A Randomized Clinical Trial of Prophylaxis for Vacuum Abortion: 3 Versus 7 Days of Doxycycline

E. Steve Lichtenberg, MD, MPH, and Susan Shott, PhD

OBJECTIVE: To estimate prospectively whether reducing oral doxycycline prophylaxis from 7 to 3 days increases the incidence of endometritis after elective rst-trimester vacuum abortion. METHODS: We randomized 800 women requesting rsttrimester abortion to two study groups: 1) 100 mg of doxycycline administered orally twice a day for 7 days or 2) 100 mg of doxycycline orally twice a day for 3 days followed by an oral placebo twice daily for the last 4 days. Doxycycline was prescribed immediately after surgery. RESULTS: There were no statistically signicant differences in age, race, gravidity, parity, number of previous abortions, current gestational age, and history of previous pelvic infection, chlamydia, or intrauterine device use. Women in the doxycycline plus placebo group were more likely to have had a history of gonorrhea (3.3% versus 0.8%, P .42) or chlamydia (7.0% versus 4.3%, P .18). The 66.3% of enrollees returning for 2-week examinations were distributed similarly between study groups and were similarly compliant in self-reported pill taking (mean 97.5%). The study groups did not differ in the incidence of postoperative symptoms or examination ndings suggesting infection. One patient in the doxycycline-only group developed endometritis and was treated as an outpatient. CONCLUSION: Shortening oral doxycycline prophylaxis from 7 to 3 days had no adverse effect on the incidence of postabortion infection. (Obstet Gynecol 2003;101: 726 31. 2003 by The American College of Obstetricians and Gynecologists.)

lones. The overall protective effect was highest in women with a history of pelvic inammatory disease or a positive chlamydia culture. Citing this evidence, The American College of Obstetricians and Gynecologists (ACOG)2 and the Royal College of Obstetricians and Gynaecologists3 currently recommend the use of antibiotic prophylaxis at the time of induced abortion. ACOG makes no recommendation about specic regimens, but the Royal College cites the combination of 1 g of rectal metronidazole at the time of abortion plus 1 week of oral doxycycline, 100 mg twice daily for 7 days, beginning after surgery. Current practice among US abortion providers appears generally to reect these recommendations. A recent survey4 of practice patterns among member facilities of the National Abortion Federation, the largest professional association of North American abortion providers, performing a third of all rst-trimester abortions, found that 91% routinely prescribe prophylactic antibiotics, with doxycycline as the favored drug (81%, 191 of 236 responding sites). Relative to 1-day (15%), 3-day (16%), and 5-day (16%) courses, a 1-week regimen of doxycycline is most commonly chosen (40%, unpublished data). We devised this randomized trial to test whether reducing the duration of doxycycline oral prophylaxis from 7 days to 3 days would increase the incidence of postabortion infection. Shorter-course prophylaxis increases patient compliance, lessens side effects, and may encourage greater physician use. MATERIALS AND METHODS We recruited 1261 consecutive women requesting rsttrimester surgical abortion from November 1995 through May 1996 after approval by the institutional ethics committee at the Albany MedicalSurgical Center, a licensed ambulatory surgical center in Chicago, Illinois. To be included, consenting women had to be no more than 1307 weeks from last menstrual period, speak English, and live within a 50-mile radius. In addition, we excluded patients who were allergic to tetracyclines, breast-feeding, on current antibiotic therapy, febrile

Periabortion antibiotic prophylaxis has been shown to reduce postabortion infection in women undergoing elective rst-trimester vacuum abortion. In a meta-analysis comprising over 5000 women, Sawaya et al1 found an overall 42% decrease in infection after analyzing results from 12 randomized trials with treatment courses from 2 hours to 2 weeks in length that encompassed regimens from ve different antibiotic classespenicillins, tetracyclines, macrolides, imidazoles, and quinoFrom The Department of Obstetrics and Gynecology, Northwestern University Medical Center; and Biostatistics Unit, The Department of Obstetrics and Gynecology, Rush Medical College, Chicago, Illinois.

726

VOL. 101, NO. 4, APRIL 2003 2003 by The American College of Obstetricians and Gynecologists. Published by Elsevier.

0029-7844/03/$30.00 doi:10.1016/S0029-7844(03)00041-3

Table 1. Reasons for Patient Exclusions (n 461) No. of patients (%) Patient refused Non-English speaking Lives 50 miles away HMO member Medical indications (eg, breast-feeding, allergy to tetracycline, current antibiotic therapy, etc) Total
HMO health maintenance organization.

242 (52.5) 85 (18.4) 54 (11.7) 52 (11.3) 28 (6.1) 461 (100)

(higher than 37.5C orally), symptomatic for pelvic infection, or unable to swallow pills. Finally, we excluded patients whose care was being funded by a health maintenance organization because these patients were often instructed by the referring site to seek follow-up care with their primary physician. In all, 461 women were excluded under these criteria (Table 1). The remaining 800 women constituted the study population. After giving consent, study subjects underwent a directed physical examination, blood sampling for hematocrit and Rh type, and an ultrasonic examination if there was a size dates discrepancy or a clinical estimate of gestation less than 7 or greater than 12 weeks from last menstrual period. During the intake interview patients identied their race or ethnicity according to the categories Caucasian (white), Hispanic, African-American, or other. We did not take cervical cultures and gave no preoperative medication. Patients were randomized into two groups of 400 patients using a table of random numbers. Patients chose local cervical or general anesthesia. Vacuum abortion was performed by one of three highly experienced obstetrician gynecologists using similar surgical techniques. The aspirate was immediately examined to verify the presence of chorionic villi and adequate tissue mass. Every effort was made to insure double masking. Postoperatively, the recovery room charge nurse dispensed antibiotic packets in identical opaque brown envelopes by individual allocation. Identication numbers assigned at the time of enrollment determined these allocations. A study coordinator kept the randomization code; she also formulated the antibiotic packets but did not perform clinical services, data analysis, or interpretation of outcomes. We used a generic form of doxycycline and similar-appearing placebo tablets. Identicalappearing packets contained either 14 doxycycline tablets or six doxycycline tablets and eight placebos. Patients receiving placebos were carefully instructed to begin these after nishing all doxycycline tablets and were not aware that some tablets were chemically inactive. During their follow-up visit or phone call, patients

gave no indication that they had guessed or even attempted to guess which treatment group they were in. Surgeons were unaware of the treatment code and did not participate in the care of patients postoperatively. Patients were encouraged to return for their 2-week follow-up visit with a promise of three cycles of oral contraceptives at no charge or a small cash award. An existing staff of advanced practice nurses specially trained in abortion care conducted all follow-up visits. At this examination, patients lled out a standard form containing a checklist of common symptoms related to infection, and vital signs were taken. The practitioner reviewed the checklist and performed and recorded a directed abdominal and bimanual pelvic examination for each returning patient. Finally, we conducted a telephone survey of a random sample of nonreturnees to inquire about possible postoperative problems. Before beginning the study, a sample size of 400 women in each group was calculated to achieve 88% power at a 5% signicance level to detect an increase in infection rates from 2% to 6.5%,1 based on the 2 test of association. The nal sample sizes of 257 in the doxycycline-only group and 273 in the doxycycline plus placebo group insured at least 83% power for detecting an increase in infection rates from 2% to 7.5%, again based on the 2 test of association (nQuery Advisor 2.0; Statistical Solutions, Saugus, MA). Comparison of groups with respect to categoric variables was done with the 2 test of association or the Fisher exact test. The pooled variance t test was used to compare groups with respect to age, and the MannWhitney test was done to compare groups with respect to the remaining noncategoric demographic variables. The statistician (SS) was masked to treatment allocation; she conducted and interpreted analyses before the treatment code was revealed. A .05 signicance level was used for all statistical tests. No one-sided tests were done. RESULTS The allocation and randomization scheme for 1261 screened women is presented in Figure 1. The mean ages of women who qualied for the study (n 800) closely matched those who were excluded (n 461) (Table 2). Although a statistically signicant difference was found between qualied and excluded patients with respect to self-reported race, the racial percentages were similar in these two groups (Table 2). There was close similarity with respect to racial or ethnic composition between the doxycycline-only (n 257) and the doxycycline plus placebo (n 273) groups who returned for follow-up examinations (Table 3). There were no statistically signicant differences between women who returned for a

VOL. 101, NO. 4, APRIL 2003

Lichtenberg and Shott

Antibiotic Prophylaxis

727

Figure 1. Enrollment and randomization scheme for 1261 screened women.

Lichtenberg. Antibiotic Prophylaxis. Obstet Gynecol 2003.

follow-up visit and women who did not return with respect to age, gravidity, parity, and the percentages with induced and spontaneous abortions (Table 3). We compiled a risk prole for pelvic infection based on a history of any of the following: chlamydia, gonorrhea, other sexually transmitted diseases, pelvic infection, ectopic pregnancy, and use of an intrauterine contraceptive device. Although nonreturning women reported a higher incidence in most of these categories than women who returned for follow-up, none of these differences reached statistical signicance (Table 4). We also analyzed the infection risk prole according to treatment allocation (Table 4). Placebo-treated women had higher rates of all historical categories for pelvic risk, but only with a history of gonorrhea did the percentage difference reach statistical signicance. We dened compliance with pill taking as self-reported adherence to the twice-daily dose schedule for at least 6 of the 7 prescribed days. Among returnees, 97.7% and 97.4% of doxycycline-only and doxycycline plus placebo patients, respectively, were compliant. Clinical infection was dened as reported pelvic pain plus an on-site oral temperature greater than 37.5C and

either uterine, adnexal, or abdominal rebound tenderness. One patient met these criteria. She presented with reported pelvic pain and an oral temperature of 37.6C. On examination she had uterine but no adnexal or abdominal rebound tenderness. Her white cell count was 11,500 mm3 (normal range 3500 11,000 mm3), and she responded rapidly to a course of outpatient treatment with 150 mg of oral clindamycin. She had received and taken a 7-day course of doxycycline. To canvass the population of 270 nonreturnees, we conducted a random telephone survey of these patients. We succeeded in obtaining responses from 32 of these (12%), 17 and 15 of whom were in the doxycycline-only and doxycycline plus placebo groups, respectively. One patient in the former group reported transient cramping without fever, and one patient in the latter group, noted to have uterine broids, gave a history of fever and cramping that abated after 1 day. A compilation of patient-reported postoperative problems among returnees and telephone contacts (n 562) is presented in Table 5. There were 11 reaspirations for suspected retained tissue or clot, six (of 257, 2.3%) in the doxycycline-only group and ve (of 273, 1.8%) in the doxycycline plus placebo group. None of these patients met the criteria for a clinical infection, and all recovered without additional antibiotics. We found no differences in outcomes for patients receiving local cervical or general anesthesia. DISCUSSION We conducted a literature search of English and foreign language clinical abortion trials published between January 1967 and April 2002 using the MEDLINE database. We used the key words abortion, induced abortion, infection, antibiotics, and prophylaxis. None of the trials thus identied were comparisons of two prophylactic dosages, making this the rst study to compare the dose of a prophylactic antibiotic among patients undergoing induced abortion.

Table 2. Characteristics of Included and Excluded Patients (n 1261) Patient characteristics Age (y) Race White Hispanic Black Other
SD standard deviation. * Pooled variance t test. 2 test of association.

Included group (n 800) (mean SD or %) 26.6 6.4 48.8 25.2 19.9 6.1

Excluded group (n 461) (mean SD or %) 26.9 6.9 43.8 19.3 24.7 12.3

Included vs excluded P .42* .001

728

Lichtenberg and Shott

Antibiotic Prophylaxis

OBSTETRICS & GYNECOLOGY

Table 3. Characteristics of Women Returning for Follow-up Examinations (n 530) Doxycycline-only group (n 257) (mean SD or %) 26.6 6.4 47.8 26.5 18.7 7.0 3.1 2.0 0.9 1.0 1.0 1.3 0.2 0.6 Doxycycline plus placebo group (n 273) (mean SD or %) 27.1 6.7 48.7 23.0 21.6 6.7 3.2 2.2 0.1 1.1 1.0 1.4 0.2 0.5 Doxycycline-only versus doxycycline/ placebo P .42* .75

Patient characteristics Age (y) Race White Hispanic Black Other Gravidity Parity No. of previous elective abortions No. of previous spontaneous abortions
SD standard deviation. * Pooled variance t test. 2 test of association. MannWhitney test.

.77 .32 .27 .86

The decision to forgo preoperative cervical screening in our patient population was based on several factors. First, the baseline incidence of sexually transmitted diseases in our contraceptive population and the incidence of postabortion infections in our surgical population are low. Second, others5 have found that a strategy of universal prophylaxis is at least as effective in preventing postabortion infection as a screen and treat strategy for Chlamydia trachomatis or Neisseria gonorrhoeae. Although a 3-day course of doxycycline is not among the regimens currently recommended for chlamydial treatment by the Centers for Disease Control and Prevention (CDC),6 1-day doxycycline regimens have proven effective in prior randomized, blinded trials of periabortion prophylaxis in which the drug was compared with a placebo. Darj and co-workers7 reported efcacy with a single oral dose of 400 mg 10 12 hours before surgery (relative risk [RR] 0.33; condence inter-

val [CI] 0.22, 0.73), and Levallois and Rioux8 reported a more striking result using 300 mg as a divided 1-day dose (RR 0.12; CI 0.08, 0.38). Efcacy of these short-course regimens in this setting may be bolstered because most abortion patients probably comply with the customary admonition that they refrain from coitus for 2 weeks after the abortion. Four prior trials710 have shown oral doxycycline to be effective in reducing postabortion infection relative to placebo. Because it is inexpensive, easily tolerated, and effective, doxycycline is the current drug of choice in 81% of member facilities of the National Abortion Federation that routinely prescribe antibiotic prophylaxis.4 Members of this association perform a total of more than 350,000 rst-trimester abortions annually, and their preferences may well reect antibiotic choices of the wider array of clinicians performing abortions in North America.

Table 4. Historical Risk Prole for Pelvic Infections Among Enrollees (n 800) and Returnees (n 530) Returnees Enrollees Returnees (n 530) [n (%)] 30 (5.7) 11 (2.1) 18 (3.4) 44 (8.3) 2 (0.4) 19 (3.6) Nonreturnees (n 270) (n (%)] 19 (7.0) 5 (1.9) 8 (3.0) 29 (10.7) 1 (0.4) 14 (5.2) Returnees vs nonreturnees P* 0.44 0.83 0.74 0.27 1 0.28 Doxycyclineonly returnees (n 257) [n (%)] 11 (4.3) 2 (0.8) 5 (1.9) 18 (7.0) 0 (0.0) 8 (3.1) Doxycycline/ placebo returnees (n 273) (n (%)] 19 (7.0) 9 (3.3) 13 (4.8) 26 (9.6) 2 (0.7) 11 (4.0) Returnees doxycycline-only vs doxycycline/ placebo P .18 .04 .07 .28 .50 .57

Positive past history Chlamydia Gonorrhea Pelvic infection History of other STD Ectopic pregnancy IUD in past

STD sexually transmitted disease; IUD intrauterine device. * 2 test of association or Fisher exact test. Statistically signicant.

VOL. 101, NO. 4, APRIL 2003

Lichtenberg and Shott

Antibiotic Prophylaxis

729

Table 5. Patient-Reported Postoperative Problems (n) Among Returnees and Telephone Contacts (n 562) Doxycycline-only Doxycycline plus group placebo group (n 257) (n 273) Total Cramping Bleeding Clots Fever Pelvic pain No menses Abdominal pain Chills Diarrhea Vomiting 12 9 2 1 3 2 0 1 0 0 16 8 1 2 0 0 1 0 1 1 28 17 3 3 3 2 1 1 1 1

The age and race or ethnicity of those who returned for follow-up did not differ signicantly from those who failed to return (Table 2). Moreover, these features closely parallel those of US women in a national survey11 who obtained induced abortions in a comparable time period, indicating that our study probably provided a representative sample of such women. Our follow-up rate of 66% exceeds that reported in recent surgical abortion studies12,13 in the United States, and our telephone survey increased the likelihood of uncovering all possible instances of infection. In addition, our experience and community visibility as a major provider of abortion services for 14 years make it likely that we would have been contacted about a noteworthy complication in a nonreturning patient. Nonetheless, it is possible that one or more such events occurred without our knowledge. Self-reported pill compliance, dened as taking all pills for at least 6 days, was about 97% for both study groups. We did not verify these reports by requiring pill counts, but other studies that have done so report nonadherence of 13% to 93% in adult outpatients.14 Wehbeh et al15 used pill bottles with an embedded microchip to monitor pill compliance without patient knowledge in a small group (n 30) of pregnant women who were given a 10-day course of erythromycin to treat cervical culture positive C trachomatis. Despite intensive counseling and their pregnant status, these women took a mean of less than 23 of the 40 tablets prescribed, with ingestion and dosing intervals faltering after day 3. Shortening antibiotic regimens is likely to result in better patient compliance, decreased drug-related side effects, and possibly more universal use by clinicians. Worldwide, the ranges of infection rates reported in large case series for rst- and second-trimester induced abortion are quite similar.16 During second-trimester induction abortion, Spence et al17 observed a statistically

signicant reduction in postabortion endometritis among 198 women randomized to cephalothin versus placebo. It is therefore possible that short-course prophylactic regimens could be safely applied to induced abortion beyond the rst trimester. The advent of rapid, highly accurate assays for diagnosing sexually transmitted diseases offers the opportunity to evaluate screen-and-treat regimens using singledose (directly observed therapy) drug regimens if cost and microbiologic issues can be overcome. The CDC6 currently recommends azithromycin (1 g) in a single oral dose for treatment of uncomplicated chlamydial infection. In the case of asymptomatic gonorrhea, guidelines of the California Department of Health Services Sexually Transmitted Disease Control Branch18 sanction azithromycin (2 g) as a single dose (with a test of cure). In our study, shortening the course of oral doxycycline prophylaxis after elective vacuum abortion to 3 days did not result in an increase in postabortion infection. Future trials are needed to compare a several-day course with briefer regimens in the rst and second trimester, and multicenter trials could add greater generalizability to our ndings.

REFERENCES 1. 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;87:88490. 2. American College of Obstetricians and Gynecologists. Antibiotic prophylaxis for gynecologic procedures. ACOG practice bulletin no. 23. Washington: American College of Obstetricians and Gynecologists, 2001. 3. Royal College of Obstetricians and Gynaecologists. The care of women requesting induced abortion. Evidencebased guideline no. 7. London: RCOG, 2000. 4. Lichtenberg ES, Paul M, Jones H. First trimester surgical abortion practices: A survey of National Abortion Federation members. Contraception 2001;64:34552. 5. Penney GC, Thomson M, Norman J, McKenzie H, Vale L, Smith R, et al. A randomized comparison of strategies for reducing infective complications of induced abortion. Br J Obstet Gynaecol 1998;105:599604. 6. Division of Sexually Transmitted Diseases. Sexually transmitted diseases treatment guidelines 2002. Atlanta: Centers for Disease Control and Prevention. Available at http://www.cdc.gov/std/treatment. Accessed 2003 Jan 17. 7. Darj E, Stralin E-B, Nilsson S. The prophylactic effect of doxycycline on postoperative infection after rst-trimester abortion. Obstet Gynecol 1987;70:7558. 8. Levallois P, Rioux J-E. Prophylactic antibiotics for suction

730

Lichtenberg and Shott

Antibiotic Prophylaxis

OBSTETRICS & GYNECOLOGY

9.

10.

11.

12.

13.

14.

15.

curettage abortion: Results of a clinical trial. Am J Obstet Gynecol 1988;158:1005. London RS, London ED, Siegelbaum M, Goldstein P. Use of doxycycline in elective rst trimester abortion. South Med J 1978;71:6723. Brewer C. Prevention of infection after abortion with a supervised single dose of oral doxycycline. BMJ 1980;281: 7801. Henshaw SK, Kost K. Abortion patients in 1994-1995: Characteristics and contraceptive use. Fam Plann Perspect 1996;28:1407, 158. Edwards J, Creinin MD. Surgical abortion for gestations of less than 6 weeks. Curr Probl Obstet Gynecol Fertil 1997; 20:119. Paul M, Lackie E, Mitchell C, Rogers A, Fox M. Is pathology examination useful after early surgical abortion? Obstet Gynecol 2002;99:56771. Sclar DA, Tartaglione TA, Fine MJ. Overview of issues related to medical compliance with implications for the outpatient management of infectious diseases. Infect Agents Dis 1994;3:26673. Wehbeh H, Fleisher JM, Coasino M, Ayoub A, Margos-

sian H, Zarou D. Erythromycin for Chlamydiasis in pregnant women. J Reprod Med 2000;45:4658. 16. Lichtenberg ES, Grimes DA, Paul M. Abortion complications: Prevention and management. In: Paul M, Lichtenberg ES, Borgatta L, Grimes DA, Stubbleeld PG, eds. A clinicians guide to medical and surgical abortion. New York: Churchill Livingstone, 1999:197216. 17. Spence MR, King TM, Burkman RT, Atienza MF. Cephalothin prophylaxis for midtrimester abortion. Obstet Gynecol 1982;60:5025. 18. California STD/HIV Prevention Training Center. California STD treatment guidelines for adults and adolescents 2002. Berkeley, California: California STD/HIV Prevention Training Center. Available at http://www.dhs.ca.gov/ps/dcdc/std/ docs/Txguidln2002.pdf. Accessed 2003 Jan 17. Reprints are not available. Address correspondence to: E. Steve Lichtenberg, MD, MPH, 5086 North Elston Avenue, Chicago, IL 60630; E-mail: fpamg@cais.com. Received June 20, 2002. Received in revised form August 18, 2002. Accepted October 17, 2002.

VOL. 101, NO. 4, APRIL 2003

Lichtenberg and Shott

Antibiotic Prophylaxis

731