Cultivable vaginal Lactobacillus is not associated with fecundability in Kenyan women attempting to conceive.

Women enrolled in the Microbiota and Preterm Birth Study in Nairobi and Mombasa, Kenya were eligible for this prospective analysis ( 17 ). Women were eligible for the parent study if they were trying to conceive, HIV-negative, ≤45 years old, and reported a menstrual period in the prior three months or recently discontinued the implant, hormonal intrauterine device (IUD), or depo medroxyprogesterone acetate (DMPA) injectable. Exclusion criteria included: not at risk for pregnancy, history of a condition associated with preterm birth, recent antibiotic use, or history of infertility care-seeking. For this fecundability analysis, additional criteria were applied ( 14 ). Participants were included if they contributed ≥1 menstrual cycle between April 18, 2017 and March 18, 2020, did not report history of ectopic pregnancy, polycystic ovarian syndrome, endometriosis, or hospitalization for PID, and reported ≤3 menstrual cycles of pre-enrollment conception attempt time ( 18 , 19 ). The study was approved by the institutional review boards of Kenyatta National Hospital-University of Nairobi and the University of Washington. Participants provided written informed consent. At enrollment, participants completed an interview including demographics, socio-behavioral characteristics, medical history, and first day of last menstrual period (LMP) ( 17 ). Clinicians collected vaginal swabs during a pelvic examination for STI detection, vaginal Gram stain, and Lactobacillus culture. At monthly preconception visits, participants underwent a urine pregnancy test (One Step Urine Pregnancy Test, EROVITA; Eros Ventures Ltd, Nairobi, Kenya), reported condomless sex frequency for the prior four weeks and first day of most recent LMP, and self-collected vaginal swabs. Beginning in July 2018, participants who missed a preconception visit(s) were asked to report all first days of LMP (“interim menstrual cycles”), not just most recent LMP, since their last visit. Participants were eligible for six months of follow-up for pregnancy except those discontinuing DMPA within three to six months of enrollment, who were eligible for nine months due to delayed return to fertility; women with a DMPA injection within three months of enrollment were not eligible for the parent study) ( 20 ). Participants with genital symptoms at any visit were treated based on Kenyan syndromic management guidelines ( 21 ). Additional treatment was provided for laboratory detected STIs from enrollment samples. Enrollment vaginal samples were tested for Neisseria gonorrhoeae, Chlamydia trachomatis, and Trichomonas vaginalis (Aptima Combo-2 CT/NG Detection System, Aptima Trichomonas vaginalis assay; Hologic Inc., San Diego, CA). Bacterial vaginosis was assessed using Nugent’s criteria ( 22 ). A vaginal sample from each visit was inoculated onto Rogosa agar. The plate was placed immediately in a candle jar and transferred within four hours to anaerobic jars for culture (37°C for 72 hours). Colonies identified on Rogosa agar were confirmed as Lactobacillus through a Gram stain. Lactobacillus colonies were collected with a sterilized wire loop and inoculated into serum vials containing skim milk and stored at −20°C. Every two months, stored colonies were sub-cultured on Rogosa agar in anaerobic jars (37°C for 72 hours) and assessed for Lactobacillus growth. Negative samples were sub-cultured again; repeat negatives were considered negative for sub-culture. Lactobacillus positive samples were sub-cultured on tetramethylbenzidine agar containing horseradish peroxidase and horse serum (TMB-Plus) in anaerobic jars (37°C for 72 hours) ( 23 ). On exposure to air, samples were considered to produce H 2 O 2 if any blue was identified by visual inspection ( 23 ). The outcome for these analyses was fecundability – the per menstrual cycle probability of pregnancy. The number of discrete menstrual cycles experienced by each participant was generated from their first days of LMPs reported at each visit, as previously described ( 14 ). For the 130 participants with visits occurring prior to the addition of the “interim menstrual cycle” question (see Study Procedures), we derived “interim menstrual cycles” using self-reported menstrual cycle duration for those reporting a menstrual cycle length (n=22 derived cycles), while 28 days was used for women reporting unknown cycle length at enrollment (n=19 derived cycles). A total of 128 cycles (9.3%) were derived (n=41) or reported (n=82). Pregnancy was detected using a urine pregnancy test collected monthly. Because pregnancy testing was conducted at variable times during the menstrual cycle, cycles were considered negative for pregnancy only if the test was negative and the participant reported a subsequent LMP at the following visit; this confirmed pregnancy did not occur later in the cycle. For derived and reported cycles and for rare missingness in time-varying follow-up data (<0.05% of cycles), missing data were imputed using data from the last visit carried forward. Menstrual cycle phase at sample collection was estimated to compare the prevalence of Lactobacillus detection by phase. Follicular phase was defined as day 1 through the day before ovulation, and luteal phase as ovulation through the last day of the cycle. Day of ovulation was estimated by subtracting 13 days from cycle length ( 24 , 25 ). For this prevalence comparison, samples from visits with a positive pregnancy test were considered pregnancy samples. To estimate the association between Lactobacillus and fecundability, we utilized discrete time proportional probabilities models to generate fecundability ratios (FR) and 95% confidence intervals (95%CI) ( 19 , 26 ). Specifically, we employed generalized linear models with a binomial distribution and log-link, robust standard errors, and an indicator variable for menstrual cycle at risk. The models incorporated delayed entry for participants with pre-enrollment conception attempt time to reduce left truncation bias (ex: participant with two cycles of pre-enrollment trying time enters analysis at cycle three) ( 18 , 19 ). The FR is interpreted as the per-menstrual cycle probability of pregnancy in cycles with versus without Lactobacillus detection. Censoring criteria included initiation of biomedical infertility treatment (n=3), participant withdrawal or loss to follow-up (n=49), no conception by end of follow-up (n=73), and halting study visits for COVID-19 (n=35). The primary exposure was presence of cultivable Lactobacillus (any growth of Lactobacillus on Rogosa agar) at the visit prior to each pregnancy test allowing for a time-varying analytic approach. We conducted several sensitivity analyses. First, we excluded participants with potential sub-fecundity ( N. gonorrhoeae, C. trachomatis, T. vaginalis or PID at enrollment; any history of PID, N. gonorrhoeae, C. trachomatis, T. vaginalis , or syphilis; self-report of fibroids or unknown uterine abnormality; DMPA use within six months of enrollment; and having a HIV-seropositive partner). Next, we excluded derived and reported interim cycles. Lastly, because women included versus excluded in this analysis had differential detection of cultivable Lactobacillus at enrollment and to assess potential bias associated with exclusion of women with longer pre-enrollment conception attempts, we repeated analyses including women reporting ≤6 months of pre-enrollment trying time. Four secondary exposures were assessed including: 1) level of Lactobacillus growth on culture (none, light growth – 2–3 colonies, moderate growth – covers half of streak, heavy growth – covers entire streak), 2) detection of cultivable H 2 O 2 -producing Lactobacillus at the visit prior (none, Lactobacillus phenotype unknown due to non-growth upon sub-culture, non-H 2 O 2 -producing Lactobacillus , and H 2 O 2 -producing Lactobacillus ), 3) Lactobacillus detected on Gram stain at the visit prior (no/yes, based on results of Lactobacillus sub-score of the Nugent score), and 4) multi-level sub-score indicating the number of Lactobacillus morphotypes detected per high-powered field on Gram stain (none, <1, 1–4, 5–29, ≥30 ( 22 )). A parsimonious set of adjustment factors were selected based on existing literature and consideration of causal relationships between variables. Age (<25, 25–29, 30–34, 35–39, 40–45) and frequency of condomless sex in the prior four weeks (none, 1–4, 5–8, ≥9; time-varying ) were included a priori in multivariable models ( 27 , 28 ). Given study site-specific differences including vaginal washing behavior, most recent contraceptive method, and education level, study site was also included a priori ( 14 ). Bacterial vaginosis was not included in the adjusted model, as BV may be on the causal pathway in the association between Lactobacillus and fecundability. We then considered selected potential confounding factors for inclusion in the adjusted models using a manual forward stepwise approach; factors associated with fecundability at p10%. None of the characteristics considered, including vaginal washing in the last month ( 28 , 29 ), type of recent contraceptive use ( 20 ), any condom use in the last month, maternal education, and household income, were retained in the final models. Sixty-five percent (458/701) of participants in the parent study were eligible for this fecundability analysis. These 458 participants experienced 1,376 menstrual cycles and 255 became pregnant, for a six-cycle cumulative pregnancy rate of 70.4% (95%CI 65.1–75.4). Most participants were 25–29 (32.1%, n=147) or 30–34 (26.4%, n=121) years old, had a prior pregnancy (93.9%, n=430), and reported no pre-enrollment conception attempt time (80.1%, n=367) ( Table 1 ). At enrollment, 65.3% (n=299) of participants had cultivable Lactobacillus , 47.4% (n=217) had cultivable H 2 O 2 -producing Lactobacillus , and 64.6% (n=296) had Lactobacillus morphotypes on Gram stain. The proportion of participants with Lactobacillus at enrollment was similar among those who were retained versus those who were lost to follow-up or withdrew (Culture: Retained 63.6%, n=231/363 vs Loss to follow-up 71.6%, n=68/95, p=0.15; Gram stain: Retained 64.5%, n=234/363 vs Loss to follow-up 65.3%, n=62/95, p=0.89). Participants in the parent study excluded from this analysis were less likely to have any cultivable Lactobacillus than those included (53.3%, n=128/240 vs 65.3%, n=299/458, p<0.002). While participants at the Nairobi site were more likely to have cultivable Lactobacillus (70.2%, n=259/369 vs 44.9%, n=40/89) and less likely to report vaginal washing (28.7%, n=106/369 vs 62.9%, n=56/89) compared to participants at the Mombasa site, there was no difference in detection of Lactobacillus by Gram stain (Nairobi 64.8%, n=239/369 vs Mombasa 64.0%, n=57/89). Of the 1,248 menstrual cycles with vaginal samples collected, 44.2% (n=551) were collected during the follicular phase, 35.4% (n=442) during the luteal phase, and 20.4% (n=255) during early pregnancy. The median gestational age at pregnancy detection was 5.9 weeks (IQR 5.0–7.3). The proportion of cycles with detection of cultivable Lactobacillus or H 2 O 2 -producing Lactobacillus were similar by phase ( Table 2 ). Lactobacillus morphotype detection by Gram stain was more common in early pregnancy samples than follicular or luteal phase samples (79.0% vs 69.3% vs 67.9%). Cultivable Lactobacillus was detected at 65.2% (n=897) of menstrual cycles. Among the 335 women contributing more than one menstrual cycle, 15.5% (n=52) had cultivable Lactobacillus detected at 0% of cycles, 14.3% (n=48) at 1%−49% of cycles, 28.1% (n=94) at 50–99% of cycles, and 42.1% (n=141) detected at 100% of cycles. In unadjusted analysis for this primary exposure, cultivable Lactobacillus at the visit prior to pregnancy testing was not associated with fecundability (FR 0.92, 95%CI 0.73–1.16) ( Table 3 ). Results were unchanged after adjustment for age, frequency of condomless sex, and study site (adjusted FR [aFR] 0.92, 95%CI 0.72–1.18). Similarly, there was no association between cultivable Lactobacillus and fecundability in sensitivity analyses excluding participants with potential sub-fecundity (aFR 0.92, 95%CI 0.71–1.20) and when excluding derived and reported menstrual cycles (aFR 0.96, 95%CI 0.75–1.22) ( Supplemental Table 1 ). When including participants with ≤6 cycles of pre-enrollment conception attempt, 511 participants contributed 1,577 menstrual cycles and 270 pregnancies. There was no association between cultivable Lactobacillus at the visit prior and fecundability (aFR 0.96, 95%CI 0.76–1.22) ( Supplemental Table 2 ). There was no clear association between level of Lactobacillus growth on culture and fecundability ( Table 3 ). Compared to cycles characterized as having no cultivable Lactobacillus , the adjusted FRs for light growth, moderate growth, and heavy growth were 0.82 (95%CI 0.55–1.30), 1.03 (95%CI 0.75–1.40), and 0.83 (95%CI 0.63–1.10), respectively. There was also no association between detection of H 2 O 2 -producing Lactobacillus and fecundability (aFR 0.92, 95%CI 0.71–1.19). Detection of Lactobacillus on Gram stain was associated with modestly higher fecundability (aFR 1.18, 95%CI 0.92–1.51), but this was not statistically significant ( Table 3 ). When considering each level of the Lactobacillus sub-score representing the number of Lactobacillus morphotypes per high-powered field on Gram stain, there was not a clear dose-response association with increasing numbers of morphotypes detected. However, the highest Lactobacillus sub-score (4, indicating ≥30 morphotypes per high-powered microscope field) was associated with the highest per-cycle probability of pregnancy (aFR 1.20, 95%CI 0.92–1.57).

Given the predominance of Lactobacillus species in the optimal vaginal environment and their association with protection against adverse reproductive outcomes, it is plausible that vaginal Lactobacillus could promote fecundity. However, in this cohort of Kenyan women, neither cultivable Lactobacillus nor cultivable H 2 O 2 -producing Lactobacillus were associated with fecundability. There was, however, a non-significant association between Lactobacillus detected on Gram stain and higher fecundability. This highlights the need for larger studies powered to detect effect sizes of ≥15–20% and underscores the need for detailed quantitative and species-specific analyses of the association between vaginal Lactobacillus and fecundity. These results may also suggest a potential role of vaginal Lactobacillus species that do not grow on Rogosa agar, such as L. iners, in promoting fecundity. It is also possible that the presence of BV-associated bacteria, rather than the absence of Lactobacillus , is associated with lower fecundability. Future studies of vaginal Lactobacillus species and fecundability would benefit from molecular testing to explore the presence, species, functionality, and immunological effects of vaginal Lactobacillus during the periconception period.

In this prospective preconception cohort of Kenyan women, contrary to our hypothesis, there was no association between cultivable Lactobacillus , or abundance of growth, and fecundability. However, while not statistically significant, fecundability was 18% higher in menstrual cycles with Lactobacillus morphotypes detected on Gram stain. We previously reported that women in this cohort had a 17% lower fecundability associated with recent BV (Nugent score ≥7) ( 14 ). The Nugent score includes three sub-scores, of which the Lactobacillus morphotype score is one ( 22 ). Notably, the Lactobacillus detected on Gram stain results presented here show a modestly higher fecundability (+18%), which is the inverse of the association observed in our earlier study of BV (−17%). Overall, the culture and Gram stain results point to a complex relationship between the presence, functionality, and species of vaginal Lactobacillus and fecundity. The difference in the associations between fecundability and Lactobacillus detected by culture on Rogosa agar versus Gram stain may reflect the strengths and limitations of these methods. Rogosa agar supports the growth of Lactobacillus species that are most consistently associated with vaginal health (i.e., L. crispatus and L. jensenii ), but does not support the growth of L. iners which is common in women of African descent ( 30 – 34 ). Lactobacillus morphotype detection on Gram stain may have detected a wider range of Lactobacillus species such as L. iners , L. gasseri , and L. vaginalis . Molecular methods, on the other hand, detect the relative abundance or absolute quantities of specific species. Few studies have assessed whether there is an absolute concentration threshold for detection of Lactobacillus on culture ( 35 ). When considering the proportion of visits with Lactobacillus detection by timing of sampling – luteal, follicular, or early pregnancy – Lactobacillus prevalence was similar when considering culture results, but approximately 10% higher in early pregnancy compared to the luteal and follicular phase when assessing Lactobacillus on Gram stain. This could reflect an early shift of the vaginal microbiota to one that is Lactobacillus dominated in pregnancy or could suggest that those who become pregnant are more likely to have peri-conceptual Lactobacillus species detected on Gram stain. Studies of reproductive tract microbiota and fertility have been conducted in populations seeking IVF. A unique feature of our study was that participants were attempting natural conception and did not have a prior infertility diagnosis. The reproductive tract microbiota of women undergoing medically assisted reproduction may be affected by infertility protocols. Moreover, studies assessing reproductive tract microbiota and IVF outcomes have primarily included White women and results may not be generalizable to African women. Nonetheless, these studies provide the only comparisons. In a study of 91 women undergoing IVF, H 2 O 2 -producing Lactobacillus detected by culture at embryo transfer had a higher live birth rate (vaginal fluid: 50% vs 21%; transfer tip: 70% vs 25%) ( 7 ). In a study of 32 women with endometrial fluid samples collected prior to the embryo transfer cycle, those with a relative abundance of Lactobacillus >90% had higher rates of implantation (60.7% versus 23.1%), pregnancy (70.6% versus 33.3%), ongoing pregnancy (58.8% versus 13.3%), and live birth (58.8% versus 6.0%) compared to women with <90% Lactobacillus ( 8 ). Similarly, when considering 150 vaginal samples collected prior to embryo transfer, women with vaginal microbiota dominated by L. crispatus were more likely to have a biochemical pregnancy, clinical pregnancy, and live birth ( 9 ). In contrast to these studies suggesting better pregnancy outcomes associated with the presence of reproductive tract Lactobacillus species in women undergoing IVF, several studies show no association ( 10 , 11 , 13 ). One study including 75 women undergoing embryo transfer within two months of vaginal sampling found no significant differences in biochemical or clinical pregnancy rates among women with three community state types defined by high relative abundance of L. crispatus , high relative abundance of L. iners , and a diverse community of bacteria (biochemical pregnancies: 44% versus 59% versus 30%; clinical pregnancies: 33% versus 53% versus 10%) ( 11 ). While these differences were non-significant, the sample size was small. Interestingly, biochemical and clinical pregnancy rates were highest in women whose vaginal bacterial communities were dominated by L. iners . A hypothesized mechanism explaining associations between vaginal and endometrial microbiota and reduced reproductive success is disruption of embryo implantation due to the immune response to pathogenic microbiota, with optimal Lactobacillus preventing this response ( 36 ). In the context of non-medically assisted reproduction, reproductive tract microbiota disruption may additionally interrupt the physiochemical properties of cervical mucus and inhibit sperm function ( 37 , 38 ). A reduction in Lactobacillus spp. abundance and increase in BV-associated bacteria contributes to degradation of cervical mucus and ascension of microorganisms into the upper reproductive tract ( 39 ). This may also disrupt sperm transport through the cervix. In addition, the inflammatory response may increase reactive oxygen species production and oxidative stress leading to reduced sperm motility and DNA fragmentation in the upper reproductive tract ( 37 , 38 ). One in vitro study found that a Lactobacillus probiotic containing L. brevis , L. salivarius , and L. plantarum protected against reactive oxygen species-induced lipid peroxidation of sperm and was associated with better sperm motility ( 40 ). These lactobacilli, however, are not typically found in the vagina. Several studies have identified Lactobacillus spp. as one of the most abundant species in semen ( 41 – 44 ), and men with normal semen parameters had higher relative abundance of seminal Lactobacillus species, suggesting a beneficial effect ( 42 , 44 ). Major strengths of this analysis were its prospective design, inclusion of women attempting to conceive naturally, monthly vaginal sampling allowing for time-varying analysis, and detection and characterization of Lactobacillus using three methods including culture on Rogosa agar, sub-culture for H 2 O 2 production, and Gram stain. This approach allowed for assessment of the presence and functionality of Lactobacillus at multiple time points during the periconception period. This study also had limitations. First, our study provides only a semi-quantitative assessment of Lactobacillus and did not employ molecular methods of detection. Lactobacillus concentrations, rather than presence versus absence, may be more strongly associated with fecundability. The results of our exploratory analyses showed a modestly higher (though non-significant) fecundability associated with the highest numbers of Lactobacillus morphotypes identified on Gram stain, while no such association was observed with the highest abundance of Lactobacillus growth in culture. Molecular methods of detection could provide data on quantity or abundance of species, whereas the culture method used in this study demonstrates viability. Second, for participants who missed preconception visits during the first year of the study, we derived interim menstrual cycles and assigned exposure data based on the most recent visit. A sensitivity analysis was performed excluding these cycles, demonstrating similar findings. Third, there was likely residual confounding by frequency of condomless sex due to monthly self-report and lack of a reliable measure of condomless sex within biologically-confirmed fertile windows. Fourth, we may have under-ascertained biochemical pregnancies due to the indiscriminate timing of pregnancy testing, but this would not have been differential by Lactobacillus status, minimizing concern for bias. Fifth, over 90% of participants reported at least one prior pregnancy so results may not be generalizable to nulligravid women.

Absence and low levels of vaginal Lactobacillus species have been associated with adverse reproductive health outcomes, including bacterial vaginosis (BV), sexually transmitted infections (STI), and preterm birth ( 1 , 2 ). Lactobacillus species contribute to an optimal vaginal environment by producing lactic acid, maintaining an acidic pH, and preventing colonization by other bacteria ( 3 ). While epidemiological studies have demonstrated associations between hydrogen peroxide (H 2 0 2 ) producing lactobacilli, such as Lactobacillus crispatus and L. jensenii, and lower risk of adverse reproductive outcomes, the mechanism is debated ( 4 – 6 ). Lactobacillus species may contribute to improved outcomes for women undergoing in-vitro fertilization (IVF) ( 7 – 13 ), but there is a paucity of data on vaginal microbiota and reproductive success in women not undergoing assisted reproduction. We recently demonstrated an association between BV, which is characterized by reductions in optimal Lactobacillus, during the preconception period and reduced fecundability in Kenyan women attempting non-medically assisted conception ( 14 ). If a vaginal microbiota dominated by optimal Lactobacillus species is important for fecundity, then interventions such as Lactin-V, a vaginal probiotic, or periodic presumptive treatment of BV among those with BV recurrence may be worth exploring for some women ( 15 , 16 ). The objective of this analysis was to test the hypothesis that cultivable vaginal Lactobacillus would be associated with higher fecundability.