Research General gynecology| Volume 204, ISSUE 4, P301.e1-301.e5, April 01, 2011

# Are antibiotics necessary after 48 hours of improvement in infected/septic abortions? A randomized controlled trial followed by a cohort study

Published:January 03, 2011

### Objective

We sought to investigate whether oral antibiotics are necessary, after 48 hours of clinical improvement, in uncomplicated septic abortion.

### Study design

In a randomized double-blind clinical trial, 56 women with uncomplicated septic abortion were treated with intravenous antibiotics, followed by uterine evacuation. On hospital discharge (day 1), patients were randomized to receive either oral doxycycline plus metronidazole or placebo, until completing 10 days of treatment. Clinical cure was defined by the absence of fever (<37.7°C), reduced vaginal bleeding, and minimal or no pelvic pain.

### Results

Cure was observed in all 56 patients. The institutional review board stopped the treatment arm as it was adding risk with no further benefit to the patients. An observational cohort with additional 75 cases was followed up in the no treatment arm and no failure was identified (probability of an adverse event, 0%; 95% confidence interval, 0–0.03).

### Conclusion

After 48 hours of clinical improvement, antibiotics may not be necessary.

## Key words

In the developing world, an estimated 5 million women are admitted to the hospital for treatment of complications from induced abortions each year. This is an average of 5.7 per 1000 women per year in all developing regions, excluding China.
• Singh S.
Hospital admissions resulting from unsafe abortion: estimates from 13 developing countries.
In Brazil, abortion is illegal, and because of this, the complications of this procedure represents the third leading cause of maternal mortality and accounts for 12% of the maternal mortality rate. Brazil's abortion rate is high at 40.8 per 1000 women, and approximately 31% of pregnancies end in an induced abortion.
• Goldman L.A.
• Garcia S.G.
• Diaz J.
• Yam E.A.
Brazilian obstetrician-gynecologists and abortion: a survey of knowledge, opinions and practices.
Patients with established infection, as indicated by fever (arbitrarily defined as >38°C), pelvic peritonitis, or tachycardia, should be hospitalized for parenteral antibiotic therapy and prompt uterine evacuation.
• Stubblefield P.G.
• Grimes D.A.
Septic abortion.
Bacteremia, which is more common in septic abortion than in other pelvic infections, may result in septic shock and adult respiratory distress syndrome. The management of severe sepsis requires eradication of the infection, and supportive care for the cardiovascular system and other involved organ systems.
• Stubblefield P.G.
• Grimes D.A.
Septic abortion.
The Brazilian Ministry of Health recommends the use of intravenous clindamycin plus gentamicin for 7-10 days, in cases of uncomplicated septic abortions.
• da Saúde Brasil Ministério
Similarly, the World Health Organization (WHO) advises that broad-spectrum intravenous antibiotic therapy should be followed by oral doxycycline (200 mg/d) for 10-14 days.
World Health Organization
Managing complications in pregnancy and childbirth: a guide for midwives and doctors.
This regimen is similar to that recommended for pelvic inflammatory disease by the Centers for Disease Control and Prevention (CDC).
• Workowski K.A.
• Berman S.M.
Sexually transmitted diseases treatment guidelines, 2006.
Contrary to prolonged treatment with oral antibiotics, French and Smaill
• French L.M.
• Smaill F.M.
Antibiotic regimens for endometritis after delivery.
demonstrated that it is unnecessary to continue antibiotic therapy in postcesarean section endometritis once the patient is afebrile for 48 hours. It could be safe and less expensive to utilize this regimen in cases of uncomplicated septic abortion. There is no evidence thus far to show that this shorter regimen is safe in such cases.
The objective of this is study was to investigate the need for oral antibiotics after 48 hours of clinical improvement, in patients with uncomplicated septic abortion who have undergone uterine evacuation and received a minimum of 48 hours of intravenous antibiotics.

## Materials and methods

### Participants

Patients admitted to Hospital de Clínicas de Porto Alegre, with a diagnosis of uncomplicated septic abortion, according to the WHO, were eligible for the study. The WHO defines the condition as abortion with the presence of one of the following signs or symptoms: chills or sweats, fever (≥37.8°C), foul-smelling vaginal discharge, abdominal rebound tenderness, hypotension (arterial systolic pressure <90 mm Hg), history of intrauterine manipulation with nonsterile probes, malaise, or pus coming from cervix or mixed with blood in the vagina.
World Health Organization
Managing complications in pregnancy and childbirth: a guide for midwives and doctors.
We also used the additional criteria of leukocytosis (>14,000 leukocytes/mL) as suggested by Lurie et al.
• Lurie S.
• Rahamim E.
• Piper I.
• Golan A.
Total and differential leukocyte counts percentiles in normal pregnancy.
Patients were treated with intravenous clindamycin 2700 mg, plus gentamicin 240 mg per day and immediate uterine evacuation (day 1). Patients received intravenous antibiotics in the hospital until 48 hours of clinical improvement was noted. Clinical improvement was defined as: no fever, normal leukogram, and a reduction in vaginal bleeding and pelvic pain.
Patients who used antibiotics within 1 week prior to hospitalization, who were not willing to participate, with a diagnosis of tubo-ovarian abscess, or with known allergy to doxycycline or metronidazole were excluded.

### Intervention

Before hospital discharge, patients were invited to participate in the study (day 1). Those meeting study criteria and providing consent were randomized to receive either oral doxycycline 100 mg twice daily plus metronidazole 250 mg twice a day, or an identical placebo, until completion of 10 days of treatment (day 10). Medications were prepared by the pharmacy of the Hospital de Clínicas de Porto Alegre, in identical coded blister packets and capsules to assure double blinding. Patients were instructed not to use drugs for pain or fever. They were instructed to return to the hospital if they experienced pain or fever.

### Outcome

At 48-72 hours after hospital discharge, patients were contacted by telephone to determine their clinical condition. After finishing the outpatient treatment (day 10), patients were seen by one of the investigators (G.S.M., R.A.C.). Adherence to the study protocol was evaluated by examining the medication packets. At this visit, a standard interview was utilized to identify the presence or absence of the primary outcome (clinical cure). Clinical cure was defined as the absence of fever (<37.8°C), reduced vaginal bleeding, and minimal or no pelvic pain. Clinical failure was defined as hospital readmission, presence of fever (≥37.8°C) after hospital discharge, no reduction of vaginal bleeding, persistence of pelvic pain, or the need to use additional medication (antibiotics or pain medication) whether prescribed or not.

### Sample size

Sample size for equivalence was calculated according to Blackwelder,
• Blackwelder W.C.
”Proving the null hypothesis” in clinical trials.
using the formula:
$n = (Zα + Zβ)2⁢[PS(1 − PS) + PT(1 − PT)](PS − PT − d)2$

where n = the sample size for each treatment group; Zα = the standard normal variate corresponding to the α significance level, ie, an α error of 0.02 = 2.326; Zβ = the standard normal variate corresponding to the tail probability of size β, ie, β error of 0.1 = 1.2816; and d = the difference between the conventional (PS) and short (PT) treatment effects that is considered to be clinically meaningful, herein = 10%. In antiinfective studies, a d of 10% was chosen as recommended by the US Food and Drug Administration.
• Wiens B.L.
Choosing an equivalence limit for noninferiority or equivalence studies.
Based on our local experience, and hospital electronic records, rate of cures with long antibiotic protocol is near 99%. The estimated rate of clinical cure with placebo treatment was 97%. These figures yield a minimum of 79 patients in each group (conventional and short treatment).
Interim analysis was planned a priori and calculated according to Fang,
• Fang J.
Design and analysis of sequential experiments.
considering 2 stages of analysis. The adjusted critical Zα value (derived from adjusted α for 0.05, and β of 0.1) was considered in a scenario where the rate of cure for the standard intervention is 100% and for the placebo is 80% (chosen arbitrarily) for possible early stopping. If a higher rate of cure in the placebo and in the antibiotic group was observed, eg, 100%, the use of antibiotics would give no additional benefit for the patients, and the study should be stopped. These figures suggested that the first interim analysis should be with 27 patients per group, in each stage, yielding a total of 108 patients, ie, 27 × 2 groups × 2 stages.
If the study was stopped due to a high rate of cure in the placebo arm (eg, 100%), an observational cohort would be carried out until reaching at least the initial sample size calculated for the placebo arm.

### Observational cohort

In the observational cohort, the same inclusion and exclusion criteria, follow-up, and outcome assessment used in the randomized clinical trial were applied.

### Randomization

After signing the informed consent, treatment allocation was dispensed according to a computer-generated randomization list, grouped in blocks of 4. The allocation list was concealed until interventions were assigned and coded. The treatment assignment was then communicated, via telephone, from a central location. All study personnel and participants were blinded to treatment assignment for the duration of the study. Treatment was given by one of the authors (G.S.M. or R.A.C.).

### Statistical methods and ethical issues

Fisher exact test and descriptive analysis were used for statistical analysis. The outcome was analyzed using rates of cure in intention to treat and per protocol with 95% confidence intervals (CIs). For 0 events, ie, 100% of cure, the method of Hanley and Lippman-Hand
• Hanley J.A.
• Lippman-Hand A.
If nothing goes wrong, is everything all right? Interpreting zero numerators.
of probability for 95% CI was used. The following formula was applied: (3/n), where n = number of cases in the group with 0 events. Adherence included exposure to treatment, availability of measurements, and absence of protocol violations. The study protocol followed the CONSORT guidelines.
• Moher D.
• Schulz K.F.
• Altman D.G.
The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials.
It was approved by the ethics committee of Hospital de Clínicas de Porto Alegre (number 05-452), and is registered at ClinicalTrials.gov, number NCT00353743. In Brazil, Data and Safety Monitoring Committees (DSMCs) are usually not appointed in studies, unless there is pharmaceutical company involvement. This function is performed by the local institutional review board instead. Furthermore, little literature on DSMCs has been published, and guidelines for managing or serving on such committees are almost nonexistent. No guide or textbook about DSMC functions has been written to date, although one such text is in production.
• Pocock S.
• Furberg C.D.
Procedures of data and safety monitoring committees.

## Results

From May 2006 through November 2007, 60 patients with a diagnosis of septic abortion were admitted to Hospital de Clínicas de Porto Alegre. Of these, 56 were eligible for the study and were randomized. The Figure depicts details of the randomization. In November 2007, the first interim analysis was conducted on 56 patients. There were no clinical failures in either arm (absolute risk difference, 0; 95% CI, 0.0–0.1). Characteristics of the sample before randomization are in Table 1. A woman from the placebo group, who returned the medication without taking it, constituted the only protocol violation. She had a clinical cure. There were no serious adverse events experienced by any patient after hospital discharge. Furthermore, 28 patients in each arm, with a 0 event, yielded a probability of failure up to 10% with a 95% CI, which was in accordance to the initial d.
TABLE 1Baseline characteristics of study population during randomization
CharacteristicRandomization
Placebo (n = 28)Doxycycline plus metronidazole (n = 28)
Age, y (mean ± SD)28.2 ± 7.725.9 ± 6.2
Ethnicity, n
Caucasian1615
Afro-Brazilian1213
Days admitted at hospital (mean ± SD)3.2 ± 1.13.0 ± 0.9
Days of intravenous antimicrobial (mean ± SD)2.6 ± 0.83.0 ± 1.0
History of intrauterine manipulation with nonsterile probes, n57
Foul-smelling vaginal discharge, n58
Pus in cervix, n10
Abdominal rebound tenderness, n21
Leukocytosis (>14,000 leukocytes/mL), n1114
Warm extremities, thready pulse, and tachycardia, n
Tachycardia defined as >110 beats/min.
64
Cyanosis and/or paleness, n68
Tachypnea (>30 respiratory movements/min), n11
Arterial hypotension (systolic arterial pressure <90 mm Hg), n1412
Oliguria, n23
Fever (≥37.8°C), n1618
Savaris. Antibiotics in infected/septic abortions. Am J Obstet Gynecol 2011.
a Tachycardia defined as >110 beats/min.
From December 2007 through June 2010, 75 consecutive patients with uncomplicated septic abortion, using the same inclusion criteria, were discharged without medication, and were reviewed 10 days later. There were no losses to follow-up. All additional observational cases had clinical cures. According to the method of Hanley and Lippman-Hand,
• Hanley J.A.
• Lippman-Hand A.
If nothing goes wrong, is everything all right? Interpreting zero numerators.
the probability of an adverse event is ≤3% (95% CI, 0–0.03). Characteristics of the observational cohort after the trial suspension are in Table 2. No serious adverse events were seen in this group.
TABLE 2Characteristics of study population during observational study
CharacteristicObservational
(n = 75)
Age, y (mean ± SD)27 ± 6.19
Ethnicity, n
Caucasian54
Afro-Brazilian21
Days admitted at hospital (mean ± SD)3.2 ± 2.3
Days of intravenous antimicrobial (mean ± SD)3.2 ± 2.3
History of intrauterine manipulation with nonsterile probes, n1
Foul-smelling vaginal discharge, n8
Pus in cervix, n8
Abdominal rebound tenderness, n1
Leukocytosis (>14,000 leukocytes/mL), n52
Warm extremities, thready pulse, and tachycardia, n
Tachycardia defined as >110 beats/min.
4
Cyanosis and/or paleness, n1
Tachypnea (>30 respiratory movements/min), n0
Arterial hypotension (systolic arterial pressure < 90 mm Hg), n4
Oliguria, n1
Fever (≥37.8°C), n30
Some patients had > 1 characteristic.
Savaris. Antibiotics in infected/septic abortions. Am J Obstet Gynecol 2011.
a Tachycardia defined as >110 beats/min.

## Comment

All (100%) of the 103 patients with uncomplicated infected/septic abortion treated with intravenous antibiotics followed by uterine evacuation had clinical cure, using a short treatment protocol.
The strengths of the study are few. It is a randomized, double-blind study reflecting a clinical practice, with an adequate randomization method, an excellent follow-up, a priori specified interim analysis, which is important to evaluate initial suppositions about rate of cure. Sample size can be increased or reduced after the first interim analysis. Special attention was given to the sample size calculation to give enough power in an equivalence trial. The reasoning of an equivalence trial is different from the superiority trials. Instead of rejecting the null hypothesis, where the percentage of both treatments is equal, researches want to prove that it is true. Due to the intrinsic properties of the formula proposed by Blackwelder,
• Blackwelder W.C.
”Proving the null hypothesis” in clinical trials.
treatment with high rate of cure (ie, 99%) in both arms would require a smaller sample size, if the d is 10%. That was the case in this study. With a sample size of 56 patients, considering a 99% rate of cure in both arms, an α error of 0.2, and a power of 90%, 26 patients would be necessary in each arm to have statistical power. With these parameters we were able to state that we are 90% sure that there is no difference >10% between both groups, with a 97% CI. However, a 10% difference should be considered too high, and this is a weakness of the study. To reduce this 10% difference, we faced another problem. The institutional review board determined that it would be unethical to continue using standard treatment. Thus, the alternative to increase sample size would be an observational cohort in the short treatment group. Thus, from November 2007 through June 2010, an additional 75 cases of uncomplicated infected/septic abortion who received no treatment after hospital discharge were followed up in the same manner as the initial 56 patients from the clinical trial. No clinical failures were observed in these patients, which was reassuring. Therefore, using these results from the short treatment protocol, we are able to state that in 95% of the cases, we expected to find a rate of cure of between 97-100%, and that seems to be an adequate CI.
The implications of the study can be quite important in terms of reducing time of hospital stay and antibiotics. The Brazilian Health Ministry recommendation for infected abortion is 7-10 days of intravenous, broad-spectrum antibiotics,
• da Saúde Brasil Ministério
while the WHO recommendation for infected abortion
World Health Organization
Managing complications in pregnancy and childbirth: a guide for midwives and doctors.
is similar to the CDC's recommendation for pelvic inflammatory disease, where treatment would last up to 14 days.
• Workowski K.A.
• Berman S.M.
Sexually transmitted diseases treatment guidelines, 2006.
The mechanisms that might explain these outcomes could be the same as those found in the study reported in the Cochrane systematic review on endometritis after delivery.
• French L.M.
• Smaill F.M.
Antibiotic regimens for endometritis after delivery.
The removal of infected tissue would be enough to reduce the period of antibiotic use. Endometrial culture was not performed because of the polymicrobial nature of these infections, as demonstrated by other authors.
• Ramsay A.M.
• Brown E.H.
• Manners S.M.
Investigation and treatment of septic abortion.
• Moberg P.J.
• Gottlieb C.
• Nord C.E.
Anaerobic bacteria in uterine infection following first trimester abortion.
• Sokolova I.E.
Antibiotic sensitivity of the microflora isolated from the uterine cavity of patients with postabortion endometritis.
For that reason, we believe that the rates of cure found in our study are comparable to those published by French and Smaill,
• French L.M.
• Smaill F.M.
Antibiotic regimens for endometritis after delivery.
in cases of postpartum endometritis after caesarean section or vaginal birth.
To our knowledge, this is the first report pertaining to the non-use of antibiotics, after hospital discharge, in patients with uncomplicated septic abortion. Hence we are unable to draw any comparison with other studies. Since the point of entry into the study was at hospital discharge, the average stay in hospital for both groups was 3 days. In light of our results, we have stopped giving antibiotics to patients with uncomplicated septic abortion after 48 hours of clinical improvement.
This trial questions the need for long periods of treatment with oral antibiotics after clinical improvement in patients with uncomplicated septic abortion. It should be used as a reference for further research.

## References

• Singh S.
Hospital admissions resulting from unsafe abortion: estimates from 13 developing countries.
Lancet. 2006; 368: 1887-1892
• Goldman L.A.
• Garcia S.G.
• Diaz J.
• Yam E.A.
Brazilian obstetrician-gynecologists and abortion: a survey of knowledge, opinions and practices.
Reprod Health. 2005; 2: 10
• Stubblefield P.G.
• Grimes D.A.
Septic abortion.
N Engl J Med. 1994; 331: 310-314
• da Saúde Brasil Ministério
in: da Saúde Brasil Ministério Urgências e emergências maternas: guia para diagnóstico e conduta em situações de risco de morte materna. MS/FEBRASGO, Brasília2000: 13
• World Health Organization
Managing complications in pregnancy and childbirth: a guide for midwives and doctors.
WHO, Geneva1994
• Workowski K.A.
• Berman S.M.
Sexually transmitted diseases treatment guidelines, 2006.
MMWR Recomm Rep. 2006; 55: 1-94
• French L.M.
• Smaill F.M.
Antibiotic regimens for endometritis after delivery.
Cochrane Database Syst Rev. 2004; 4 (CD001067)
• Lurie S.
• Rahamim E.
• Piper I.
• Golan A.
Total and differential leukocyte counts percentiles in normal pregnancy.
Eur J Obstet Gynecol Reprod Biol. 2008; 136: 16-19
• Blackwelder W.C.
”Proving the null hypothesis” in clinical trials.
Control Clin Trials. 1982; 3: 345-353
• Wiens B.L.
Choosing an equivalence limit for noninferiority or equivalence studies.
Control Clin Trials. 2002; 23: 2-14
• Fang J.
Design and analysis of sequential experiments.
in: Fang J. Medical statistics and computer experiments. World Scientific Publishing Co Pte Ltd, Singapore2005: 563-578
• Hanley J.A.
• Lippman-Hand A.
If nothing goes wrong, is everything all right? Interpreting zero numerators.
JAMA. 1983; 249: 1743-1745
• Moher D.
• Schulz K.F.
• Altman D.G.
The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials.
Lancet. 2001; 357: 1191-1194
• Pocock S.
• Furberg C.D.
Procedures of data and safety monitoring committees.
Am Heart J. 2001; 141: 289-294
• Ramsay A.M.
• Brown E.H.
• Manners S.M.
Investigation and treatment of septic abortion.
Br Med J. 1955; 2: 1239-1243
• Moberg P.J.
• Gottlieb C.
• Nord C.E.
Anaerobic bacteria in uterine infection following first trimester abortion.
Eur J Clin Microbiol. 1982; 1: 82-86
• Sokolova I.E.
Antibiotic sensitivity of the microflora isolated from the uterine cavity of patients with postabortion endometritis.
Antibiot Med Biotekhnol. 1986; 31 ([in Russian]): 687-690