- Email: [email protected]
Antibiotic prophylaxis to prevent post-abortal upper genital tract infection in women with bacterial vaginosis: randomised controlled trial
British Journal of Obstetrics and Gynaecology April 2001, Vol. 108, pp. 396±402
Antibiotic prophylaxis to prevent post-abortal upper genital tract infection in women with bacterial vaginosis: randomised controlled trial Tessa Crowley a,*, Nicola Low b, Andrew Turner c, Ian Harvey b, Ken Bidgood d, Patrick Horner a Objective To determine the prevalence of bacterial vaginosis in women undergoing ®rst trimester suction termination of pregnancy and to evaluate the ef®cacy of metronidazole in reducing the risk of post abortal pelvic infection in women with bacterial vaginosis. Design Randomised double-blind placebo-controlled trial. Setting Two teaching hospitals and one district general hospital. Sample Two hundred and seventy-three women with bacterial vaginosis undergoing termination of pregnancy. Methods Women with bacterial vaginosis, diagnosed using modi®ed Spiegel's criteria, were individually randomised to receive either a 2 g metronidazole suppository or identical placebo per-operatively. Participants, doctors and investigators were blinded to treatment allocation. Participants were asked to complete a questionnaire about post-operative symptoms, visits to the general practitioner, antibiotic treatment, readmission to hospital, contraception and emotional response after one month. Results The prevalence of bacterial vaginosis was 29.3% (326/1111). Intention-to-treat analysis showed that post-operative upper genital tract infection developed in 12/142 (8.5%) women allocated to metronidazole and 21/131 (16.0%) women randomised to placebo, a difference of 7.6% (95% con®dence intervals ±15.4 to 10.2%; relative risk 0.53, 0.27 to 1.03, P 0.055). The effect of prophylaxis was similar when the analysis was restricted to women receiving the allocated treatment and with complete follow up. There was no difference in the risk of readmission to hospital and the frequencies of self reported symptoms in the two groups were similar. Conclusion This randomised placebo-controlled trial among women with bacterial vaginosis provides weak evidence that metronidazole decreases the risk of upper genital tract infection after ®rst trimester suction termination of pregnancy but a chance ®nding could not con®dently be excluded. Large well-conducted randomised trials with long term outcome measures are now needed to determine the most effective antibiotic combinations and strategy for prevention of post-abortal pelvic infection.
INTRODUCTION Bacterial vaginosis is characterised by a malodorous vaginal discharge in which the normally predominant lactobacilli are replaced by large numbers of a mixed group of bacteria 1,2. It is the most common cause of vaginal discharge in women of reproductive age 3 and has been reported to be present in 28% of women request-
a
Department of Sexual Health Medicine, Bristol Royal In®rmary, UK b MRC Health Services Research Collaboration, Department of Social Medicine, University of Bristol, UK c Genitourinary Infection Reference Laboratory, Public Health Laboratory Service, Bristol Royal In®rmary, UK d Department of Obstetrics and Gynaecology, Musgrove Park Hospital, Taunton, UK * Correspondence: Dr T. Crowley, Milne Sexual Health Centre, Bristol Royal In®rmary, Lower Maudlin Street, Bristol BS2 8HW, UK. q RCOG 2001 British Journal of Obstetrics and Gynaecology PII: S03 06-5456(00)0009 1-7
ing termination of pregnancy 4. Bacterial vaginosis is associated with adverse obstetric outcomes such as premature rupture of membranes, late miscarriage and post-caesarean endometritis 5±7, and with genital tract infections following gynaecological operations, including surgical termination of pregnancy 8±12. There is no agreed policy in the UK about the prevention of post-abortal upper genital tract infection. In 1996 a study group of the Royal College of Obstetricians and Gynaecologists recommended that women undergoing termination of pregnancy should either be given prophylactic antibiotics active against bacterial vaginosis and Chlamydia trachomatis or screened and treated if infections are found (screen and treat policy) 13. These recommendations used the ®ndings of a meta-analysis of 12 randomised double-blind and placebo-controlled trials 12. Five trials used a nitroimidazole drug but only one included women with proven bacterial vaginosis, and reported a threefold reduction in the incidence of postabortal infection among women allocated a ten-day course of metronidazole 11. www.bjog-elsevier.com
ANTIBIOTIC PROPHYLAXIS TO PREVENT GTI 397
The aims of this study were to determine the prevalence of bacterial vaginosis in women undergoing ®rst trimester suction termination of pregnancy and to evaluate the ef®cacy of a single per-operative 2 g dose of metronidazole per rectum in reducing the incidence of clinically diagnosed upper genital tract infection in the ®rst four post-operative weeks. METHODS The study was an individually randomised doubleblind, placebo-controlled trial conducted at three hospitals in the South West of England, Pregnancy Advisory Service Clinic, St. Michael's Hospital, Bristol; Southmead Hospital, Bristol; and Musgrove Park Hospital, Taunton. An unrestricted randomisation list allocating active treatment or placebo was prepared using random number tables by one investigator (I.H.) who was not involved in the assignment process. The central pharmacy prepared suppositories containing two grams of metronidazole and placebos that were identical in appearance and texture. These were sealed in plastic bags numbered sequentially according to the randomisation list. Each bag was placed in a large brown opaque envelope with a correspondingly numbered study consent form, information sheet, questionnaire and stamped addressed envelope. Packages were stored in a fridge near the operating theatre at each centre. The randomisation list was stored in the central pharmacy and the code was broken only after all data had been entered. Patients, surgeons, general practitioners and the investigator collecting follow up data (T.C.) were all unaware of treatment allocation. Ethical approval for the trial was obtained at each centre. Because women requesting termination of pregnancy might be under emotional stress consent for enrolment into the prevalence study and for randomisation were carried out in separate stages to allow women time to consider their decision to participate. All women requesting termination of pregnancy between October 1996 and December 1998 were eligible for entry into the study. The study was explained and demographic and clinical data on women who wished to participate were recorded on a standard form. During vaginal speculum examination a high vaginal swab was taken, rolled onto a glass slide and air-dried. The swab was placed in charcoal medium to be cultured for Candida albicans and examined for Trichomonas vaginalis. Endocervical and urethral swabs were taken for detection of C. trachomatis using an ampli®ed enzyme immunoassay (IDEIA, Dako Diagnostics Ltd.). All reactive specimens were con®rmed as positive by direct immuno¯uorescence 14. Endocervical and urethral swabs for Neisseria gonorrhoeae were taken at both centres in Bristol but not in Taunton. All samples were q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
stored at 48C before being transported to the Public Health Labouratory in Bristol. At the laboratory slides were Gram-stained and read separately by two investigators (T.C. and A.T.), blinded to the patient's identity. The pattern of vaginal ¯ora was described using modi®ed Spiegel's criteria described by Hay 5, Grade I±Normal, predominantly lactobacillus morphotypes; Grade II±Intermediate, reduced lactobacillus mixed with other morphotypes; Grade III±Abnormal, few or no lactobacilli, greatly increased Gram variable morphotypes. Bacterial vaginosis was de®ned as Grade III ¯ora and these women were eligible for randomisation. A diagnosis was reached by consensus for discrepant results. The laboratory contacted the trial co-ordinator (T.C.) with the patient identi®cation number when a diagnosis of bacterial vaginosis was made. The patient was assigned the next available trial number at each centre. Because of time delays in transporting and reading slides some women had already undergone surgery before the diagnosis was available and were not randomised. Women with bacterial vaginosis were told their diagnosis on the day of the procedure and asked again if they wished to participate in the study. Women who gave written consent were given a two gram suppository per rectum by the surgeon while still under anaesthetic. They were given a questionnaire about the duration and extent of post-operative bleeding, offensive discharge, abdominal pain, post-operative visits to the general practitioner, post-operative antibiotic treatment, readmission to hospital, contraception and emotional response. Participants were asked to complete the questionnaire one month after the termination of pregnancy and to return it in a prepaid envelope. Consent was also requested to contact them by telephone or to liase with their general practitioner if the questionnaire was not returned within six weeks of the procedure. If consent to contact the general practitioner was not given hospital records were searched for details of readmission. The primary outcome was a clinical diagnosis of infection of the upper genital tract diagnosed within four weeks of suction termination of pregnancy. This was de®ned as prescription for antibiotics by the patient's general practitioner for any combination of two or more of the following symptoms: fever, lower abdominal pain, heavy vaginal bleeding, offensive or bloody vaginal discharge, or readmission to hospital with a clinical diagnosis of pelvic in¯ammatory disease. Readmission to hospital and self-reported symptoms were considered as secondary outcomes. Statistical analysis A power calculation was performed using the results of Larsson et al. 11. We estimated that we needed to enrol around 360 women overall where the incidence of upper
398 T. CROWLEY ET AL.
genital tract infection was 3.6% in the treatment arm and 12.2% in the placebo arm with power of 80% and two tailed alpha of 0.05. Clinical data were entered into a database using the Statistical Package for the Social Sciences (SPSS for Windows, Version 7.0). Intention-to-treat analysis of the primary outcome was performed, including women in the group to which they had been allocated. In the main analysis, women without follow up data were assumed not to have developed upper genital tract infection. Sensitivity analyses were performed for scenarios assuming that all women lost to follow up developed upper genital tract infection and including only women who received the allocated treatment. The effect of antibiotic prophylaxis was calculated as both the absolute difference in risk and relative risk (with 95% con®dence intervals). Statistical analyses were carried out using STATA (Version 6.0, STATA Corporation, Austin, Texas, USA). RESULTS Recruitment was slower than expected, and was termi-
nated before the target was reached due to lack of resources. The numbers of patients consenting to screening for bacterial vaginosis and randomisation, and their subsequent progress are summarised in Fig. 1. During the study period 1143 women agreed to be screened for bacterial vaginosis and 178 declined. The median age of women screened was 25 years (range 14 to 45 years). Thirty two Gram-stained slides could not be graded because of lack of material. The pattern of vaginal ¯ora was classi®ed as Grade I in 706/1111 (63.5%), Grade II in 79 (7.1%) and Grade III, bacterial vaginosis, in 326 (29.3%). C. trachomatis was detected in 47/1073 (4.4%) and N. gonorrhoeae in one of 572 women tested. C. trachomatis was detected more frequently in women with bacterial vaginosis 25/308 (8.1%) than those without (19/747 2.5%, relative risk 3.2, 95% con®dence intervals 1.9 to 5.7). Of the 326 women eligible for randomisation, results were received after surgery in 23 and 30 decided not to continue in the study. Thus, 273 women gave written consent to be randomised, 130 (47.6%) at Southmead, 91 at St Michael's, and 52 at Musgrove Park Hospitals. Characteristics of these women (Table 1) shows that
Fig. 1. Trial pro®le showing progress of patients through screening for bacterial vaginosis, randomisation and follow up.
q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
ANTIBIOTIC PROPHYLAXIS TO PREVENT GTI 399 Table 1. Characteristics of women with bacterial vaginosis at baseline, according to allocation to treatment and placebo groups. Values are given as n (%), unless otherwise shown. PID pelvic in¯ammatory disease. Metronidazole (n 142) Hospital Southmead St Michael's Musgrove Park Age in years Median [range] Ethnic group White Other Not known Contraception Hormonal Non-hormonal Barrier (condome or diaphragm) Emergency contraception None History of PID Chlamydia trachomatis detected
Placebo (n 131)
68 (48) 45 (32) 29 (20)
62 (47) 46 (35) 23 (18)
24 [16±43]
25 [16±44]
128 (90) 7 (5) 7 (5)
114 (87) 7 (5) 10 (8)
24 (17) 6 (4) 60 (42) 3 (2) 49 (35) 9 (6) 10 (7)
28 (21) 5 (4) 56 (43) 6 (5) 36 (28) 5 (4) 11 (8)
randomisation resulted in balanced groups. Sixty three women (31 assigned to metronidazole and 32 to placebo) did not receive the allocated suppositories because they continued with the pregnancy, the suppositories were lost or the surgeon forgot to put them in (Fig. 1). We obtained clinical outcome data on 202/273 (74%) women (Fig. 1). Information was available from questionnaires for 142 women, by telephone for 23 and from general practitioners for 35 women. A computer search showed that two of the remaining 75 women had been readmitted to hospital in the four weeks post-operatively. Among those with follow up data 49 women visited their general practitioner during the four week follow up period with symptoms of fever, abnormal vaginal bleeding, lower abdominal pain, offensive or bloody vaginal discharge and 29 of these were prescribed antibiotics.
Eleven of these women were also readmitted to hospital and a further four women were readmitted without having seen their general practitioner. Thus, 33 women were classi®ed as having upper genital tract infection requiring treatment. Intention-to-treat analysis, assuming that women lost to follow up did not develop the outcome, showed that post-operative upper genital tract infection developed in 12/142 (8.5%) women allocated to metronidazole and 21/ 131 (16.0%) women randomised to placebo, a difference of 7.6% (95% con®dence intervals ±15.4 to 10.2%; relative risk 0.53, 0.27 to 1.03, P 0.055). Similar results were obtained in analysis restricted to women who were followed up (risk difference ±9.6%, 95% con®dence intervals ±19.7 to 10.6%, P 0.066) and in on-treatment analysis among women with follow up who actually received the allocated treatment (Table 2). In a worst case scenario, which assumed that all those lost to follow up developed the outcome the risk of upper genital tract infection was 4.5% (±16.1 to 17.0%, P 0.44) lower in those receiving metronidazole (Table 2). Information about post-operative symptoms was available for 165 women. Fever was experienced by 45/160 (28%) women, and abnormal or irregular vaginal bleeding by 42/157 (27%) and 22/150 (15%), respectively. The median duration of post-operative bleeding was 11 days (range 0±35). Post-operative pain was reported by 120 (75%) women. On a numerical scale where one represents no pain and a score greater than seven represents severe or very severe pain 45/137 (33%) graded their pain as severe or very severe. There were no differences in the frequency of symptoms between women randomised to receive metronidazole or placebo (Table 3). Only 31/160 (19%) women described themselves as having experienced a physical problem after the termination of pregnancy whereas 84/157 (54%) said they had been depressed afterwards. Seventeen women were ®tted with intrauterine contraceptive devices immediately post-operatively, twelve in
Table 2. Incidents of primary endpoint; post-abortal upper genital tracts infection in women with bactewrial vaginosis, according to allocation to metronidazole or placebo.
Intention-to-treat analysis Treated for upper genital tract infection a On-treatment analysis Treated for upper genital tract infection b Worst-case scenario Treated for upper genital tract infection c a b c
Metronidazole n (%)
Placebo n (%)
RR (95% CI)
RR (95%CI)
12/142 (8.5)
21/131 (16.0)
±7.6 (±15.4 to 0.2)
0.53 (0.27 to 1.03)
11/100 (11.0)
18/90 (20.0)
±9 (±19.3 to 1.3)
0.55 (0.27 to 1.10)
51/142 (35.9)
53/131 (40.5)
±4.5 (±16.1 to 7)
Assumes that all those lost to follow up do not develop outcome. Includes only women who received allocated treatment and who had follow up data available. Assumes that all those lost to follow up develop outcome.
q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
0.89 (0.66 to 1.20)
400 T. CROWLEY ET AL. Table 3. Incidence of secondary endpoints: self-reported post-post-operative symptoms in women with bacterial vaginosis, according to allocation to metronidazole or placebo. Symptom Fever Vaginal bleeding Very heavy or with blood clots Abnormal bleeding Irregular bleeding Vaginal discharge Offensive odour Irritant Blood stained Abdominal pain at 4±6 weeks Any physical problems Depressed mood IUCD inserted at time of operation
Metronidazole n (%)
Placebo n (%)
RR (95% CI)
RR (95%CI)
24/88 (27.3)
21/72 (29.2)
±1.9 (±15.9 to 12.1)
0.94 (0.54 to 1.54)
30/85 (35.3) 22/88 (25.0) 10/84 (11.9)
23/62 (37.1) 20/69 (29.0) 12/66 (18.2)
±1.7 (±17 to 14) ±4 (±10 to 18) ±6.3 (±5.3 to 17)
0.95 (0.62 to 1.47) 0.86 (0.51 to 1.45) 0.65 (0.3 to 1.42)
10/88 (11.4) 10/88 (11.4) 6/88 (6.8) 8.79 (10.1) 19/88 (21.6) 44/86 (51.2) 12/90 (13.3)
11/72 (15.3) 4/72 (5.6) 11/72 (15.3) 10/65 (15.4) 12/72 (16.7) 40/71 (56.3) 5/74 (6.8)
±3.9 (±6.7 to 14.5) 5.8 (±2.7 to 14.3) ±8.5 (±1.3 to 18.3) ±5.3 (±5.7 to 16.3) 4.9 (±7.2 to 17) ±5.1 (±10.5 to 20.8) 6.5 (-2.5 to 15.6)
0.74 (0.33 to 1.65) 2.05 (0.67 to 6.25) 0.45 (0.17 to 1.15) 0.66 (0.28 to 1.57) 1.16 (0.6 to 2.24) 0.91 (0.68 to 1.21) 1.97 (0.73 to 5.35)
the metronidazole group and ®ve in the placebo group. These women were more likely to visit their general practitioner with abdominal pain or abnormal vaginal bleeding (10/17, 59%) than those without intrauterine devices (30/141 21%) and were 3.2 (1.6±6.6) times more likely to receive antibiotics. C. trachomatis was not detected in any of these women. C. trachomatis was detected in 3/40 (8%) women under 20 and 14/132 (11%) women under 25 years. All women with chlamydia were treated pre-operatively and none developed upper genital tract infection. Of 142 women providing information about sexual activity since the termination of pregnancy 98 (69%) had had sexual intercourse, including nine with chlamydia (only one of whom used condoms consistently). DISCUSSION Intention-to-treat analysis of this randomised doubleblind placebo-controlled study among women with bacterial vaginosis found that the incidence of postoperative upper genital tract infection requiring treatment was 7.6% (±15.4 to 10.2%, P 0.055) lower in those allocated to receive a two g metronidazole suppository at the time of suction termination of pregnancy compared with those assigned placebo. The results therefore include the possibility that this ®nding was due to chance. The effect of prophylaxis was similar when the analysis was restricted to women receiving the allocated treatment and with complete follow up. There was no difference in the risk of readmission to hospital post-operatively and the frequencies of self reported symptoms in the two groups were similar. The prevalence of bacterial vaginosis in this study (29%) was similar to that reported by Blackwell et al. 4. The power of the study to detect an effect of the size
observed with con®dence was limited for two reasons. First, the achieved sample size was lower than planned and a quarter of women enrolled could not be followed up. The high proportion of participants lost to follow up (27%) was disappointing but was similar in each group. Follow up by questionnaire was thought to be the most acceptable method since post-operative visits following termination of pregnancy are not routine in the UK. Many of those who did not return a questionnaire also declined to give permission for their general practitioner or themselves to be contacted by telephone, perhaps because of embarrassment or to preserve con®dentiality. In the only other UK based trial, in which follow up data were collected at a post-operative, visit 37% of women did not attend but information was obtained from general practitioners 15. Secondly, the effect size was lower than assumed in the power calculation. Our study was analysed according to intention-to-treat and this may explain why the observed treatment effect was more modest than that found by Larsson et al. in which a quarter of patients were excluded from analysis after randomisation 11. The case de®nition for pelvic infection also differed between the two trials. Post-abortal upper genital tract infection is dif®cult to de®ne because of the poor speci®city of the symptoms associated with pelvic infection and their overlap with symptoms that follow surgical termination of pregnancy 16,17. This was shown by the frequent reporting of heavy vaginal bleeding, fever and pain. Our de®nition, based on the prescription of antibiotics for symptoms may overestimate the incidence of infection but it takes into account the fact that most cases are managed in primary care without access to blood tests or laparoscopy. Even in trials using strict case de®nitions the incidence of post-abortal infection has been found to vary from 5% 18 to 23% 19. Differences in the ef®cacy of the antibiotic regimen might also have contributed to the difference q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
ANTIBIOTIC PROPHYLAXIS TO PREVENT GTI 401
in effect size. We used a single prophylactic dose of metronidazole, compared with the prolonged treatment regimen used by Larsson et al. 11. The Royal College of Obstetricians and Gynaecologists recommend that either antibiotic prophylaxis with antibiotics active against both anaerobes and C. trachomatis or a screen and treat policy are acceptable alternatives 13. The suggested regimen, combining a single dose of metronidazole and seven day course of doxycycline 13 has not, however, been evaluated against placebo or any single regimen. Larsson et al. 11 studied the effect of metronidazole in women with bacterial vaginosis and four other randomised controlled trials included in the published metaanalysis used a nitroimidazole derivative in an unselected patient group 12. The pooled odds ratio for all six trials, including ours, is 0.50 (0.33±0.75) (P 0.57 for heterogeneity), suggesting that the bene®cial effect is similar in women with or without bacterial vaginosis. Randomised trials have shown no bene®t in treating partners of women with bacterial vaginosis so contact tracing is not indicated 20,21 and therefore universal antibiotic prophylaxis against anaerobes appears to be an appropriate strategy for the prevention of post-abortal pelvic infection. The role of C. trachomatis in pelvic in¯ammatory disease is well established and the reported incidence of post-abortal infection in women with genital chlamydia ranges from 10 to 63% 22,4. Bacterial vaginosis may facilitate the attachment and entry to the upper genital tract of C. trachomatis through the activity of enzymes such as mucinase and sialidase 23. The disadvantage, however, of prophylaxis without testing for chlamydia is that infected women do not have access to partner noti®cation procedures that would reduce the risk of re-infection. In this study over half the women for whom we had a sexual history had been sexually active with the same or a new partner since the termination and few used condoms. Blackwell et al. 4 found that the costs of hospital admission for post-abortal infection were more than double the cost of screening for C. trachomatis combined with prophylaxis against chlamydia and bacterial vaginosis. Universal prophylaxis against these infections has been reported to be more cost-effective than a screen and treat policy, although this evaluation did not include the costs of chlamydia-associated complications due to re-infection from untreated partners 15. Universal prophylaxis may therefore not prevent the late consequences of chlamydia-associated post-abortal infection. CONCLUSION This randomised double-blind placebo-controlled trial in women with bacterial vaginosis provides weak q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
evidence that metronidazole decreases the risk of upper genital tract infection after ®rst trimester suction termination of pregnancy. This trial, however, could not con®dently exclude a chance ®nding. One other published trial in this patient group found metronidazole to be bene®cial but was not analysed according to intention-to-treat 11. Meta-analysis suggests that metronidazole reduces the risk of infection in women with and without bacterial vaginosis. The effectiveness of antibiotic combinations and comparisons of universal prophylaxis with prophylaxis combined with screening for chlamydia should now be evaluated in large well-conducted randomised trials with long term outcome measures. Acknowledgements The authors would like to acknowledge all the help and cooperation from all the staff at the Pregnancy Advice Service, Central Health Clinic, Bristol, Southmead Hospital; and Bristol and Musgrove Park Hospital, Taunton, in particular Dr S. Bodard, Dr J. Yin Thu and D. Morgan. This study was presented at the conference STIs at the Millennium: Past, Present, and Future; Baltimore, May 3±7, 2000: abstract number 005. This study was supported by a grant of £15,000 from the Special Trustees for the United Bristol Hospitals. References 1. Easmon CSF, Hay PE, Ison CA. Bacterial vaginosis: a diagnostic approach. Genitourin Med 1992;68:134±138. 2. Eschenbach DA, Hillier S, Critchlow C, Stevens C, De Rouen T, Holmes KK. Diagnosis and clinical manifestations of bacterial vaginosis. Am J Obstet Gynecol 1988;158:819±828. 3. Hay PE, Taylor-Robinson D. De®ning bacterial vaginosis: to BV or not to BV, that is the question. Int J STD AIDS 1996;7:233±235. 4. Blackwell AL, Thomas PD, Wareham K, Emery SJ. Health gains from screening for infection of the lower genital tract in women attending for termination of pregnancy. Lancet 1993;342:206±210. 5. Hay PE, Lamont RF, Taylor-Robinson D, Morgan DJ, Ison C, Pearson J. Abnormal bacterial colonisation of the genital tract and subsequent preterm delivery and late miscarriage. BMJ 1994;308: 295±298. 6. Lamont RF, Taylor-Robinson D, Newman M, Wigglesworth J, Elder MG. Spontaneous early preterm labour associated with abnormal genital bacterial colonisation. Br J Obstet Gynaecol 1986;93:4± 10. 7. McGregor JA, French JI, Jones W, et al. Bacterial Vaginosis is associated with prematurity and vaginal ¯uid mucinase and sialidase: results of a controlled trial of topical clindamycin cream. Am J Obstet Gynecol 1994;170:1048±1059. 8. Larsson P-G, Platz-Christensen J-J, Fursom U, Pahlson C. Clue cells in predicting infections after abdominal hysterectomy. Obstet Gynecol 1991;77:450±452. 9. Soper DE, Bump RC, Hurt WG. Bacterial vaginosis and trichomonas vaginitis are risk factors for cuff cellulitis after abdominal hysterectomy. J Clin Microbiol 1990;163:1016±1023. 10. Sweet RL. Role of bacterial vaginosis in pelvic in¯ammatory disease. Clin Infect Dis 1995;20:S271±S275. 11. Larsson P-G, Platz-Christensen JJ, Thejls H, Forsum U, PaÈhlson C.
402 T. CROWLEY ET AL.
12. 13. 14. 15. 16. 17. 18.
Incidence of pelvic in¯ammatory disease after ®rst trimester legal abortion in women with bacterial vaginosis after treatment with metronidazole: a double blind randomized study. Am J Obstet Gynecol 1992;166:100±103. Sawaya GF, Grady D, Kerlikowske K, Grimes DA. Antibiotics at the time of induced abortion: the case for universal prophylaxis based on a meta analysis. Obstet Gynecol 1996;87:884±890. Templeton A, editor. The Prevention of Pelvic Infection. London: RCOG Press, 1996. Caul EO, Paul ID. Monoclonal antibody based ELISA for detecting Chlamydia trachomatis. Lancet 1985;1:279. Penney GC, Thompson M, Norman J, et al. A randomised comparison of strategies for reducing infective complications of induced abortion. Br J Obstet Gynaecol 1998;105:599±604. Jacobson L, Westrom L. Objectivized diagnosis of acute pelvic in¯ammatory disease. Am J Obstet Gynecol 1969;105:1088±1098. Bevan CD, Johal BJ, Mumtaz G, et al. Clinical, laparoscopic and microbiological ®ndings in acute salpingitis: report on a United Kingdom cohort. Br J Obstet Gynaecol 1995;102:407±414. Levallois P, Rioux JE. Prophylactic antibiotics for suction curettage
19.
20. 21. 22. 23.
abortion: results of a clinical controlled trial. Am J Obstet Gynecol 1988;158:100±105. Heisterberg L, Gnarpe H. Preventive lymecycline therapy in women with a history of pelvic in¯ammatory disease undergoing ®rst -trimester abortion: A clinical controlled trial. Eur J Obstet Gynecol Reprod Biol 1998;28:241±247. Vejtorp M, Bollerup AC, Vejtorp L, et al. Bacterial vaginosis: a double-blind randomised trial of the effect of treatment of the sexual partner. Br J Obstet Gynaecol 1988;95:920±926. Moi H, Erkkola R, Jerve F. Should male consorts of women with bacterial vaginosis be treated? Genitourin Med 1989;65:263±268 Accepted 31 July 2000. Osser S, Persson K. Postabortal infectious morbidity after antibiotic treatment of chlamydia-positive patients. Sex Transm Dis 1989;16:84± 87. Howe L, Wiggins R, Millar MR, Horner PJ, Caul®eld AP. Mucinase and sialidase activity of the vaginal micro¯ora: implications for the pathogenesis of preterm labour. Int J STD AIDS 1999;10:442±447.
Accepted 7 July 2000
q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
Antibiotic prophylaxis to prevent post-abortal upper genital tract infection in women with bacterial vaginosis: randomised controlled trial Tessa Crowley a,*, Nicola Low b, Andrew Turner c, Ian Harvey b, Ken Bidgood d, Patrick Horner a Objective To determine the prevalence of bacterial vaginosis in women undergoing ®rst trimester suction termination of pregnancy and to evaluate the ef®cacy of metronidazole in reducing the risk of post abortal pelvic infection in women with bacterial vaginosis. Design Randomised double-blind placebo-controlled trial. Setting Two teaching hospitals and one district general hospital. Sample Two hundred and seventy-three women with bacterial vaginosis undergoing termination of pregnancy. Methods Women with bacterial vaginosis, diagnosed using modi®ed Spiegel's criteria, were individually randomised to receive either a 2 g metronidazole suppository or identical placebo per-operatively. Participants, doctors and investigators were blinded to treatment allocation. Participants were asked to complete a questionnaire about post-operative symptoms, visits to the general practitioner, antibiotic treatment, readmission to hospital, contraception and emotional response after one month. Results The prevalence of bacterial vaginosis was 29.3% (326/1111). Intention-to-treat analysis showed that post-operative upper genital tract infection developed in 12/142 (8.5%) women allocated to metronidazole and 21/131 (16.0%) women randomised to placebo, a difference of 7.6% (95% con®dence intervals ±15.4 to 10.2%; relative risk 0.53, 0.27 to 1.03, P 0.055). The effect of prophylaxis was similar when the analysis was restricted to women receiving the allocated treatment and with complete follow up. There was no difference in the risk of readmission to hospital and the frequencies of self reported symptoms in the two groups were similar. Conclusion This randomised placebo-controlled trial among women with bacterial vaginosis provides weak evidence that metronidazole decreases the risk of upper genital tract infection after ®rst trimester suction termination of pregnancy but a chance ®nding could not con®dently be excluded. Large well-conducted randomised trials with long term outcome measures are now needed to determine the most effective antibiotic combinations and strategy for prevention of post-abortal pelvic infection.
INTRODUCTION Bacterial vaginosis is characterised by a malodorous vaginal discharge in which the normally predominant lactobacilli are replaced by large numbers of a mixed group of bacteria 1,2. It is the most common cause of vaginal discharge in women of reproductive age 3 and has been reported to be present in 28% of women request-
a
Department of Sexual Health Medicine, Bristol Royal In®rmary, UK b MRC Health Services Research Collaboration, Department of Social Medicine, University of Bristol, UK c Genitourinary Infection Reference Laboratory, Public Health Laboratory Service, Bristol Royal In®rmary, UK d Department of Obstetrics and Gynaecology, Musgrove Park Hospital, Taunton, UK * Correspondence: Dr T. Crowley, Milne Sexual Health Centre, Bristol Royal In®rmary, Lower Maudlin Street, Bristol BS2 8HW, UK. q RCOG 2001 British Journal of Obstetrics and Gynaecology PII: S03 06-5456(00)0009 1-7
ing termination of pregnancy 4. Bacterial vaginosis is associated with adverse obstetric outcomes such as premature rupture of membranes, late miscarriage and post-caesarean endometritis 5±7, and with genital tract infections following gynaecological operations, including surgical termination of pregnancy 8±12. There is no agreed policy in the UK about the prevention of post-abortal upper genital tract infection. In 1996 a study group of the Royal College of Obstetricians and Gynaecologists recommended that women undergoing termination of pregnancy should either be given prophylactic antibiotics active against bacterial vaginosis and Chlamydia trachomatis or screened and treated if infections are found (screen and treat policy) 13. These recommendations used the ®ndings of a meta-analysis of 12 randomised double-blind and placebo-controlled trials 12. Five trials used a nitroimidazole drug but only one included women with proven bacterial vaginosis, and reported a threefold reduction in the incidence of postabortal infection among women allocated a ten-day course of metronidazole 11. www.bjog-elsevier.com
ANTIBIOTIC PROPHYLAXIS TO PREVENT GTI 397
The aims of this study were to determine the prevalence of bacterial vaginosis in women undergoing ®rst trimester suction termination of pregnancy and to evaluate the ef®cacy of a single per-operative 2 g dose of metronidazole per rectum in reducing the incidence of clinically diagnosed upper genital tract infection in the ®rst four post-operative weeks. METHODS The study was an individually randomised doubleblind, placebo-controlled trial conducted at three hospitals in the South West of England, Pregnancy Advisory Service Clinic, St. Michael's Hospital, Bristol; Southmead Hospital, Bristol; and Musgrove Park Hospital, Taunton. An unrestricted randomisation list allocating active treatment or placebo was prepared using random number tables by one investigator (I.H.) who was not involved in the assignment process. The central pharmacy prepared suppositories containing two grams of metronidazole and placebos that were identical in appearance and texture. These were sealed in plastic bags numbered sequentially according to the randomisation list. Each bag was placed in a large brown opaque envelope with a correspondingly numbered study consent form, information sheet, questionnaire and stamped addressed envelope. Packages were stored in a fridge near the operating theatre at each centre. The randomisation list was stored in the central pharmacy and the code was broken only after all data had been entered. Patients, surgeons, general practitioners and the investigator collecting follow up data (T.C.) were all unaware of treatment allocation. Ethical approval for the trial was obtained at each centre. Because women requesting termination of pregnancy might be under emotional stress consent for enrolment into the prevalence study and for randomisation were carried out in separate stages to allow women time to consider their decision to participate. All women requesting termination of pregnancy between October 1996 and December 1998 were eligible for entry into the study. The study was explained and demographic and clinical data on women who wished to participate were recorded on a standard form. During vaginal speculum examination a high vaginal swab was taken, rolled onto a glass slide and air-dried. The swab was placed in charcoal medium to be cultured for Candida albicans and examined for Trichomonas vaginalis. Endocervical and urethral swabs were taken for detection of C. trachomatis using an ampli®ed enzyme immunoassay (IDEIA, Dako Diagnostics Ltd.). All reactive specimens were con®rmed as positive by direct immuno¯uorescence 14. Endocervical and urethral swabs for Neisseria gonorrhoeae were taken at both centres in Bristol but not in Taunton. All samples were q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
stored at 48C before being transported to the Public Health Labouratory in Bristol. At the laboratory slides were Gram-stained and read separately by two investigators (T.C. and A.T.), blinded to the patient's identity. The pattern of vaginal ¯ora was described using modi®ed Spiegel's criteria described by Hay 5, Grade I±Normal, predominantly lactobacillus morphotypes; Grade II±Intermediate, reduced lactobacillus mixed with other morphotypes; Grade III±Abnormal, few or no lactobacilli, greatly increased Gram variable morphotypes. Bacterial vaginosis was de®ned as Grade III ¯ora and these women were eligible for randomisation. A diagnosis was reached by consensus for discrepant results. The laboratory contacted the trial co-ordinator (T.C.) with the patient identi®cation number when a diagnosis of bacterial vaginosis was made. The patient was assigned the next available trial number at each centre. Because of time delays in transporting and reading slides some women had already undergone surgery before the diagnosis was available and were not randomised. Women with bacterial vaginosis were told their diagnosis on the day of the procedure and asked again if they wished to participate in the study. Women who gave written consent were given a two gram suppository per rectum by the surgeon while still under anaesthetic. They were given a questionnaire about the duration and extent of post-operative bleeding, offensive discharge, abdominal pain, post-operative visits to the general practitioner, post-operative antibiotic treatment, readmission to hospital, contraception and emotional response. Participants were asked to complete the questionnaire one month after the termination of pregnancy and to return it in a prepaid envelope. Consent was also requested to contact them by telephone or to liase with their general practitioner if the questionnaire was not returned within six weeks of the procedure. If consent to contact the general practitioner was not given hospital records were searched for details of readmission. The primary outcome was a clinical diagnosis of infection of the upper genital tract diagnosed within four weeks of suction termination of pregnancy. This was de®ned as prescription for antibiotics by the patient's general practitioner for any combination of two or more of the following symptoms: fever, lower abdominal pain, heavy vaginal bleeding, offensive or bloody vaginal discharge, or readmission to hospital with a clinical diagnosis of pelvic in¯ammatory disease. Readmission to hospital and self-reported symptoms were considered as secondary outcomes. Statistical analysis A power calculation was performed using the results of Larsson et al. 11. We estimated that we needed to enrol around 360 women overall where the incidence of upper
398 T. CROWLEY ET AL.
genital tract infection was 3.6% in the treatment arm and 12.2% in the placebo arm with power of 80% and two tailed alpha of 0.05. Clinical data were entered into a database using the Statistical Package for the Social Sciences (SPSS for Windows, Version 7.0). Intention-to-treat analysis of the primary outcome was performed, including women in the group to which they had been allocated. In the main analysis, women without follow up data were assumed not to have developed upper genital tract infection. Sensitivity analyses were performed for scenarios assuming that all women lost to follow up developed upper genital tract infection and including only women who received the allocated treatment. The effect of antibiotic prophylaxis was calculated as both the absolute difference in risk and relative risk (with 95% con®dence intervals). Statistical analyses were carried out using STATA (Version 6.0, STATA Corporation, Austin, Texas, USA). RESULTS Recruitment was slower than expected, and was termi-
nated before the target was reached due to lack of resources. The numbers of patients consenting to screening for bacterial vaginosis and randomisation, and their subsequent progress are summarised in Fig. 1. During the study period 1143 women agreed to be screened for bacterial vaginosis and 178 declined. The median age of women screened was 25 years (range 14 to 45 years). Thirty two Gram-stained slides could not be graded because of lack of material. The pattern of vaginal ¯ora was classi®ed as Grade I in 706/1111 (63.5%), Grade II in 79 (7.1%) and Grade III, bacterial vaginosis, in 326 (29.3%). C. trachomatis was detected in 47/1073 (4.4%) and N. gonorrhoeae in one of 572 women tested. C. trachomatis was detected more frequently in women with bacterial vaginosis 25/308 (8.1%) than those without (19/747 2.5%, relative risk 3.2, 95% con®dence intervals 1.9 to 5.7). Of the 326 women eligible for randomisation, results were received after surgery in 23 and 30 decided not to continue in the study. Thus, 273 women gave written consent to be randomised, 130 (47.6%) at Southmead, 91 at St Michael's, and 52 at Musgrove Park Hospitals. Characteristics of these women (Table 1) shows that
Fig. 1. Trial pro®le showing progress of patients through screening for bacterial vaginosis, randomisation and follow up.
q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
ANTIBIOTIC PROPHYLAXIS TO PREVENT GTI 399 Table 1. Characteristics of women with bacterial vaginosis at baseline, according to allocation to treatment and placebo groups. Values are given as n (%), unless otherwise shown. PID pelvic in¯ammatory disease. Metronidazole (n 142) Hospital Southmead St Michael's Musgrove Park Age in years Median [range] Ethnic group White Other Not known Contraception Hormonal Non-hormonal Barrier (condome or diaphragm) Emergency contraception None History of PID Chlamydia trachomatis detected
Placebo (n 131)
68 (48) 45 (32) 29 (20)
62 (47) 46 (35) 23 (18)
24 [16±43]
25 [16±44]
128 (90) 7 (5) 7 (5)
114 (87) 7 (5) 10 (8)
24 (17) 6 (4) 60 (42) 3 (2) 49 (35) 9 (6) 10 (7)
28 (21) 5 (4) 56 (43) 6 (5) 36 (28) 5 (4) 11 (8)
randomisation resulted in balanced groups. Sixty three women (31 assigned to metronidazole and 32 to placebo) did not receive the allocated suppositories because they continued with the pregnancy, the suppositories were lost or the surgeon forgot to put them in (Fig. 1). We obtained clinical outcome data on 202/273 (74%) women (Fig. 1). Information was available from questionnaires for 142 women, by telephone for 23 and from general practitioners for 35 women. A computer search showed that two of the remaining 75 women had been readmitted to hospital in the four weeks post-operatively. Among those with follow up data 49 women visited their general practitioner during the four week follow up period with symptoms of fever, abnormal vaginal bleeding, lower abdominal pain, offensive or bloody vaginal discharge and 29 of these were prescribed antibiotics.
Eleven of these women were also readmitted to hospital and a further four women were readmitted without having seen their general practitioner. Thus, 33 women were classi®ed as having upper genital tract infection requiring treatment. Intention-to-treat analysis, assuming that women lost to follow up did not develop the outcome, showed that post-operative upper genital tract infection developed in 12/142 (8.5%) women allocated to metronidazole and 21/ 131 (16.0%) women randomised to placebo, a difference of 7.6% (95% con®dence intervals ±15.4 to 10.2%; relative risk 0.53, 0.27 to 1.03, P 0.055). Similar results were obtained in analysis restricted to women who were followed up (risk difference ±9.6%, 95% con®dence intervals ±19.7 to 10.6%, P 0.066) and in on-treatment analysis among women with follow up who actually received the allocated treatment (Table 2). In a worst case scenario, which assumed that all those lost to follow up developed the outcome the risk of upper genital tract infection was 4.5% (±16.1 to 17.0%, P 0.44) lower in those receiving metronidazole (Table 2). Information about post-operative symptoms was available for 165 women. Fever was experienced by 45/160 (28%) women, and abnormal or irregular vaginal bleeding by 42/157 (27%) and 22/150 (15%), respectively. The median duration of post-operative bleeding was 11 days (range 0±35). Post-operative pain was reported by 120 (75%) women. On a numerical scale where one represents no pain and a score greater than seven represents severe or very severe pain 45/137 (33%) graded their pain as severe or very severe. There were no differences in the frequency of symptoms between women randomised to receive metronidazole or placebo (Table 3). Only 31/160 (19%) women described themselves as having experienced a physical problem after the termination of pregnancy whereas 84/157 (54%) said they had been depressed afterwards. Seventeen women were ®tted with intrauterine contraceptive devices immediately post-operatively, twelve in
Table 2. Incidents of primary endpoint; post-abortal upper genital tracts infection in women with bactewrial vaginosis, according to allocation to metronidazole or placebo.
Intention-to-treat analysis Treated for upper genital tract infection a On-treatment analysis Treated for upper genital tract infection b Worst-case scenario Treated for upper genital tract infection c a b c
Metronidazole n (%)
Placebo n (%)
RR (95% CI)
RR (95%CI)
12/142 (8.5)
21/131 (16.0)
±7.6 (±15.4 to 0.2)
0.53 (0.27 to 1.03)
11/100 (11.0)
18/90 (20.0)
±9 (±19.3 to 1.3)
0.55 (0.27 to 1.10)
51/142 (35.9)
53/131 (40.5)
±4.5 (±16.1 to 7)
Assumes that all those lost to follow up do not develop outcome. Includes only women who received allocated treatment and who had follow up data available. Assumes that all those lost to follow up develop outcome.
q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
0.89 (0.66 to 1.20)
400 T. CROWLEY ET AL. Table 3. Incidence of secondary endpoints: self-reported post-post-operative symptoms in women with bacterial vaginosis, according to allocation to metronidazole or placebo. Symptom Fever Vaginal bleeding Very heavy or with blood clots Abnormal bleeding Irregular bleeding Vaginal discharge Offensive odour Irritant Blood stained Abdominal pain at 4±6 weeks Any physical problems Depressed mood IUCD inserted at time of operation
Metronidazole n (%)
Placebo n (%)
RR (95% CI)
RR (95%CI)
24/88 (27.3)
21/72 (29.2)
±1.9 (±15.9 to 12.1)
0.94 (0.54 to 1.54)
30/85 (35.3) 22/88 (25.0) 10/84 (11.9)
23/62 (37.1) 20/69 (29.0) 12/66 (18.2)
±1.7 (±17 to 14) ±4 (±10 to 18) ±6.3 (±5.3 to 17)
0.95 (0.62 to 1.47) 0.86 (0.51 to 1.45) 0.65 (0.3 to 1.42)
10/88 (11.4) 10/88 (11.4) 6/88 (6.8) 8.79 (10.1) 19/88 (21.6) 44/86 (51.2) 12/90 (13.3)
11/72 (15.3) 4/72 (5.6) 11/72 (15.3) 10/65 (15.4) 12/72 (16.7) 40/71 (56.3) 5/74 (6.8)
±3.9 (±6.7 to 14.5) 5.8 (±2.7 to 14.3) ±8.5 (±1.3 to 18.3) ±5.3 (±5.7 to 16.3) 4.9 (±7.2 to 17) ±5.1 (±10.5 to 20.8) 6.5 (-2.5 to 15.6)
0.74 (0.33 to 1.65) 2.05 (0.67 to 6.25) 0.45 (0.17 to 1.15) 0.66 (0.28 to 1.57) 1.16 (0.6 to 2.24) 0.91 (0.68 to 1.21) 1.97 (0.73 to 5.35)
the metronidazole group and ®ve in the placebo group. These women were more likely to visit their general practitioner with abdominal pain or abnormal vaginal bleeding (10/17, 59%) than those without intrauterine devices (30/141 21%) and were 3.2 (1.6±6.6) times more likely to receive antibiotics. C. trachomatis was not detected in any of these women. C. trachomatis was detected in 3/40 (8%) women under 20 and 14/132 (11%) women under 25 years. All women with chlamydia were treated pre-operatively and none developed upper genital tract infection. Of 142 women providing information about sexual activity since the termination of pregnancy 98 (69%) had had sexual intercourse, including nine with chlamydia (only one of whom used condoms consistently). DISCUSSION Intention-to-treat analysis of this randomised doubleblind placebo-controlled study among women with bacterial vaginosis found that the incidence of postoperative upper genital tract infection requiring treatment was 7.6% (±15.4 to 10.2%, P 0.055) lower in those allocated to receive a two g metronidazole suppository at the time of suction termination of pregnancy compared with those assigned placebo. The results therefore include the possibility that this ®nding was due to chance. The effect of prophylaxis was similar when the analysis was restricted to women receiving the allocated treatment and with complete follow up. There was no difference in the risk of readmission to hospital post-operatively and the frequencies of self reported symptoms in the two groups were similar. The prevalence of bacterial vaginosis in this study (29%) was similar to that reported by Blackwell et al. 4. The power of the study to detect an effect of the size
observed with con®dence was limited for two reasons. First, the achieved sample size was lower than planned and a quarter of women enrolled could not be followed up. The high proportion of participants lost to follow up (27%) was disappointing but was similar in each group. Follow up by questionnaire was thought to be the most acceptable method since post-operative visits following termination of pregnancy are not routine in the UK. Many of those who did not return a questionnaire also declined to give permission for their general practitioner or themselves to be contacted by telephone, perhaps because of embarrassment or to preserve con®dentiality. In the only other UK based trial, in which follow up data were collected at a post-operative, visit 37% of women did not attend but information was obtained from general practitioners 15. Secondly, the effect size was lower than assumed in the power calculation. Our study was analysed according to intention-to-treat and this may explain why the observed treatment effect was more modest than that found by Larsson et al. in which a quarter of patients were excluded from analysis after randomisation 11. The case de®nition for pelvic infection also differed between the two trials. Post-abortal upper genital tract infection is dif®cult to de®ne because of the poor speci®city of the symptoms associated with pelvic infection and their overlap with symptoms that follow surgical termination of pregnancy 16,17. This was shown by the frequent reporting of heavy vaginal bleeding, fever and pain. Our de®nition, based on the prescription of antibiotics for symptoms may overestimate the incidence of infection but it takes into account the fact that most cases are managed in primary care without access to blood tests or laparoscopy. Even in trials using strict case de®nitions the incidence of post-abortal infection has been found to vary from 5% 18 to 23% 19. Differences in the ef®cacy of the antibiotic regimen might also have contributed to the difference q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
ANTIBIOTIC PROPHYLAXIS TO PREVENT GTI 401
in effect size. We used a single prophylactic dose of metronidazole, compared with the prolonged treatment regimen used by Larsson et al. 11. The Royal College of Obstetricians and Gynaecologists recommend that either antibiotic prophylaxis with antibiotics active against both anaerobes and C. trachomatis or a screen and treat policy are acceptable alternatives 13. The suggested regimen, combining a single dose of metronidazole and seven day course of doxycycline 13 has not, however, been evaluated against placebo or any single regimen. Larsson et al. 11 studied the effect of metronidazole in women with bacterial vaginosis and four other randomised controlled trials included in the published metaanalysis used a nitroimidazole derivative in an unselected patient group 12. The pooled odds ratio for all six trials, including ours, is 0.50 (0.33±0.75) (P 0.57 for heterogeneity), suggesting that the bene®cial effect is similar in women with or without bacterial vaginosis. Randomised trials have shown no bene®t in treating partners of women with bacterial vaginosis so contact tracing is not indicated 20,21 and therefore universal antibiotic prophylaxis against anaerobes appears to be an appropriate strategy for the prevention of post-abortal pelvic infection. The role of C. trachomatis in pelvic in¯ammatory disease is well established and the reported incidence of post-abortal infection in women with genital chlamydia ranges from 10 to 63% 22,4. Bacterial vaginosis may facilitate the attachment and entry to the upper genital tract of C. trachomatis through the activity of enzymes such as mucinase and sialidase 23. The disadvantage, however, of prophylaxis without testing for chlamydia is that infected women do not have access to partner noti®cation procedures that would reduce the risk of re-infection. In this study over half the women for whom we had a sexual history had been sexually active with the same or a new partner since the termination and few used condoms. Blackwell et al. 4 found that the costs of hospital admission for post-abortal infection were more than double the cost of screening for C. trachomatis combined with prophylaxis against chlamydia and bacterial vaginosis. Universal prophylaxis against these infections has been reported to be more cost-effective than a screen and treat policy, although this evaluation did not include the costs of chlamydia-associated complications due to re-infection from untreated partners 15. Universal prophylaxis may therefore not prevent the late consequences of chlamydia-associated post-abortal infection. CONCLUSION This randomised double-blind placebo-controlled trial in women with bacterial vaginosis provides weak q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402
evidence that metronidazole decreases the risk of upper genital tract infection after ®rst trimester suction termination of pregnancy. This trial, however, could not con®dently exclude a chance ®nding. One other published trial in this patient group found metronidazole to be bene®cial but was not analysed according to intention-to-treat 11. Meta-analysis suggests that metronidazole reduces the risk of infection in women with and without bacterial vaginosis. The effectiveness of antibiotic combinations and comparisons of universal prophylaxis with prophylaxis combined with screening for chlamydia should now be evaluated in large well-conducted randomised trials with long term outcome measures. Acknowledgements The authors would like to acknowledge all the help and cooperation from all the staff at the Pregnancy Advice Service, Central Health Clinic, Bristol, Southmead Hospital; and Bristol and Musgrove Park Hospital, Taunton, in particular Dr S. Bodard, Dr J. Yin Thu and D. Morgan. This study was presented at the conference STIs at the Millennium: Past, Present, and Future; Baltimore, May 3±7, 2000: abstract number 005. This study was supported by a grant of £15,000 from the Special Trustees for the United Bristol Hospitals. References 1. Easmon CSF, Hay PE, Ison CA. Bacterial vaginosis: a diagnostic approach. Genitourin Med 1992;68:134±138. 2. Eschenbach DA, Hillier S, Critchlow C, Stevens C, De Rouen T, Holmes KK. Diagnosis and clinical manifestations of bacterial vaginosis. Am J Obstet Gynecol 1988;158:819±828. 3. Hay PE, Taylor-Robinson D. De®ning bacterial vaginosis: to BV or not to BV, that is the question. Int J STD AIDS 1996;7:233±235. 4. Blackwell AL, Thomas PD, Wareham K, Emery SJ. Health gains from screening for infection of the lower genital tract in women attending for termination of pregnancy. Lancet 1993;342:206±210. 5. Hay PE, Lamont RF, Taylor-Robinson D, Morgan DJ, Ison C, Pearson J. Abnormal bacterial colonisation of the genital tract and subsequent preterm delivery and late miscarriage. BMJ 1994;308: 295±298. 6. Lamont RF, Taylor-Robinson D, Newman M, Wigglesworth J, Elder MG. Spontaneous early preterm labour associated with abnormal genital bacterial colonisation. Br J Obstet Gynaecol 1986;93:4± 10. 7. McGregor JA, French JI, Jones W, et al. Bacterial Vaginosis is associated with prematurity and vaginal ¯uid mucinase and sialidase: results of a controlled trial of topical clindamycin cream. Am J Obstet Gynecol 1994;170:1048±1059. 8. Larsson P-G, Platz-Christensen J-J, Fursom U, Pahlson C. Clue cells in predicting infections after abdominal hysterectomy. Obstet Gynecol 1991;77:450±452. 9. Soper DE, Bump RC, Hurt WG. Bacterial vaginosis and trichomonas vaginitis are risk factors for cuff cellulitis after abdominal hysterectomy. J Clin Microbiol 1990;163:1016±1023. 10. Sweet RL. Role of bacterial vaginosis in pelvic in¯ammatory disease. Clin Infect Dis 1995;20:S271±S275. 11. Larsson P-G, Platz-Christensen JJ, Thejls H, Forsum U, PaÈhlson C.
402 T. CROWLEY ET AL.
12. 13. 14. 15. 16. 17. 18.
Incidence of pelvic in¯ammatory disease after ®rst trimester legal abortion in women with bacterial vaginosis after treatment with metronidazole: a double blind randomized study. Am J Obstet Gynecol 1992;166:100±103. Sawaya GF, Grady D, Kerlikowske K, Grimes DA. Antibiotics at the time of induced abortion: the case for universal prophylaxis based on a meta analysis. Obstet Gynecol 1996;87:884±890. Templeton A, editor. The Prevention of Pelvic Infection. London: RCOG Press, 1996. Caul EO, Paul ID. Monoclonal antibody based ELISA for detecting Chlamydia trachomatis. Lancet 1985;1:279. Penney GC, Thompson M, Norman J, et al. A randomised comparison of strategies for reducing infective complications of induced abortion. Br J Obstet Gynaecol 1998;105:599±604. Jacobson L, Westrom L. Objectivized diagnosis of acute pelvic in¯ammatory disease. Am J Obstet Gynecol 1969;105:1088±1098. Bevan CD, Johal BJ, Mumtaz G, et al. Clinical, laparoscopic and microbiological ®ndings in acute salpingitis: report on a United Kingdom cohort. Br J Obstet Gynaecol 1995;102:407±414. Levallois P, Rioux JE. Prophylactic antibiotics for suction curettage
19.
20. 21. 22. 23.
abortion: results of a clinical controlled trial. Am J Obstet Gynecol 1988;158:100±105. Heisterberg L, Gnarpe H. Preventive lymecycline therapy in women with a history of pelvic in¯ammatory disease undergoing ®rst -trimester abortion: A clinical controlled trial. Eur J Obstet Gynecol Reprod Biol 1998;28:241±247. Vejtorp M, Bollerup AC, Vejtorp L, et al. Bacterial vaginosis: a double-blind randomised trial of the effect of treatment of the sexual partner. Br J Obstet Gynaecol 1988;95:920±926. Moi H, Erkkola R, Jerve F. Should male consorts of women with bacterial vaginosis be treated? Genitourin Med 1989;65:263±268 Accepted 31 July 2000. Osser S, Persson K. Postabortal infectious morbidity after antibiotic treatment of chlamydia-positive patients. Sex Transm Dis 1989;16:84± 87. Howe L, Wiggins R, Millar MR, Horner PJ, Caul®eld AP. Mucinase and sialidase activity of the vaginal micro¯ora: implications for the pathogenesis of preterm labour. Int J STD AIDS 1999;10:442±447.
Accepted 7 July 2000
q RCOG 2001 Br J Obstet Gynaecol 108, pp. 396±402