation is undermined by the failure of the corresponding intervention. The entry's existing Ascending Infection and Adverse Pregnancy Outcomes node is the attachment point, and the treatment-failure argument is exactly the sort of constraint this entry already handles well elsewhere (see its retained REFUTE items).

Other outcome leads with real effect sizes:

• PID: PMID:34396403 (Longitudinal Study of Vaginal Flora, N=2956, prospective) — Nugent-BV aHR 1.53 (1.05–2.21), symptomatic Amsel-BV aHR 2.15 (1.23–3.75), vaginal douching aHR 1.47 (1.03–2.09). Note this cuts against the entry's retained REFUTE item (PMID:35086915, no association between a Gardnerella-dominated microbiome and subsequent PID). The two are not necessarily inconsistent — one measures a taxon, the other measures the syndrome — but curating both would materially improve the PID arm, and the distinction between them is itself the finding.
• HIV acquisition: PMID:18614873 is already curated. Additional: PMID:22745608 (female-to-male transmission, prospective African couples), PMID:21358808 (IPD meta-analysis of intravaginal practices).
• Infertility: PMID:23543384 (Hum Reprod 2013 meta-analysis) — BV more prevalent in infertility than antenatal controls (OR 3.32, 1.53–7.20), and more prevalent in tubal infertility specifically (OR 2.77, 1.62–4.75), but not associated with decreased conception rates (OR 1.03, 0.79–1.33) though associated with preclinical pregnancy loss. The negative conception result is the more useful curation target, because it constrains the mechanism.

3. Mechanism — the sexual-transmission hypothesis has a named model

The entry curates recurrence as two hypotheses (biofilm_persistence_relapse CANONICAL, sexual_reinfection EMERGING). The sweep found that the transmission account is not only a recurrence hypothesis but a primary-pathogenesis model with an explicit published statement:

• PMID:31369673 — Muzny et al., "An Updated Conceptual Model on the Pathogenesis of Bacterial Vaginosis" (J Infect Dis 2019). States that BV epidemiology supports sexual transmission and that the central debate is primary pathogen versus sexually transmitted polymicrobial consortium. Its update responds to the objection that G. vaginalis is found in virginal women, by proposing that the genus contains 13 species and that healthy women carry non-pathogenic ones while virulent strains cause BV. Also names Prevotella bivia and Atopobium vaginae as model components.
• PMID:24511102 — the earlier (2014) version of the same conceptual model.
• PMID:30001418 — phenotypic characterisation of G. vaginalis subgroups suggesting they differ in virulence potential.

Why this matters for the entry as written. The entry currently records that P. bivia does not adhere (from PMID:39162399) and treats the consortium as the pathogenic unit. The Muzny model is a genuinely competing framing — a virulent-strain model, where identity within Gardnerella is doing the work that the entry attributes to the community. That is a candidate ALTERNATIVE mechanistic_hypotheses group, and it also bears on the entry's HUMAN_MODEL_MISMATCH, since a strain-resolved model changes what an animal or organ-chip system would have to reproduce.

4. Host genetics — an absent section

The entry has no genetic: section. There is real, if modest, literature:

• PMID:17314118 — IL-1B −511 CC (OR 1.5, 1.03–2.14) and +3954 TT (OR 2.8, 1.37–5.88) associated with BV in non-pregnant Italian women; the −511/+3954 T-C haplotype protective (OR 0.7).
• PMID:17123692 — TNFA −308G>A and the TNF-α response to altered vaginal flora.
• PMID:23021866, PMID:19200604 — TLR variants and BV / Gardnerella–Atopobium carriage.
• PMID:32723796 — host genetic factors associated with vaginal microbiome composition (Kenyan women).

Recommendation: curate with restraint or not at all. These are small single-population candidate-gene association studies of the kind that replicate poorly, and no GWAS surfaced. If curated, relationship_type: SUSCEPTIBILITY with the population named in the description, and the absence of replication stated. A KNOWLEDGE_GAP recording that BV has no GWAS-scale host-genetic evidence may be worth more than the individual associations.

5. Immune / barrier mechanism — one strong addition

• PMID:32094253 (Infect Immun 2020) — host matrix metalloproteinases secreted in response to BV-associated bacteria disrupt endocervical epithelial polarisation and increase HIV-1 transmigration through the epithelium; an MMP inhibitor reduces the effect, and cervicovaginal fluid with higher BV-associated MMP concentrations increases transmigration.

This is a host-derived barrier-damage mechanism running in parallel to the bacterial glycosidase mechanism the entry already curates (sialidase/fucosidase stripping the glycocalyx). The entry's barrier arm is currently entirely bacterial-enzyme-driven; this would add the host-protease arm and would sit naturally upstream of the HIV-susceptibility node. Strongest single mechanistic addition found by this sweep.

Supporting: PMID:24403560 (species-specific effects of vaginal taxa on innate immunity and barrier properties), PMID:32515473 (Prevotella species modulate barrier function in a 3D endometrial epithelial model — relevant to the entry's HUMAN_MODEL_MISMATCH, since it is a human 3D model rather than an animal one), PMID:18403235 (Atopobium vaginae innate immune response in vitro).

6. Diagnostics — molecular assays

The entry curates Amsel and Nugent as definitions. Molecular assays are absent: PMID:22535982 (validated semiquantitative multitarget PCR), PMID:20814710 (qPCR vs Gram stain accuracy), PMID:38061216 (real-time PCR for intermediate Nugent scores). A molecular definition would be a genuine addition, because the entry already argues that BV's prevalence range is partly an artefact of diagnostic method — a third method with its own operating characteristics is the evidence for that argument.

7. Sialidase source — already correctly curated

PMID:39186657 (PNAS 2024, "Prevotella are major contributors of sialidases in the human vaginal microbiome") is already cited in the entry. The sweep confirms this is the current state of the art on sialidase provenance and that the entry's choice to treat sialidase as a consortium property rather than a Gardnerella property is the correct reading.

8. Datasets

No datasets: block exists on the entry. This sweep did not run a repository search; just discover-datasets Bacterial_Vaginosis and just verify-datasets are the right tools and were not run here. Flagged as an open item rather than answered.

Summary of what this sweep changed

Acted on in this revision: section 1 (three treatments added), plus the review's other two blocking findings.

Recommended follow-ups, in descending value:

1. Host-MMP barrier-disruption arm (PMID:32094253) — section 5.
2. Preterm-birth association-vs-intervention controversy as a discussion — section 2.
3. Muzny virulent-strain conceptual model as an ALTERNATIVE hypothesis — section 3.
4. PID prospective cohort alongside the existing REFUTE item — section 2.
5. Astodrimer sodium as a biofilm/adhesion-targeting agent — section 1.
6. Molecular diagnostic definition — section 6.
7. Host genetics, with restraint, or as a knowledge gap — section 4.
8. Dataset discovery — section 8.

These are deliberately not all done in this PR. The PR is a review-response on an entry that is already approved-in-substance; folding eight new arms into it would make the diff unreviewable and would re-open questions the reviewer has already settled. They belong in a follow-up pass.

References surfaced by this sweep

All PMIDs below were returned by PubMed E-utilities during this sweep and their titles read back from the NCBI record.