Intravenous co-amoxiclav to prevent infection after operative vaginal delivery: the ANODE RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 13/96/07. The contractual start date was in March 2015. The draft report began editorial review in February 2019 and was accepted for publication in July 2019. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Oliver Rivero-Arias, Ursula Bowler, Edmund Juszczak, Marian Knight, Louise Linsell and Julia Sanders report receipt of funding from the National Institute for Health Research (NIHR) outside the submitted work. Edmund Juszczak reports Clinical Trials Unit infrastructure support funding received from NIHR and active membership of the Health Technology Assessment (HTA) Commissioning Board and the HTA General Board while the study was being undertaken.

Copyright statement

© Queen’s Printer and Controller of HMSO 2019. This work was produced by Knight et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2019 Queen’s Printer and Controller of HMSO

Objective

The objectives of this research were to investigate whether or not a single dose of prophylactic antibiotic following operative vaginal birth is clinically effective for preventing confirmed or presumed maternal infection and to investigate the associated impact on health-care costs.

Design

The A randomised controlled trial of prophylactic ANtibiotics to investigate the prevention of infection following Operative vaginal DElivery (ANODE) trial was a multicentre, randomised, blinded, placebo-controlled trial conducted in the UK.

Ethics approval and research governance

The ANODE trial protocol was approved by the Health Research Authority National Research Ethics Service Committee South Central – Hampshire B (study reference number 15/SC/0442).

Local approval and site-specific assessments were obtained from each NHS hospital site.

Patient and public involvement

The research question was initially prioritised by the user advisory group to the NIHR ‘Beyond maternal death’ programme, which generated the initial information about sepsis risk in association with operative vaginal birth. 9 To obtain the perspective of a group more representative of the wider maternity population, we contacted the PRIME (Public and Researchers Involvement in Maternity and Early pregnancy) group. PRIME is a patient and public involvement group that was set up in collaboration with the University of Birmingham Collaborations for Leadership in Applied Health Research and Care (CLAHRC). PRIME group members helped design the trial processes and materials, particularly assisting with designing our approach to consent. The group was represented among the co-applicant group to continue to advise throughout the trial.

Participants

Setting

The trial was conducted in 27 consultant-led obstetric units in England and Wales (see Appendix 1).

Informed consent and recruitment

Information about the trial was made widely available throughout the maternity units in the form of posters and leaflets [with QR (Quick Response) codes to the trial website]. Written information about the trial was available to all women at participating centres during their pregnancy through different routes, depending on the centre, for example at their antenatal booking visit, as part of their hand-held notes or at their 19- to 21-week scan visit.

On admission, women in labour or admitted for induction were reminded about the trial by their health-care professional and information about the trial could be provided if not previously seen. After the clinical decision for operative vaginal birth had been made, the following approaches were used by the woman’s midwife, obstetrician or anaesthetist to obtain informed consent, depending on the clinical circumstances (Figure 1):

1. If there was no time constraint (e.g. in cases of operative vaginal birth for delayed second-stage progress or if birth was already completed), the health-care professional discussed the trial with the woman and provided her with the participant information leaflet. If she was happy to join the trial, then informed written consent was obtained.

2. If there was a time or other constraint (e.g. in cases of operative vaginal birth for suspected fetal compromise, or if an approach for written consent was considered inappropriate by the health professionals in attendance), women were approached to give verbal consent. It is possible that urgent deliveries are associated with a lower standard of asepsis, and so it was particularly important that these women were able to participate in the trial. If the attending obstetrician or midwife felt that it was appropriate, the woman was provided with verbal information about the trial and asked if she was willing to participate, in principle; if she agreed, she was randomised. Verbal consent was documented by the clinician recruiting the woman and countersigned by a witness. All women enrolled under this procedure were approached before discharge by trial midwives to give full written consent for inclusion of their data in the trial and for participation in the planned follow-up.

FIGURE 1.

Consent and randomisation processes. Reproduced from Knight et al. 28 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/. The text includes minor additions and formatting changes to the original text.

Intervention

Centres were supplied with sealed sequentially numbered indistinguishable packs containing trial treatment [a single dose of intravenous co-amoxiclav (1 g of amoxicillin/200 mg of clavulanic acid)] or placebo (a single dose of intravenous sterile saline), as designated. Bottles of 1000 mg/200 mg of co-amoxiclav, in the form of sterile powder for solution, were supplied for making up as an injection reconstituted with sterile water, which was also supplied. The placebo (0.9% saline) was supplied as 20 ml single-use vials of clear liquid. Reconstitution was not required.

As co-amoxiclav, when reconstituted, has a distinct colour and odour, it was impossible to blind those preparing and checking the intervention to women’s allocation. The investigational medicinal product (IMP) (co-amoxiclav or placebo) was made up in an opaque-coloured syringe so that the woman herself remained blinded to allocation. To ensure that awareness of allocation could not influence outcomes, it was specified that the research nurse/midwife conducting telephone follow-up should not have prepared or checked the intervention. Sites were instructed to give the intervention as soon as possible after women had given birth, and no later than 6 hours after birth of the baby.

Randomisation, blinding and code-breaking

A randomisation list was generated by the Senior Trials Statistician at the National Perinatal Epidemiology Unit Clinical Trials Unit (NPEU CTU) using permuted blocks of variable size to ensure balance and unpredictability overall. Pack numbers were added by the Senior Trials Programmer at the NPEU CTU, who liaised directly with the packaging and distribution company. Sites were provided with a series of sequential packs and stocks were replenished from the global sequence as required. Women were randomised by the allocation of the next sequentially numbered indistinguishable pack once consent and eligibility were established. Pack use was recorded by the recruiting site and reviewed by NPEU CTU.

An emergency code-breaking procedure was not required. As only a single dose of co-amoxiclav was administered, there was no need to code-break if further antibiotics were required. Centres were advised that if a woman had an anaphylactic reaction, she should be treated as if she had been given the active drug.

Internal pilot

We conducted an internal pilot during the first 9 months of the trial, when 1034 recruits were predicted, to test our recruitment and retention assumptions. The predefined stop–go criteria were:

• If recruitment was ≥ 75% (n ≥ 775), then the target sample size was clearly achievable and Trial Steering Committee (TSC) recommendation to Health Technology Assessment (HTA) programme would be to continue directly with the main trial.

• If recruitment was between 50% and 75% (517 ≤ n ≤ 775), then the TSC recommendation to the HTA programme would be to recruit more centres and review again in 6 months.

• If recruitment was < 50% (n < 517), then urgent discussions were required between the Project Management Group and the TSC to undertake a detailed review of options to subsequently recommend to the HTA programme.

Outcomes

Sample size

Observational studies of operative vaginal birth estimate postnatal infection to occur in between 2% and 16% of women. 29 The single existing trial of antibiotic prophylaxis at operative vaginal birth found a 4% rate of postnatal infection. 10 We therefore used this more conservative estimate of the maternal infection rate following operative vaginal birth. We assumed an estimated relative risk reduction of 50% in this rate with antibiotics to 2% in the treatment arm. The single trial relating to operative vaginal birth suggests a greater reduction than this, but this rate of reduction is based on that seen in the more robust antibiotic prophylaxis for caesarean section trials. 6 We used the absolute difference (i.e. a reduction from 4% to 2%) to calculate the sample size. To detect such a difference with 90% statistical power at the two-sided 5% level of significance required 1626 participants per group; with an estimated 5% loss to follow-up, the trial required 1712 participants per group, which was a total of 3424 women.

Statistical analyses

Statistical analyses were carried out according to a prespecified statistical analysis plan finalised and agreed prior to unblinding. In summary, demographic and clinical data were summarised with counts and percentages for categorical variables, means [standard deviations (SDs)] for normally distributed continuous variables and medians (with interquartile or simple ranges) for other continuous variables. Women were analysed in the groups to which they were randomly assigned, comparing the outcome of all women allocated to active treatment with all those allocated to placebo, regardless of deviation from the protocol or treatment received (referred to as the intention-to-treat population). Binary outcomes were analysed using risk ratios, while continuous outcomes were analysed using either a mean or a median difference, as appropriate. As randomisation did not involve stratification or minimisation, the primary analysis was based on unadjusted estimates of effect. Two-sided statistical testing was performed throughout. A 5% level of statistical significance was used for analyses of the primary outcome, and 1% for secondary outcomes. 95% CIs are presented for analyses of the primary outcome and 99% CIs for secondary outcomes.

Sensitivity analyses

Four planned sensitivity analyses were carried out:

1. examining the primary outcome restricted to women who had not received antibiotics in the 7 days prior to birth, in case any masking of a prophylactic effect occurred by inclusion of pre-treated women

2. examining the primary outcome excluding women prescribed antibiotics (other than the trial intervention) within 24 hours of birth in case these women were already infected prior to administration of the intervention

3. a repeat analysis of the primary outcome restricted to women whose primary outcome was obtained based on data obtained between 6 and 10 weeks after women had given birth

4. a sensitivity analysis including centre as a random effect.

Data collection

Data were collected at hospital discharge by extraction of information from a woman’s clinical records by the research midwife and at 6 weeks post partum by telephone interview with a research midwife to obtain information on the primary outcome. Following this, each woman was sent a postal or online questionnaire (as preferred by each woman) for collection of data on secondary outcomes.

Text reminders for completion of the questionnaire were sent and women were offered the option for telephone completion in the event of a delayed response to ensure a high response rate. Information about any hospital re-admissions and results of any microbiological investigations when a woman indicated that she had a suspected or confirmed infection were collected from hospital records by the site research midwife.

Basic demographic, medical and obstetric details were collected for all women, including details of any antibiotic treatment in the 7 days before women gave birth and the indication for antibiotic prescription.

Data on maternal anaphylaxis were collected up until hospital discharge. Data on other secondary outcomes were collected at 6 weeks post partum using standard instruments, when possible, as detailed below.

Adverse event reporting

The safety reporting window for this trial was from administration of intervention to 6 hours post administration or discharge (whichever was sooner). Non-serious adverse events were not routinely recorded as the IMP is a licensed product being given at a standard dose. However, adverse events that were part of the trial outcomes were recorded in the case report form.

All serious adverse events (SAEs) were reported immediately, at least within 24 hours, except the following SAEs (which were not considered to be causally related to the trial intervention because these events occurred prior to the trial intervention being administered):

• birth defect/congenital anomaly

• hypertensive disorder of pregnancy (e.g. pre-eclampsia/eclampsia)

• post-partum haemorrhage with onset before the intervention.

Health-care resource use and cost analysis

As an additional analysis, not specified in the trial protocol, we conducted a within-trial comparison of health-care resource use and associated costs between trial arms. The perspective of the analysis was that of the UK NHS and the following categories of resource use were collected using the telephone interview and postal questionnaire at 6 weeks post delivery: antibiotic use (co-amoxiclav prophylaxis and new prescriptions), health-care professional visits [general practitioner (GP), nurse, midwife, health visitor and district nurse], outpatient hospital visits and all-cause hospital re-admissions.

Unit costs for co-amoxiclav were collected and estimated using the British National Formulary 2017. 38 We did not have details of the antibiotics used for new prescriptions and, therefore, assumed that all new prescriptions were co-amoxiclav, a low-cost generic antibiotic, so that our estimate of costs was conservative. Although some women may have received more than one course of antibiotics, we assumed that all women who had a new prescription of antibiotics had a single 7-day course of co-amoxiclav with three daily doses, again so that our estimate of costs was conservative. Unit costs for GP visits and nurse/midwife services at general practice were collected from Unit Costs of Health and Social Care 2017. 39 For other health-care resource categories, the unit cost was taken from NHS Reference Costs 2017/18. 40 More details on sources and associated estimates of unit costs (expressed in 2017/18 Great British pounds) for the different categories of resource use are presented in Appendix 2, Tables 17 and 18. The costs associated with each category of resource use were estimated by multiplying resource use by unit costs.

We calculated the mean and SD of resource uses (number of visits for health-care professional and outpatient hospital visits, number of days for re-admissions) and costs for each category by trial arms. We also calculated overall mean (SD) costs at 6 weeks following delivery by adding up all individual cost categories, together with the cost for intervention (a single dose of intravenous co-amoxiclav). Mean differences in health-care resource use and costs were calculated and associated 99% parametric CIs were estimated.

The main analysis was based on complete cases among women who returned the postal questionnaire at 6 weeks post delivery and completed questions on usage for each specific health-care resource category. We compared the characteristics of women who returned the postal questionnaire with those who did not using the t-test for continuous variables and Pearson’s chi-squared test for categorical variables. A sensitivity analysis using multiple imputation with chained equations was conducted to evaluate the impact of missing data on our complete-case results. Fifty values were imputed for each missing data point by specifying separate regression models for each variable with missing data. Within each model, the remaining variables with missing data were used as predictors along with selected patient characteristics [maternal age, gestational age at randomisation in weeks, ethnic group, body mass index (BMI), any previous pregnancies of ≥ 22 weeks’ gestation, labour induction, episiotomy in current delivery, perineal tear in current delivery]. Imputation was performed using prediction mean matching (using five nearest neighbours) and for each trial arm separately.

Governance and monitoring

A monitoring plan for the trial, including responsibilities, was developed prior to the start of recruitment. Remote monitoring was carried out, with at least one monitoring assessment of each site. Site monitoring visits were carried out if remote monitoring identified any discrepancies.

The trial was supervised on a day-to-day basis by a Project Management Group. A TSC was convened comprising an independent chairperson, four other independent members, a patient and public involvement representative, the NPEU CTU Director and the chief investigator. A Data Monitoring Committee (DMC), independent of the applicants and of the TSC, reviewed the progress of the trial annually and provided the TSC with advice on the conduct of the trial.

Summary of changes to the trial protocol

The initial planned primary outcomes were defined as follows.

Confirmed or suspected maternal infection within 6 weeks of delivery, as defined by one of:

• a new prescription of antibiotics

• confirmed systemic infection on culture

• endometritis as defined by the US CDC. 31

Following an interim review of the data by the DMC in February 2017, a change to the primary outcome was recommended to exclude women who were prescribed antibiotics for unrelated indications. This was because of concerns that the noise arising from antibiotic prescriptions for unrelated indications would mask the effect of the intervention. The primary outcome was therefore revised to the following.

Confirmed or suspected maternal infection within 6 weeks of delivery, as defined by one of:

• a new prescription of antibiotics for presumed perineal wound-related infection, endometritis or uterine infection, urinary tract infection with systemic features or other systemic infection

• confirmed systemic infection on culture

• endometritis as defined by the US CDC. 31

Indications for antibiotic prescription were coded independently as described in Data collection by two clinically qualified staff, based on responses to the telephone questionnaire, the infection form or the postal questionnaire. The revised primary outcome could, therefore, be derived for all women already recruited into the trial and no additional processes were needed.

A summary of the other changes made to the original protocol is presented in Appendix 3.

Recruitment and retention

Between March 2016 and June 2018, 3427 women were randomised: 1719 to the antibiotic arm and 1708 to the placebo arm (Figure 2).

FIGURE 2.

Flow of participants. Figure modified from Knight et al. 30 This article is available under the terms of the Creative Commons Attribution License (CC BY). You may copy and distribute the article, create extracts, abstracts and new works from the article, alter and revise the article, text or data mine the article and otherwise reuse the article commercially (including reuse and/or resale of the article) without permission from Elsevier. You must give appropriate credit to the original work, together with a link to the formal publication through the relevant DOI and a link to the Creative Commons user license above. You must indicate if any changes are made but not in any way that suggests the licensor endorses you or your use of the work.

Figure modified from Knight et al. 30 This article is available under the terms of the Creative Commons Attribution License (CC BY). You may copy and distribute the article, create extracts, abstracts and new works from the article, alter and revise the article, text or data mine the article and otherwise reuse the article commercially (including reuse and/or resale of the article) without permission from Elsevier. You must give appropriate credit to the original work, together with a link to the formal publication through the relevant DOI and a link to the Creative Commons user license above. You must indicate if any changes are made but not in any way that suggests the licensor endorses you or your use of the work

The number of participants recruited at each site varied from 22 to 388 (Table 1).

Characteristics of participants

Characteristics of participants were similar between the two trial arms (Table 2) and appeared to be representative of the general population of women undergoing operative vaginal birth. Women had a mean age of 30 years, approximately half were of normal BMI and more than four-fifths were of white ethnicity. Overall, 77% of women were primiparous, and 49% had labour induced. Sixty-five per cent of births were assisted by forceps and 35% were assisted by vacuum extraction. Eighty-eight per cent of women had an episiotomy, 31% had a perineal tear and > 99% of women had suturing of a perineal wound.

Outcomes

Primary outcome data were available for 3225 out of 3420 women (94.3%). There were no substantive differences between women for whom primary outcome data were available (Table 3). Questionnaires were received for 2593 women (75.8%).

Adverse events and side effects

Only three women reported side effects of the intervention: 2 out of 1715 in the antibiotic arm (0.12%) and 1 out of 1705 (0.06%) in the placebo arm (risk difference 0.06%, 95% CI –0.14% to 0.25%). The woman in the placebo arm reported a skin rash, and the women in the antibiotic arm reported other reactions (e.g. itching, swollen throat). There were no cases of anaphylaxis. Three SAEs were reported (Table 6), but only one was thought to be causally related to the intervention.

Adherence

The intervention was administered a median of 3 hours after women had given birth (Table 7). Thirty-three women (1.0%) received the intervention > 6 hours after giving birth. Overall, 33 telephone follow-up interviews (1.0%) were conducted by staff who had prepared or checked the ANODE intervention and may, theoretically, have been unblinded to allocation. Similarly, as antibiotic prescription was part of the primary outcome, we checked whether or not the person prescribing antibiotics was also the person who had checked or prepared the ANODE intervention and would therefore have been unblinded. Only one woman, in the antibiotic arm, was prescribed further antibiotics by a member of staff who may have been unblinded.

Sensitivity analyses

There were no material changes to the primary outcome with any of the sensitivity analyses (Tables 8–11 and Figure 3).

FIGURE 3.

Forest plot showing the results of the sensitivity analyses.

Health-care resource use and cost analysis

Among the 2593 women who returned the questionnaire, we found evidence that, at 6 weeks post delivery, women randomised to the antibiotic arm consumed fewer NHS health-care resources than those in the placebo arm (Table 12). The mean difference in all categories of resource use favoured the antibiotic arm, with the number of visits to the GP (p < 0.001), nurse or midwife home visits (p < 0.001) and outpatient hospital visits (p < 0.001) being statistically significantly different between groups. No statistically significant mean differences were detected in the length of stay for all-cause hospital re-admissions.

Similar results were found for costs. Among different health-care categories, the highest cost was estimated for hospital re-admissions, followed by nurse or midwife home visits and outpatient hospital visits (Table 13). Compared with the women allocated to the placebo arm, the health-care costs for women in the antibiotic arm were, on average, a significant £0.40 (99% CI £0.20 to £0.50) less for new prescriptions of antibiotic, £3.90 (99% CI £1.60 to £6.20) less for GP visits, £12.10 (99% CI £4.00 to £20.10) less for nurse or midwife home visits and £15.40 (99% CI £5.20 to £25.50) less for outpatient hospital visits. The mean cost of hospital re-admissions was also found to be lower in the antibiotic arm (£50.30 vs. £81.60), but the difference was not statistically significant (p = 0.164). The total mean cost at 6 weeks following delivery was estimated to be £102.50 (SD £652.40) per woman in the antibiotic arm and £155.10 (SD £497.40) per woman in the placebo arm. The mean difference in the total health-care cost was –£52.60 (99% CI –£115.10 to £9.90) per woman, which was not statistically significantly different at the 1% level (p = 0.030).

On the basis of the estimated number of doses of co-amoxiclav used for the costs analysis (see Table 13), we estimated the potential effect of a policy of universal prophylaxis with a single dose of co-amoxiclav after operative vaginal birth on overall antibiotic use. In the placebo arm, 19% of women had a confirmed or suspected infection and 1297 women received an estimated 5166 doses of co-amoxiclav to treat infection (three daily doses for 7 days in 246 women). In the antibiotic arm, 11% of women had a confirmed or suspected infection and 1296 women received an estimated 3003 doses of co-amoxiclav to treat infection (three daily doses for 7 days in 143 women), as well as 1296 prophylactic doses; therefore, a total of 4299 doses. A policy of universal prophylaxis would therefore be associated with an estimated net reduction of 867 doses of antibiotic (17% reduction).

There were statistically significant differences in the characteristics of women who had missing secondary resource outcomes compared with those who did not (Table 14). However, the comparisons of post-delivery health-care resource use and costs after multiple imputation were similar to those generated by using complete cases (Tables 15 and 16). Compared with women allocated to the placebo arm, women in the antibiotic arm had significantly fewer GP visits, nurse or midwife visits at home, and outpatient hospital visits. The total mean costs after multiple imputation were estimated to be £117.30 (SE £20.23) in the co-amoxiclav group and £168.30 (SE £15.07) in the placebo group. The mean difference in the total cost was –£50.90 (99% CI–£114.70 to £12.90; p = 0.040).

Summary of main findings

This trial showed clear evidence of benefit of a single dose of intravenous co-amoxiclav administered to women a median of 3 hours after operative vaginal birth. Women in the antibiotic arm had a 42% relative reduction, from 19% to 11%, in the risk of suspected or confirmed infection. This was principally driven by the prescription of antibiotics for presumed perineal wound-related infection, endometritis or uterine infection, urinary tract infection with systemic features or other systemic infection, but women in the antibiotic arm also had a statistically significant 56% reduction in the risk of confirmed systemic infection on culture, from 1.5% to 0.6%. Secondary outcomes also favoured the active (co-amoxiclav) arm, with significant reductions in rates of both deep and superficial perineal infection, perineal pain, wound breakdown and the need for additional perineal care.

At 6 weeks post delivery, women who were randomised to the antibiotic arm consumed fewer NHS health-care resources than women randomised to the placebo arm, with significantly fewer visits to the GP, midwife/nurse home visits and outpatient hospital visits. The mean total health-care cost per woman was £52.60 lower in the antibiotic arm than in the placebo arm, but this was not statistically significantly different at the 1% level.

Limitations

We took the pragmatic approach in this trial of defining suspected or confirmed maternal infection using a composite outcome, including a new prescription of antibiotics for confirmed or suspected infection, which can be interpreted as equating to a clinical diagnosis of infection. This clinical diagnosis of infection drove the overall outcome and resulted in a substantially higher than anticipated event rate, which may suggest overprescription of antibiotics in the postnatal period. The use of this clinical definition rather than microbiologically confirmed infection could be regarded as a limitation; however, we observed a statistically significant decrease in the rate of microbiologically confirmed systemic infection following culture from a sterile site, which supports the assumption that the findings represent a genuine decrease in infection, despite our pragmatic primary outcome definition.

We achieved a 76% follow-up rate for the majority of our secondary outcomes, which represents the main limitation of this trial. There were statistically significant differences in the characteristics of the women who completed the 6-week questionnaire and those who did not. Most notably, a greater proportion of women who returned the 6-week questionnaire were primiparous. It is possible that their consultation behaviour differed from multiparous women; thus, rates of some of the reported secondary outcomes may be higher than would be seen in the general population. This difference in characteristics between those with follow-up data and those without is, however, unlikely to account for the magnitude of difference we observed between the antibiotic and placebo arms.

The population of women included in the trial was representative of the population of women undergoing operative vaginal birth and thus appear broadly generalisable. The trial was limited to women who were not penicillin-allergic and, therefore, is not directly applicable to women who are allergic to penicillin. However, it would be anticipated that another antibiotic with a similar spectrum of activity would have a similar protective effect in women for whom co-amoxiclav is contraindicated. Options for penicillin-allergic women that offer a comparable spectrum of activity to co-amoxiclav would include cefuroxime with metronidazole or clindamycin with gentamicin. Similarly, we included only women giving birth at ≥ 36 weeks’ gestation. However, there are no clear reasons why the results would not also be generalisable to women who had an instrumental vaginal birth at lower gestations than this.

We asked sites to administer the intervention as soon as possible after women had given birth, and no later than 6 hours after women had given birth. In practice, it was administered a median of 3 hours after women had given birth. It is possible that administration this length of time after women had given birth made the co-amoxiclav less effective than it might have been, given that caesarean section trials suggest that pre-delivery administration is more effective at preventing wound infection and endometritis. 41 However, as noted earlier, the perineal wound, which is highly likely to become contaminated, is very different from the caesarean section wound, and later administration may have allowed for a longer duration of protective effect and greater efficacy. Further analyses are needed to investigate the mechanism of effect, as described in Implications for research.

Comparison with existing literature

The single previous trial of antibiotic prophylaxis after operative vaginal birth42 reported on endometritis only, noting a rate of 4% in the no antibiotic arm. This is considerably lower than the rate of suspected or confirmed infection we observed in the ANODE trial but, interestingly, using the strict CDC surveillance definition,31 we observed a lower endometritis rate. The estimate of effect we observed for endometritis (relative risk 0.65, 95% CI 0.34 to 1.24) in the ANODE trial is compatible with the effect estimate in the Heitmann and Benrubi trial42 (relative risk 0.07, 95% CI 0.00 to 1.17). Combining the results of the two trials using Mantel–Haenszel fixed-effect meta-analysis gives an overall relative risk 0.50 (95% CI 0.27 to 0.93) for endometritis.

Women in the ANODE trial who received placebo had a 19% rate of confirmed or suspected infection at 6 weeks post partum. This is higher than the infection rate reported in most other studies of complications following instrumental vaginal birth. 29 As noted previously, very few of these observational studies followed women beyond discharge, yet post-partum wound infections and endometritis have been reported to occur at a peak of 7 days post discharge in large data-linkage studies. 43 The ANODE trial shows very clearly a significant burden of confirmed or suspected infection and, most notably, both superficial and deep perineal wound infection after initial hospital discharge.

Although a single dose of co-amoxiclav almost halved the infection rate, 11% of women still had a confirmed or suspected infection after receiving antibiotic prophylaxis. The question therefore arises as to whether or not other, non-pharmaceutical interventions might reduce this further. We did not collect information about the aseptic techniques used at the time of operative vaginal birth, nor whether or not gloves and instrument trollies were changed between birth and perineal repair; the RCOG guidelines simply state that good standards of hygiene and aseptic technique are recommended. 15 It is possible that further attention to aseptic technique at the time of birth may influence later outcomes. A Cochrane review identified no difference in infection rates in surgical wounds cleansed with water versus other solutions (saline, procaine spirit) or no cleansing. 44 The review identified only one small trial of wound-cleansing post episiotomy, which randomised 100 women to cleanse their episiotomy wounds with either water or procaine spirit;45 the authors report no difference in infection rates, but did not give exact figures. They note that women cleaned their wounds an average of five times per day and that all wounds were healed well by 14 days post partum. Therefore, there may be a place for further investigation of wound-cleansing after operative vaginal birth to see if infection rates are lower.

There is an apparent discrepancy between the proportion of women who were diagnosed with systemic sepsis in the ANODE trial, defined by the modified SIRS criteria for pregnancy, and those with a microbiologically confirmed systemic infection. There are several possible reasons for this. The modified SIRS criteria in pregnancy3,32 have never been validated against microbiological or other diagnostic tests, and work is ongoing to establish a clear and actionable definition of maternal sepsis. 46,47 It is possible that women with microbiologically confirmed systemic infection are false negative cases of maternal sepsis when using the modified SIRS criteria. Conversely, with the substantially raised awareness of maternal sepsis, which has followed the UK Confidential Enquiries into Maternal Deaths and Morbidity48 and the work of the UK Sepsis Trust [https://sepsistrust.org (accessed 12 December 2018)], women with microbiologically confirmed systemic infection may have been diagnosed and treated early, before progression to infection sufficiently severe to meet the modified SIRS criteria.

Women reported a range of other complications after operative vaginal birth, the majority of which were also less frequent among women who were randomised to receive antibiotic prophylaxis. A very high proportion of women reported perineal pain at 6 weeks after giving birth (46% in the antibiotic arm and 55% in the placebo arm), although a much lower proportion reported use of pain relief for their perineal pain (8% in the antibiotic arm vs. 11% in the placebo arm). A recent US study of women undergoing vaginal birth noted that operative vaginal birth was associated with opioid prescription in the 7 days after discharge (adjusted odds ratio 1.52, 95% CI 1.49 to 1.54), indicating a significant degree of pain. A trial of restrictive versus routine use of episiotomy at operative vaginal birth noted very similar rates of perineal pain at 6 weeks post partum (44%), but this had decreased to 18% at 1 year post partum. 49 Notably, women reported higher levels of pain with restrictive use of episiotomy compared with routine use; however, almost 90% of women included in the ANODE trial had an episiotomy, suggesting that restricted use is unlikely to be a potential underlying reason for the high rates of perineal pain observed.

One in five women in the placebo arm and one in 10 in the antibiotic arm reported that they had experienced perineal wound breakdown. Although a previous feasibility study reported that most women whose perineal wound had dehisced had healed by 6–8 weeks,35 women described long-term impacts 6–9 months later,37 including psychosexual morbidity. It is therefore probable that the almost 50% reduction in wound breakdown reported in the antibiotic arm is associated with longer-term benefit on sexual function, even though we observed no difference in dyspareunia between the groups at 6 weeks post partum. Only 40% of women had resumed intercourse, which may also mask any potential beneficial effect of the active intervention, but our results highlight the high levels of ongoing maternal morbidity following childbirth.

The mean total health-care cost per woman was £52.60 lower in the antibiotic arm than in the placebo arm, but this was not statistically significantly different at the 1% level. However, considering the probable long-term persistence of some complications, and the likelihood that the intervention dominates, it is probable that co-amoxiclav prophylaxis will prove cost-effective as well as clinically effective in the longer term. In the light of current concerns over antimicrobial stewardship and the emergence of antimicrobial resistance,21 an assessment of the impact of the single prophylactic dose on overall antibiotic use is important. The additional health-care cost and resource analysis conducted for the ANODE trial estimates that, for each additional 100 doses of antibiotic used in prophylaxis, 168 treatment doses will be avoided, representing a 17% overall reduction in antibiotic use with a policy of universal prophylaxis.

Sensitivity analyses

None of the sensitivity analyses we conducted materially affected the assessment of the primary outcome. A number of women were prescribed antibiotics in labour, for indications such as prophylaxis against group B streptococcal infection or maternal pyrexia. Exclusion of these women from the analysis did not change the estimate of effect of antibiotic prophylaxis, indicating that the effect we observed was not caused by a protective effect of prior antibiotic administration. We additionally examined whether or not exclusion from the primary outcome of any women who were prescribed antibiotics within the first 24 hours after giving birth had any impact, to test if the effect was solely due to prevention of infection among women who were already infected at or around the time of giving birth. As this had no substantial effect on the results, it suggests that the protective effect of a single dose of antibiotic is not caused by treatment of early or already established infections alone.

We also examined whether or not the exclusion of data that were obtained < 6 weeks or > 10 weeks after women had given birth had an impact on the results. There was no major influence, suggesting that the results were not biased by any over-reporting of outcomes from data returned at a later time point or under-reporting of outcomes in data returned at an earlier time point.

As a final sensitivity analysis we included site as a random effect; there was no suggestion that the results differed by site.

Implications for practice

Current national guidance on operative vaginal birth in the UK,15 the USA,50 Australia and New Zealand51 either do not mention or do not recommend antibiotic prophylaxis after instrumental vaginal birth. The WHO guidelines on prevention of maternal infection explicitly state that antibiotic prophylaxis is not recommended after instrumental vaginal birth based on a lack of evidence of effectiveness. The ANODE trial provides clear evidence of effectiveness of a single dose of prophylactic intravenous co-amoxiclav on prevention of confirmed or suspected maternal infection, a range of secondary outcomes and a potential impact on health service costs. On the basis of this evidence, reconsideration of official policy/guidance about whether or not routine prophylaxis with a single dose of intravenous antibiotic after operative vaginal birth should be considered.

Implications for research

Conclusions

The ANODE trial has shown clear evidence of the benefit of a single intravenous dose of prophylactic co-amoxiclav after operative vaginal birth; these results indicate the need for reconsideration of official policy/guidance. Further analysis of the mechanism of action of this single dose of antibiotic is needed to investigate whether earlier, pre-delivery or repeated administration could be more effective. Until these analyses are completed, there is no indication for administration of more than a single dose of prophylactic antibiotic, or for pre-delivery administration.

Sponsorship

The ANODE trial was sponsored by the University of Oxford.

Independent Trial Steering Committee

Fiona Denison (chairperson), Simon Cousens, Timothy Overton, Rachel Plachinski, Donna Southam and Matthew Wilson.

Independent Data Monitoring Committee

Siladitya Bhattacharya (chairperson), Carrol Gamble and Michael Millar.

ANODE Trial Collaborative Group

• Marian Knight, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford.

• Helen Enderby, lay member, PRIME group.

• Derek Tuffnell, Bradford Teaching Hospitals NHS Foundation Trust.

• Kim Hinshaw, Sunderland Royal Hospital, Tyne and Wear.

• Ranee Thakar, Croydon University Hospital.

• Abdul H Sultan, Croydon University Hospital and St George’s, University of London.

• Julia Sanders, Cardiff University.

• Dharmintra Pasupathy, King’s College London.

• Philip Moore, Birmingham Women’s Hospital.

• James Gray, Birmingham Women’s and Children’s NHS Foundation Trust.

• Oliver Rivero-Arias, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford.

• Ed Juszczak, NPEU CTU, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford.

• Louise Linsell, NPEU CTU, National Perinatal Epidemiology Unit, Nuffield Department of Population Health, University of Oxford.

• Aethele Khunda, James Cook University Hospital, Middlesbrough.

Contributions of authors

Marian Knight wrote the first draft of the manuscript with contributions from Virginia Chiocchia, Oliver Rivero-Arias and Xinyang Hua.

Virginia Chiocchia, Christopher Partlett, Louise Linsell and Edmund Juszczak developed the analysis plan and conducted the analyses.

Marian Knight, Virginia Chiocchia, Christopher Partlett, Oliver Rivero-Arias, Xinyang Hu, Ursula Bowler, James Gray, Shan Gray, Kim Hinshaw, Aethele Khunda, Philip Moore, Linda Mottram, Nelly Owino, Dharmintra Pasupathy, Julia Sanders, Abdul H Sultan, Ranee Thakar, Derek Tuffnell, Louise Linsell and Edmund Juszczak contributed to the development and conduct of the trial and edited and approved the final version of the manuscript.

All members of the collaborative group contributed to the development and conduct of the trial.

Publications

Knight M, Mottram L, Gray S, Partlett C, Juszczak E, ANODE collaborative group. Prophylactic antibiotics for the prevention of infection following operative vaginal delivery (ANODE): study protocol for a randomised controlled trial. Trials 2018;19:395.

Knight M, Ciocchia V, Partlett C, Rivero-Arias O, Hua X, Hinshaw, et al. Prophylactic antibiotics in the prevention of infection after operative vaginal delivery (ANODE): a multicentre randomised controlled trial. Lancet 2019;393:2395–403.

Data-sharing statement

Data will be shared in accordance with the National Perinatal Epidemiology Unit Data Sharing policy. Requests for access to the data will be considered by the National Perinatal Epidemiology Unit Data Sharing committee. Access to anonymised data can be requested from general@npeu.ox.ac.uk.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.

Bradford Royal Infirmary

Derek Tuffnell (Consultant Obstetrician and Gynaecologist).

Jennifer Syson (Local Research Midwife).

Burnley General Hospital

Manisha Golash (Consultant Obstetrician).

Rebekah McCrimmon (Local Research Midwife).

Croydon University Hospital

Ranee Thakar (Subspecialist in Urogynaecology and Consultant Obstetrician and Gynaecologist).

Hana Kleprlikova (Local Research Midwife).

Darlington Memorial Hospital

Poornima Ranka (Consultant Obstetrician).

Jacqui Jennings (Local Research Midwife).

Derriford Hospital

Alex Taylor (Consultant Obstetrician).

Heidi Hollands (Local Research Midwife).

East Surrey Hospital

Mahalakshmi Gorti (Consultant Obstetrician and Gynaecologist).

Linda Bailey (Local Research Midwife).

Guy’s and St Thomas’ Hospital

Dharmintra Pasupathy (Consultant and Reader in Maternal and Fetal Medicine).

Cathryn Marsh (Local Research Midwife).

John Radcliffe Hospital

Christos Ioannou (Consultant in Obstetrics and Fetal Medicine).

Clare Edwards (Local Research Midwife).

Liverpool Women’s Hospital

Mark Clement-Jones (Consultant Obstetrician).

Siobhan Holt (Local Research Midwife).

Northumbria Specialist Emergency Care Hospital

Jonathan Brady (Consultant).

Helen Howlett (Local Research Midwife).

Princess Anne Hospital

James Hounslow (Consultant Obstetrician and Gynaecologist).

Fiona Walbridge/Nicki Martin (Local Research Midwife).

Queen Elizabeth Hospital

Vaideha M Deshpande (Consultant Obstetrician/Gynaecologist and Reproductive Medicine Specialist).

Barbara Lynam (Local Research Midwife).

Royal Berkshire

Patrick Bose (Consultant).

Fidelma Lee (Local Research Midwife).

Royal Devon and Exeter Wonford Hospital

Tracey Kay (Consultant Obstetrician and Gynaecologist).

Jacqui Tipper (Local Research Midwife).

Royal Stoke University Hospital

Simon Cunningham (Consultant Obstetrician and Gynaecologist).

Donna Brayford (Local Research Midwife).

Royal Victoria Infirmary

Suzanne Jackson (Consultant Obstetrician).

Michelle Perkins (Local Research Midwife).

Singleton Hospital – Swansea/Princess of Wales – Bridgend

Madhuchanda Dey (Consultant Obstetrician and Gynaecologist).

Sharon Jones (Local Research Midwife).

South Tyneside District Hospital

Umo Esen (Consultant Obstetrician and Gynaecologist).

Judith Ormonde (Local Research Midwife).

Stoke Mandeville Hospital

Nutan Mishra (Consultant Obstetrician).

Julie Tebbutt (Local Research Midwife).

Sunderland Royal Hospital

Kim Hinshaw (Consultant Obstetrician and Gynaecologist).

Lesley Hewitt (Local Research Midwife).

The James Cook University Hospital

Aethele Khunda (Consultant Obstetrician and Urogynaecologist).

Hazel Alexander (Research Practitioner).

University Hospital of North Durham

Tadala Saukila (Consultant Obstetrician).

Vicki Atkinson (Local Research Midwife).

University Hospital of North Tees

Mihraban Bapir (Consultant Obstetrician and Gynaecologist).

Sharon Gowans (Local Research Midwife).

University Hospital of Wales

Julia Sanders (Consultant Midwife).

Anouk Ridgway (Local Research Midwife).

Warrington Hospital

Rita Arya (Consultant Obstetrician and Gynaecologist).

Rachel Crone/Lindsay Roughley (Local Research Midwife).

Whittington Health

Meg Wilson (Consultant).

Osi Egole (Local Research Midwife).

Exclusion criteria (page 16)

The exclusion criterion was amended to exclude all participants who have the contraindications listed in the Summary of Medicinal Product Characteristics for co-amoxiclav as requested by the Medicines and Healthcare products Regulatory Agency (MHRA).

Reporting serious adverse events and procedure for immediate reporting of serious adverse events (pages 25–26)

The section was amended in response to a request made by the MHRA on reviewing the initial ANODE trial Clinical Trials Authorisation application.

A list of SAEs that were not considered to be causally related to the trial intervention were agreed with Professor Bhattacharya (chairperson of the ANODE trial DMC). The SAE reporting procedure was also amended to include the ability to report SAEs via the Clinical Database OpenClinica© (Enterprise edition, OpenClinica LLC, Waltham, MA, USA) on page 26.

Listed below are all edits reviewed by the MHRA prior to the Clinical Trials Authorisation being granted (ANODE trial protocol version 2.0):

1. The Ethics Reference number was added to page 1.

2. Social care visits were removed from the secondary outcomes; they had been included in error.

3. The spelling of amoxicillin has been made consistent throughout the document.

4. The duration of the study on page 16 was amended to reflect changes made to the recruitment start date (changed from 1 September 2015 to 1 December 2015).

5. A reference was deleted on page 34 of the protocol because it was duplicated in error.

The following changes have been made to create protocol version 3.0 3 December 2015.

Exclusion criteria (page 16)

In the point ’Note that receiving antenatal or postnatal antibiotics e.g. for maternal Group B Streptococcal carriage or prolonged rupture of membranes, is not a reason for exclusion if there is no indication for ongoing antibiotic prescription post-delivery’, the words ’or postnatal’ have been removed because this wording was incorrect and contradicts the previous sentence.

Reporting serious adverse events and procedure for immediate reporting of serious adverse events (pages 25–26)

Wording was amended to improve consistency and to make it clear that events that commence prior to the administration of the trial intervention do not require reporting as a SAE.

Randomisation, blinding and code-breaking (page 20)

Text edited regarding balance and unpredictability from ’within centre’ to ’overall’ by trial statistician.

Text edited to show that an emergency code-breaking procedure will not be required; as only a single dose of co-amoxiclav will be administered, there is no need to code-break if further antibiotics are required. If a woman was to have an anaphylactic reaction, she would be treated as if she had been given the active drug.

Other edits to the protocol in version 3.0 are listed below

1. The list of investigators was removed from the cover page to make the protocol clearer and to ensure that the ANODE trial team at the Clinical Trials Unit were approached with any protocol queries in the first instance rather than a co-investigator. The investigators will be listed on the ANODE trial website.

2. The confidentiality statement was removed from page 2 as the protocol is no longer confidential and is available publicly.

The following changes have been made to create protocol version 4.0 22 April 2016.

Section 7.0 trial design (page 15)

Edits to this section were made to specify exactly who would not be blinded to allocation to clarify that this also includes the person responsible for checking the intervention. The people not blinded to intervention were the person who prepared the trial intervention and the person who checked the intervention prior to administration. Training was provided to all unblinded staff on the importance of maintaining blinding and unblinded staff were not involved in the collection of outcomes information.

Other edits to the ANODE trial protocol

• Page 1 – ISRCTN added and signature blocks for the chief investigator and the statistician removed; they were documented separately and filed with the protocol in the trial master file.

• Global edit – where Hospital Episode Statistics (HES) are referenced, ’or NHS Wales Informatics Service’ was added. HES were accessed for participants recruited in England; NHS Wales Informatics Service were accessed for those participants recruited in Wales.

• Pages 15–16 and appendix A – duration of study edited to reflect the change in start date agreed with the HTA programme.

• Page 19 – edited to clarify that the original signed consent forms would be sent the co-ordinating centre and a copy retained at site.

• Page 20 – section 9.5 edited to clarify the roles of the senior trials statistician and the senior trial programmer with regard to their responsibilities regarding the randomisation schedule generation.

The following changes have been made to create protocol version 5.0 30 November 2017.

Amendment to the definition of the primary outcome [pages 7, 9 (in flow chart), 14 and 20]

Primary outcome refined to the amended text below:

• A new prescription of antibiotics for presumed perineal wound-related infection, endometritis or uterine infection, urinary tract infection with systemic features or other systemic infection.

• Confirmed systemic infection on culture.

• Endometritis as defined by the US CDC. 31

Trial timeline updated to reflected changes following an extension to the duration of the trial (pages 16 and 39)

Description of statistical methods (page 27)

The statistics section of the protocol was updated to make it consistent with the current strategy detailed in the statistical analysis plan as requested by the ANODE trial DMC in a meeting held on 27 November 2017.