Ecological Effects of Perorally Administered Pivmecillinam on the Normal Vaginal Microflora

Abstract

The knowledge of the effects of antimicrobial agents on the normal vaginal microflora is limited. The objective of the present study was to study the ecological impact of pivmecillinam on the normal vaginal microflora. In 20 healthy women, the estimated day of ovulation was determined during three subsequent menstrual cycles. Microbiological and clinical examinations were performed on the estimated day of ovulation and on day 3 in all cycles and also on day 7 after ovulation in cycles 1 and 2. Anaerobic and facultative anaerobic gram-positive rods, mainly species of lactobacilli and actinomycetes, dominated the microflora. One woman was colonized on the third day of administration with a resistant Escherichia coli strain, and Candida albicans was detected in one woman on days 3 and 7 in cycle 2. No other major changes in the normal microflora occurred during the study. Administration of pivmecillinam had a minor ecological impact on the normal vaginal microflora.

The vaginal microflora is influenced by a number of endogenous and exogenous factors (10). In women of reproductive age, transient changes occur over the menstrual cycle, with increased rates of growth of lactobacilli and decreased rates of recovery of other gram-positive rods and Prevotella species (6, 15). Intercourse, the numbers of sexual partners, and the contraceptive methods used have also been shown to influence the normal vaginal microflora (3, 4, 9, 22). Vaginal ecological changes occur after menopause and include a reduction in the numbers of lactobacilli (2, 21). It is well known that antimicrobial agents cause ecological disturbances in the intestinal and oropharyngeal microfloras (24). However, the knowledge of the effects of antimicrobial agents on the vaginal microflora is limited. The dominant species in the vaginal flora is lactobacilli, and hydrogen peroxide-producing strains of lactobacilli have been implicated as being particularly important in controlling the microenvironment (5, 16). Disturbances of the ecological balance may result in bacterial vaginosis, vaginal candidiasis, and urinary tract infections. Pivmecillinam is a prodrug of amdinocillin and belongs to the class of beta-lactam antimicrobial agents. The antimicrobial spectrum includes in particular aerobic gram-negative rods. One indication for treatment with pivmecillinam is urinary tract infections. In studies on the ecological effects of pivmecillinam on the normal oropharyngeal, intestinal, and skin microfloras, the main impact has been decreased numbers of intestinal Escherichia coli isolates (23). The impact of pivmecillinam on the vaginal flora remains unknown. The aim of the present study was therefore to investigate the ecological impact of orally administered pivmecillinam on the vaginal normal microflora in healthy, normally menstruating women.

MATERIALS AND METHODS

Subjects.

Twenty-seven healthy, normally menstruating women volunteered for the study. A physical examination was carried out for each woman before inclusion in the study. All of them had normal blood pressure (<90/<140). Clinical laboratory tests for kidney, thyroid, and liver functions were performed. The women did not use any contraceptives and had regular menstrual cycles with mean intervals of 30 days (range, 26 to 32 days). Each of the women had only one sexual partner. None of the women had taken any antimicrobial agent within the 3 months preceding the study, and no other medication except the study drug was allowed during the investigation period. The Local Ethics Committee at Karolinska University Hospital Huddinge, Karolinska Institutet, Stockholm, Sweden, approved the study.

Prestudy screening.

A gynecological examination was performed, including a Pap smear and registration of the color and the amount of vaginal and cervical discharge. Microbiological sampling was performed for detection of pathogens, and an endocervical swab sample was taken for the diagnosis of candidiasis, the detection of Trichomonas vaginalis, and determination of the numbers of cervical white blood cells per high-power field (1).

Clinical examinations and sampling procedures.

The estimated day of ovulation was determined during three subsequent menstrual cycles by using observations of cervical secretion and vaginal ultrasonography (Aloka SSD1400; Aloka Co., Ltd., Tokyo, Japan) for identification and measurement of the diameter of the leading follicle. Ovo-sticks (Clearplan; Unipath Limited, Bedford, United Kingdom) were used for identification of peak luteinizing hormone levels in urine, which were later confirmed by determination of luteinizing hormone, estradiol, and progesterone levels in blood, when required. Vaginal and cervical swab samples for cultivation of the normal microflora were taken on the day of ovulation (day 1) in each cycle. In the first two cycles sampling was also performed on days 3 and 7, and in the third cycle sampling was performed on day 3, as described in Table 1. At each visit, an examination that included observation of the appearance of the vaginal and the cervical epithelia, inspection of discharge, measurement of vaginal pH (pH.Fix 3.6-6.1; Macherey-Nagel, Düren, Germany), and photographic imaging (Nikon CoolPix 800; Nikon Corp., Tokyo, Japan) of the cervix was carried out. Samples from the cervical canal were collected with Copan swabs (Copan Italia, Brescia, Italy), and samples from the vaginal posterior fornix were collected with preweighed cotton swabs. All samples were immediately transported to the microbiology laboratory. The cervical swab specimens were cultured directly, while the vaginal swab samples were reweighed and diluted in prereduced broth and stored at −70°C.

TABLE 1.

Study design for evaluation of ecological impact of pivmecillinam on normal vaginal microflora

Procedure	Procedure performance on the indicated daya of the following:
	Control cycle				Treatment cycle			Follow-up cycle
	0	1	3	7	1	3	7	1	3
Prestudy screening	x
Estimation of ovulation day	x	x			x			x
Clinical examination	x	x	x	x	x	x	x	x	x
pH determination		x	x	x	x	x	x	x	x
Cervical swab culture		x	x	x	x	x	x	x	x
Vaginal swab culture		x	x	x	x	x	x	x	x
Start of administration					x
End of administration							x
Antimicrobial concentration determination							x

^aDay 0, before ovulation; day 1, day of ovulation; days 3 and 7, days after the day of ovulation.

Administration of antimicrobial agent.

Pivmecillinam (Selexid; LEO Pharma, Ballerup, Denmark) was administered perorally at 200 mg three times daily for 7 days. The administration started on the day of ovulation in the second menstrual cycle.

Compliance.

The women kept diary cards with daily notes of tablet intake. The cards were checked at every visit at the clinic.

Assays of amdinocillin concentrations.

The concentrations of amdinocillin in cervical and vaginal secretions and serum were determined on day 7 of the treatment cycle by the agar diffusion method. Cervical and vaginal secretions were collected with preweighed cotton swabs and were diluted 1:3 or 1:5 in distilled water, depending on the weight of the sample. Serum samples were used undiluted. The test medium was antibiotic medium 1 (Difco, Detroit, Mich.). E. coli ATCC 25922 was used as the reference strain. Samples were run in duplicate, and on each plate a concomitant standard series was included (range, 0.125 to 64 mg/liter). The plates were incubated for 18 h at 37°C. Amdinocillin concentrations were determined in relation to the diameters of the inhibition zones caused by the known concentrations from the linear standard series.

Microbiological procedures.

The cervical swab specimens were inoculated undiluted on blood agar plates (for aerobic and anaerobic incubation) and on hematin agar. The vaginal swab specimens were serially diluted in prereduced broth to 10⁻⁶ and inoculated on the following media: blood agar (Colombia base agar II; Acumedia, Baltimore, Md.) with 0.01% tryptophan (Merck, WWR International AB, Stockholm, Sweden) and 5% citrated horse blood as a nonselective medium for detection of aerobic and anaerobic microorganisms; hematin agar (GC-agar; Acumedia) with 1% hemoglobin and 1% IsoVitaleX as a nonselective medium for facultative anaerobic bacteria; CLED agar (LabMKemila, Bury, United Kingdom) for detection of members of the family Enterobacteriaceae; Enterococcosel agar (BBL Microbiology Systems, Cockeysville, Md.) for detection of enterococci; Sabouraud agar (Difco) for detection of yeasts; Rogosa-SL agar (Difco) for cultivation of lactobacilli; BL agar (Difco) for cultivation of bifidobacteria; kanamycin-vancomycin-blood agar (Colombia agar base II; Acumedia) for detection of Prevotella and Bacteroides spp.; neomycin-vancomycin-blood agar for cultivation of fusobacteria; human blood bilayer agar (Colombia CNA-agar; BBL) with 2 μg of Fungizone (Bristol-Myers Squibb, Bromma, Sweden) per ml for detection of Gardnerella vaginalis; and Veillonella agar (Difco) for cultivation of Veillonella cocci. The plates were incubated aerobically, aerobically with 5% CO₂ supplementation, and anaerobically. The aerobic plates were incubated for 48 h at 37°C, and the anaerobic plates were incubated for 72 h at 37°C. Anaerobic plates were incubated for an additional 48 h and checked a second time for the growth of new species. The anaerobic plates were placed in an anaerobic gas chamber (Concept 400; Maltec ApS, Karlslunde, Denmark) and in anaerobic jars (GasPak; BBL). After incubation, all different colony types were counted and isolated in pure cultures. Gram staining was performed, followed by biochemical tests for identification of the organsims to the genus level (19). The API 20 STREP test kit (Biomérieux SA, Marcy l'Etoile, France) and a test for the synthesis of extracellular dextran were used for the identification of streptococci (8). The API 20E system (Biomérieux SA) was used for the identification of members of the family Enterobacteriaceae, and the API ID32C system (Biomérieux SA) was used for the identification of fungi. Candida isolates were tested for color change on Chrom-agar (Chromagar Microbiology, Paris, France) and for growth at 43°C for differentiation between Candida albicans and Candida dubliniensis. All strains of staphylococci were tested for growth on novobiocin agar (Colombia base agar; Acumedia), with novobiocin (Sigma, Sigma-Aldrich, Stockholm, Sweden) at 1.75 μg/ml used for the detection of Staphylococcus saprophyticus. Anaerobic bacteria were identified to the genus level by gas-liquid chromatography of metabolites from glucose (19). The lower limit of detection was 10² to 10³ CFU/ml of vaginal fluid, depending on the weight of the sample.

Assays for lactobacillus H₂O₂ production.

Ten colonies of each lactobacillus morphological type on the Rogosa agar plates at each sampling occasion were tested for the production of H₂O₂. H₂O₂ production was determined by plating the lactobacilli onto MRS agar (BBL Microbiology Systems, Sparks, Md.) containing tetramethylbenzidine (Sigma, Sigma-Aldrich) and horseradish peroxidase (Sigma, Sigma-Aldrich) (18). After anaerobic incubation for 3 days at 37°C, the isolates were exposed to ambient air for ≤30 min. H₂O₂-producing colonies formed a blue pigment. The plates were used within 2 days of preparation. A vaginal strain of Lactobacillus (strain AS 1) with strong H₂O₂ production was used as a reference strain in the evaluation of the method and in the later determinations.

Identification of H₂O₂-producing reference strain and strains of C. dubliniensis.

The genomic DNA from the bacteria was prepared by use of a DNA mini kit (Qiagen, Valencia, Calif.), and the genomic DNA from the fungal cells was prepared by using a MagNA Pure system (Roche Applied Sciences, Stockholm, Sweden). A region of the 16S rRNA gene for the bacteria and a region of the 18S rRNA gene for the fungus were amplified by PCR with respective universal primers and AmpliTaq Gold DNA polymerase. Primers and free nucleotides from the PCR products were removed by using a QIAquick-spin PCR purification kit (Qiagen). The purified PCR products were processed for DNA sequencing with a Big-Dye terminator with capillary electrophoresis technology in an ABI 310 genetic analyzer (Applied Biosystems, Foster City, Calif.). Both strands of the PCR-amplified fragments were sequenced to avoid sequencing errors (7, 12). The DNA sequence was then analyzed with DNA software and by a search of sequences in databases with the BLAST algorithm for bacterial and fungal identification and typing (7, 13).

Antibiotic susceptibility tests.

The amdinocillin MICs for strains of enterobacteria were determined by the agar dilution method, according to NCCLS document M7-A6 for aerobic bacteria (20). The breakpoint values used were ≤8 mg/liter for susceptibility, 16 mg/liter for intermediate resistance, and ≥32 mg/liter resistance. The reference strain was E. coli ATCC 25922.

Statistical analyses.

Quantitative alterations in clinical and microbiological variables between equivalent days in the three menstrual cycles were compared by using the Wilcoxon signed rank test for paired samples. P values ≤0.05 were considered statistically significant and were adjusted for the multiple analyses.

RESULTS

Subjects.

Twenty of 27 women fulfilled the study entry requirements. The mean age of the women fulfilling the study entry requirements was 34 years (range, 24 to 40 years). One woman had a body mass index of 35.1; all others had body mass indices less than 26 (mean, 23.3; standard deviation, 3.8; range, 18.2 to 35.1). Eleven women had had one to three live births, and six women had had one to three abortions. Of the seven women who dropped out of the study, two never started the study, one woman did not menstruate within a reasonable time in the treatment cycle, one woman dropped out due to poor compliance, two women had to take antimicrobial agents for reasons that had no connection with the study, and one woman had an allergic reaction to pivmecillinam.

Clinical parameters.

No noticeable changes in appearance of the vaginal or the cervical epithelium or in the volume or the color of the discharge occurred during the investigation period. No statistically significant differences in vaginal pH between the cycles or between days within each cycle were found. The pH remained acidic, with a mean ± standard deviation pH of 4.4 ± 0.7 (range, 3.6 to 6.1). Three women had pHs >5.0 at each occasion.

Concentrations of amdinocillin in cervical and vaginal secretions and serum.

Amdinocillin was detected in the sera of 12 of 20 women and also in vaginal secretions in 4 women. The mean ± standard deviation concentration in serum was 0.7 ± 0.6 mg/liter (range, 0.2 to 1.9 mg/liter), and that in vaginal secretions was 2.1 ± 1.7 mg/liter (range, 0.6 to 4.6 mg/liter). One woman had detectable levels of amdinocillin in her cervical secretions as well (5.3 mg/liter).

Impact of amdinocillin on the aerobic vaginal microflora.

The effect of pivmecillinam administration on the aerobic vaginal microflora is shown in Fig. 1. Corynebacterium was the most prevalent aerobic species and was isolated from 19 women. Enterococci were isolated from 13 women: Enterococcus faecalis was isolated from 10 women, Enterococcus durans was isolated from 2 women, and Enterococcus faecium was isolated from 1 woman. Alpha-hemolytic streptococci were detected in 19 women. The majority of Streptococcus strains were identified as Streptococcus mitis or Streptococcus intermedius, while other strains from the anginosus, mitis, salivarius, and mutans groups of streptococci were identified sporadically. Coagulase-negative staphylococci were detected in the majority of the women (n = 19). Staphylococcus aureus was detected in samples from one woman, and no woman was colonized with Staphylococcus saprophyticus. Micrococci were isolated in low numbers from 18 women. Enterobacteria were identified in samples from eight women. Four women were colonized in the first cycle only, two women were colonized in the first two cycles (in cycle 2 on day 1), and one woman was colonized in all three cycles. In one woman, enterobacteria were identified in one only sample, i.e., on day 3 in the second cycle during treatment, and were present at log₁₀ 3.0 CFU/ml of vaginal fluid. The majority of strains of enterobacteria were identified as E. coli, and one woman also harbored Klebsiella pneumoniae. Candida species were detected in six women: four were colonized in all three cycles and one was colonized only in cycle 3. One woman was colonized with Candida (log₁₀ 4.3 to 4.4 CFU/ml of vaginal fluid) on days 3 and 7 during the administration of pivmecillinam. C. albicans was identified in five women, and C. dubliniensis was identified in one. Species of Haemophilus and Neisseria were each identified in four women.

FIG. 1.

Ecological impact of pivmecillinam, administered at 200 mg three times daily for 7 days during the second of three menstrual cycles, on the aerobic vaginal microflora of 20 women. The dotted lines show the median values.

Impact of amdinocillin on vaginal anaerobic microflora.

The impact of pivmecillinam on the anaerobic vaginal microflora is shown in Fig. 2. The anaerobic microflora was dominant, and the most frequently isolated species were strains of lactobacilli followed by actinomycetes. Lactobacilli were identified in the vaginal microfloras of all women and were predominant in all samples from 15 women. Species of actinomycetes and anaerobic cocci were isolated from 14 women. There was a statistically significant difference (P = 0.003) in the numbers of anaerobic cocci on day 7 between cycles 1 and 2. Gardnerella vaginalis was isolated from samples from 14 women. Other organisms identified from a minority of women were species of the genera Veillonella (n = 8), Bacteroides (n = 8), Prevotella (n = 7), Eubacterium (n = 4), Bifidobacterium (n = 4), Propionibacterium (n = 3), and Fusobacterium (n = 3). The administration of pivmecillinam did not influence the numbers of anaerobic species to any major extent.

FIG. 2.

Ecological impact of pivmecillinam, administered at 200 mg three times daily for 7 days during the second of three menstrual cycles, on the anaerobic vaginal microflora of 20 women. The dotted lines show the median values.

H₂O₂ production by lactobacilli.

There were both intra- and interindividual variations in the proportion of H₂O₂-producing strains of lactobacilli, but the actual numbers remained relatively stable during the study period (Fig. 3). Two women who were colonized with lactobacilli at all sampling occasions, and three women who were sporadically colonized had no detectable H₂O₂-producing lactobacilli. H₂O₂-producing strains were detected from three women on only one or two occasions, while the numbers of H₂O₂-producing strains from the majority of women (n = 12) did not change to any major extent.

FIG. 3.

Ecological impact of pivmecillinam, administered at 200 mg three times daily for 7 days during the second of three menstrual cycles, on the numbers of hydrogen peroxide-producing lactobacilli of 20 women. The dotted line shows the median values.

Impact of amdinocillin on cervical microflora.

The cervical microflora mirrored the vaginal microflora but was sparser both in numbers and in the diversity of species. Occasionally, new species that had not been identified in the vaginal microflora of the same woman were detected. Anaerobic cocci were detected in cervical samples from three women, and Haemophilus species were detected in cervical samples from two women. Other species detected in low numbers from one woman each were bacillus, eubacteria, bacteroides, and fusobacteria. No major changes in the cervical microflora occurred during the study.

Identification of suspected C. dubliniensis strains and H₂O₂ reference strain.

DNA sequence analysis showed that the H₂O₂-producing reference strain was a Lactobacillus sp. A fungal strain isolated from one woman was C. dubliniensis.

In vitro activities of amdinocillin against strains of enterobacteria.

The MIC at which 50% of isolates were inhibited (MIC₅₀) and the MIC₉₀ of amdinocillin for 42 strains of enterobacteria (11 strains from cervical swab samples and 31 strains from vaginal swab samples) were 0.125 and 8.0 mg/liter, respectively (range, 0.125 to 64.0 mg/liter). E. coli isolates (4 CFU) isolated from one woman on day 3 during the antimicrobial treatment were intermediate resistant (MICs = 16.0 mg/liter) or resistant (MICs ≥ 32 mg/liter) to amdinocillin.

Side effects.

On the second day of pivmecillinam administration one woman experienced itching and stinging around the mouth. The doctor medically responsible for the study subjects prescribed clemastine and referred the woman to the emergency department, where she was treated with betamethasone tablets (0.5 mg) until the symptoms disappeared. Four women reported slightly looser stools during treatment, and two women reported slightly harder stools during treatment.

DISCUSSION

Amdinocillin is an amidinopenicillin that binds specifically to penicillin-binding protein PBP 2 and is effective mainly against aerobic gram-negative rods. Only a minority of the women were colonized by enterobacteria, and administration of pivmecillinam did not influence the normal vaginal microflora to any major extent. However, one woman was colonized with intermediate or resistant strains of E. coli (MICs ≥ 16 mg/liter) in the second menstrual cycle during the administration of pivmecillinam. Since it was not possible to coordinate the intake of pivmecillinam with the sampling for amdinocillin concentrations, no connection between the levels of amdinocillin and the development of resistant strains could be identified. In a recent international survey (14), the amdinocillin resistance rates of E. coli strains from women with symptoms of uncomplicated urinary tract infection have been shown to vary from 0 to 2.2%. The prevalence of resistant strains has been stable since the introduction of pivmecillinam in Scandinavia (11).

Lactobacilli dominated the vaginal microflora of the majority of the women, even though five subjects were colonized only sporadically. The production of lactic acid and H₂O₂ has been regarded as the basis for the protective role of vaginal lactobacilli against infections, but the inhibitory effect of H₂O₂ on G. vaginalis is pH dependent (16). In this study, no clear correlations between the growth of lactobacilli, H₂O₂-producing lactobacilli, and G. vaginalis could be identified. However, in samples from the three women with pHs >5 on all sampling occasions, lactobacilli were detected in only one sample from each woman, and the majority of the samples from these women were colonized with G. vaginalis and Prevotella species.

A relationship between the prevalence of vaginal lactobacilli and Candida species due to estrogen levels has been observed earlier, and species of Candida have been shown to cocolonize with H₂O₂-producing lactobacilli (17). In the present study, Candida species were detected only in women in whom lactobacilli dominated the microflora, and the lactobacillus strains from five of the six women produced H₂O₂. In one woman, the detection of C. albicans was probably caused by the administration of pivmecillinam. However, as is known from other microbial habitats, Candida is a common member of the normal microflora. Candida colonizes individuals in low numbers, i.e., below the detection limit, and causes no infections when there is an ecological balance. In conclusion, administration of pivmecillinam had a minor ecological effect on the normal vaginal microflora.