Association of vaginal dysbiosis with cervical lesions among women with non-16/18 high-risk HPV infections: a cross-sectional study of 2,264 women in Eastern China

Abstract

Background

To investigate whether vaginal dysbiosis is associated with an increased risk of cervical lesions infected with non-16/18 high-risk human papillomavirus (HR-HPV) genotypes and to inform risk-stratified cervical cancer screening strategies.

Methods

This cross‑sectional study included 2,264 women attending Jiaxing Maternal and Child Health Hospital between June 2023 and October 2024. HPV genotyping, vaginal microecological assessment, and Thinprep cytologic test (TCT) were performed. HR- HR-HPV infections were classified as HPV16/18-positive or non-16/18 HR-HPV infections. Logistic regression models were used to evaluate associations, adjusting for age. Multiplicative interaction was assessed using Wald tests for product terms. Benjamini‑Hochberg FDR correction was applied for multiple comparisons.

Results

The overall HPV prevalence was 49.87%, with HR-HPV detected in 36.13%. HR‑HPV infection was significantly associated with cervical lesions in both normal (OR = 11.45) and dysbiotic (OR = 4.40) vaginal microbiota (both P < 0.001). Stratified analyses showed higher point estimates for HPV58 and HPV59 in the dysbiosis group (OR = 10.98 and 9.51, respectively) than in the normal group (OR = 7.50 and 6.44, respectively). However, formal multiplicative interaction tests did not reach statistical significance (Wald P = 0.526 and 0.559, respectively; FDR‑adjusted q = 0.559 for both). No HPV subtype remained significantly associated with ASC‑US after FDR correction.

Conclusions

In this cross‑sectional study, vaginal dysbiosis is specifically associated with cervical lesions among women with non‑16/18 HR‑HPV infections, particularly HPV58 and HPV59, which showed elevated point estimates in the presence of dysbiosis. However, formal multiplicative interaction tests did not reach statistical significance. These findings are exploratory and hypothesis‑generating; causality cannot be inferred due to the cross‑sectional design. Prospective studies are needed to validate the observed associations before any clinical application can be considered. Our results should be interpreted with caution and do not support changes to current screening practice.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12905-026-04714-6.

Background

Cervical cancer remains the fourth most common cancer among women globally, with an estimated 604,000 new cases and 342,000 deaths in 2020 [1]. More than 85% of cervical cancer deaths occur in developing countries [2–4]. In China, the incidence and mortality of cervical cancer have shown an increasing trend from 1990 to 2019 [5–9], with approximately 110,000 new cases in 2020 [1].

Base on the classification of HPV genotypes strictly according to the latest International Agency for Research on Cancer (IARC, 2026) classification system [10], HPV was classified into three groups: Group 1 (Carcinogenic to humans): HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59; Group 2 A (Probably carcinogenic to humans): HPV 68; Group 2B (Possibly carcinogenic to humans): HPV 26, 30, 34, 53, 66, 67, 69, 70, 73, 82, 85, 97; Group 3 (Not classifiable as to its carcinogenicity to humans): HPV 6, 11. Over 200 HPV types have been identified, classified into high-risk and low-risk categories based on carcinogenic potential [11, 12].

Persistent infection with high-risk human papillomavirus (HR-HPV) is the primary cause of cervical cancer and its precursors [13]. HPV16/18-positive are well-established carcinogens, but the pathogenic risks of non-16/18 HR-HPV types (e.g., HPV31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82) vary geographically and remain less understood [11, 14, 15].

According to the IARC statistics [16], there were an estimated 109,700 new cases and 59,100 deaths from cervical cancer among Chinese women in 2020, making it the third most common cancer among women aged 15–44 years in China. HPV 16 and 18 pose the highest risk for HSIL. Based on the 2021 IARC Global and China HPV-Related Disease Reports [16], the distribution of HPV genotypes differs notably between Chinese and global populations. Among cytologically normal women, the most prevalent HPV genotypes in China are HPV52, 16, 58, 18, and 33, whereas the global predominant types are HPV16, 52, 53, 31, and 18. For Chinese women with precancerous lesions, the genotype distribution also exhibits distinct characteristics: in the low-grade squamous intraepithelial lesion (LSIL) group, the three most common genotypes were HPV52 (20.31%), HPV16 (16.81%), and HPV58 (14.44%); in the high-grade squamous intraepithelial lesion (HSIL) group, the predominant genotypes were HPV16 (45.69%), HPV58 (15.50%), and HPV52 (11.74%) [17]. By contrast, a global meta-analysis revealed a different pattern: in the LSIL group, the most frequent genotypes were HPV16 (25.1%), HPV31 (9.5%), and HPV18 (8.7%); in the HSIL group, the leading genotypes were HPV16 (47.5%), HPV31 (11.0%), and HPV18 (9.6%) [18].

The vaginal microbiota serves as a critical barrier for female reproductive health. In healthy states, Lactobacillus species dominate, maintaining an acidic environment (pH 3.8–4.5) [14, 19, 20]. Microecological disturbances, such as bacterial vaginosis (BV), elevated pH, and Vaginal dysbiosis, have been associated with increased HPV infection and persistence [20, 21]. Recent evidence suggested an association between vaginal dysbiosis and persistent HR-HPV as well as cervical lesions, potentially mediated by local inflammatory microenvironments [22]. However, whether the association between HPV16/18‑positive versus non‑16/18 HR‑HPV infections and cervical lesions varies by vaginal microbiota status remains largely unexplored.

This study therefore aims to evaluate the role of vaginal microbiota status in modifying the association between different HR-HPV types and cervical lesions, using a large sample of women in Eastern China, to provide evidence for risk-stratified cervical cancer screening.

Methods

Study population

A total of 2,264 women aged 18–55 years who attended Jiaxing Maternal and Child Health Hospital between June 2023 and October 2024 were consecutively enrolled. All participants underwent HPV genotyping, vaginal secretion testing, and liquid-based cytology (Thinprep cytologic test, TCT).

Based on HPV results, participants were categorized as:

• High-risk HPV (HR-HPV) group: positive only for HR-HPV types (16, 18 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 26, 53, 66, 73, 82). Divided into two groups: participants with HPV16/18-positive infections and participants with non-16/18 HR-HPV infections, positive for HR-HPV types other than 16/18 (31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 68, 26, 53, 66, 73, 82);

• Low-risk HPV (LR-HPV) group: positive only for LR-HPV types (6, 11, 42, 43, 81);

• Mixed infection group: co-infection with HR-HPV and LR-HPV.

Inclusion and exclusion criteria

• Inclusion criteria: (1) age 18–55 years with sexual experience; (2) no sexual intercourse within 48 h before examination; (3) no vaginal douching or intravaginal medication use within 3 days; (4) no antibiotic use within 2 weeks; (5) voluntarily agreed to HPV typing, vaginal secretion, and TCT testing.

• Exclusion criteria: (1) pregnant or lactating women; (2) women with severe immune system diseases or malignant tumors.

• The study was approved by the Ethics Committee of Jiaxing Maternal and Child Health Hospital (approval number: [2026-Y-052]).

Specimen collection

• Cervical secretions: a cervical brush was rotated clockwise 3–4 times at the cervix and placed into a collection tube with cell preservation solution for HPV genotyping.

• Vaginal secretions: two sterile swabs were rotated in the posterior vaginal fornix and placed into sterile tubes for microecological assessment.

• Cervical cytology: a cervical brush was used to collect cells for TCT.

Laboratory testing

HPV genotyping

HPV genotyping was performed using a human papillomavirus genotyping kit (Sansure Biotech Inc., China) for nucleic acid extraction and PCR amplification. Twenty-three HPV types were detected: 18 h-HPV types (16, 18, 26, 31, 33, 35, 39, 45, 51, 52, 53, 56, 58, 59, 66, 68, 73, 82) and 5 LR-HPV types (6, 11, 42, 43, 81).

Vaginal microecological assessment

Wet mount microscopy was used to assess cleanliness, clue cells, Trichomonas vaginalis, spores, and hyphae. Dry chemical enzyme assays measured hydrogen peroxide, leukocyte esterase, sialidase, β-glucuronidase, N-acetylglucosaminidase, and pH. Normal reference ranges: pH 3.8–4.5, leukocyte esterase ≤+, negative for other enzymes.

The vaginal cleanliness grade was assessed based on routine wet-mount microscopy of vaginal secretions, following the Chinese national diagnostic criteria for vaginitis (WS/T 322–2010). The grading criteria are as follows: Grade I: Normal vaginal flora (Lactobacillus predominance, no leukocytes, clean); Grade II: Normal (Lactobacillus predominant, few leukocytes, clean); Grade III: Mild-to-moderate dysbiosis (mixed flora, reduced Lactobacillus, increased leukocytes and bacteria); Grade IV: Severe dysbiosis (predominance of pathogenic bacteria or clue cells, abundant leukocytes). Grade I–II was defined as normal vaginal microbiota, and Grade III–IV was defined as vaginal dysbiosis.

The assay detects the presence of H₂O₂ produced by vaginal Lactobacillus species via an enzyme-catalyzed colorimetric reaction. Importantly, the interpretation of this test is inverse to what the nomenclature might suggest: Negative result (no color change): Indicates the presence of detectable H₂O₂, suggesting sufficient Lactobacillus metabolic activity and a normal vaginal microbiota; Positive result (color development): Indicates the absence or depletion of H₂O₂, suggesting reduced Lactobacillus abundance and a state of vaginal dysbiosis.

Bacterial vaginitis (BV) was diagnosed using Amsel criteria (≥ 3 of: thin homogeneous discharge, vaginal pH > 4.5, positive whiff test, clue cells). Vulvovaginal candidiasis (VVC) was diagnosed by microscopic detection of fungi, and trichomonal vaginitis (TV) by detection of Trichomonas vaginalis.

Vaginal microbiota status was defined as:

• Normal: pH 3.8–4.5, cleanliness grade I/II, negative hydrogen peroxide, and negative for BV/VVC/TV;

• Dysbiosis: any deviation from the above. These conditions were combined into a single composite category because the retrospective data did not allow reliable subtype‑specific classification for all cases, and the sample size for individual subtypes was insufficient for separate analyses. We acknowledge that this composite definition may obscure differential effects of specific dysbiosis subtypes.

Cervical cytology

TCT results were reported using the Bethesda System (TBS) as: negative for intraepithelial lesion or malignancy (NILM), atypical squamous cells of undetermined significance (ASC-US), low-grade squamous intraepithelial lesion (LSIL), high-grade squamous intraepithelial lesion (HSIL), and atypical squamous cells cannot exclude HSIL (ASC-H). Cervical lesions were defined as ≥ ASC-US, including ASC-US, LSIL, HSIL and ASC-H.

Statistical analysis

Statistical analyses were performed using SPSS 26.0. Categorical variables were expressed as frequencies (%) and compared using the chi-square test. Multivariate logistic regression was used to evaluate associations between HPV infection, vaginal microbiota status, and cervical lesions, adjusting for age groups (18–25, 26–35, 36–45, 46–55 years). Other potential confounders (smoking, parity, sexual behaviour, contraceptive use, menopausal status, and HPV vaccination history) were not available in the retrospective dataset and could not be included. We acknowledge this limitation in the Discussion. Interaction terms were included in the logistic regression models to assess whether the association between HPV infection and cervical lesions differed by microbiota status. Stratified analyses were conducted to calculate odds ratios (ORs) for different HPV types by microbiota status. P < 0.05 was considered statistically significant. Graphs were generated using GraphPad Prism 9.0. To account for multiple comparisons, we applied the Benjamini–Hochberg false discovery rate (FDR) correction separately to each family of analyses. An FDR‑adjusted q‑value < 0.10 was considered statistically significant for all exploratory analyses. Fisher’s exact test was used for comparisons with expected cell counts < 5.

Results

Participant characteristics and HPV prevalence

A total of 2,264 women were included. HPV-negative participants accounted for 1,135 (50.13%), and HPV-positive for 1,129 (49.87%). Among HPV-positive women, 818 (36.13%) had only HR-HPV, 166 (7.33%) had only LR-HPV, and 145 (6.4%) had mixed HR-HPV and LR-HPV infections (Table 1). HPV prevalence showed a bimodal distribution by age, with peaks in the 18–25 and 46–55 age groups (57.64% and 54.73%, respectively) (Fig. 1). The highest prevalence of LR-HPV alone was observed in the 46–55 age group (9.28%).

Table 1.

Age-specific prevalence of HPV infection

Group	N	LR-HPV only	HR-HPV only	Mixed (LR + HR)	Total HPV-Positive
Age Group (years)		n (%) [95% CI]	n (%) [95% CI]	n (%) [95% CI]	n (%) [95% CI]
18–25	203	12(5.91%) [3.41%~10.05%]	84(41.38%) [34.83%~48.26%]	21(10.34% ) [6.87%~15.30%]	117(57.64%) [50.76%~64.23%]
26–35	806	54(6.70%) [5.17%~8.64%]	291(36.10%) [32.86%~39.48%]	55(6.82%) [5.28%~8.78%]	400(49.63%) [46.20%~53.08%]
36–45	727	51(7.02%) [5.38%~9.11%]	243(33.43% ) [30.09%~36.94%]	29(3.99% ) [2.79%~5.67%]	323(44.43%) [40.86%~48.06%]
46–55	528	49(9.28%) [7.09%~12.06%]	200(37.88%) [33.85%~42.10%]	40(7.58% ) [5.61%~10.15%]	289(54.73%) [50.47%~58.93%]
Total	2264	166(7.33%) [6.33%~8.48%]	818(36.13% ) [34.18%~38.13%]	145(6.40% ) [5.47%~7.49%]	1129(49.87%) [47.81%~51.93%]

Fig. 1.

Prevalence rates of LR-HPV, HR-HPV and co-infection of LR-HPV and HR-HPV in different age groups

HPV type distribution

Among 963 women with HR-HPV infection, 397 had multiple infections (234 with dual infections, 163 with three or more types). The most common HR-HPV types were HPV52 (0.97%), HPV58 (0.90%), HPV59 (0.82%), HPV53 (0.70%), and HPV16 (0.61%). Among multiple infections, HPV58 (1.36%), HPV52 (1.34%), and HPV59 (1.16%) were most frequent (Fig. 2).

Fig. 2.

Prevalence of single and multiple infections with different HPV subtypes

Association between HPV infection and vaginal microbiota

The mean vaginal pH was slightly but significantly lower in HPV-positive women compared to HPV-negative women (4.81 ± 0.466 vs. 4.85 ± 0.457, P = 0.0393). No significant differences were observed in cleanliness grade. BV and TV were more frequent in HPV-positive women, but the difference of BV did not reach statistical significance (P = 0.0539) (Table 2).

Table 2.

Analysis of vaginal microecology differences between HPV-infected group and HPV-negative group

Variable	HPV positive (n = 1129)	HPV negative (n = 1135)	χ2	P
	Case (%)	Case (%)
pH value	4.81 ± 0.466	4.85 ± 0.457		0.0393
I-II	805(71.30%)	809(71.28%)	0.0053	0.9427
III-IV	324(27.70%)	326(28.72%)	0.0053	0.9427
BV	89(7.88%)	66(5.81%)	3.7182	0.0539
VVC	77(6.82%)	86(7.58%)	0.5238	0.4695
TV	9(0.80%)	0(0.00%)	-	0.0021*

BV bacterial vaginitis, VVC vulvovaginal candidiasis, TV trichomonal vaginitis; *Fisher’s exact test was used for TV due to an expected cell count < 5

Association between HR-HPV types and cervical lesions

After adjusting for age, logistic regression models were constructed for ASC-US, LSIL, and HSIL (Tables 3, 4 and 5). In the exploratory genotype-specific analyses, several HPV subtypes showed nominal associations with ASC-US in the uncorrected model. However, these associations did not survive adjustment for multiple comparisons, suggesting that they may be chance findings or that the statistical power was insufficient for robust detection (Table 3). Future studies with larger sample sizes are warranted to clarify whether these specific HPV genotypes are independently associated with low-grade cytological abnormalities. In the analysis of LSIL, HPV56 (OR = 2.292, 95% CI: 1.217–4.317, P = 0.010) and HPV58 (OR = 1.735, 95% CI: 1.120–2.686, P = 0.014) remained significantly associated with LSIL after FDR correction for multiple comparisons (q = 0.060 for both, Table 4). HPV51 showed a nominal association in the uncorrected analysis (OR = 1.875, P = 0.030), but this did not retain significance after FDR adjustment (q = 0.108, Table 4). In the analysis of HSIL, HPV16 (OR = 2.458, P = 0.039) and HPV31 (OR = 3.314, P = 0.031) showed nominal associations in the uncorrected analysis. However, after applying FDR correction for multiple comparisons, neither association retained statistical significance (both q-values > 0.10, Table 5). HPV33 also showed a borderline nominal association (P = 0.093, q = 0.403). These findings should be interpreted with caution due to the limited number of HSIL cases and the exploratory nature of the subtype-specific analyses.

Table 3.

Logistic regression analysis of HR-HPV infection and ASC-US

HPV Genotype	B	SE	WaldX2	P	FDR-adjusted q	OR(95%CI)
HPV16	-0.429	0.317	1.828	0.176	0.453	0.651(0.350–1.213)
HPV18	-0.302	0.529	0.327	0.567	0.705	1.353(0.480–3.815)
HPV31	-0.277	0.608	0.208	0.648	0.705	1.320(0.401–4.347)
HPV33	-0.349	0.372	0.880	0.348	0.579	0.706(0.341–1.462)
HPV35	0.272	0.608	0.199	0.655	0.705	1.312(0.398–4.320)
HPV39	-0.789	0.382	4.259	0.039	0.216	0.454(0.215–0.961)
HPV45	-0.715	0.362	3.896	0.048	0.216	0.489(0.241–0.995)
HPV51	-0.168	0.389	0.186	0.666	0.705	0.846(0.394–1.813)
HPV52	− 0.0321	0.270	1.416	0.234	0.468	0.725(0.427–1.231)
HPV56	0.990	0.728	1.846	0.174	0.453	2.690(0.645–11.216)
HPV58	-0.343	0.276	1.544	0.214	0.468	0.710(0.413–1.219)
HPV59	-0.038	0.313	0.014	0.905	0.905	0.963(0.522–1.779)
HPV68	-0.742	0.323	5.274	0.022	0.198	0.476(0.253–0.897)
HPV26	-0.509	0.549	0.860	0.354	0.579	0.601(0.205–1.763)
HPV53	-0.668	0.280	5.688	0.017	0.198	0.513(0.296–0.888)
HPV66	-0.466	0.344	1.836	0.175	0.453	0.628(0.320–1.231)
HPV73	0.416	0.605	0.473	0.492	0.705	1.517(0.463–4.968)
HPV82	-0.220	0.485	0.205	0.651	0.705	0.803(0.310–2.078)

B regression coefficient, SE standard error; P values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method across 18 HPV genotypes. A q-value < 0.10 was considered statistically significant

Table 4.

Logistic regression analysis of HR-HPV infection and LSIL

HPV Genotype	B	SE	WaldX2	P	FDR-adjusted q	OR(95%CI)
HPV16	0.249	0.281	0.785	0.376	0.565	1.283(0.740–2.225)
HPV18	0.531	0.336	2.495	0.114	0.342	1.700(0.880–3.286)
HPV31	-0.185	0.484	0.146	0.702	0.791	0.831(0.322–2.147)
HPV33	0.088	0.338	0.067	0.796	0.796	1.092(0.562–2.119)
HPV35	0.229	0.423	0.293	0.588	0.706	1.257(0.549–2.879)
HPV39	-0.032	0.416	0.006	0.938	0.938	0.968(0.429–2.187)
HPV45	0.071	0.371	0.037	0.848	0.848	1.074(0.519–2.222)
HPV51	0.629	0.290	4.713	0.030	0.108	1.875(1.063–3.309)
HPV52	0.187	0.234	0.641	0.423	0.565	1.206(0.762–1.908)
HPV56	0.829	0.323	6.588	0.010	0.060	2.292(1.217–4.317)
HPV58	0.551	0.223	6.090	0.014	0.060	1.735(1.120–2.686)
HPV59	0.246	0.248	0.984	0.321	0.565	1.278(0.787–2.077)
HPV68	0.215	0.316	0.462	0.497	0.638	1.240(0.667–2.305)
HPV26	0.021	0.546	0.002	0.969	0.969	1.021(0.350–2.978)
HPV53	0.261	0.262	0.992	0.319	0.565	1.298(0.777–2.169)
HPV66	0.049	0.326	0.022	0.881	0.881	1.050(0.554–1.991)
HPV73	-0.330	0.480	0.471	0.492	0.638	0.719(0.280–1.844)
HPV82	0.299	0.399	0.564	0.453	0.565	1.349(0.618–2.946)

B regression coefficient, SE standard error; P values were adjusted using the Benjamini–Hochberg false discovery rate (FDR) method across 18 HPV genotypes. A q-value < 0.10 was considered statistically significant

Table 5.

Logistic regression analysis of HR-HPV infection and HSIL

HPV Genotype	B	SE	WaldX2	P	FDR-adjusted q	OR(95%CI)
HPV16	0.899	0.435	4.280	0.039	0.253	2.458(1.049–5.763)
HPV18	0.069	0.741	0.009	0.926	0.994	1.071(0.251–4.573)
HPV31	1.198	0.556	4.648	0.031	0.403	3.314(1.115–9.852)
HPV33	0.837	0.498	2.828	0.093	0.403	2.309(0.871–6.125)
HPV35	-0.344	1.027	0.112	0.738	0.959	0.709(0.095–5.304)
HPV39	-	-	-	-	–	NA
HPV45	-0.009	0.740	0.000	0.991	0.994	0.991(0.233–4.225)
HPV51	-0.260	0.738	0.124	0.725	0.959	0.771(0.182–3.276)
HPV52	-0.357	0.537	0.442	0.506	0.959	0.700(0.244–2.005)
HPV56	-	-	-	-	–	NA
HPV58	-0.600	0.610	0.967	0.326	0.959	0.549(0.166–1.814)
HPV59	0.320	0.458	0.487	0.485	0.959	1.377(0.561–3.382)
HPV68	0.152	0.615	0.061	0.805	0.959	1.164(0.349–3.884)
HPV26	-	-	-	-	–	NA
HPV53	0.294	0.492	0.356	0.551	0.959	1.341(0.511–3.518)
HPV66	-0.375	0.740	0.258	0.612	0.959	0.687(0.161–2.927)
HPV73	-	-	-	-	–	NA
HPV82	-	-	-	-	0.959	NA

B regression coefficient, SE standard error, NA no positive results for the corresponding HPV genotype; P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method across all available HPV genotypes (HPV39, HPV56, HPV26, and HPV73 were not estimable due to zero cases in the HSIL group and were excluded from the correction). An FDR-adjusted q-value < 0.10 was considered statistically significant

Interaction between HPV type and vaginal microbiota

Stratified by vaginal microbiota status, the association between HPV and cervical lesions varied. In women with normal microbiota, HR-HPV or mixed infection was associated with a high risk of cervical lesions (OR = 11.45, 95% CI 7.20–18.21, P < 0.001, FDR-adjusted q < 0.001). LR-HPV showed a nominal association in the uncorrected analysis (OR = 2.89, P = 0.039), but this did not retain significance after FDR correction (q = 0.078). In women with dysbiosis, both LR-HPV (OR = 2.92, P < 0.001) and HR-HPV/mixed infection (OR = 4.40, P < 0.001) remained significantly associated after FDR correction (both q < 0.001) (Table 6). In sensitivity analyses stratified by cytological grade, the associations were consistently stronger for LSIL and HSIL than for ASC‑US (Supplementary Table S1 and S2).

Table 6.

Correlation between HPV infection types and cervical lesions under different flora states

Vaginal Microecology		Normal Cervix Group (NLSM)	Abnormal Cervix Group (≥ ASC-US)	OR (95%CI)	P	FDR-adjusted q
Normal group	HPV negative group	678	47	1.00
	LR-HPV group	25	5	2.89 (1.06 ~ 7.88)	0.039	0.078
	HR-HPV and mixed infection group	68	54	11.45 (7.20 ~ 18.21)	< 0.001	< 0.001
Dysbiosis Group	HPV negative group	382	28	1.00
	LR-HPV group	112	24	2.92 (1.63 ~ 5.25)	< 0.001	< 0.001
	HR-HPV and mixed infection group	636	205	4.40 (2.90 ~ 6.65)	< 0.001	< 0.001

P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method across the four comparisons presented in this table (two HPV types × two microbiota strata). An FDR-adjusted q-value < 0.10 was considered statistically significant. After FDR correction, HR-HPV/mixed infection remained significantly associated with abnormal cervical cytology in both normal and dysbiosis groups (both q < 0.001), whereas the association for LR-HPV in the normal group did not retain significance (q = 0.078)

When HR-HPV was further classified, both HPV16/18 genotypes and non-16/18 HR-HPV genotypes were associated with increased cervical lesion risk compared to HPV-negative women in both normal and dysbiotic microbiota environments, with HPV16/18-positive HR-HPV type showing higher value of OR (Table 7). These findings indicate that HPV‑associated risk is not materially diminished by the presence of dysbiosis, supporting the continued importance of HPV genotyping across different vaginal ecological states.

Table 7.

Risk of cervical lesions associated with hpv16/18-positive and Non-16/18 HR-HPV infections under microbiota stratification

HPV Genotype	Vaginal Microecology	Cervical lesions / Total (%)	Adjusted OR (95%CI)1	P	FDR-adjusted q
HPV16/18- genotypes	Normal	18/53 (33.96%)	12.38 (6.52–23.52)	< 0.001	< 0.001
	Abnormal	34/102 (33.33%)	11.56 (6.98–19.15)	< 0.001	< 0.001
Non-16/18 HR-HPV genotypes	Normal	51/212 (24.06%)	7.25 (4.48–11.74)	< 0.001	< 0.001
	Abnormal	101/453 (22.30%)	6.83 (4.32–10.81)	< 0.001	< 0.001

¹Taking HPV-negative status as the reference, stratify the age groups (18–25, 26–35, 36–45, 46–55 years old). P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method across the four comparisons presented in this table (two HPV categories × two microbiota strata). An FDR-adjusted q-value < 0.10 was considered statistically significant

In the subtype-specific analyses, all four non-16/18 HR-HPV types (HPV52, HPV53, HPV58, and HPV59) were significantly associated with cervical lesions in both microbiota strata, with adjusted ORs ranging from 5.71 to 16.32 (all P < 0.001, except HPV59 in the normal group: P = 0.001). After FDR correction for multiple comparisons, all associations remained significant (all q < 0.001, Table 8). Notably, the strongest effects were observed for HPV58 in the dysbiosis group (OR = 10.98, 95% CI: 5.58–21.59, q < 0.001) and HPV53 in the normal group (OR = 16.32, 95% CI: 4.20–63.43, q < 0.001). In the multiplicative interaction analysis, none of the non‑16/18 HR‑HPV infections showed statistically significant interaction with microbiota status (all Wald P > 0.05, Table 9). Although the stratified ORs for HPV58 and HPV59 were higher in the dysbiosis group than in the normal group, the interaction terms did not reach statistical significance.

Table 8.

Risk of cervical lesions of major non-16/18 HR-HPV genotypes under microbiota stratification

HPV Genotype	Vaginal Microecology	Cervical lesions / Total (%)	Adjusted OR (95%CI)1	P	FDR-adjusted q
HPV52	Normal	17/56 (30.4)	9.56 (4.00–22.84)	< 0.001	< 0.001
	Abnormal	16/84 (19.0)	5.71 (2.73–11.95)	< 0.001	< 0.001
HPV53	Normal	5/15 (33.3)	16.32 (4.20–63.43)	< 0.001	< 0.001
	Abnormal	18/84 (21.4)	8.20 (4.10–16.41)	< 0.001	< 0.001
HPV58	Normal	9/36 (25.0)	7.50 (2.86–19.65)	< 0.001	< 0.001
	Abnormal	23/82 (28.0)	10.98 (5.58–21.59)	< 0.001	< 0.001
HPV59	Normal	6/30 (20.0)	6.44 (2.10–19.80)	0.001	0.001
	Abnormal	22/88 (25.0)	9.51 (4.81–18.80)	< 0.001	< 0.001

¹The Mantel-Haenszel stratified analysis method was adopted to adjust for age (18–25, 26–35, 36–45, 46–55 years old), with the population testing entirely negative for HPV in the same stratum serving as the reference group. P values were adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) method. An FDR‑adjusted q‑value < 0.10 was considered statistically significant

Table 9.

Interaction analysis of major non-16/18 HR-HPV genotype

HPV Genotype	β (interaction)	SE	Wald χ²	P (Wald)	FDR‑adjusted q
HPV52	-0.516	0.582	0.787	0.375	0.754
HPV53	-0.688	0.778	0.781	0.377	0.701
HPV58	0.381	0.601	0.402	0.526	0.559
HPV59	0.391	0.669	0.341	0.559	0.559

Multiplicative interaction was assessed using multivariable logistic regression models including product terms (HPV genotype × microbiota status) for each non‑16/18 HR‑HPV genotype, adjusted for age. Wald P‑values for the interaction coefficients are reported. P‑values were adjusted for multiple comparisons using the Benjamini‑Hochberg false discovery rate (FDR) method across the four non‑16/18 HR‑HPV genotypes examined. An FDR‑adjusted q‑value < 0.10 was considered statistically significant. After FDR correction, none of the interaction terms reached statistical significance (all q > 0.10)

Discussion

This cross-sectional study of 2,264 women in Eastern China provides evidence of an association between vaginal dysbiosis and an increased risk of cervical lesions. The findings suggest that the observed associations of certain HR‑HPV types with cervical lesions may vary according to the local vaginal microenvironment, although this observation requires confirmation in prospective studies. The overall HPV prevalence in this study was 49.87%, which is higher than that reported in population-based screening studies [23]. This is likely due to the hospital-based nature of the sample, as most participants attended due to gynecological symptoms. The bimodal age distribution of HPV infection is consistent with previous studies in China [23].

In line with earlier reports [24], we found that HPV56 and HPV58 were associated with LSIL after FDR correction for multiple comparisons. However, the associations with HSIL did not retain statistical significance after correction, likely due to the limited number of HSIL cases. These findings, together with the observation that no HPV subtype remained significantly associated with ASC US, underscore the importance of using more specific cytological endpoints and accounting for multiple comparisons in subtype specific risk assessment. The limited number of HSIL cases in our cohort resulted in sparse data for several genotype subgroups, leading to non-estimable odds ratios. Thus, our findings regarding genotype-specific associations with HSIL are preliminary and require replication in larger studies with greater statistical power.

We acknowledge that direct comparison of ORs across microbiota strata should be interpreted with caution, as ORs are derived from different reference populations and are influenced by differences in baseline risk. The higher OR observed in the normal microbiota group reflects the lower baseline lesion rate in HPV‑negative women with normal microbiota, rather than indicating a higher absolute risk. Absolute risk differences—rather than ORs alone—should be considered when interpreting the clinical significance of stratified findings.

A key finding of this study is that the association between HPV infection and cervical lesions differs by vaginal microbiota status. While HPV16/18-positive types conferred a high risk regardless of microbiota status, the risk associated with non-16/18 HR-HPV types was significantly stronger in the presence of dysbiosis. Our stratified analyses showed higher point estimates for HPV58 and HPV59 in the dysbiosis group; however, the multiplicative interaction terms were not statistically significant. This suggests that while dysbiosis may be associated with elevated risks for these subtypes, the evidence that these associations differ significantly by microbiota status is not robust in our dataset and warrants further investigation in larger cohorts.

The apparent discrepancy between the overall and stratified analyses reflects a typical pattern of effect modification by vaginal microbiota. While the unstratified analysis showed no robust association, the stratified analysis revealed that the risk was specifically concentrated in women with dysbiosis. This underscores the importance of considering microbiota status when evaluating HPV subtype‑specific associations.

The prevalence of HPV58 in our cohort is consistent with previous reports indicating that HPV58 is more common in China populations than in global populations [16]. This regional variation underscores the importance of including non‑16/18 HR‑HPV infections in screening and vaccine strategies tailored to specific populations. In China, HPV58 has been identified as the common HR‑HPV type in cervical cancer cases [25], and our findings further highlight its clinical relevance.

The elevated ORs we observed for HPV58 and HPV59 in the dysbiosis group are broadly consistent with studies suggesting that depletion of Lactobacillus‑dominant microbiota may facilitate HPV persistence and progression [20, 21]. However, direct comparison with previous literature is challenging, as most studies have focused on HPV16/18 genotypes and few have examined the interaction between vaginal microbiota and non‑16/18 HR‑HPV genotypes separately. It was reported that bacterial vaginosis was associated with increased HPV persistence, but did not differentiate by HPV genotype [26]. Our finding of subtype‑specific patterns—particularly for HPV58 and HPV59—extends this literature by suggesting that the vaginal environment may have differential associations with specific HPV types.

A major limitation of this study is the broad composite definition of vaginal dysbiosis. We combined BV, VVC, TV, abnormal pH, and altered cleanliness into a single exposure category. These conditions have distinct microbiological aetiologies and may exert differential effects on HPV persistence and cervical lesion progression. By grouping them together, we may have masked subtype‑specific associations—for example, the effect of BV may differ from that of VVC or TV, and pH abnormalities may have a distinct biological pathway. This aggregation could dilute or obscure signals that would be detectable in analyses of individual dysbiosis subtypes. Owing to the retrospective design and the very limited number of events in each specific subtype, we were unable to perform well‑powered subtype‑specific analyses. Consequently, our findings reflect the overall effect of a generally disturbed vaginal environment rather than the effect of any particular pathogen. Future prospective studies with standardized and detailed microbiological diagnostics—including molecular characterization of specific microbial taxa—are urgently needed to clarify whether specific dysbiosis subtypes are differentially associated with HPV‑related cervical abnormalities.

The cross-sectional design of this study precludes any causal interpretation. All reported associations reflect statistical correlations at a single time point, and the temporal direction between vaginal microecology, HPV infection, and cervical cytological changes cannot be determined. Therefore, our results should be interpreted as hypothesis-generating rather than confirmatory. Prospective cohort studies or longitudinal intervention trials are needed to establish temporal sequences and potential causality.

The mechanisms underlying this interaction are not fully understood but may involve dysbiosis-induced chronic inflammation, disruption of epithelial barrier integrity, and altered local immune responses [20, 21]. These factors may create a permissive environment for HPV persistence and progression, particularly for non‑16/18 HR‑HPV types, for which the associations may show stronger variation with microenvironmental factors.

Several limitations should be acknowledged. First, the cross‑sectional design precludes causal inference regarding the temporal relationship between dysbiosis and HPV pathogenesis. Second, and importantly, our multivariable analyses adjusted only for age. We did not account for several potential confounders, including smoking, parity, sexual behaviour, contraceptive use, menopausal status, and HPV vaccination history. This limitation stems from the unavailability of these variables in the retrospective data source, which may have introduced residual confounding. For example, smoking and parity are known risk factors for cervical lesions, and sexual behaviour is closely linked to HPV acquisition; their absence from our models may have led to overestimation or underestimation of the true associations. Third, the definition of vaginal dysbiosis was based on composite clinical criteria rather than molecular profiling (e.g., 16 S rRNA sequencing), which could provide more granular information on microbial composition. Fourth, the hospital‑based sample may limit generalisability to the general population. Future prospective studies with comprehensive data collection on lifestyle, reproductive history, and vaccination status, as well as microbiome analysis, are needed to confirm these findings and to control more adequately for confounding.

Despite these limitations—particularly the composite definition of vaginal dysbiosis, which may mask subtype‑specific effects—the study has notable strengths, including a large sample size, detailed HPV genotyping, and a standardized clinical‑microscopic assessment of vaginal microecology. The finding that dysbiosis shows differential associations with non‑16/18 HR‑HPV genotypes may have implications for future research. It was hypothesis‑generated that women with dysbiosis and non-16/18 HR-HPV infections may benefit from closer follow-up, and that microbiota-informed risk stratification could help avoid unnecessary interventions in low-risk women. However, given the cross‑sectional design and the exploratory nature of the subtype‑specific analyses, these findings should be interpreted as exploratory and hypothesis‑generating, with prospective validation required before any clinical application. Prospective cohort studies are needed to validate the observed associations before considering microbiota‑informed screening or follow‑up strategies.

Conclusions

This cross‑sectional study of 2,264 women in Eastern China demonstrated that vaginal dysbiosis and non-16/18 HR-HPV infections (particularly HPV58 and HPV59) are associated with an increased prevalence of cervical lesions. In stratified analyses, HPV58 and HPV59 showed substantially elevated point estimates in the presence of vaginal dysbiosis. However, formal multiplicative interaction tests did not reach statistical significance, and the cross‑sectional design precludes any causal interpretation.

Although it was suggested that the combination of vaginal dysbiosis and non-16/18 HR-HPV infections may identify a subgroup with elevated lesion prevalence in this study, prospective cohort studies are needed to validate these associations before considering microbiota‑informed screening or follow‑up strategies. The results underscore the importance of subtype‑specific HPV assessment and the potential value of vaginal ecological evaluation in future research. Although these findings require validation in prospective studies, they provide additional evidence supporting further investigation of the associations between vaginal dysbiosis and HR-HPV genotype-specific cervical lesion risk.

Abbreviations

ASC-US

Atypical squamous cells of undetermined significance

BV

Bacterial vaginosis

CI

Confidence interval

HPV

Human papillomavirus

HR-HPV

High-risk human papillomavirus

HSIL

High-grade squamous intraepithelial lesion

LR-HPV

Low-risk human papillomavirus

LSIL

Low-grade squamous intraepithelial lesion

OR

Odds ratio

TCT

Thinprep cytologic test

TV

Trichomonal vaginitis

VVC

Vulvovaginal candidiasis

Authors’ contributions

SY designed and analyzed of the work, and drafted the manuscript. DW analyzed of the work. KH interpretated of data. YL substantively revised the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by grants from the Zhejiang Provincial Medical and Health Science and Technology Program (2024KY450).

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

In accordance with the Declaration of Helsinki, the study was approved by the Ethics Committee of Jiaxing Maternal and Child Health Hospital (approval number: [2026-Y-052]). As this was a retrospective study analyzing de‑identified data from routine clinical practice, ethics approval and waiver of informed consent were obtained prior to data analysis. Written informed consent was obtained from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.