Dynamic, transition and variation of cervicovaginal microbiome and HPV infection and cervical dysplasia and cancer: a systematic review

Abstract

Background

Cervical cancer is the fourth most common malignancy in women worldwide, with approximately 660,000 new cases and 350,000 deaths annually. The burden falls disproportionately on low- and middle-income countries. Although persistent infection with high-risk HPV (hrHPV) is the necessary cause, most infected women clear the virus spontaneously, implicating additional cofactors, including the cervicovaginal microbiome in determining oncogenic outcomes.

Methods

PubMed was searched through September 10, 2024, to identify longitudinal studies assessing cervicovaginal microbiota in relation to HPV infection or cervical lesion outcomes at two or more time points. Methodological quality was evaluated using the Newcastle-Ottawa Scale (NOS). Given the substantial heterogeneity, a structured thematic synthesis was performed across three predefined domains: (a) baseline microbiome composition and clinical outcomes; (b) community state type (CST) dynamics and temporal stability; and (c) microbiome changes following treatment.

Results

Twelve studies enrolling 1,663 women across 11 countries met inclusion criteria. NOS scores ranged from 4 to 9. Lactobacillus-dominated CSTs at baseline were consistently associated with HPV clearance and CIN regression, while Lactobacillus-depleted states showed higher transition rates and unfavourable outcomes. Prior L.iners (CST III) dominance was repeatedly linked to favourable outcomes, although evidence on this species remains conflicting. Cervicovaginal dysbiosis frequently preceded HPV persistence or lesion progression.

Conclusion

Sustained Lactobacillus-dominated CST stability, rather than dominance by any single species, is the most consistent microbiome factor associated with favourable HPV and cervical lesion outcomes. Standardized longitudinal designs incorporating metagenomic sequencing, frequent sampling intervals, and rigorous confounder adjustment are needed to advance mechanistic understanding.

Clinical trial registration number

Not applicable.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13027-026-00777-0.

Introduction

Cervical cancer ranks as the fourth most common cancer in women worldwide, accounting for an estimated 660,000 incident cases and 350,000 deaths per year [1, 2]. Its burden is disproportionately concentrated in low- and middle-income countries, particularly in sub-Saharan Africa, South Asia, and Latin America, where access to vaccination and organized screening remains limited [1]. Persistent infection with hrHPV, most notably HPV-16 and HPV-18, which collectively account for approximately 70% of cases, is the necessary cause of virtually all cervical cancers [3]. Despite this, most hrHPV-infected women clear the infection within two years through immune-mediated mechanisms, and only a minority progress to CIN or invasive carcinoma [4, 5]. This discrepancy underscores the role of modifying cofactors in disease progression.

Among the cofactors implicated in hrHPV persistence and cervical carcinogenesis, including smoking, hormonal contraception, immunosuppression, and genetic susceptibility [6, 7], the cervicovaginal microbiome has attracted growing attention. The vaginal tract of reproductive-age women is characteristically dominated by Lactobacillus species, principally L. crispatus, L. iners, L. gasseri, and L. jensenii, which sustain an acidic cervicovaginal pH, produce bacteriocins and hydrogen peroxide, and maintain a relatively anti-inflammatory mucosal environment [8–10].

Disruption of this dominance by anaerobic genera such as Gardnerella, Prevotella, Atopobium, and Sneathia constitutes the dysbiotic state associated with bacterial vaginosis and increased susceptibility to sexually transmitted infections [11, 12]. Cross-sectional evidence consistently links Lactobacillus-depleted communities to higher rates of hrHPV prevalence, persistence, and neoplastic progression [7, 13–15]. A systematic review and network meta-analysis of 11 studies established that CST IV (low lactobacilli) and L. iners-dominated states carry higher odds of hrHPV detection, dysplasia, and cancer relative to L. crispatus dominance; however, only three of those studies were longitudinal, limiting causal inference [15]. A subsequent review reported a temporal relationship between cervicovaginal dysbiosis and cervical carcinogenesis [16] but did not systematically analyze CST transition dynamics across serial timepoints or formally appraise study quality.

To address this gap, the aims: the temporal behaviour of the cervicovaginal microbiome in relation to HPV infection and cervical lesion evolution has not been comprehensively synthesized. To address this gap, the aims of this systematic review are: (a) to synthesize longitudinal evidence on the influence of cervicovaginal microbiome dynamics on HPV clearance or persistence and CIN regression or progression; (b) to identify bacterial taxa and CST patterns consistently associated with these outcomes across studies; and (c) to critically appraise methodological heterogeneity and its implications for future research design.

Methods and materials

Search strategy

PubMed was searched from inception through September 10, 2024, using the following MeSH terms and free-text keywords: “cervicovaginal microbiota”, “vaginal microbiome”, “human papillomavirus”, “HPV”, “cervical intraepithelial neoplasia”, “cervical dysplasia”, “cervical cancer”, “longitudinal study”, “cohort”, and “community state types”. The search was restricted to English-language publications in humans. Reference lists of eligible articles were manually screened to identify additional studies. Restriction to a single database (PubMed) is acknowledged as a limitation.

Eligibility criteria

Studies were included if they: (1) assessed cervicovaginal microbiota using 16 S rRNA gene sequencing or an equivalent culture-independent method; (2) measured microbiota composition and HPV infection status or cervical lesion grade at two or more timepoints in the same participants; and (3) employed a prospective or retrospective longitudinal cohort design. Studies with a single microbiota assessment, purely cross-sectional designs, and non-human studies were excluded. One potentially eligible study was excluded because outcomes were reported only at aggregate level by cumulative HPV viral load, precluding individual-level clinical classification [17].

Data extraction and screening

Title, abstract, and full-text screening were performed by one reviewer, with supplementary screening of reference lists. Extracted variables included: first author, publication year, country, study design, sample size, population characteristics, follow-up duration and sampling intervals, HPV classification method, cervical lesion staging, microbiota characterisation technique and sequenced 16 S rRNA region, alpha-diversity metrics. Single-reviewer screening represents a methodological limitation and is acknowledged in the Limitations section.

Quality assessment

The methodological quality of all included studies was evaluated using the Newcastle-Ottawa Scale (NOS) for cohort studies. The NOS assesses three domains: selection (maximum 4 stars), comparability (maximum 2 stars), and outcome ascertainment (maximum 3 stars), yielding a total maximum score of 9. Studies scoring 7 or above were classified as high quality, 5–6 as moderate, and 4 or below as low quality (Table S2).

Data synthesis

Due to substantial clinical and methodological heterogeneity in study populations, follow-up intervals, CST classification schemes, sequencing methods, and outcome definitions, meta-analysis was not feasible. Findings were synthesized using a structured thematic approach organized around three predefined themes aligned with the review objectives: (1) baseline microbiome composition and clinical outcomes; (2) CST dynamics, transition patterns, and temporal stability; and (3) microbiome changes following cervical lesion treatment. Within each theme, the consistency of associations is summarized semi-quantitatively in Table S1: ++ denotes a consistent association across three or more studies; + in one to two studies; +/- denotes conflicting evidence; and - denotes absent or negative association. Given its recurring and contested associations across all three themes, the role of L. iners is addressed in a dedicated subsection.

Results

Search results and study characteristics

The PubMed search retrieved 1,430 records. Titles and abstracts screening led to 60 full-text assessments and 12 studies meeting inclusion criteria (Fig. 1). One additional potentially eligible study was excluded due to lack of individual-level outcome data, with individual study samples ranging from 14 to 448 participants. Seven studies examined the natural history of HPV infection or cervical lesions [18–25], and sevem assessed microbiome changes before and after cervical lesion treatment [26–29]. Study characteristics are summarised in Table S1, whereas detailed methodological features and principal findings of the included longitudinal studies are presented in Table 1. Key descriptive features are presented in Figure S1. Studies were conducted in the USA (n = 3) [19, 20, 25], China (n = 3) [23, 26, 29], and one each in the Netherlands [22], Sweden [18], South Africa [24], Costa Rica [21], the United Kingdom [27], and Italy [28]. Follow-up durations ranged from 16 weeks [19] to three years [23]; one study did not report this parameter [29]. Alpha-diversity was quantified using the Shannon index in sevem studies [18, 21, 23, 26, 29] and species richness or Chao1 in one [27]; beta-diversity was assessed using Bray-Curtis dissimilarity in three studies [18, 22, 25].

Fig. 1.

PRISMA Flow chart summarizing the identification, inclusion, and exclusion of research articles

Table 1.

Summary of 12 selected studies

Author Year Country	Duration of sample collections and assessments	Sample/Population studied	Methods	Main findings
Brotman et al. [19] United States of America	Twice weekly for 16 weeks (32 times)	-39 reproductive-age (18 or older) women. -During the follow up period, 5 were always HPV negative, 2 were positive for 1 HPV subtype, and 25 were positive for 2 or more HPV subtypes.	-Self-collected midvaginal swabs. -The V1-V2 hypervariable regions of 16 S rRNA genes -Samples were screened for 37 HPV types: all 13 women > 40 years old tested positive for HPV	-CST II (L. gasseri–dominated) was associated with the fastest HPV remission rate (HPV positive to no detection) and CST IV-B was associated with the slowest HPV remission rate compared to the CST I (L. crispatus-dominated. -Among the 11 persistent HPV women, 4 were most often in CST IV-B, 1 in L. crispatus–dominated CST I, and 6 in L. iners–dominated CST III.
Berggrund et al. [18] Sweden	5–6 months for HPV 16 clearance and 18 months for persistent HPV 16 infection and CIN2+ (2 times)	-96 women aged 30–49 years old. -32 HPV negative (single self-sample with a negative HPV) -26 transient HPV 16 infection -38 persistent HPV16 infection and diagnosed CIN2+	- Vaginal self-sampling using the Rovers Viba-brush -16 S RNA gene sequencing with variable regions V2, V3, V4, V6, V7, V8 and V9	-Alpha diversity did not differ between three groups in baseline neither for second visit. -Women with persistent HPV16 infection (34%) reported more transition between cluster than those who cleared the HPV16 infection (19%)
Mitra et al. [20] United Sates of America	Baseline, 12 and 24 months (3 times)	-87 women aged between 16 to 26 years old with histologically confirmed, untreated CIN2 -comparison groups: regression and non-regression of CIN2	-Cervical samples with a cytobrush and spatula - Illumina MiSeq sequencing of 16 S rRNA gene amplicons was used with V1–V2 hypervariable regions	- Lactobacillus-depleted vaginal microbiota (VMB) at baseline had a tendency to regress slower than lactobacillus-dominant, CST I or III. - At 12 months, the most stable CST in non-regressors was CST I and for the 12–24-month CST I was the most stable VMB in regressors and CST IV most stable in non-regressors. At 12 the most frequent transition was from CST IV to CST III in regressors compared to CST III to CST IV in non-regressors, but at 12–24 months the most frequent transition was from CST III to CST IV in regressors, and CST I to CST III in non-regressors.
Moscicki et al. [25] United Sates of America	4–6-month intervals for 8 consecutive visits (8 times)	− 14 healthy women aged 16–25 years old with normal cytology, who had acquired HPV 16 during follow-ups. -grouped as HPV 16 persistent, HPV16 clearance and HPV16 persistent negative	-The hypervariable regions V1-V3 of the 16 S ribosomal RNA (rRNA) gene were amplified -Cervical washes were obtained at each visit for microbiota analysis	-During HPV infection the L. iners-dominated abundance was reduced and then increased last post-clearance. -Women infected by HPV 16 were more likely to transit between the 3 states than those with no history of HPV and had a higher probability of remaining in not lactobacillus-dominated CST. -Post-clearance visits showed increased alpha diversity when compared to the other states. -Enterococcus had an inverse relationship with G. vaginalis which has increased in relative abundance at immediate pre-clearance.
Van de Wijgert et al. [24] South Africa	A median of 16-month intervals of time	-448 women HIV+ aged 25–50 years old enrolled in the two sub-studies. -The hrHPV sub-study (N = 304) -The CIN sub-study (N = 144) included women with CIN2 or higher (CIN2+)	- The 16 S rRNA gene sequencing with V3-V4 regions -Vaginal swabs was collected for microbiota analysis	-L. crispatus or L. jensenii-dominated or high diversity BV-anaerobes VMB types at baseline were most likely to have the same VMB type at end-line. -Not significant difference was observed between baseline VMB compositions of women who acquired or cleared hrHPV during the study and the persistent hrHPV-negative controls. -hrHPV + at both visits compared to persistent hrHPV-negative controls was associated with baseline anaerobic dysbiosis. -Be CIN2 + at least once compared to persistent hrHPV-positive controls was associated with endline anaerobic dysbiosis, but when compared to persistent hrHPV-negative controls, CIN2 + at least once was associated with anaerobic dysbiosis at both baseline and end-line.
Usyk et al. [21] Costa Rica	-Average sampling interval was 1.5 SD: ±0.9 years - For progression, visit 2(V2) was the closest visit before diagnosis of CIN2	-273 Women aged 18–25 years old with an incident HR-HPV infection Grouped in hrHPV clearance, hrHPV persistent and CIN2 + progression groups	-Cervical brush samples were collected for HPV and microbiome analysis -The V4 variable region of the 16 S rRNA gene sequencing	-There was no difference in alpha diversity at baseline visit between groups, but at second visit there was significant, increased microbial diversity associated with progression to CIN2+. -At baseline visit, the relative abundance of G. vaginalis and L. iners had the highest positive correlation to progression and clearance respectively. -A significant protective effect of Lactobacillus (genus) abundance at baseline visits and microbial diversity at second visit (mediated effect of baseline visit Gardnerella abundance) was a significant risk factor for CIN2 + progression.
Molina et al. [22] Netherlands	Baseline and 6 months	-141 hrHPV positive women with normal cytology at first visit. -grouped into negative for Intraepithelial lesion or malignancy (NILM), non-progression group, squamous intraepithelial lesions (SIL) at 6 months visit.	- Cervical smears in PreservCyt were sequenced for microbiome profiling -The 16 S and 23 S rRNA genes were performed for cervicovaginal microbiome analysis.	-Women in both non-progression and progression groups did not exhibit a significant change in microbial diversity from baseline to second visit. -L. crispatus dominance at both visits is associated with non-progression to SIL, whereas those with high diversity and Lactobacillus depletion at first visit have a higher risk of progression to SIL. - Gardnerella vaginalis and Atopobium vaginae at both visit is associated with developing SIL. Women with Lactobacillus-dominated at baseline had a significantly more similar microbiome at second visit than women with Lactobacillus-depleted at baseline.
Zhang et al. [23] China	Baseline, 2 and 3 consecutive years (3 times)	-72 women with HPV persistence infection were included in the study. -Grouped into HPV persistent, CIN progression, and CIN regression groups.	-Cervical samples collected with cervical brush -The 16 S rRNA gene sequencing V3-V4 variable region	- Reduced alpha diversity at first and second visits in the CIN regression group compared to that in the persistence and progression groups. -Change in cervicovaginal microbiota may precede the CIN disease. -Atopobium, Sneathia, and Enterococcus were predictive of persistent HPV infection; CIN progression was associated with the enrichment of Acinetobacter, Pseudomonas, Stenotrophomonas, and Klebsiella; and CIN regression was associated with Streptococcus and Lactobacillus spp
			Cervical lesion or cancer treatment studies
Caselli et al. [28] Italy	Baseline and 6 months after LEEP	-85 patients aged 30–50 years diagnosed with CIN2 and CIN3 represented 48.2% and 51.8% respectively.	-LEEP by monopolar energy diathermic loop was used for lesion removal. -Cervicovaginal swabs samples and vaginal washing samples were collected using sterile rayon swabs before and after LEEP treatment respectively -Bacterial 16 S rRNA gene sequencing	- At baseline, there was increase of CST-IV in CIN3 compared to CIN2 patients, and a concomitant decrease of CST-I. - Compared to baseline the post-treatment visit showed a significant increase of CST-I (L. crispatus dominance) and concomitant decrease of CST-IV. CST-II (L. gasseri dominant) and CST-V (L. jensenii dominant) were unvaried, whereas CST-III (L. iners dominant) was increased, but not significant. - The CST-I increase was higher in CIN3 than in the CIN2 group, while the reverse was observed for CST-III, which was more prevalent in CIN3 compared to CIN2 patients.
Mitra et al. [27] United Kingdom	Baseline and 6 months after local excision	-103 pre-menopausal, non-pregnant women aged 18 to 45 with CIN (CIN1 15/ 103, 15%; CIN2 40/103, 39%; CIN3 48/103, 46%) -39 women with normal cytology as control. -Post-treatment: HSIL:2 LSIL:3, borderline nuclear changes (BNC):17 and normal:81 (11 HPV + and 70 HPV-)	-Treatment: local excisional treatment for cervical intraepithelial neoplasia -The V1–V3 hypervariable regions of 16 S rRNA genes -Swabs samples were collected for microbiota analysis	- No difference of richness or diversity between pre-vs post-treatment in the three CSTs identified. -Most patients (58/103, 56%) kept a stable vaginal microbiota CST before and after treatment. -Women with CST III (L. iners) were most likely to have the same CST before and after treatment (22/37, 59%), followed by CST I (L. crispatus) (26/45, 58%) and finally CST IV (10/21, 48%). -Of the 45 women who switched to another CST, 56% switched between Lactobacillus sp.-dominant states, 20% from a Lactobacillus sp.-dominant CST to CST IV and 24% from CST IV to a Lactobacillus spp.-dominant CST.
Li et al. [26] China	Samples collected 3 months before and 3 months after treatment	91 women classified into 4 groups [healthy control women (n = 25), HPV16- infected women with CIN I (n = 26), HPV16-infected women with CIN II/III (n = 34), and HPV16-infected women with SCC (n = 6)]	-Endocervical swabs with cervicovaginal samples were collected 3 months prior to treatment and at 3 months post-treatment -Treatment received: recombinant human interferon α2a vaginal suppository for CIN I, cervical excision by LEEP for CIN II/III, and radical hysterectomy for SCC patients -16 S rRNA gene pyrosequencing with the V3 and V4 hypervariable regions for microbiota identification	-In the healthy control L. crispatus (CST I) and L. iners (CSTIII) were dominated CST at first visit and L. crispatus, L. iners and CSTIV were dominated at second visit. -In the CIN I and CIN II/III groups, the most abundant cluster was CST IV and CST III (L. iners) before and after treatment respectively. -In the SCC group, the most abundant clusters were Burkholderia (CST II) and CST IV before treatment and L. iners (CST III) after treatment. -Treatments appear to have altered the CSTs toward the healthy types (from CST II and CST IV to CST I and CST III). -The microbiota of HPV16-infected women with CIN or SCC was more associated with CST IV and less associated with CST II.
Wang et al. [29] China	Not specified	125 vaginal microbial samples from patients with cervical cancer (of stage IB–IVB cervical cancer) before and after	-Treatment: concurrent chemoradiotherapy (CCRT) - vaginal swabs collected - The V3 region of bacterial 16 S rRNA gene sequencing	-There was difference in the vaginal microbial community before and after CCRT was found. -L. iners was most abundant in the recurrent group before CCRT. Based on these results. -The vaginal microbial community in the non-recurrence group was significantly more stable than in the recurrence group during the follow ups.

Quality assessment results

NOS scores ranged from 4 to 9 (Table S2). Four studies achieved high quality (score 7 or above) [20, 23, 25, 27], reflecting adequate cohort definition, clear outcome ascertainment, and adjustment for at least one key confounder. Seven studies scored 5–6, indicating moderate quality, primarily reflecting limited confounder adjustment. One studies scored 4 or below [29], mainly due to inadequate reporting of follow-up procedures and absence of confounder control.

CST classification heterogeneity

Studies classified cervicovaginal communities into 3 to 7 CSTs. The most commonly applied scheme (four studies) [20, 23, 25, 27], distinguished three categories: CST I (L. crispatus dominant), CST III (L. iners dominant), and CST IV (Lactobacillus-depleted). Four-CST and five-CST schemes were each used in two studies [18, 21] and [22, 28], respectively; six- and seven-CST classifications appeared in one study each [19, 26]. This granularity variation, compounded by differing sequenced 16 S rRNA regions, limits direct cross-study taxon-level comparisons.

Theme 1: Baseline microbiome composition and clinical outcomes

Lactobacillus dominance and favorable outcomes. Lactobacillus-dominated CSTs at baseline were consistently associated with HPV clearance and CIN regression ( + + evidence for both outcomes). Low alpha-diversity and high L. crispatus abundance at baseline and follow-up were associated with a reduced risk of SIL progression in hrHPV-positive women [22]. Women with CIN2 whose baseline microbiota was Lactobacillus-dominated were more likely to achieve regression at 12 and 24 months compared to those with Lactobacillus-depleted communities [20, 21, 24]. Furthermore, women with incident or persistent CIN2 + displayed higher anaerobic dysbiosis markers at both baseline and endpoint compared to hrHPV-positive and persistently HPV-negative groups [24].

Specific dysbiotic taxa were associated with adverse outcomes across multiple studies. Baseline colonization with Gardnerella vaginalis and Prevotella amnii predicted subsequent CIN2 progression in the Costa Rica cohort [21]. In a Chinese prospective cohort, Atopobium, Sneathia, and Enterococcus predicted persistent HPV infection [29], while Acinetobacter, Pseudomonas, Stenotrophomonas, and Klebsiella were associated positively with CIN progression; conversely, Lactobacillus spp. and Streptococcus spp. were associated with CIN regression in the same cohort [29]. These findings suggest that the prognostic signal from the baseline microbiome extends beyond the broad Lactobacillus/non-Lactobacillus dichotomy to specific anaerobic taxa with distinct clinical correlates.

Alpha-diversity findings were heterogeneous (+/- evidence). Studies found no baseline diversity difference between HPV-positive and HPV-negative states [18, 20] or between CIN progressors and non-progressors [22]. In contrast, elevated Shannon diversity at the second visit was significantly associated with CIN2 + progression in Costa Rica (p = 0.027) [21], and serial reductions in alpha-diversity characterised the CIN regression group in a Chinese cohort [23]. In treatment studies, one study found no richness differences before versus after LEEP [27], while another reported lower post-treatment diversity [26], and elevated pre-treatment diversity predicted disease recurrence after concurrent chemoradiotherapy [29].

Theme 2: CST dynamics, transition patterns, and temporal stability

Lactobacillus-dominated CSTs consistently exhibited greater temporal stability, while Lactobacillus-depleted communities showed higher transition frequencies ( + + evidence). In hrHPV-positive women progressing to SIL, microbiome composition was significantly less similar between baseline and follow-up than in women who remained SIL-negative [22]. Women with HPV-16 persistence and CIN2 + exhibited significantly more inter-cluster transitions over 18 months than those who cleared the infection [18]. Moscicki et al. [25] demonstrated that HPV-16-infected women were significantly more likely to transition between CSTs and showed a higher probability of remaining in non-Lactobacillus-dominated states compared to HPV-negative controls. Critically, the directionality of transitions diverged between clinical outcome groups: CIN2 regressors most frequently transitioned from CST IV (Lactobacillus-depleted) to CST III (L. iners-dominated) at both follow-up intervals, whereas non-regressors predominantly transitioned in the reverse direction during the same periods [20]. This divergence in transition direction is among the most informative longitudinal findings in this review.

Van de Wijgert et al. [24] found that L. iners-dominated, mixed Lactobacilli/BV-anaerobes, and BV-anaerobe-dominated VMB types were the least stable in women living with HIV. Brotman et al. [19], using biweekly self-sampling over 16 weeks, found that L. crispatus and L. gasseri-dominated CSTs showed fewer transitions than other types; women with CST IV-B (Atopobium-enriched) microbiota frequently transitioned to CST III (L. iners), though rarely to CST I (L. crispatus). Collectively, these data support L. crispatus dominance as the most stable cervicovaginal state, with CST III serving as a frequent transitional state.

Theme 3: Microbiome changes following treatment

Five studies evaluated microbiome dynamics before and after cervical lesion treatment [26–29]. In CIN patients undergoing LEEP, 56% maintained the same CST between the treatment date and six-month follow-up; among those who transitioned, the majority shifted between Lactobacillus-dominant states (56%), while (24%) moved from CST IV to Lactobacillus-dominant and 20% transitioned in the opposite direction [27]. In Chinese patients receiving cervical excision or interferon-based therapy, CIN I-III groups transitioned predominantly from CST IV to CST III (L. iners-dominated), and SCC patients shifted from CST II (Burkholderia-dominant) and CST IV toward CST III [26].

In Italian patients, L. crispatus abundance increased significantly after LEEP in both CIN2 and CIN3 groups, with a proportionally greater rise in CIN3 (from 16% to 29%) than in CIN2 (from 39% to 45%) [28]. Women without disease recurrence after concurrent chemoradiotherapy maintained significantly more stable microbial communities during follow-up, while L. iners was the most abundant taxon in the recurrence group before treatment [29]. Taken together, treatment appears to facilitate a partial short-term shift toward Lactobacillus-enriched communities, though whether this represents sustained ecological restoration or a transient response to tissue repair remains unresolved.

The contested role of L. iners across outcomes

Prior L. iners dominance was positively associated with HPV clearance in the Costa Rica cohort [21] and with the highest CST stability in treated CIN patients [27]. In women with HPV-16 infection, L. iners relative abundance increased during the pre-infection period, declined during active infection, and remained lower than in HPV-negative controls at post-clearance visits [25]. Additionally, Brotman and colleagues [19] evaluated the HPV incidence and showed that CSTs with the greatest proportion of HPV-positive samples were the low-Lactobacillus groups (CST IV) and the L. iners–dominated (CST III), when compared to CSTs dominated by L. crispatus and L. gasseri [19]. Post-treatment, L. iners dominance was the most common endpoint state across multiple studies [26, 28, 29]. In contrast, cross-sectional meta-analyses link L. iners-dominated CSTs to elevated HPV risk relative to L. crispatus [15], and van de Wijgert et al. [24] identified L. iners-dominated VMB types among the least stable in HIV-positive women. Evidence consistency for all microbiome-outcome relationships is summarized in Table S2 and Figure S2.

Discussion

This systematic review of 12 longitudinal studies across 11 countries provides the most comprehensive synthesis to date of cervicovaginal microbiome dynamics in relation to HPV infection and cervical lesion outcomes. Three overarching patterns emerge from the evidence. First, Lactobacillus-dominated CSTs, particularly L. crispatus-dominant states, are consistently associated with HPV clearance, CIN regression, and temporal microbiome stability across multiple cohorts and geographic settings. Second, Lactobacillus-depleted communities show higher transition rates and unfavorable clinical outcomes. Third, cervicovaginal microbiome shifts appear to precede, rather than merely accompany, changes in HPV status or cervical lesion grade in several cohorts, lending temporal plausibility to a modulating rather than merely reactive role.

The role of L. iners is one of the most contested questions in this field, and the present synthesis does not fully resolve it. Cross-sectional meta-analyses associate L. iners dominance with elevated HPV risk relative to L. crispatus [15], yet several longitudinal studies in this review identify positive associations between prior L. iners abundance and HPV persistence or incidence or CIN recurrence/progression [19, 20, 25, 28, 29]. These observations are not inherently contradictory. Ecologically, L. iners occupies an intermediate position: it survives under anaerobic and pH-variable conditions more readily than L. crispatus, enabling it to colonise the cervicovaginal niche after disturbance and potentially bridge the transition toward more stable Lactobacillus dominance [30, 31]. The observation that CIN2 regressors characteristically transition from CST IV to CST III before potential consolidation in CST I is consistent with this stepping-stone model [20] (Figure S2).

Nonetheless, L. iners has also been identified as a marker of ecological instability in other contexts [24, 32], and recent functional evidence suggests it may suppress host defence peptides that support L. crispatus survival under specific mucosal conditions [33]. Resolving this question will likely require strain-level metagenomic analysis, given that L. iners encompasses genomically diverse lineages with potentially divergent metabolic capacities [34]. The present evidence reflects genuine biological ambiguity: the ecological role of L. iners appears context-dependent, varying with host immune status, concomitant HPV genotype, and surrounding community composition.

Lactobacillus-produced lactic acid maintains vaginal pH below 4.5 and has been shown in vitro to suppress HPV E6/E7 oncoprotein expression [35]. Bacteriocins and hydrogen peroxide secreted by L. crispatus and L. jensenii further limit anaerobic colonization. Conversely, BV-associated species produce sialidase and proteases that degrade cervicovaginal mucus and reduce secretory leukocyte protease inhibitor (SLPI), generating a pro-inflammatory cytokine milieu, characterized by elevated IL-1beta, IL-6, IL-8, and TNF-alpha, that may impair innate immune clearance of HPV-infected epithelium [33, 35]. Biofilm formation by Gardnerella may additionally protect HPV virions from immune effectors while sustaining local mucosal inflammation [33]. Functional metagenomic profiling of the CC-associated microbiome has identified enrichment in pathways related to fatty acid biosynthesis, folate metabolism, and oxidative phosphorylation [16], suggesting that dysbiotic communities may confer metabolic advantages to transformed epithelial cells that extend beyond their pro-inflammatory effects.

The HPV-microbiome relationship is increasingly recognized as bidirectional. HPV itself may suppress Lactobacillus dominance by interfering with host defence pathways: the HPV E7 oncoprotein has been shown to disrupt NF-kappaB and Wnt/beta-catenin signaling, reducing expression of host defense peptides that support Lactobacillus survival [33]. This feedback mechanism, whereby dysbiosis favours HPV persistence which in turn perpetuates dysbiosis, may help explain why microbiome changes precede progressive disease in several cohorts reviewed here [21, 24, 29].

The substantial heterogeneity across included studies directly constrains the strength of this synthesis. CST classification granularity ranged from 3 to 7 subtypes, collapsing or subdividing potentially distinct ecological states in non-comparable ways: three-CST schemes aggregate all Lactobacillus-depleted states into a single CST IV category, masking distinctions between Gardnerella-dominated and highly diverse anaerobic communities. Variable 16 S rRNA sequencing regions (V1-V3, V3-V4, V4) further compromise taxon-level cross-study comparisons, as different regions yield different resolution profiles for key genera.

Sampling interval heterogeneity additionally limits temporal inference. Studies with annual sampling [20, 21, 23] may miss intermediate transitions captured by biweekly self-sampling [19]. The single HIV-positive cohort [24] introduces confounding by systemic and mucosal immune dysregulation, which profoundly shapes both microbiome composition and HPV clearance capacity. Most critically, only two studies formally adjusted for key confounders such as ethnicity, sexual activity, and antibiotic use [19, 24]. Given established ethnicity-related variation in cervicovaginal microbiome composition [36] and the strong behavioral determinants of both microbiome and HPV outcomes, residual confounding remains a plausible alternative explanation for the associations observed in most included studies.

Based on the identified evidence patterns, future longitudinal studies should: (1) adopt standardized sampling intervals of three months or less to capture clinically relevant microbiome dynamics; (2) employ shotgun metagenomic sequencing to enable strain-level identification and functional pathway analysis beyond 16 S rRNA compositional profiling; (3) systematically report and adjust for key confounders, including sexual activity, antibiotic exposure, hormonal contraception, ethnicity, and HIV status; (4) adopt a harmonized CST classification framework to enable cross-study comparison and eventual meta-analytic synthesis; and (5) design large, multi-ethnic cohort studies to improve generalizability and allow meaningful subgroup analyses.

Regarding clinical translation, the current evidence does not support microbiome-targeted interventions, including probiotics, prebiotics, or microbiome transplantation, as adjuncts to HPV management or CIN surveillance outside of clinical trial settings. Although mechanistic and early-phase trial data demonstrate biological plausibility for Lactobacillus-based interventions [33, 37], randomized controlled trials with validated cervical lesion endpoints are absent, and translational claims should be framed accordingly.

Limitations

This review has several limitations. The literature search was restricted just to PubMed; EMBASE, Cochrane Library, and Web of Science were not searched, introducing potential selection bias. A single reviewer, increasing the risk of inadvertent exclusion, performed title and abstract screening. Meta-analysis was not feasible due to clinical and methodological heterogeneity, precluding quantitative effect pooling. All included studies relied on 16 S rRNA gene sequencing, which lacks the strain-level resolution and functional inference capacity of metagenomic approaches. Primary study quality was heterogeneous (Table S2), and most studies did not adjust for key confounders, constraining causal conclusions. These limitations should be considered when interpreting the synthesized evidence.

Conclusion

This systematic review demonstrates that cervicovaginal microbiome composition is temporally associated with HPV infection and cervical lesion outcomes. In response to the stated objectives: (a) sustained Lactobacillus-dominated CST stability is the most consistent microbiome factor associated with HPV clearance and CIN regression across studies; (b) among specific taxa, L. crispatus dominance shows the strongest and most consistent protective signal, while Gardnerella, Prevotella, and Atopobium are recurrently associated with adverse outcomes; and (c) methodological heterogeneity, particularly in CST classification schemes, sampling intervals, and confounder adjustment, precludes definitive causal conclusions and highlights the need for standardized designs. Microbiome dysbiosis appears to precede HPV persistence and CIN progression in several cohorts, supporting a temporally antecedent modulating role. Advancing from the current associative evidence to mechanistic understanding and clinical application will require interdisciplinary collaboration combining rigorous longitudinal epidemiological designs with next-generation metagenomic technologies and prospective evaluation of microbiome-targeted interventions in well-powered randomized trials.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

MGH and MM conceptualised and designed the study. MGH drafted the manuscript. MM and NWAA contributed to data interpretation and critically revised the manuscript for important intellectual content. All authors approved the final version of the manuscript and agree to be accountable for all aspects of the work.

Funding

This research received no specific grant from any funding agency.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.