Effects of tampons and menses on the composition and diversity of vaginal microbial communities over time


Correspondence: Dr LJ Forney, Life Sciences South 441A, University of Idaho, Moscow, ID 83844, USA. Email lforney@uidaho.edu



To investigate the influence of menses on the vaginal microbiota and determine whether tampons that differ in material composition influence these bacterial communities in different ways.


A single-centre trial with randomised, complete block design.


Procter & Gamble facility.


Seven self-declared healthy, female volunteers of reproductive age.


Volunteers used a pad and two types of tampons during the study, one product exclusively each month for three sequential menstrual cycles. During menses and once each mid-cycle, vaginal bacterial community composition was characterised by cultivation-independent methods based on pyrosequencing of V1–V2 variable regions of 16S ribosomal RNA genes.

Main outcome measures

Changes in the species composition, abundance and diversity in vaginal bacterial communities over time and between treatments.


The vaginal microbiotas of all seven women were dominated by Lactobacillus spp. at mid-cycle, and the compositions of those communities were largely consistent between cycles. Community dynamic patterns during menses varied considerably and were more or less individualised. In three of the seven women the community diversity during pad use was significantly different from at least one tampon cycle.


Changes in the composition of the vaginal microbiota during menses were common, but the magnitude of change varied between women. Despite these changes, most communities were capable of resuming a composition similar to previous mid-cycle sampling times following menstruation. Overall we conclude that the two tampons tested do not significantly impact the vaginal microbiota in different ways; however, larger studies should be performed to confirm these findings.


Bacterial communities indigenous to the human vagina are thought to play an important role in protecting the host against infectious disease and maintaining health. Historically, a high abundance of Lactobacillus spp. has been considered a hallmark of a healthy vaginal microbiota because they create an acidic environment that is hostile to invading pathogens.[1] However, more recent studies performed using cultivation-independent methods have demonstrated that the composition of the vaginal microbiota can vary widely among healthy women in terms of the species present and their relative abundances.[2-8] Although the majority of healthy women of reproductive age have vaginal bacterial communities dominated by Lactobacillus spp., non-Lactobacillus species dominate the communities of up to 40% of women in some racial groups. Healthy vaginal microbiota often harbour other lactic acid bacteria such as Streptococcus or Atopobium as well as a diverse array of strict anaerobes including Prevotella, Dialister and Megasphaera.[4, 7]

Although we are developing a more complete understanding of bacterial diversity in the vaginal microbiota, the temporal dynamics of these communities and their responses to perturbations are not well understood. In other ecological environments, changes in the species composition and function of bacterial communities can, and often do, occur as a result of changes in the physical or chemical properties of the environment.[9] Therefore, it is reasonable to assume that differences in vaginal community composition and structure could be significant if communities respond differently to changes in the vaginal environment related to menses or hormonal cycles. Furthermore, if the vaginal microbiota within a woman changes over time, it could vary in its overall function and, potentially, its ability to maintain host health. Few studies have been performed to systematically explore this issue, and those that have done so have relied primarily on cultivation-based methods.

Most investigators have reported that although the vaginal microbiota remains relatively constant throughout most of the menstrual cycle, changes in community composition often occur during menstruation. However, reported findings differ somewhat in their details, probably as a result of slight differences in the methodologies used. Several studies noted fluctuations in the relative levels of aerobic and facultative bacteria in relation to strictly anaerobic bacteria, but the data were often opposing as to which group became more or less abundant during menses.[10-13] It was also common to observe an increase in variability or diversity of the vaginal microbiota during menses,[14-18] even when overall microbial growth was reduced.[12, 19] With respect to Lactobacillus, Keane et al.[16] and Eschenbach et al.[20] observed decreased relative abundance during menses and a subsequent replenishment thereafter. Other studies determined that changes occurring during menses were not clinically significant with respect to the numerically dominant populations of bacteria, which usually remained present throughout the entire menstrual cycle,[21] but one team noted that the techniques employed in their study may have underestimated the true dynamics of the vaginal microbiota.[22] In any case, there were almost always exceptions to the rule, suggesting that there is no simple explanation for the behaviour of the vaginal microbiota during menstruation. In a similar vein of research, others have investigated whether tampons affect the vaginal microbiota in clinically relevant ways. Aside from one group that observed the number of coagulase-negative staphylococci becoming more prevalent in relation to other bacteria during menses with the use of tampons,[23] all other studies concluded that tampons do not significantly influence the vaginal microbiota compared with the normal dynamics of these communities.[13, 19, 24-26]

Whereas some common themes emerge from these findings, the ability to draw reliable conclusions is limited because all of these studies employed cultivation-based methods that fail to account for many of the bacterial populations now known to exist in the vaginal microbiota.[2-4, 7, 27] An important example of this is the fact that Lactobacillus iners, now known to be one of the most prevalent bacterial members of the vaginal microbiota, does not grow on culture media normally used for the cultivation and enumeration of Lactobacillus.[28] Similarly, many highly fastidious strict anaerobes now known to be common members of vaginal communities may not have been recovered using traditional cultivation techniques. Given that our knowledge is constrained by methods used in the past, it is important to re-examine fundamental questions about vaginal microbiology using cultivation-independent methods to gain a more comprehensive understanding of these communities. Our study employed such methods, namely pyrosequencing of 16S ribosomal RNA (rRNA) genes, to evaluate whether the vaginal microbiota changes throughout menstruation with the use of two tampons differing in material composition.

In the study described herein, our primary objective was to determine whether and how menses influenced the structure and composition of vaginal bacterial communities over the course of a woman's menstrual period. Additionally, we sought to establish whether tampons of different material composition affected the communities in different ways, if at all. The study design also allowed us to evaluate changes in composition across sequential menstrual cycles to determine whether the structure and composition of microbial communities were significantly different from one cycle to the next. In this study we employed high-throughput pyrosequencing technology to generate thousands of 16S rRNA sequence reads per sample so that the vaginal microbial communities could be characterised more comprehensively than in previous similarly oriented studies.


Subjects and clinical study design

This was a single centre, randomised, controlled study designed in collaboration with and conducted by The Procter & Gamble Company (Cincinnati, OH, USA). Eight healthy, menstruating women between the ages of 18 and 45 years were enrolled in the study. Subjects were eligible for enrolment if they had regular menstrual cycles (21–35 days) and menstruated for at least 3 days; were in good general and gynaecological health (self-reported at each scheduled visit); were tampon users; did not bathe or shower within 2 hours of their scheduled visit; refrained from douching or applying powders, perfumes, wipes or emollients to the genital area for 48 hours before their scheduled visit; refrained from vaginal sexual intercourse for 48 hours before their scheduled visit; were willing to refrain from using medications containing nonsteroidal anti-inflammatories for 48 hours before their scheduled visit; and were willing to comply with all other protocol requirements. Subjects were not eligible if they had a self-reported or physician-documented history of toxic shock syndrome or symptoms consistent with toxic shock syndrome; self-reported difficulty wearing tampons or had experienced irritation from the use of a tampon product; had a self-reported pregnancy; self-reported staphylococcal or streptococcal infection within 3 months of the start of the study; used antibiotics or antifungal medication(s) for 6 weeks before their scheduled visit; a recent history of prescription anti-inflammatory drug or steroid use; self-reported vaginal boils or lesions; self-reported diabetes or active yeast infection; body piercing(s) in the genital region; medical condition such as cancer, anaemia, leucopenia, leucocyte function deficiency or malnutrition that might compromise immune system functions; or a history of immunosuppressive drug therapy, chemotherapy or radiation therapy. The study protocol and informed consent document were approved by Procter & Gamble's Institutional Review Board. Informed consent from all women was documented before participation in the study.

All women were asked to use three types of catamenial products during the study, one product exclusively per month. These included a pad (P), tampon A (TA) and tampon B (TB). The difference between tampons A and B was the presence of calcium citrate malate in tampon B; this compound aids drying in tampon manufacturing. Product use was randomised as indicated in Table 1. Tampon and pad usage were per the woman's normal routine, but we ensured that a tampon had been worn for 2 hours before sampling. Women had the option of using a menstrual pad or panty-liner provided by the investigators with or without a tampon, but tampons were not used during an assigned pad cycle. Subject eligibility and compliance was confirmed with each scheduled visit. Women were given written and oral instructions on collection of the self-obtained vaginal swab. Before self-sampling, women were instructed to remove the tampon if they were using tampons during that specific menstrual cycle. After self swab collection, the swab was immediately placed in a freezer at − 70 ± 10°C until they were shipped on dry ice to the University of Idaho for analysis.

Table 1. Randomised assignment of catamenial products over three menstrual cycles
CycleProduct type
Part 1Part 2Part 3Part 4Part 5Part 6Part 7
  1. P, pad; TA, tampon A; TB, tampon B.


Testing was conducted during three menstrual cycles. Subjects were scheduled for visits to the test facility as required by the protocol; this also included weekend visits. It was anticipated that some of the women might have been unable to be scheduled within consecutive menstrual cycles. The maximum amount of time that a woman would be enrolled in the study was six menstrual cycles. All women underwent vaginal microbiota sampling at their first visit, on each day of menstruation, and at mid-cycle between cycles 1 and 2 and 2 and 3, and at the end of the study. Mid-cycle was defined as approximately day 14 of the menstrual cycle, where day 1 is the start of menstruation. The number of swabs collected during each menstrual period differed among and within women because of variable period lengths. A schematic of the sample collection strategy is shown in the Supplementary material (Figure S1). One woman was dropped from the study following menstrual cycle 1 because she failed to meet the inclusion criteria of “has a regular menstrual cycle and menstruates for at least 3 days.” Self-obtained vaginal swabs collected from this woman before dropout were not analysed.

Genomic DNA extraction from vaginal swabs

We extracted and purified genomic DNA from 111 vaginal swabs using a method modified from that described by Li et al.[29] Before extraction, vaginal swabs were thawed on ice and vortexed vigorously in pre-reduced anaerobic sterilised liquid dental transport medium every 5 minutes for 15 minutes. A 0.5-ml aliquot was transferred to a sterile 2.0-ml bead-beating tube and kept on ice. Cell lysis was initiated by adding 50 μl lysozyme (10 mg/ml; Sigma-Aldrich, St Louis, MO, USA), 6 μl mutanolysin (25 000U/ml; Sigma-Aldrich), 3 μl lysostaphin (4000 U/ml in sodium acetate; Sigma-Aldrich) and 41 μl TE50 buffer (10 mmol/l Tris–HCl, 50 mmol/l EDTA, pH 8.0). Following a 1-hour incubation at 37°C, the microbial cells were lysed further by adding ~750 mg of 0.1-mm zirconium silica beads (BioSpec, Bartleseville, OK, USA) to each sample and mechanically disrupting the mixtures in a mini-bead beater (BioSpec) at 6.0 m/second for 1 minute. The lysates were processed using a QIAamp DNA Mini Kit (Qiagen, Germantown, MD, USA) according to the manufacturer's protocol with omission of the lysis steps. DNA was eluted twice through QIAmp spin columns provided in the kit with 50 μl hot (56°C) molecular biology-grade water. DNA concentrations were measured on a NanoDrop spectrophotometer (Thermo Scientific, Waltham, MA, USA).

Pyrosequencing of barcoded 16S rRNA gene amplicons

Polymerase chain reaction (PCR) was carried out on the genomic DNA samples using universal barcoded eubacterial primers 27F and 338R (Sigma-Aldrich). These primers flank the variable regions V1 and V2 of the 16S rRNA gene as characterised by Neefs et al.[30] The PCR products were quantified using gel electrophoresis and spectrophotometry (on a NanoDrop), and approximately 100 ng of product from each sample was shipped on dry ice to the Institute for Genome Sciences, University of Maryland School of Medicine for pyrosequencing on a Roche 454 GS-FLX DNA sequencer as previously described.[7] The partial 16S rRNA gene sequences were binned on the basis of the unique sequence barcodes associated with the unique primer used for each sample. Those that were sufficiently long and met other quality control criteria were classified to the genus level, and resulting sequences belonging to the genus Lactobacillus were further classified as species as previously described.[7] A small proportion of Lactobacillus phylotypes could not be classified to the species level using the methods employed here and are reported as Lactobacillus spp.

Statistical analyses

Descriptive and statistical analyses were performed to evaluate changes in vaginal bacterial community composition and diversity over three menstrual cycles in seven women. First, vaginal bacterial community composition data were normalised as described in the Supplementary material, Appendix S1 and summarised in the Supplementary material, Table S1. We performed descriptive analyses to assess broad-scale similarities and differences as well as to characterise temporal patterns in the vaginal bacterial communities. First, nonmetric multidimensional scaling (NMDS) and hierarchical clustering were carried out to visually assess similarities among all the vaginal communities. To do this the community abundance (read count) data were transformed using the Hellinger method,[31] and a distance matrix was computed on the transformed data using Euclidean distance. This distance matrix was used to perform both NMDS and hierarchical clustering. For NMDS, random starts and axis scaling (centring, principle component rotation and half-change scaling) were used to find a stable solution. Ward's minimum variance criterion[32] was used to compute the hierarchical cluster dendrogram. We also constructed bar plots from the untransformed data to compare the relative proportions of taxa across all samples from each individual. All statistical analyses and resulting figures were generated in R[33] using packages vegan,[34] MASS,[35] multcomp[36] and distrEx,[37] as well as the graphics packages RColorBrewer[38] and moduleColor.[39].

The Shannon diversity index, commonly used in ecological studies, is a quantitative measure of diversity that accounts for both the number and relative abundance of taxa in a community. We calculated Shannon indices[40] with the ‘diversity’ function in vegan,[34] and these were used in the remaining statistical analyses. Fixed effects linear models were tested to identify variables significantly associated with differences in Shannon indices across samples from all seven women; these details can be found in the Supplementary material, Appendix S1 and Table S2. Following the model testing, we used pairwise comparisons to determine the effect of product type on the community diversity for each woman individually while controlling for time. “TAP” instances and mid-cycle samples were excluded from the analysis. P-values were adjusted using the Benjamini–Hochberg method.[41] Treatments were considered significantly different if the resulting P-value was less than 0.05. Q-Q and residual plots associated with the chosen model were assessed for equality of variance and normality. Conforming to the normality assumption was further assessed using the Shapiro–Wilk test as implemented in R.[42]


The bacterial composition of vaginal bacterial communities in seven asymptomatic women of reproductive age during menses and mid-cycle were characterised over three menstrual cycles. Each woman provided 13–19 self-collected vaginal swabs (average 16), totalling 111 samples and 84 that were collected during menses. The average length of sampling during a menstrual period was 4 days, with a minimum of 3 days and maximum of 5 days. The bacterial diversity of these communities was determined by pyrosequencing the V1–V2 variable regions of 16S rRNA genes amplified from total genomic DNA that had been isolated from each sample. Excluding sample 7-TB-2 which was removed from the dataset (see Methods), an average of 3233 sequence reads with length >200 base pairs were obtained from each sample (range 495–7739; median 3004). Further excluding taxon reads constituting <1% of the median total read count, the average read count was 3182 (range 423–7700; median 2971).

Temporal dynamics of vaginal bacterial communities

We performed hierarchical clustering and NMDS to obtain visual representations of the overall similarity among vaginal bacterial community samples. Figure 1 shows the hierarchical clustering dendrogram of the samples based on the relative abundances of the bacterial taxa in each community. Many samples clustered closely by participant, but it was also common for community composition to be different between samples collected on different days from the same woman. For instance, samples from participants 6 and 7 cluster closely within participant, but samples from participants 2 and 8 are spread across the dendrogram. A similar trend was observed in Figure 2, which shows the NMDS plot of all samples based on community composition. Both results indicate that women experienced changes in community composition during menses to varying degrees, but there does not appear to be any pronounced effect related to a specific catamenial product.

Figure 1.

Hierarchical clustering dendrogram of all samples. Coloured bars below the dendrogram correspond to the different participants. Height refers to the Euclidean distance between samples. In the sample names the first number refers to the participant (2 through 8); the middle letter(s) indicates whether a pad (P), tampon A (TA) or tampon B (TB) was used, or if it was a nonmenstrual mid-cycle (MC) sample; and the last number refers to the sequence of sample collection within each treatment. In participants 5 and 8, ‘TAP’ refers to a pad used during an assigned tampon A cycle. For mid-cycle samples, ‘MC-0′ is the baseline sample collected before any product treatments, and ‘MC-1’ to ‘MC-3’ are the mid-cycle samples collected during each of the three study cycles.

Figure 2.

NMDS plot of all samples. NMDS was performed using the Euclidean distance matrix computed from community abundance data transformed using the Hellinger method.[31] Sample labels are the same as in Figure 1 and are coloured by participant.

Consistent with previous research[8] we found that dynamic patterns in vaginal bacterial communities varied largely among women and were more or less “personalised” to each woman. Figure 3 shows the bar plots comparing the relative proportions of taxa at each sampling time in the seven women. Most of the participants experienced some degree of change in community composition during menses, but no overall trends were apparent. The vaginal microbiotas of all seven women were typically dominated by Lactobacillus spp. throughout most of the study, particularly L. crispatus and L. iners, and to a lesser extent, L. gasseri and L. jensenii. Lactobacillus salivarius and L. vaginalis were also present in some samples from three participants (2, 6 and 7), but they constituted very low proportions of these communities (<0.05). Even in women with similar community composition at mid-cycle, dynamic patterns during menses could be quite different. Two particularly interesting dynamic patterns were observed for participants 4 and 8. The vaginal microbiota of participant 4 during mid-cycle was populated by L. iners, L. crispatus and, to a lesser extent, L. jensenii. During menses, L. crispatus initially increased in relative abundance, but Streptoccocus began to increase quite markedly shortly thereafter. The vaginal microbiota of participant 8, on the other hand, was dominated almost entirely by L. crispatus during mid-cycle, but during each menstrual period the relative abundance of L. iners increased. Overall, despite the various kinds of changes that women's vaginal microbiotas experienced during menses, a coarse view across the mid-cycle samples of each woman reveals that the communities were capable of resuming a composition similar to previous mid-cycle samplings. This suggests that there is some degree of resilience in vaginal communities following menses, but this result should be interpreted with caution because the women were sampled at only a single mid-cycle time-point during each menstrual cycle and it is therefore unknown how the communities may have fluctuated throughout the entire cycle.

Figure 3.

Bacterial community composition of all vaginal swab samples from seven participants. Bar plots were generated from proportions of 27 taxa in each sample. The x-axis shows sample names as described in Figure 1.

Comparison of community diversity across catamenial product types within participants

Vaginal bacterial community diversity varied among and within participants. Figure 4 shows the box plots of the Shannon index values for the mid-cycle vaginal swabs as well as the three product treatments within each participant. Although there was considerable variability in Shannon indices within and across participants, no consistent patterns emerged. For example, diversity was sometimes higher during menses relative to mid-cycle (participant 7), roughly the same (participant 6), or somewhere in between. To investigate this further, we performed hypothesis testing to determine whether vaginal communities associated with the three catamenial products were significantly different from each other. Results of the pairwise t-tests comparing product treatments within each participant are listed in Table 2. Five comparisons in three participants were found to be significantly different: pad versus tampon A (= 0.0134) and pad versus tampon B (= 0.0134) in participant 3; pad versus tampon A (= 0.0134) in participant 4; and pad versus tampon A (= 0.0005) and pad versus tampon B (= 0.0066) in participant 5. In participant 3, the samples collected during pad use were more diverse than either of the tampon treatments; however, the opposite trend was observed in participants 4 and 5 (i.e. diversity was higher on average in both tampon cycles). In all three of these women, the Shannon indices were more similar between the two tampons than between either tampon and the pad, supporting our hypothesis that the two tampons do not affect the vaginal microbiota in different ways. Furthermore, the remaining four participants showed no significant difference in community diversity among all three product treatments. The outcome of fixed effects model comparison is shown in the Supplementary material, Table S3; the best-fit model was one that identified participant, product, time (linear), participant-time and product–participant interactions as significant.

Table 2. P-values of linear hypothesis tests comparing catamenial product treatments by participant
P vs TAP vs TBTA vs TB
  1. P, pad; TA, tampon A; TB, tampon B.

  2. P-values were adjusted using the Benjamini–Hochberg method and are significant at *α = 0.05, **α = 0.01, or ***α < 0.001.

Figure 4.

Boxplots of Shannon index values by treatment per participant. The y-axis of each plot indicates the value of the Shannon index. The x-axis labels indicate nonmenstrual mid-cycle samples (MC) or samples collected during catamenial product use (P, pad; TA, tampon A; TB, tampon B). The first seven panels (left to right, top to bottom) are organised by individual participant, whereas the last panel summarises the Shannon indices across all seven participants.


Our results paint a complex picture of vaginal bacterial community dynamics in healthy women of reproductive age. Unlike most previous studies that described community composition on a cross-sectional basis during menses, the longitudinal sampling employed in our experimental design allowed us to more closely follow the day-to-day changes that the vaginal microbiota experience during menstruation with the use of different catamenial products. The results showed that changes were common and fluctuations can occur on a short time scale. The nature of changes in community composition differed among individuals and sometimes between menstrual periods of a single individual. In several women, changes in community structure were relatively insignificant, whereas in some the shifts were more pronounced. There were also instances in which the relative abundance of otherwise minor members of the community quickly increased to constitute a high proportion of the community during menses. For example, in one woman the relative abundance of Streptococcus spp., virtually undetected at each mid-cycle, rose to high abundance during menses and then decreased to baseline levels by the following mid-cycle. In another woman, L. crispatus, highly prevalent at each mid-cycle, gradually gave way to elevated levels of L. iners during each menstrual period. In contrast, the microbiota of another woman was comparatively invariable over the course of the study. The monthly mid-cycle sampling also allowed us to evaluate whether communities were likely to “rebound” after communities had been altered during menses. The communities we evaluated appeared to be very capable of doing so, but we do not know if this pattern is sustained over longer periods of time.

The patterns of changes in vaginal community composition were similar to those of previous cultivation-based studies that reached somewhat different conclusions. Our common observation of minor fluctuations in some genera during menses echoed the various degrees of fluctuation seen in both aerobic and anaerobic populations in previous studies,[10, 12, 14, 15] but a direct comparison is not possible because our methods are inherently qualitative whereas theirs were primarily quantitative. The different patterns we observed with respect to the relative abundance of Lactobacillus spp. can be seen in previous studies, which observed either general declines in this organism during menses[11, 16, 20] or no change at all throughout the menstrual cycle.[22]

Recently, Gajer et al.[8] published findings that revealed wide-scale variability in temporal dynamic patterns among the vaginal microbiomes of 32 women who self-sampled twice weekly for 16 weeks. They found that patterns were very complex and somewhat individualised, and communities were dynamic not only during menses but in many cases throughout the menstrual cycle. Given this, it is not surprising that the community dynamics we observed during menses in just seven women were also diverse. It could well be that changes in community composition during menses are the norm, but that the type and magnitude of changes differ among individuals. Future studies should seek to understand the importance and cause of variability in the vaginal microbiota not only during menses, but throughout the menstrual cycle.

Overall, our findings indicate that changes in community composition and diversity are not strongly influenced by tampon use. However, because of the limited scale of our experimental design, there was insufficient power to test whether tampon use significantly affected specific taxa during menses. Instead, we tested whether community diversity represented by the Shannon index was significantly affected by catamenial product type. In two women, community diversity during both assigned tampon cycles was significantly different from the assigned pad cycle; in another woman, only tampon A was different from the pad. Even though we identified significant differences between these treatments, different trends were observed. In two women the tampon-associated microbiota had higher diversity, whereas in the other woman the pad-associated microbiota was more diverse. None of the differences between tampon A and tampon B were significant among the seven women, indicating that the two tampons probably did not impact the vaginal microbiota in different ways. Taken together with visual interpretations of the similarities and differences among samples based on community composition, our data suggest that tampon use does not significantly alter the normal dynamics of vaginal communities. A larger analysis with more individuals and repeated treatments of products would allow for multivariate testing to determine whether specific taxa are significantly affected by the treatments.

Our conclusion is consistent with the findings of previous studies that evaluated the effects of different tampons on the vaginal microbiota. Some of the earlier, most extensive investigations were published by Onderdonk et al.[13, 19, 21] in the mid-to-late 1980s. Women in these studies used tampons of various combinations of cotton and rayon fibres, and they were sampled over an average of seven[19, 21] or nine[13] menstrual cycles, during which vaginal swabs were collected on days 2, 4 and 21 of each cycle. Both studies concluded that the tampons evaluated did not significantly alter the vaginal microbiota in terms of microbial growth or abundance of the numerically dominant microbial populations, regardless of fibre type.[13, 19, 21] Three more recent studies conducted by Shehin et al.[24] and Chase et al.[25, 26] evaluated tampons that differed in shape[24] or material[25, 26] over two menstrual cycles and concluded that none of the tampons tested were associated with significant changes in the vaginal microbiota either. As was mentioned previously, however, all of these studies employed culture-based methods that preclude the detection of rare or fastidious organisms. Therefore, there was still considerable motivation to determine whether tampons could exert a more fine-scale influence, and our results help to bridge this knowledge gap.

To our knowledge, our study is the first to employ cultivation-independent methods to characterise vaginal microbiota and analyse samples from each day of menses for multiple menstrual cycles. A few previous studies had collected samples daily throughout menses and the rest of the menstrual cycle, but they were limited in their ability to characterise the bacteria present because they relied on Gram stain scoring of vaginal smears.[16-18] Most other studies describing the effects of menses on the vaginal microbiota evaluated women on only one or two days during menses, and only three studies spanned three or more menstrual cycles.[12, 13, 19] Regarding the effects of tampons on the vaginal microbiota during menses, the three studies with experimental designs[24] or material[25, 26] most similar to ours all employed cultivation-based methods that preclude the more fine-scale characterisation we were able to obtain based on partial 16S rRNA gene sequences obtained from each community sampled.

Nonetheless, there are certain limitations to our study that impede any broad conclusions. The number of individuals enrolled in this pilot study was small, and each catamenial product was tested during only one menstrual period in each woman, so limiting our ability to conduct rigorous statistical analyses. Additionally, the cultivation-independent methods employed here cannot produce estimates of the total numbers or concentrations of bacteria present at a given time, so we are unable to compare estimates of quantitative changes obtained in previous studies. Furthermore, we suspect that many factors may influence the dynamics of vaginal communities during menstruation, such as the volume of menstrual flow, frequency of changing tampons, elements of the host immune system and mucosal environment, various sexual and hygienic practices,[43] and so forth. To assess the importance of these variables on vaginal community dynamics, larger studies with replication of treatments over consecutive menstrual periods are necessary.


We investigated the temporal dynamics and diversity of vaginal microbiota in seven self-declared healthy, women of reproductive age over the course of three menstrual cycles while using a different catamenial product for each menstrual period. Changes in the composition of vaginal microbiota during menses were common among the participants, but the magnitude of change varied between women. Despite these changes, our data suggest that the communities were capable of resuming a composition similar to previous mid-cycle samples following menstruation. Community diversity during pad use was significantly different from at least one tampon cycle in three women, but the trend differed among individuals. Overall, we conclude that the two tampons tested do not exert a notable impact on vaginal bacterial community composition. However, because of the small sample size of our study and the variety of community dynamic patterns observed in the data, larger studies enrolling more participants for longer periods of time are warranted.

Disclosure of interests

LJF and ZA consult for the Procter & Gamble Company. All other authors have nothing to disclose.

Contribution to authorship

RJH and XZ processed the samples and analysed and interpreted data; RJH along with LJF interpreted the findings and wrote the manuscript. ZA performed statistical analyses and wrote sections of the manuscript pertaining to those analyses. KN and MH designed the clinical study, prepared the clinical protocol and conducted the clinical study. TWO and FW developed the test products; TWO contributed to the clinical study design.

Details of ethics approval

The clinical protocol for this study (PPCT-07018-FC) was reviewed and approved by the Procter & Gamble Institutional Review Board on 8 June 2007.


This research was supported by a grant from the Procter & Gamble Company.


We wish to thank Jacques Ravel and Pawel Gajer at the Institute for Genome Sciences of the University of Maryland School of Medicine for their assistance in classification of partial 16S rRNA gene sequences. We also thank Rebecca Patterson at Procter & Gamble for her help in preparing the test products as well as Julie Erb at Procter & Gamble for monitoring the study.

Commentary on ‘Effects of tampons and menses on the composition and diversity of vaginal microbial communities over time’

In the study reported by Hickey et al., we wished to determine whether and how the composition and diversity of vaginal bacterial communities were influenced by both menses and tampons that differed in composition. Evaluation of differences in community composition was achieved using hierarchical clustering and nonmetric dimensional scaling. These descriptive approaches are useful for identifying and visualising overall patterns of similarity among samples, but we were also interested in determining whether changes to the microbiota were statistically significant between menstrual periods. This required two steps. First, we needed to simplify our data set from many variables (bacterial taxa) to one (community diversity) because the number of samples in our study precluded robust multivariate approaches. This was achieved by calculating the Shannon index (Shannon, CE (1948) A mathematical theory of communication. Bell System Technical Journal, 27:379-423.) for each sample, which is a quantitative measure of community diversity commonly used in ecological studies. Second, we employed a model comparison approach to identify variables contributing significantly to observed differences in diversity across menstrual periods.

Model comparison is commonly used to identify the significant effects of variables under study in an experiment. The variables (sometimes referred to as main effects) we tested included participant, catamenial product type and time. Relationships between these variables can be formalised as mathematical equations, or statistical models. Model comparison begins by identifying a complex model that includes all possible variables of interest as well as their interactions (i.e. combined, non-additive effects of multiple variables). The model is ‘fitted’ to the data to explain the variability observed, and each variable or term in the model accounts for part of the total variability. Inevitably, a portion of that variability is not explainable by the model; this indicates the stochastic nature or variability due to ‘noise’ in the data. Starting with the complex model, we followed a systematic approach of eliminating variables, starting with complex interactions and narrowing down to the main experimental effects (e.g. participant, product and time). If removal of a chosen variable adds too much variability back to the noise component, then this variable is deemed useful in explaining some of the variability and is kept in the model. If it does not, then it is deemed expendable and can be removed. This comparison between models is accomplished formally using either a chi-square test or a likelihood ratio test.

In our study we determined the best-fit model to be one that identified participant, product, time, participant–time interactions and product–participant interactions as significant explanatory variables. Pairwise comparisons of Shannon diversity indices between catamenial product treatments revealed that community diversity during pad use was significantly different from at least one tampon cycle in three women, but the trend differed among these individuals. Overall we concluded that the two tampons tested did not exert a notable impact on vaginal bacterial community composition. However, because of the small sample size of our study and the variety of community dynamic patterns observed in the data, larger studies enrolling more participants for longer periods of time are warranted.

  • RJ Hickey, Z Abdo, LJ Forney

  • University of Idaho, Department of Biological Sciences, Moscow, ID, USA

Reviewer's commentary on ‘Effects of tampons and menses on the composition and diversity of vaginal microbial communities over time’

The report from Hickey et al. in this month's issue is an important addition to the growing literature on the use of cultivation-independent methods in the study of the human vaginal microbiota. The primary objective of this study was to evaluate the effect of two catamenial products (tampons) on the vaginal microbiota throughout the menstrual cycle, using high-throughput molecular technology based on polymerase chain reaction (PCR) and pyrosequencing of barcoded 16S ribosomal RNA (rRNA) gene amplicons. Earlier reports on the same topic relied on cultivation-based methods.

Molecular gene amplification techniques offer significant advantages over culture-based methods. They can be performed directly on samples, avoiding culture biases. This is particularly important because <10% of the microorganisms can typically be cultured when studying natural microbial systems (Amann et al. Microbiol Rev 1995;59:143–69) Second, any microbe can be detected as long as its nucleic acid is intact because all bacteria possess the 16S rRNA gene, which includes both highly conserved and variable regions unique to each species. Amplicons of the latter regions are sequenced to identify the bacterial species present in a given sample.

Molecular techniques also have some drawbacks. False negatives can occur because of inhibitors present in vaginal secretions or amplicons that impede the gene amplification protocol. Furthermore, broad-range PCR primers used to amplify 16S rDNA are not universal and may occasionally overlook a whole phylum. More often, ‘conserved’ positions in the 16S rRNA gene in individual species scattered among the phyla may also be polymorphic to some extent, and amplification occurs at reduced efficiency, resulting in loss of detection of some species. Increasing the number of amplification cycles may overcome this problem but may result in detection of low-level background bacterial DNA contaminants that are present in laboratory reagents, DNA extraction kits and PCR reagents, for instance, the Taq polymerase. Furthermore, the number of 16S rDNA molecules detected by PCR may not be representative of the degree of microbe colonisation. Bias may occur if the DNA extraction method yields a greater amount of rDNA from a particular species. Bacteria have differing numbers of rDNA operons per genome (1–18). Those with a higher number of operons would tend to be over-represented by molecular techniques.

With these limitations in mind, the data generated by the molecular techniques employed in this study fulfil predictions from the culture-based literature. However, a direct comparison between this study and previous studies that relied on culture-based methods is not possible for several reasons. First, Lactobacillus iners and Atopobium vaginae are not readily detectable by culture. Second, comparative studies suggest that a number of species such as Mobiluncus, Mycoplasma and Provatella are under-represented in vaginal flora from healthy women or those with bacterial vaginosis when using PCR with universal primers compared with culture (Kalra et al. Curr Infect Dis Rep 2007;9:485–500). Third, the culture-independent methods employed here, unlike quantitative culture techniques, measure proportions of individual populations, irrespective of individual variations in the total bacterial load.

Although confined to a small number of women, this study achieves its objectives of demonstrating dynamic changes in the vaginal microbiota throughout the menstrual cycles as well as the effect of catamenial products on such changes. As the investigators pointed out, larger studies are needed to allow generalisability of their conclusions. The era of culture-independent methods continues to greatly improve our understanding of environmental ecosystems and microbial diversity, as applied to the vagina as well as other body sites.

Disclosure of interests

There are no conflicts of interests that I should disclose.

  • M Genc

  • Department of Obstetrics and Gynecology, New York Presbyterian Hospital—Weill Cornell Medical Center, New York, NY, USA