Copper intrauterine device increases vaginal concentrations of inflammatory anaerobes and depletes lactobacilli compared to hormonal options in a randomized trial

Abstract

Effective contraceptives are a global health imperative for reproductive-aged women. However, there remains a lack of rigorous data regarding the effects of contraceptive options on vaginal bacteria and inflammation. Among 218 women enrolled into a substudy of the ECHO Trial (NCT02550067), we evaluate the effect of injectable intramuscular depot medroxyprogesterone acetate (DMPA-IM), levonorgestrel implant (LNG), and a copper intrauterine device (Cu-IUD) on the vaginal environment after one and six consecutive months of use, using 16S rRNA gene sequencing and multiplex cytokine assays. Primary endpoints include incident BV occurrence, bacterial diversity, and bacterial and cytokine concentrations. Secondary endpoints are bacterial and cytokine concentrations associated with later HIV seroconversion. Participants randomized to Cu-IUD exhibit elevated bacterial diversity, increased cytokine concentrations, and decreased relative abundance of lactobacilli after one and six months of use, relative to enrollment and other contraceptive options. Total bacterial loads of women using Cu-IUD increase 5.5 fold after six months, predominantly driven by increases in the concentrations of several inflammatory anaerobes. Furthermore, growth of L. crispatus (MV-1A-US) is inhibited by Cu²⁺ ions below biologically relevant concentrations, in vitro. Our work illustrates deleterious effects on the vaginal environment induced by Cu-IUD initiation, which may adversely impact sexual and reproductive health.

Fig. 1. Contraceptive use significantly alters the composition and diversity of vaginal bacterial communities across 218 participants.

a Cross-sectional comparisons of clinical Nugent scores by randomization arm Cu-IUD (n = 47); DMPA-IM (n = 36); LNG implant (n = 46). b Cross-sectional comparison of the effect of each contraceptive method on vaginal bacterial Shannon diversity Cu-IUD (n = 55); DMPA-IM (n = 45); LNG implant (n = 53). c PCoA ordination of relative abundance transformed taxonomic abundance using Bray-Curtis distance. Samples are colored by CST. d PCoA ordination of relative abundance transformed taxonomic abundance using Bray-Curtis distance. Samples are colored by randomization arm. e Network analysis of samples visualized using a Bayesian directed acyclic graph (DAG) and colored by CST. f DAG colored by randomization arm. DAGs were generated using Euclidean distance on centered log ratio transformed taxonomic abundance. Network inference was performed using the PC algorithm. Two-tailed P values were calculated using an ANCOVA model for multiple comparisons of Nugent scores and Shannon diversity, and a permutational ANOVA for multiple comparisons in the beta-diversity analysis. All hypothesis tests included baselines values as a covariate. Boxplots limits define the first and third quartiles, the center line represents the median, and whiskers define 1.5x interquartile range.

Fig. 2. Contraceptive use drives CST transition and shifts in bacterial abundance.

a CSTs are indicated at each time point and alluvials track the progression of a given participant across time. Plots are faceted by randomization arm. b Fold changes in bacterial abundance after 1 and 6 months of contraceptive use, relative to enrollment abundance. Fold changes are log₂ transformed and were calculated using the default (two-sided) ANCOM-BC model for two-group comparisons. Vertical dashed lines indicate a 0.5-fold change. Each point represents the arithmetic mean log₂ fold change and solid horizontal lines represent the standard error. Taxa with P < 0.05 after adjustment for multiple comparisons (Benjamini and Hochberg) are shown; Cu-IUD (n = 55); DMPA-IM (n = 45); LNG implant (n = 53).

Fig. 3. Cu-IUD use induces significant increases in the concentrations of several cervicovaginal bacteria and cytokines.

a Absolute 16S rRNA gene copy number assessed by broad-range qPCR at the study enrollment visit and after 6 months of the assigned contraceptive use. P values were calculated using a Student’s t test and corrected using the Benjamini and Hochberg method; Cu-IUD (n = 54); DMPA-IM (n = 44); LNG implant (n = 52). b Increases in absolute bacterial abundance after 6 months of Cu-IUD use. Only those taxa whose abundance remained significant after adjustment for multiple comparisons (Benjamini and Yekutieli) are shown. Data are presented as mean values +/− SEM. c The additive concentration of bacteria altered by Cu-IUD use correlates strongly with Nugent score. d The additive concentration of bacteria altered by Cu-IUD use correlates strongly with the concentration of several key cytokines after 6 months of contraceptive use. Two-tailed P values were derived from the t-value of the generalized linear model. Cytokine and bacterial concentrations were log₂ transformed prior to analysis. The class of cytokine is indicated above each subplot. Participant data points are colored by randomization arm and are from data collected after 6 months of contraceptive use. Boxplot center lines indicate the median, while the hinges indicate the first and third quartiles, and whiskers extend to 1.5 * IQR from the given hinge. Models with p < 0.05 after adjustment for multiple comparisons (Benjamini and Yekutieli) are shown. Regression lines are indicated in blue and the shaded regions around regression lines represent the 95% confidence interval.

Fig. 4. Growth curves of Lactobacillus crispatus, Streptococccus agalactiae, and Escherichia coli over 30 h in the presence of media alone (yellow) or increasing concentration of Copper (Cu²⁺) ions.

Data represent triplicates of each sample at each Cu²⁺ dilution and are represented as the arithmetic mean (point) and standard error (vertical line). Experiments were performed three times.

Fig. 5. Integrative analysis of bacterial taxa and cytokines differentiating cases (n = 22) from matched controls (n = 94) in the substudy.

a Boxplots displaying the distribution of scores along the fourth sparse principal component (SPC4), which was distinct for cases (n = 20) and controls (n = 93). b Loadings scores for SPC4, which segregates cases from controls. Bars are colored according to which group had higher median scores for SPC4. c The distribution of SPC4 scores grouped by contraceptive arm Cu-IUD (n = 29); DMPA-IM (n = 38); LNG implant (n = 46). d The distribution of SPC4 scores grouped by CST (n = 113 participants). Boxplots display the first and third quartiles and the median value. Two-tailed P values were calculated with a Wilcoxon Rank Sum test and corrected using the method of Benjamini and Hochberg. Bacterial and cytokine absolute concentrations were log₂ transformed and mean-centered prior to analysis.