Lactic acid gel versus metronidazole for recurrent bacterial vaginosis in women aged 16 years and over: the VITA RCT

Primary care (general practices)

1. Opportunistic identification of women presenting with BV in general practices that were VITA trial recruiting centres with trained research staff on site. Participants were identified, consented, randomised and prescribed study treatment at the practice. These research-ready sites required on-site availability of trained research nurses and facilities to directly consent and randomise patients.

2. Opportunistic identification and referral of women with BV attending general practices without on-site research staff (participant identification centres) to local participating VITA trial recruiting centres for invitation to participate in the trial.

3. Pre-identification of women with a history of BV by general practitioners (GPs) from electronic patient records/primary care databases. GPs would provide potential participants with information on the trial via telephone or letter and invite them to attend a local recruiting centre for consent if they developed BV and were interested in participating.

In addition, GP practices could use computerised ‘pop-up’ alerts to support the identification and recruitment of suitable women when they presented with possible BV symptoms.

Secondary care

1. Opportunistic identification of women presenting with BV in sexual health centres that were VITA trial recruiting centres with trained research staff on site. Participants were identified, consented, randomised and dispensed study treatment at the centre. These research-ready sites required on-site availability of trained research nurses and facilities to directly consent and randomise patients.

2. Opportunistic identification of women presenting with BV in gynaecology clinics that either were VITA trial recruiting centres with trained research staff on site where participants would be identified, consented, randomised and prescribed study treatment within the clinic; or acted as VITA trial referral clinics (participant identification centres) where women presenting with BV could be referred to a nearby participating recruiting sexual health centre for invitation to participate in the trial.

Inclusion criteria

Individuals had to meet all of the following inclusion criteria to be included in the trial:

1. aged ≥ 16 years

2. clinical diagnosis of BV based on patient-reported symptoms of discharge with an unpleasant (typically fishy) odour, with or without positive microscopy according to local site practice

3. history of at least one previous episode of BV in the past 2 years (clinically diagnosed or patient reported) that had been resolved with treatment

4. willing to use either intravaginal lactic acid gel or oral metronidazole tablets for the management of BV

5. willing to take their own vaginal samples

6. willing to avoid vaginal douching during treatment

7. willing to provide contact details and be contacted for the purpose of collecting follow-up information

8. willing to avoid sexual intercourse or use effective contraception for the 7-day duration of study treatment (condoms were not considered to be effective contraception owing to a potential interaction with the lactic acid gel)

9. access to the internet and e-mail and willing to complete web-based follow-up questionnaires in English

10. written informed consent.

Exclusion criteria

Individuals were excluded from the trial if they met any of the following exclusion criteria:

1. contraindications or allergy to lactic acid gel or metronidazole tablets

2. pregnant or breastfeeding

3. patients currently trying to conceive and not willing to avoid sexual intercourse or use effective contraception for the 7-day duration of study treatment

4. using oral antibiotics (other than the study treatment) or antifungal agents concurrently, within the last 2 weeks or planned use within the next 2 weeks

5. using topical vaginal antibiotics, antifungals or acidifying products (other than the study treatment) concurrently, within the last 2 weeks or planned use within the next 2 weeks

6. previous participation in this study

7. current participation in another trial involving an investigational medicinal product (IMP).

Contraindications and concomitant medications

Screening

Women either pre-identified by or presenting to referring or recruiting general practices, sexual health centres or gynaecology clinics who had symptoms of BV (and a history of one or more episodes within the previous 2 years that resolved following treatment) were invited (by telephone or letter) or approached by a member of the site research team to determine whether or not they were interested in participating in the trial. If they were interested, they were given a participant information sheet.

Women who were identified in general practices and gynaecology clinics that were acting as referral centres were introduced to the trial and directed to a local recruiting practice, sexual health centre or gynaecology clinic for consent if they were interested in participating in the trial.

A screening log was maintained at each recruiting site detailing all patients approached about the study, the number of patients agreeing to participate, the reasons for not participating and, where relevant, the route of referral.

Consent

Women were given time to read the participant information sheet and had the opportunity to ask the site research team any questions about the trial prior to consent. Written informed consent was requested during the same clinic visit by the principal investigator (PI) or the delegated study doctor or nurse prior to performing any trial-related procedure. A copy of the completed informed consent form was given to the participant, a copy was filed in the medical notes and the original copy was placed in the investigator site file.

The consent process included optional consent to be approached by researchers from the University of Warwick for a qualitative telephone interview. When participants who had given this optional consent were contacted to arrange an interview, verbal consent that they remained willing to take part was recorded by the researcher before the interview began.

Participants recruited from sexual health centres and gynaecology clinics were also asked for their optional consent to inform their GP that they were taking part in the trial.

Analysis of outcome measures

A full statistical analysis plan was developed and agreed to prior to database lock and unblinding of the analysing statistician, and all analyses were carried out using Stata^® version 15.1 (StataCorp LP, College Station, TX, USA). Continuous variables were summarised in terms of the mean, standard deviation (SD), median, lower and upper quartiles, minimum, maximum and number of observations. Categorical variables were summarised in terms of frequency counts and percentages. Descriptive statistics of demographic and clinical measures were used to assess the balance between the treatment arms at baseline, but no formal statistical comparisons were made.

The primary approach to between-arm comparative analyses was by modified intention to treat, that is it included all participants who were randomised and without imputation of missing outcome data. Sensitivity analyses were conducted to investigate the impact of missing data, additional baseline variables and adherence to allocated treatment.

Evaluation of the primary outcome was performed using a generalised estimating equation for the binary outcome that included factors used in the minimisation, with site as the panel variable. This was changed prior to database lock (and before finalising the statistical analysis plan and unblinding the trial statistician) from the originally planned mixed-effects model owing to model non-convergence because some sites had very small numbers of participants. It had been planned to also include whether or not vaginal douching had occurred in the 3 months prior to randomisation, but model convergence issues resulted in this being excluded from the model. The comparison of lactic acid gel with oral metronidazole was presented using the risk difference of the proportion of participants who reported resolution of symptoms at week 2, along with the 95% confidence interval (CI). The adjusted risk ratio and 95% CI were also presented. Owing to non-convergence of several of the models, sensitivity analyses also used a random-effects logit model including the same factors, presenting the odds ratio and 95% CI.

Secondary outcomes were analysed using appropriate regression models dependent on data type (e.g. binary, continuous, count and survival), and included factors used in the minimisation and baseline value of the outcome where measured. The analyses of secondary outcomes were considered supportive to the primary outcomes, and estimates and p-values, where presented, were interpreted in this light.

Presentations of quantitative tolerability data were descriptive. Frequency counts and percentages of the proportion of participants reporting nausea, vomiting, taste disturbance, vaginal irritation, abdominal pain and diarrhoea were presented by treatment arms.

Planned subgroup analyses

The primary analysis for symptom resolution was investigated to determine whether or not treatment effectiveness differed according to the following subgroups:

• presence of concomitant STI (yes/no)

• BV confirmed by positive microscopy (yes/no)

• number of episodes of BV in the 12 months before baseline (0, 1–3 and > 3)

• total time with BV in the 12 months before baseline (< 2 weeks, ≥ 2 weeks and < 3 months, ≥ 3 months).

A further subgroup analysis was planned to determine whether or not treatment effectiveness differed according to the type of centre at which the participant presented (sexual health vs. GP/other clinics); however, this was not performed given that no gynaecology clinics and only one GP practice took part in the trial. Between-group treatment effects were provided for each subgroup, but interpretation of any subgroup effects was based on the treatment–subgroup interaction and 95% CI, estimated by fitting an appropriate interaction term in the regression models. Given that the trial was powered to detect overall differences between the groups rather than interactions of this kind, these subgroup analyses were regarded as exploratory. An interaction term could not be fitted when investigating the subgroup for the presence of concomitant STI at baseline owing to the small number of participants with a STI.

Feasibility

There was no planned interim analysis of treatment efficacy. However, an assessment of recruitment and adherence to treatment was performed using data from the first 6 months of participant recruitment. This was to determine how feasible it was that the trial would be able to adequately address its primary and secondary objectives.

The TSC and Data Monitoring Committee (DMC) used the following criteria as a guide to determine whether or not the trial should progress:

• Review of the number of participants completing their week 2 assessment against the following targets –

  • > 90% – continue the trial

  • 65–90% – review recruitment and retention procedures to identify underlying problems and put in place strategies to address these, with review in 6 months

  • 35–65% – review recruitment and retention procedures to identify underlying problems and put in place strategies to address these. Ongoing review over 6 months and terminate the trial if the recruitment trajectory does not indicate that full recruitment can occur within an acceptable recruitment period

  • < 35% – terminate the trial.

• Review of adherence to lactic acid gel and metronidazole against the following predefined targets –

  • median adherence 5–7 days per week – continue the trial

  • median adherence 3–4 days per week – review data from the qualitative interviews on adherence and tolerability to identify underlying problems and put in place strategies to address these, with review in 6 months

  • median adherence < 3 days per week – terminate the trial.

Power calculation/sample size calculation

Assuming that 80% of participants receiving oral metronidazole would achieve resolution of symptoms,24,30–32 1710 participants (855 in each treatment arm) were required for analysis to detect a 6% increase in response rate to 86% in those receiving lactic acid gel (risk ratio 1.08) at the 5% significance level (two-sided) with 90% power. To allow for loss to follow-up of 10% (i.e. non-collection of the primary outcome), a total of 1900 participants were required to be recruited.

Questionnaires and telephone calls

The completion of data via web-based questionnaires and telephone calls was similar between the two treatment arms (see Appendix 5, Tables 33 and 34). The web-based questionnaire response rates were 318 out of 515 (62%) at week 2, 219 out of 512 (43%) at 3 months and 176 out of 512 (34%) at 6 months (see Appendix 5, Table 33). Further key data were obtained by telephone, with week 2 primary outcome information on BV resolution provided by an additional 88 out of the 202 participants for whom contact was attempted (success rate of 44%); secondary outcomes on recurrence were given by an additional 29 out of the 105 participants for whom contact was attempted at 6 months (success rate of 28%).

The median (minimum–maximum) time from randomisation to return of the week 2 questionnaires was 15 (14–55) days, with most of the questionnaires being returned within 28 days. For week 2 telephone data, the median time for obtaining these was 55.5 (29–155) days. This latter time was much longer than that for the questionnaires because the decision to collect data by telephone was made part-way through the trial to try to improve response rates. In addition, participants were given up to 28 days after randomisation to return questionnaires before a telephone call was attempted. The median time to return of the 3- and 6-month questionnaires was 93 (89–113) days and 186 (181–208) days, respectively. Six-month telephone data were collected after a median (minimum–maximum) time of 231 (203–265) days.

Week 2 vaginal samples

Week 2 vaginal samples were received by the central laboratory from 301 (58%) participants and numbers were similar between the treatment arms (see Appendix 5, Table 35). Overall, 280 out of 515 (54%) participants reported a primary outcome at week 2 and returned their week 2 samples, with similar numbers in both treatment arms (see Appendix 5, Table 35).

Nucleic acid amplification test analysis for sexually transmitted infections

Kits were provided for participants to take vaginal swab samples at baseline and week 2 for NAAT analysis of STIs. At each time point, one swab sample was placed into a tube containing a transport fluid and the tube was shipped directly to the central laboratory that was responsible for performing the analysis. On 16 January 2019, an issue was identified at three sites that some kits had been given to participants that contained time-expired NAAT sample tubes. The manufacturer of the tubes confirmed that there were no stability data past the point of expiry; therefore, it was decided that any samples received in expired tubes would be considered as void. An investigation took place into the expiry dates of existing stock at all sites and it became apparent that use of expired sample tubes was a wider issue.

Given that the expiry dates of sample tubes were not recorded by the manufacturer, site or NCTU, an audit was conducted at the central laboratory on 4 June 2019 of all NAAT samples received to determine whether or not sample tubes had expired prior to the date that each sample was taken. This audit revealed that the central laboratory had discarded some residual samples in error; therefore, for some participants it was not possible to determine whether or not the sample tubes had expired before use. For those baseline NAAT samples that were available at the central laboratory and for which expiry dates could be checked, a total of 339 samples were confirmed as being valid and within the expiry date, and a further 32 samples were confirmed as being void (i.e. the tube had expired prior to use) (see Appendix 5, Table 36). Expiry status could not be confirmed for 138 baseline samples because, although the central laboratory records indicated that these had been received, the residual samples had since been discarded. The remaining nine baseline samples either were not received by the central laboratory according to their records (n = 7) or were received but without accurate identifiable information (n = 2). For those week 2 NAAT samples that were available at the central laboratory and for which expiry dates could be checked, a total of 224 samples were confirmed as being valid and within the expiry date, and a further 19 samples were confirmed as being void (i.e. the tube had expired prior to use) (see Appendix 5, Table 36). Expiry status could not be confirmed for 58 week 2 samples because, although the central laboratory records indicated that these had been received, the residual samples had since been discarded. The remaining 214 week 2 samples either were not received by the central laboratory according to their records (n = 213) or were received but without accurate identifiable information (n = 1).

Sites followed their usual guidelines for treatment of suspected STIs during the participant’s baseline visit and clinical care of participants was not dependent on these NAAT results, which were taken purely for trial purposes.

Time to first recurrence of bacterial vaginosis

Among those participants who reported symptom resolution by week 2 (metronidazole arm, n = 143; lactic acid gel arm, n = 97), data on recurrence at 3 months were available for 73 (51%) and 50 (52%) participants, respectively, and at 6 months for 72 (50%) and 46 (47%) participants, respectively (Table 9). Of these participants, 37 out of 73 (51%) in the metronidazole arm and 23 out of 50 (46%) in the lactic acid gel arm reported recurrence by 3 months, whereas 51 out of 72 (71%) and 32 out of 46 (70%) reported recurrence by 6 months, respectively (see Table 9). The median [standard error (SE)] times to recurrence, allowing for censored times (when there was no reported recurrence up to 6 months, the 6-month data were used to calculate the censored time; if there was no reported recurrence up to 3 months and 6-month data were missing, the 3-month data were used; times were not available for recurrences reported by telephone at 6 months), were 92 (34.6) days in the metronidazole arm and 124 (SE not calculable) days in the lactic acid gel arm. It was possible to calculate only the lower limits of the 95% CIs (as there were not enough uncensored ‘events’) and these were 71 and 74 days, respectively (see Table 9). Including only those who had a time to recurrence gave median times of 54 (n = 43) days and 66 (n = 25) days in the metronidazole and the lactic acid gel arms, respectively. Times to first recurrence were not compared between treatment arms using statistical tests given that they comprised only those who had symptom resolution within 2 weeks and any comparison would, therefore, not be between randomised arms. A post hoc investigation showed that baseline characteristics were similar between treatment arms for the subset of participants resolving by week 2 (see Appendix 5, Table 38).

A Kaplan–Meier plot of recurrence against time is shown in Figure 5.

FIGURE 5.

Kaplan–Meier plot of recurrence of BV symptoms against time.

There were few participants (n = 13) who resolved, had time to recurrence data and reported that additional treatment was taken during the first 2 weeks; therefore, the difference between those with and those without additional treatment could not be assessed (see Appendix 5, Table 39).

Time to recurrence was censored at 6 months if data were available at both 3 months and 6 months, but no recurrence had been reported. Of the times used in the analysis, 30 (41%) in the metronidazole arm and 25 (50%) in the lactic acid gel arm were censored times.

Including only those with data at both 3 months and 6 months gave the percentage recurring at 6 months as 60% (32/53) in the metronidazole arm and 56% (18/32) in the lactic acid gel arm. This removes inflation of the recurrence rate resulting from needing an episode at only one time for recurrence, but data at both times to record no recurrence.

Number of participant-reported bacterial vaginosis episodes over 6 months

The number of participants whose symptoms resolved by week 2 and who had complete episode data available at both 3 months and 6 months was small (metronidazole arm, n = 48; lactic acid gel arm, n = 29). There was little difference between treatment arms in the number of subsequent episodes of BV between week 2 and 6 months for those who had resolved by week 2. Both treatment arms had a median of one episode over the 6-month period, with a maximum of six episodes in the metronidazole arm and 10 episodes in the lactic acid gel arm (Table 10). The adjusted incidence rate ratio was 0.97 (95% CI 0.56 to 1.69). The analysis was adjusted for site, number of BV episodes in the 12 months before baseline and female partner in the 12 months before baseline (yes/no). Although vaginal douching could be included in the negative binomial model, it made little difference to the estimates (incidence rate ratio 1.00, 95% CI 0.58 to 1.72) and it was more consistent with the other analyses not to include this variable. The incidence rate is defined as the number of episodes per time at risk, that is if both treatment arms were followed up for the same length of time (in this case 6 months), the ratio of episodes in the metronidazole arm compared with the lactic acid gel arm was estimated as 0.97.

Number of participant-reported bacterial vaginosis treatment courses over 6 months

Data on the number of BV treatment courses over the 6 months of the trial (excluding the study treatment) for those who had resolved by week 2 were available for 59 out of 143 (41%) participants in the metronidazole arm and 35 out of 97 (36%) participants in the lactic acid gel arm (see Appendix 5, Table 40). For those resolving by week 2, the median number of BV treatment courses received between week 2 and 6 months was similar between the treatment arms (metronidazole arm, median = 1; lactic acid gel arm, median = 1), with an adjusted incidence rate of 1.03 (95% CI 0.53 to 2.01) (see Appendix 5, Tables 40 and 41), and are explored further in the health economic analysis (see Chapter 5).

Microbiological resolution of bacterial vaginosis on microscopy of vaginal smears at week 2

The number of participants with central laboratory microbiological confirmation of BV (Ison–Hay grade 3) on vaginal smears taken at baseline was 138 participants in the metronidazole arm and 128 participants in the lactic acid gel arm (Table 11). Of those participants, 77 (56%) in the metronidazole arm and 73 (57%) in the lactic acid gel arm had central laboratory BV results available at week 2. Microbiological resolution of BV at week 2 in those with confirmed BV at baseline was higher in the metronidazole arm (59/77 participants; 77%) than in the lactic acid gel arm (31/73 participants; 42%), with an adjusted risk difference of –34.3% (95% CI –49.1% to –19.5%) (see Table 11). Microbiological resolution was defined as having Ison–Hay grade 3 at baseline followed by Ison–Hay grades 0, 1, 2 or U at week 2.

Of those participants with microbiological confirmation of BV at baseline, resolution of symptoms occurred in 83 out of 111 (75%) participants in the metronidazole arm and 49 out of 101 (49%) participants in the lactic acid gel arm.

When rates of resolution were compared between microbiological results and participant-reported symptomatic results, they were similar (60% indicating microbiological resolution and 61% indicating symptomatic resolution) (see Appendix 5, Table 42). However, these results were not consistent in about one-third of cases with data; that is, 25 out of 145 (17%) samples did not indicate microbiological resolution while the participant did report symptomatic resolution, and a further 24 out of 145 (17%) samples showed microbiological resolution without symptomatic resolution.

Time to participant-reported resolution of bacterial vaginosis symptoms

The median time to participant-reported resolution of BV symptoms (overall, not just up to week 2) was 14 days in both treatment arms (see Appendix 5, Table 43), with missing times in the first 2 weeks being substituted by 14 days [40/142 (28%) and 22/97 (23%) of times were substituted in the metronidazole and lactic acid gel arms, respectively], and times for those not recording resolution censored at the latest time without resolution (adjusted difference 0, 95% CI –1.9 to 1.9; 134 values were censored). Without censoring, the median times were 7 days in the metronidazole arm and 8.5 days in the lactic acid gel arm (Figure 6; see Appendix 5, Table 43), and without censoring or substitution median times were 5 days in both arms.

FIGURE 6.

Kaplan–Meier plot of time to resolution of BV symptoms at any time up to 6 months.

The majority of participants did not report using additional treatment, but, of those who did, median times were slightly longer than for those without (Figure 7; see Appendix 5, Table 43).

FIGURE 7.

Kaplan–Meier plot of time to resolution of BV symptoms at any time up to 6 months, split by additional treatment. Lines plotted are for oral metronidazole and intravaginal lactic acid gel either with or without additional treatment.

Tolerability of study treatments assessed by participant reporting of side effects

The description of the tolerability data included 516 participants summarised according to the study treatment that they received: 258 in each treatment arm. One participant in each arm did not receive any study treatment [one in the metronidazole arm because they received clotrimazole (Canesten^®, Bayer) for thrush diagnosed post randomisation and did not provide any safety data; one withdrew from the lactic acid gel arm prior to receiving any study treatment] (see Figure 3). Of the 516 participants included in the tolerability analyses, two in each arm received the other study treatment from that allocated (see Figure 3). In addition, of the 258 participants in the metronidazole arm who were included in the tolerability analyses, two withdrew (one of whom was allocated lactic acid gel), whereas no participants withdrew of the 258 participants in the lactic acid gel arm who were included in the tolerability analyses.

In those participants returning week 2 questionnaires, there was a higher reported incidence of side effects in the metronidazole arm than in the lactic acid gel arm, particularly of nausea (32% vs. 8%, respectively), taste changes (18% vs. 1%, respectively) and diarrhoea (20% vs. 6%, respectively) (Table 12).

Further details of these side effects are available (see Appendix 5, Tables 44–51). The proportion of participants with vaginal irritation at week 2 was lower in the group who did not have irritation at baseline than in the group who did (see Appendix 5, Tables 48 and 49).

Participant-reported adherence to treatment

Of the 318 participants who returned a week 2 questionnaire, 316 (99%) reported that they took at least some of their study treatment (Table 13). A total of 294 (92%) participants took at least 85% of their course: 146 out of 157 (93%) participants in the metronidazole arm compared with 148 out of 161 (92%) participants in the lactic acid gel arm (see Table 13). For 113 out of 157 (72%) participants in the metronidazole arm, the antibiotic was easy or very easy to take, whereas 142 out of 161 (88%) participants in the lactic acid gel arm reported that their treatment was easy or very easy to use. The brands of lactic acid gel that were used are reported in Appendix 5, Table 53.

Prevalence of concurrent sexually transmitted infections at baseline and week 2

Prevalence of concurrent STIs at both baseline and week 2 was very low (Table 14; see Appendix 5, Tables 54–56). Out of the samples from unexpired kits (both known and uncertain expiry), there was none positive for gonorrhoea at baseline or week 2, between 1% and 5% positive for chlamydia, and 0–2% positive for trichomoniasis.

Out of the 339 baseline samples that were confirmed as being valid, none had gonorrhoea, 10 (3%) had chlamydia and five (1%) had trichomoniasis (see Appendix 5, Tables 54–56). Out of the 224 week 2 samples confirmed as being valid, none had gonorrhoea, five (2%) had chlamydia and one (< 1%) had trichomoniasis (see Appendix 5, Tables 54–56).

Overall, there were 12 participants in the metronidazole arm and three in the lactic acid gel arm with a diagnosed STI at baseline. By week 2, some participants had acquired a STI, whereas in others the infection had resolved (Table 15).

An analysis of the prevalence of STIs at week 2, adjusting for baseline STIs and site, showed little difference between the treatment arms, with an odds ratio of 0.15 (95% CI 0.01 to 1.60) (Table 16).

Time with bacterial vaginosis recurrence after first resolution of symptoms

The total time with BV recurrence symptoms (in weeks) if they resolved within 2 weeks was reported by a subset of participants. Over the 6-month trial period, this was similar between the two treatment arms (see Appendix 5, Table 57). At 3 months, 37 out of 143 (26%) participants in the metronidazole arm reported recurrence compared with 23 out of 97 (24%) participants in the lactic acid gel arm. At 6 months, 51 out of 143 (36%) participants in the metronidazole arm reported recurrence compared with 32 out of 97 (33%) participants in the lactic acid gel arm.

Status of symptoms for those without resolution at week 2

The status of symptoms for those without resolution at week 2 was ‘better but not cleared/disappeared’ or ‘improved initially but worsened again’ in more than half of the participants and similar in both treatment arms (see Appendix 5, Table 58). In total, 39 out of 60 (65%) participants in the metronidazole arm reported ‘better’ or ‘improved’ compared with 64 out of 106 (60%) participants in the lactic acid gel arm.

Symptoms at week 2

Around one-third of participants reported the presence of typical BV symptoms at week 2, including genital discharge, offensive vaginal smell and vaginal irritation (see Appendix 5, Table 59). Slightly fewer participants (33%) in the metronidazole arm than in the lactic acid gel arm (41%) reported genital discharge at week 2. This was similar for those having an offensive vaginal smell (27% in the metronidazole arm vs. 40% in the lactic acid gel arm). Fewer participants reported vaginal irritation at week 2 in the lactic acid gel arm than in the metronidazole arm (23% vs. 30%, respectively).

Symptoms at recurrence compared with typical symptoms

Where reported, recurrence symptoms were mostly typical of usual BV symptoms in both treatment arms (see Appendix 5, Table 60). At 3 months, 28 out of 37 (76%) participants in the metronidazole arm reported recurrence symptoms that were ‘always’ typical of usual symptoms, compared with 15 out of 23 (65%) participants in the lactic acid gel arm. The corresponding results at 6 months were 22 out of 51 (43%) participants in the metronidazole arm and 13 out of 32 (41%) participants in the lactic acid gel arm. There was only one report of recurrence symptoms seldom being typical of usual symptoms, and this was in the lactic acid gel arm.

Additional medication for bacterial vaginosis

Nearly half of all participants in both treatment arms reported taking additional medication for BV after the first 2 weeks. Metronidazole tablets and lactic acid gel were the most frequently used additional medication, with similar proportions of each being taken in both treatment arms (see Appendix 5, Tables 61 and 62). A total of 52 out of 111 (47%) participants in the metronidazole arm and 51 out of 108 (47%) participants in the lactic acid gel arm took additional medication within 3 months, whereas 44 out of 92 (48%) and 39 out of 84 (46%) participants, respectively, took additional medication within 6 months.

Antibiotics for other conditions/illness

Around 10% of participants in both treatment arms took antibiotics for other conditions or illness in each reporting time period (see Appendix 5, Tables 63 and 64). In total, 15 out of 157 (10%) participants in the metronidazole arm and 21 out of 161 (13%) participants in the lactic acid gel arm took antibiotics for other conditions within the first 2 weeks, 8 out of 111 (7%) and 12 out of 108 (11%), respectively, took antibiotics within 3 months, and 9 out of 92 (10%) and 10 out of 84 (12%), respectively, took antibiotics within 6 months.

Vaginal thrush post randomisation

In the first 2 weeks post randomisation, of those returning a week 2 questionnaire, 42 out of 157 (27%) participants in the metronidazole arm and 27 out of 161 (17%) participants in the lactic acid gel arm reported developing thrush, for which around half of the participants took treatment (see Appendix 5, Table 65). Between week 2 and 3 months, of those returning a 3-month questionnaire, 20 out of 111 (18%) participants in the metronidazole arm and 26 out of 108 (24%) participants in the lactic acid gel arm developed thrush. Of those returning a 6-month questionnaire, 23 out of 92 (25%) and 20 out of 84 (24%) participants developed thrush between 3 and 6 months, respectively, in the two arms. The overall incidence over the 6 months was 60 out of 259 (23%) participants in the metronidazole arm and 58 out of 259 (22%) participants in the lactic acid gel arm (where no incidence over 6 months meant that ‘no’ was reported on all three questionnaires).

Sexual contact post baseline

Having sex, use of condoms and new sexual partners were similar between the two arms (see Appendix 5, Table 66). Nearly half of all participants reported having sex between the start of study treatment and the week 2 assessment in both treatment arms [71/157 (45%) in the metronidazole arm vs. 67/161 (42%) in the lactic acid gel arm]. Around three-quarters of participants did not use condoms [52/71 (73%) in the metronidazole arm vs. 48/67 (72%) in the lactic acid gel arm].

Vaginal douching post randomisation

Few participants (< 9%) reported vaginal douching during the trial, with similar percentages in each treatment arm (see Appendix 5, Table 67). At week 2, 6 out of 157 (4%) participants in the metronidazole arm reported that vaginal douching had been performed, compared with 6 out of 161 (4%) participants in the lactic acid gel arm. At 3 months, 6 out of 111 (5%) compared with 8 out of 108 (7%) participants, respectively, reported vaginal douching. At 6 months, 6 out of 92 (7%) compared with 7 out of 84 (8%), respectively, reported vaginal douching.

Participant-reported sexually transmitted infections diagnosed from 2 weeks post baseline

The incidence of participant-reported STIs diagnosed from 2 weeks post baseline was low (see Appendix 5, Table 68). In the metronidazole arm, there were five episodes of gonorrhoea, three of chlamydia, four of trichomoniasis and three of PID. In the lactic acid gel arm, there was one episode of gonorrhoea, four of chlamydia, zero of trichomoniasis and five of PID.

Other reasons reported by participants for not returning a week 2 sample

Other reasons reported by participants for not returning a week 2 sample are given in Appendix 5, Table 69. These are summarised when a sample was not received by the laboratory; often the questionnaire was completed before taking the sample. There were eight in total, five of which were because of the participant having their period (three in the metronidazole arm and two in the lactic acid gel arm).

Symptoms over the 6 months

A post hoc investigation looked at the number of participants whose symptoms resolved and remained so for 6 months compared with those whose symptoms did not resolve or who had symptoms at some point in the 6 months (see Appendix 5, Table 70). Of those with data reported at all three time points over the 6 months, 21 out of 91 (23%) participants in the metronidazole arm and 14 out of 88 (16%) participants in the lactic acid gel arm resolved and remained so for the 6 months.

Additional safety data

There were no participant-reported hospitalisations in the 2 weeks from randomisation, either for BV or for treatment-related side effects, and no SAEs or suspected unexpected serious adverse reactions were reported. There were two known pregnancies, both in the lactic acid gel arm. The estimated conception dates were 20 days pre randomisation (participant chose to have an induced abortion) and 63 days after randomisation (outcome unknown as the site was unable to make contact with the participant).

Health outcomes

The cost-effectiveness analysis reflected the primary outcome of the trial, which was participant-reported resolution of BV symptoms at week 2. The cost–utility analysis assessed health-related quality of life (HRQoL) using the SF-12 health survey (version 2), which was completed by participants at baseline, week 2, 3 months and 6 months. The SF-12 was selected as a measure of HRQoL because previous research had suggested that the derived utility scores [Short Form questionnaire-6 Dimensions (SF-6D)] were more efficient at detecting differences in self-reported health status in a related area (maternal health). 37 An algorithm developed by Brazier and Roberts38 was used to predict the SF-6D preference-based score from the SF-12 health survey. The values were obtained from a survey of the UK population to derive a utility-based algorithm. The SF-6D is a preference-based measure that is composed of a six-dimensional health state classification and uses information from the SF-12 health survey. 38 The SF-6D values take into account societal preferences for health states and, therefore, can be used to obtain quality-adjusted life-years (QALYs). Following the trapezium rule, these data were used to calculate QALY gain at 6 months. 39

Resource use and costs

For the economic evaluation, the perspective that was adopted was that of the NHS (health-care perspective), meaning that only direct health-care costs were considered. Self-reported web-based questionnaires were distributed to all participants at each follow-up point to collect resource use data in primary and secondary care settings, such as treatment use, GP visits, clinic visits, additional medication for BV and thrush, and other health-care resource use to estimate the costs associated with the administration of both treatments. A separate analysis was conducted to consider any ‘out-of-pocket’ expenses incurred by participants, including the cost of non-prescribed medications as part of the sensitivity analysis.

The following assumptions were made in relation to resource use:

• Where participants reported that a health resource was utilised without specifying the number of health-care contacts, a minimum number of contacts was assumed. For example, if a GP visit was reported without further data, it was assumed that there had been one GP visit. Although it was planned to use multiple imputation to account for missing data, this was not possible owing to the extent of missing data. 40

• If thrush was reported and treatment obtained, it was assumed that the participant had obtained one course of the treatment specified.

Analysis

Questionnaire response rate

As reported in Chapter 3, 518 participants were randomised into the trial, of whom 294 (57%) completed the resource use questionnaire at week 2: 143 in the metronidazole arm and 151 in the lactic acid gel arm. A total of 123 (24%) participants completed the resource use questionnaire at all time points (week 2, 3 months and 6 months) and were included in the base-case analysis at 6 months. The number of participants who responded to the resource use questionnaire and SF-12 health survey at different time points is provided for each treatment arm (see Appendix 5, Table 72). The baseline characteristics for those who completed resource use questionnaires at week 2 and at all time points are available (see Appendix 5, Table 73). It had been planned to use multiple imputation to generate estimates of missing values based on the distribution of observed data;48,49 however, given that resource use questionnaire data were missing for > 40% of participants at different time points, multiple imputation could not be used and a complete-case analysis was adopted. 40

Health outcomes

The mean SF-6D scores at each time point for each treatment arm are presented in Table 17. The results indicate that the differences in the mean values between the two treatment arms were not significant. The QALY gains were slightly lower for the lactic acid gel arm than for the metronidazole arm, with a mean difference of –0.003 QALYs (95% CI –0.013 to 0.009 QALYs). However, this difference was not statistically significant.

Cost and resource use

Base-case analysis

Sensitivity analysis

The results of the deterministic sensitivity analyses in which we explored the impact of varying the assumptions and for different scenarios are presented for both the cost-effectiveness and the cost–utility analyses in Tables 24 and 25. Using alternative cost estimates for metronidazole and including the cost of non-prescribed (out-of-pocket) medications administered at different time points did not affect the overall results. Including additional costs of the management of thrush at 3 and 6 months and a further course of treatment for those who developed symptoms post study treatment did slightly increase the total mean cost for both arms, but intravaginal lactic acid was still more costly than oral metronidazole.

Considering the cost of AMR associated with the use of metronidazole led to an increased mean cost for the metronidazole arm, but the average cost per participant with resolved BV remained higher for intravaginal lactic acid gel across all estimates for the potential cost of AMR that was used.

The results of 5000 bootstrap replications plotted on the cost-effectiveness plane for the primary analysis (Figure 8) suggest that lactic acid gel is likely to be less clinically effective than metronidazole because all replicates are to the left of the origin (worse health outcome). However, there is some uncertainty around the difference in costs between the treatment arms, as replicates are distributed almost equally across the north-west quadrant (less clinically effective and greater cost) and the south-west quadrant (less clinically effective and lower cost). Accordingly, there is uncertainty around whether intravaginal lactic acid gel was more or less costly than oral metronidazole at week 2.

FIGURE 8.

Cost-effectiveness plane of the primary outcome at week 2 (oral metronidazole vs. intravaginal lactic acid gel).

For the cost–utility analysis, the cost-effectiveness plane (Figure 9) shows that the majority of the bootstrap replicates were located in the north-west quadrant, suggesting that intravaginal lactic acid gel is more costly and less clinically effective than oral metronidazole at 6 months. However, some uncertainty was apparent, as some replicates were found in the north-east quadrant (more effective and more costly) and in the south-west quadrant (less clinically effective and less costly).

FIGURE 9.

Cost-effectiveness plane of the secondary outcome at 6 months (oral metronidazole vs. intravaginal lactic acid gel).

Principal findings

The results of the primary analysis suggest that intravaginal lactic acid gel is less clinically effective and slightly more costly than oral metronidazole in resolving BV symptoms by week 2. At week 2, lactic acid gel was associated with an additional cost of £7.38 (95% CI –£20.15 to £36.38) per participant. The additional cost was mainly attributed to the utilisation of secondary care resources, mainly sexual health clinics. The total cost might have been lower if more participants from primary care (general) practices had been recruited into the trial. The cost per participant with resolved BV was £147.00 in the lactic acid gel arm, compared with £86.94 in the metronidazole arm. The cost-effectiveness plane that incorporates the uncertainty around each point estimate in the results shows that, relative to oral metronidazole, intravaginal lactic acid gel is likely to be less effective. However, there is uncertainty around the difference in costs, suggesting that intravaginal lactic acid gel could be either more or less costly than oral metronidazole.

The cost–utility analysis showed that intravaginal lactic acid gel was dominated by oral metronidazole at 6 months, although there was some uncertainty around the differences in costs and outcomes.

Strengths and limitations

The strength of the economic evaluation is that it was based on a multicentre RCT exploring the potential cost-effectiveness of intravaginal lactic acid gel compared with the current standard treatment of oral metronidazole. To our knowledge, this is the first economic evaluation conducted alongside a trial that has explored intravaginal lactic acid gel as an alternative treatment for BV. Detailed data on resource use and health outcomes were collected at different time points using web-based questionnaires that were completed by participants.

The main limitation of the analysis was the small sample size and the level of missing data on resource use and quality of life at all time points. Accordingly, multiple imputation methods could not be employed. This could have introduced bias and affected the final results. Another limitation was that the follow-up was for 6 months only, which may not capture the longer-term effects of the treatments given that the trial focused on participants with recurring BV. Some pragmatic assumptions were required to address missing data relating to resource use and thrush management, but the impact associated with these assumptions was explored as part of the sensitivity analysis.

The economic cost of AMR was explored as part of the sensitivity analysis. However, there is considerable uncertainty around the role of AMR in relation to recurrent BV50 and our exploration of the potential costs associated with AMR was, therefore, necessarily limited.

Design

This qualitative substudy was completed during recruitment into the VITA trial and the results have been published. 57 The lead qualitative researcher (JAW) worked independently of the trial team and was blinded to the post-treatment results. All participants who were included in the interview subgroup had given optional consent to be contacted about the interview at the time of randomisation to the trial. Those who gave consent and had completed study treatment were consecutively sampled from both treatment arms. An invitation to take part in the interview was e-mailed to participants, which was followed by a telephone call invitation if they did not respond to the e-mail. Interviews were scheduled for a time that was convenient to the participants and were expected to take approximately 20 minutes. It was planned to interview approximately 30 participants (15 in each arm) and recruitment continued until data saturation was reached.

Data collection and analysis

Semistructured interviews took place from January 2018 to May 2018, which was part-way through recruitment into the trial, and were conducted over the telephone using a qualitative interview schedule (see Appendix 6). Only the qualitative researcher (JAW) and the participant were involved, and consent was verbally confirmed at the start of the interview. Participants were given a unique interview code to identify all data collected from them. All interviews were audio-recorded, transcribed and anonymised, and lasted between 10 and 45 minutes. Qualitative data were managed using NVivo 11 (QSR International, Warrington, UK) and all files were stored on a secure server at the University of Warwick. Data from the interviews were not analysed or triangulated with those from the web-based questionnaires.

Data were coded thematically58 and then comparatively by two researchers, with the codes being based on interview questions and emergent themes. Word-spotting of descriptive adjectives assisted in summarising all contextual data describing BV symptoms. Data were synthesised using a framework approach by comparison with the acceptability of health interventions framework55 to reflect the extent to which women receiving a health-care intervention consider treatment to be appropriate based on their anticipated or experiential cognitive and emotional responses. We considered constructs, such as perceived effectiveness and burden, according to women’s perspectives on BV treatments.

The analysis of the data required a sound narrative of the participants’ history of BV, including previous and current treatments. This helped to ensure that the analysis reflected the participants’ voices as faithfully as possible. To achieve a prolonged and intense engagement with the totality of the data, we concurrently analysed data. This helped to ensure that questions were being asked in a way that captured participants’ descriptions of their experience of BV, possible triggers of BV, perceived effectiveness of both treatments and treatment outcomes.

The quotations included fairly represent the participants’ opinions reflecting similarities, differences and nuances of opinion, experience and perception.

Theme 1: context of bacterial vaginosis

Theme 2: experiences and perspectives around the acceptability and use of treatment for bacterial vaginosis

Theme 3: treatment preference for bacterial vaginosis

Participants described their individual preferences based on the effectiveness of treatment (what they think worked) and based on personal choice (what they would rather take given individual lifestyle factors or beliefs). Treatment preference was usually dependent on the severity of symptoms, the time interval between perceived recurrences and the advantages and disadvantages of each treatment (Table 27).

Theme 4: participants’ views on which treatment they would suggest that other women take

Participants were asked which treatment they would recommend, if any. Lactic acid gel was considered a milder option and a useful ‘top-up’-type treatment for minor cases of BV:

I would recommend to try it, because it’s better than nothing, definitely. But I’m very interested in trying the gel to see if that has a better impact. I think it’s socially better as well.

V23, aged 23 years: metronidazole tablets

Taking metronidazole was regarded as a safer option to clear up moderate to severe BV, with longer time periods between episodes:

Probably the antibiotics, because it seemed to get rid of it for longer.

V9, aged 37 years: lactic acid gel

One participant said that she would like to take both treatments to test the foolproof method and ensure that ‘it does not come back’ (V7, aged 26 years: lactic acid gel). Others felt that lactic acid gel worked well enough and that they would be happy to be on it for longer periods of time.

Limitations

Many of the participants had previous experience in taking both treatment options, and this meant that these experiences merged when recalling when and what they had taken, especially when taken back to back.

Implications for health care

Although metronidazole proved to be more effective at treating BV in the short term in the VITA trial, recurrence within 6 months in a subset of participants who had initial resolution and were available for follow-up was similar in both treatment arms. In many instances, women were likely to still want to use lactic acid gel for its soothing effect and to avoid recurrent use of antibiotics. Participants were willing to purchase lactic acid gel themselves, although some found that this was not affordable. To achieve better informed and shared decision-making about treatments, health professionals need an understanding of what women think about the use of antibiotics or intravaginal gel when weighing up the pros and cons of the available options.

Recommendations for research

The overall findings of the substudy are not straightforward to communicate given the higher treatment effectiveness of metronidazole but preference, despite this, in some women for lactic acid gel. Given that women already struggle to understand BV and its cause, research is needed to understand how best to communicate treatment recommendations.

Current treatment of bacterial vaginosis

Current international guidelines recommend the use of antibiotics as first-line treatment for BV. In practice, this is usually metronidazole (given either orally or as a topical intravaginal gel) and less often clindamycin. 10,33,64 Metronidazole has a broad spectrum of actions against anaerobic bacteria often associated with BV and a minimal effect on commensal vaginal lactobacilli. 65 In those with frequent BV recurrences, short courses of antibiotics to treat each symptomatic episode are recommended or regular antibiotic therapy over several weeks or months as prophylaxis.

Treatment of BV using these antibiotic regimens leads to cure within 2–4 weeks in 51–82% of patients,66 but BV symptoms recur in the majority (69–80%) of patients over the next 12 months. 12,67 The cause of initial treatment failure and subsequent frequent relapse is not known, but has been postulated to be because of AMR, failure to re-establish the normal lactobacilli-dominated vaginal flora,68 the development of a treatment-resistant bacterial biofilm on the vaginal mucosa69 and/or reinfection from a sexual partner. 70 The use of metronidazole is associated with drug side effects including vaginal candidiasis, nausea, vomiting, change in taste perception and diarrhoea,71,72 which can limit acceptability of and adherence to treatment. Patients also dislike taking multiple courses of antibiotics to treat BV and are concerned that they may develop antibiotic resistance. 57 Preventing and reducing AMR is a priority in many countries, and improving antibiotic stewardship through optimisation of and reduction in antibiotic use is a central theme of the UK’s 5-year plan to address this issue. 16

Antibiotic therapy and resistance

Following treatment, resistance to clindamycin commonly develops in the vaginal bacteria that are associated with BV,73 and resistance to metronidazole is also found in Gardnerella vaginalis74,75 and Atopobium vaginae,76 which are commonly associated with BV. Repeated antibiotic use fails to eradicate BV-associated bacteria in a substantial proportion of patients with recurrent symptoms, which also suggests that resistance might be important. 76,77 In addition to their effects on the vaginal microbiome, metronidazole and clindamycin frequently cause disruption of the gut microbial flora, with associated gastrointestinal side effects and a risk of progression to pseudomembranous enterocolitis. 78–80

Alternative bacterial vaginosis treatment approaches

The limited initial efficacy, frequent subsequent recurrences, risk of AMR and unpleasant side effects of antibiotics, in addition to patient preference in this group of trial participants, suggest that alternative approaches for the treatment of BV are needed.

A number of different approaches have been explored to try to improve the outcome of BV treatment. These include high-dose metronidazole,81 antibiotic combination therapy,82 extended release clindamycin,83 agents to disrupt the bacterial biofilm84 and probiotics. 85,86 At best, these have provided a modest improvement in efficacy compared with standard treatment87 and have not been added as recommended therapy into treatment guidelines.

Lactic acid is produced by commensal lactobacilli in the vagina and the resulting low vaginal pH helps prevent the anaerobic bacterial overgrowth that characterises BV. 88 The loss of lactobacilli that occurs in women with BV is associated with a rise in pH, and the therapeutic use of intravaginal lactic acid gel aims to restore the level of acidity to normal and, thus, inhibit the growth of BV-associated bacteria. Previous studies have been small and evaluated a variety of different intravaginal acid gels or pessaries for the treatment of BV. These have led to a variable response, with between 18% and 100% of patients achieving resolution of their BV. 13,18–23,89,90 The regimen used most often in previous trials was once per day application for 1 week, and increasing the frequency of dosing did not appear to improve the response rate (efficacy 23–93% for once per day compared with 18–100% for twice per day).

Efficacy of lactic acid gel for the treatment of bacterial vaginosis

In the VITA trial, which included a large sample of symptomatic participants with frequent recurrent BV, we found that lactic acid gel was less effective in clearing BV symptoms than oral metronidazole. However, a substantial proportion of participants in both arms did not respond to either treatment (30% of those receiving metronidazole and 53% of those receiving lactic acid gel). Moreover, in those who did respond, the limited data we had suggested that the subsequent risk of recurrent BV was very similar in both treatment arms.

In a post hoc analysis, we found no major differences in the demographic, behavioural or disease characteristics of those who responded to metronidazole compared with those who responded to lactic acid gel. Specifically, the proportion with ‘severe’ BV (based on the number of episodes or duration of symptoms in the past year) was the same in both arms. A greater proportion of participants in the lactic acid gel arm had positive microscopy for BV at week 2 (28% vs. 10% in the metronidazole arm). Overall treatment success over the 6 months of the trial was poor in both arms, with only 23% of those in the metronidazole arm and 16% of those in the lactic acid gel arm having symptom resolution and no recurrences over the 6 months following treatment, although these estimates are based on < 50% of the original trial population.

Metronidazole was associated with notably more side effects than lactic acid gel in the 2 weeks after taking study treatment, including one-quarter of participants who reported vaginal candidiasis (27% in the metronidazole arm vs. 17% in the lactic acid gel arm). However, adherence to both therapies was nevertheless high.

Our previously published qualitative study undertaken in a subset of VITA trial participants57 indicates that women dislike taking repeat courses of metronidazole to treat BV and many prefer lactic acid gel, despite perceiving it to be less effective. A preference for the lactic acid gel was based on a desire to apply treatment directly to the site of the BV, the gel having an immediate soothing effect, concerns that frequent use of antibiotics would lead to resistance and the ready availability of lactic acid gel without requiring a medical consultation or prescription. Avoiding the need to take antibiotic tablets twice per day and being able to drink alcohol were also reported as being benefits of using lactic acid gel, whereas greater efficacy and tablets being more convenient than using intravaginal lactic acid gel were advantages of metronidazole. Participants’ preferences for treatment were, thus, not based solely on short-term effectiveness but also related to ease of use, side effects, the individual’s value system and beliefs, and the possible long-term consequences of treatment.

Health economics

The health economic analysis found lactic acid gel to be more costly than oral metronidazole for the treatment of BV, which was mostly linked to reduced efficacy, leading to increased utilisation of secondary care. This difference persisted in a number of sensitivity analyses but there was uncertainty around the difference in costs, suggesting that intravaginal lactic acid gel could be either more or less costly than oral metronidazole. We attempted to include a ‘penalty’ cost for the risk of AMR associated with the use of metronidazole. This reduced the cost difference, but a robust methodology to measure the cost of resistance is not available and it is difficult to know how accurate our estimates of the cost of resistance were.

Safety, tolerability and adherence

Oral metronidazole is recognised to cause systemic side effects71 and participants who received it reported more gastrointestinal symptoms than those in the lactic acid gel arm, although these were generally of mild severity. Vaginal irritation occurred in 21% of those receiving lactic acid gel, which was similar to those receiving metronidazole (28%). Antibiotics are known to increase the risk of vaginal candidiasis,91 and thrush was reported to occur after treatment in 27% of participants in the metronidazole arm compared with 17% of those in the lactic acid gel arm in the 2 weeks after initiating treatment.

No SAEs or suspected unexpected serious adverse reactions were reported.

Rates of adherence to metronidazole for the treatment of BV have been reported to be only 50–68%. 92 However, in our clinical trial setting adherence was high, with > 90% of participants in both arms finishing their treatment course. Participants found that both treatments were easy to take, and the most common reason for missing a dose was forgetting to take it. However, these data were based on the 61% of participants who returned their week 2 questionnaire. It is not known whether or not those who did not return their questionnaires took their study treatment.

Clinical and microscopy diagnosis of bacterial vaginosis

Entry into the VITA trial and assessment of treatment response were based on the presence or absence of participant-reported BV symptoms to mirror ‘real life’ in most clinical primary care practice. Participants were symptomatic at the time of randomisation and had a history of at least one other episode of BV within the past 2 years that had resolved following treatment. Vaginal discharge can have multiple causes, although BV is typically associated with a characteristic fishy odour. A pre-planned analysis that was restricted to participants who had positive microscopic confirmation of BV29 at baseline showed a similar response rate to that seen in the primary analysis: resolution of BV symptoms in 75% of participants receiving metronidazole and 49% for those receiving lactic acid gel. Resolution at week 2, confirmed via microscopy, occurred in 77% of participants receiving metronidazole compared with 42% of those receiving lactic acid gel, suggesting that metronidazole may have a greater effect on the pattern of vaginal bacteria seen on microscopy than lactic acid gel. However, the clinical relevance of greater microbiological resolution of BV (compared with clinical symptom resolution) is uncertain because, following a clinical treatment response (resolution of symptoms) in either treatment arm, no differences were detected in the frequency or timing of BV recurrences over the next 6 months. Caution is, however, required in interpreting the limited data that were available over the 6-month follow-up period.

We chose resolution of BV symptoms as the primary end point to maximise the trial’s relevance to clinical practice where, in many settings, microscopy is not available and management is based on history and examination. In addition, microbiological evidence of BV can be present without symptoms93 and cyclical changes in the vaginal flora (including BV-type flora) can occur in the absence of treatment,94 which makes interpretation of microbiological cures difficult. Microscopy diagnosis of BV at week 2 was included as a secondary outcome in the trial and was based on the criteria described by Ison and Hay,29 as recommended in the UK National Guideline for the management of Bacterial Vaginosis. 10 Diagnosis using Ison and Hay criteria29 correlates with the two other commonly used approaches to microbiologic BV diagnosis: Amsel’s criteria29 and Nugent’s score. 95,96

From a patient perspective, the symptoms of discharge and malodour are the main causes of the physical and emotional distress associated with BV, have an impact on self-esteem and restrict sexual activity. 56,57 Our findings are, therefore, of direct relevance to patients and those providing treatment for BV.

Other factors that might have affected treatment efficacy

The VITA trial enrolled participants with severe recurrent BV. Women with frequent recurrent BV respond less well to treatment81 and this was evident in our trial as the response rate was lower in both treatment arms in those with multiple recent episodes of BV. However, the frequency of prior BV did not predict symptom resolution differently between the two treatment arms. In those successfully treated with either metronidazole or lactic acid gel, the same proportion (around one-third) reported having BV on more than three occasions in the past year.

Participants were advised to pause using lactic acid gel if they were menstruating heavily, reflecting usual practice, but overall adherence to therapy was high with few missed doses.