: Dr E. Eason, Box 803, 501 Smyth Road, Ottawa, ON, Canada K1H 8L6.
Objective To assess whether infectious morbidity after total abdominal hysterectomy is decreased by the addition of 20 cc povidone–iodine gel at the vaginal apex after the usual vaginal preparation with povidone–iodine solution.
Study design Randomised controlled trial.
Setting Fifteen secondary and tertiary hospitals in Canada.
Sample A total of 1570 women undergoing planned total abdominal hysterectomy.
Methods Computer-generated randomisation using a centralised telephone service was stratified by study centre with variable block size. In the operating room, a swab for bacterial vaginosis was taken before vaginal antisepsis. Study group remained concealed until the standard surgical preparation in the operating room was complete. Then povidone-iodine gel 20 cc was placed at the vaginal apex in the intervention group only. Participants were followed for one month post-operative.
Main outcome measures The primary outcome was post-operative infectious morbidity during the 30 days after surgery, defined as: febrile morbidity with hospital stay greater than five days or antibiotic treatment, or infection requiring readmission to hospital or additional visit. Other outcomes included abdominal wound infection, pelvic abscess and other pelvic infections.
Results Post-operative infectious morbidity within 30 days occurred in 128/780 (16%) women receiving povidone–iodine gel preparation and 142/790 (18%) women not receiving gel (RR 0.9, 95% CI 0.7 to 1.1). Pelvic abscess was diagnosed in 0 patients in the gel group and in seven patients in the control group (P < 0.01). No significant difference was found in pelvic cellulitis (eight in each group) or abdominal wound infection (51 in the gel group and 58 in the control group, P= 0.5).
Conclusion Povidone–iodine vaginal gel antisepsis led to a 9% relative decrease in overall infectious morbidity after abdominal hysterectomy, which was not statistically significant. Povidone–iodine vaginal gel decreased the risk of pelvic abscess after total abdominal hysterectomy.
Bacterial contamination from the vaginal vault is a major cause of febrile morbidity and infectious complications such as vaginal cuff cellulitis and pelvic abscess after abdominal hysterectomy.1,2 It leads to increased investigations, antibiotic treatment and hospital stay. Prophylactic antibiotics decrease serious post-operative infections after abdominal hysterectomy from 21% to 9%,3 but even with prophylactic antibiotics, infection is still much more frequent than after surgery classified as ‘clean’.4 This higher infection rate is caused by contamination of the surgical field by bacteria from the vagina.2 To lessen the spill of vaginal bacteria into the operative site, a standard text, TeLinde's Operative Gynaecology, has recommended a vaginal douche and thorough bath with hexachlorophene or povidone–iodine the evening before surgery, and thorough povidone–iodine cleaning of the vagina in the operating room.5 On the other hand, several authorities6–8 have noted the ‘essentially intuitive’ nature of vaginal cleaning practices and the need for proper controlled clinical trials.7 Our review of studies of vaginal antisepsis highlighted the paucity and methodological weakness of research in this area9: no randomised controlled trials of the utility, timing or type of vaginal antisepsis have been adequately powered to rule out clinically important differences in outcomes. Monif et al.10 showed that bacterial counts returned to near baseline levels within 30 minutes after vaginal painting with povidone–iodine solution; but vaginal povidone–iodine gel lowered bacterial counts much more, with antibacterial activity persisting for at least 3 hours. Because vaginal incision at abdominal hysterectomy usually occurs more than 30 minutes after surgical preparation of the vagina, we carried out a pilot study in which vaginal povidone–iodine gel was inserted after the standard vaginal povidone–iodine solution preparation for abdominal hysterectomy in 158 women.1 Compared with 317 women having hysterectomy in the immediately preceding period, the odds of febrile morbidity in gel-treated women were reduced (odds ratio 0.5, 95% confidence interval [CI] 0.3 to 0.9).1 Based on this result, we undertook a multicentred, randomised, controlled trial to determine whether infectious complications of abdominal hysterectomy could be reduced by vaginal povidone–iodine gel (20 mL) after the standard surgical preparation with povidone–iodine solution.
This randomised controlled trial in women undergoing abdominal hysterectomy was carried out between April 1998 and August 1999 at 15 communities and teaching hospitals in Canada. Approval for the study was obtained from the ethics committees at all participating hospitals and each woman gave informed consent. Women were eligible to participate if they were over 18 and expected to undergo a total abdominal hysterectomy. We included women undergoing extended or radical hysterectomy or planned additional ‘clean’ surgical procedures. We excluded women undergoing hysterectomy in conjunction with bowel resection or as an emergency procedure, and women with active pelvic infection or allergy to povidone–iodine. Because the decision to do a subtotal or total hysterectomy was sometimes made intra-operatively, we obtained consent for study participation from women whose type of hysterectomy was unspecified pre-operatively. Only those who actually had a total abdominal hysterectomy were eligible to be included in the analysis, as vaginal antisepsis is not a concern in subtotal hysterectomy.
The computer-generated randomisation was stratified by study centre in variable block size of 8 or 10. Study numbers were assigned at the time of enrolment for each subject, using a centralised telephone service, but the study group assignment remained concealed until the standard surgical preparation with povidone–iodine solution in the operating room was completed. Because the group to which the patient was randomised was masked until after induction of anaesthesia, women who withdrew from the study pre-operatively were excluded from further follow up. After induction of anaesthesia and antiseptic preparation of the abdomen, a high vaginal swab for bacterial vaginosis was smeared onto a glass slide and fixed. Then the vagina was cleansed with povidone–iodine solution. The study packet labelled with that subject's study number was unsealed to disclose the study group. If the patient was in the gel group, 20 mL of povidone–iodine gel (Betadine vaginal gel, Purdue Pharma, Pickering, Canada) was drawn into a sterile vaginal applicator and placed at the vaginal apex. If the cap of the tube of gel was glued shut and covered with a label reading ‘do not use’, the patient was in the control group. For each case, the operating room nurse completed a checklist indicating whether povidone–iodine gel was actually used. She inserted the checklist and the vaginal smear into an envelope labelled with the study number, which was sealed and returned to the co-ordinating centre. The checklist was used to determine compliance with the study intervention. No entry was made in the medical record to indicate whether the gel was used. Thus, patients, gynaecology ward staff, local research nurses and investigators were blind to the patient's randomisation group until the analysis was carried out. It was not possible to blind surgeons, as the gel is clearly visible in the vagina when the uterine cervix is resected. (From the pilot study, we knew that patients and ward staff rarely detected drainage of povidone–iodine gel during the post-operative period.) Decisions regarding use of prophylactic antibiotics and closure of the vaginal vault were left to the individual surgeon.
Data about the surgery, post-operative course, re-admissions and clinic or emergency visits within 30 days were extracted from the hospital chart. The surgeon completed a questionnaire about any office visits or surgical complications within 30 days post-op. If the surgeon did not see the patient post-operatively, the study nurse carried out a structured telephone interview with the patient to identify complications or visits to other physicians or hospitals, and contacted any other physician seen. The vaginal smears were Gram stained and scored using Nugent's criteria11 at the microbiology laboratory of the Ottawa Hospital, General Campus. Research nurses at each site conducted periodic audits of study procedures in the operating room. Duplicate data extraction from the hospital records, with resolution of inconsistencies directly from the record, was carried out at most centres by the principal investigator or a research nurse from a different centre.
The primary outcome was post-operative infectious morbidity during the 30 days after surgery, defined as: febrile morbidity with hospital stay greater than five days or antibiotic treatment, or infection requiring readmission to hospital or additional visit. Febrile morbidity was defined as temperature ≥38°C on two days, excluding the first 24 post-operative hours.12 Oral temperatures were taken by electronic thermometer every 6 hours while in the hospital. Other outcomes were defined as follows: abdominal wound infection was considered definite if the wound showed erythema and culture-positive drainage, or purulent drainage, or was deliberately opened and not reclosed13; and probable if there was erythema extending at least 2 cm from the incision in any direction or if infection was diagnosed by a physician but the above criteria for definite infection were not fulfilled.13 Pelvic abscess was diagnosed only if there was operative or spontaneous drainage of pus from a collection documented by laparoscopy, laparotomy, ultrasound or clinical examination. Ultrasound and computerised tomography have been recommended to diagnose infected haematoma or abscess5,14–16 but sonographically diagnosed complex fluid collections after hysterectomy are also common in patients without infectious morbidity or fever.17 We recorded the clinical diagnosis of vaginal cuff or pelvic cellulitis, although this diagnosis is highly subjective18 and may represent a diagnosis of exclusion. Urinary tract infection was recorded as such only if urine culture was positive and treatment was instituted. All suspected adverse reactions to povidone–iodine were recorded.
The sample size was chosen to have 80% power to detect a decrease from 15% to 10% in the rate of infectious complications of hysterectomy, with a two-sided alpha of 0.05. These assumptions were based on outcomes in the pilot study, in which 15% of patients had prolonged hospital stay with fever, fever and antibiotic treatment, or post-discharge infectious complications; and there was 35% less febrile morbidity in the women in whom vaginal povidone–iodine gel was used.
Baseline characteristics of the two study groups were compared. Compliance was assessed by centre and by study group. All outcomes were analysed by intention to treat. All randomised women undergoing total abdominal hysterectomy (except those who withdrew consent before disclosure of randomisation group) were included in the analysis. Study outcomes were compared using the χ2 test, Fisher's exact test and Mantel–Haenszel test for categorical outcomes, and Student's t test for continuous outcomes. Logistic regression was used to adjust for potentially confounding covariates. Risk ratios, odds ratios (for logistic regressions) and risk differences with corresponding 95% confidence intervals were used to estimate risk. SAS version 8 statistical package was used.
The flow diagram for the study is shown in Fig. 1. Study entry characteristics of the two groups are compared in Table 1. The use of prophylactic antibiotics varied by centre from 3% to 99%, and bacterial vaginosis varied by centre from 6% to 32%, but there was no significant difference between the gel and control groups. Non-compliance by the surgical team occurred in 33 (4%) subjects in the gel group and in 7 (1%) in the control group. The study results and risk ratios are shown in Table 2. Adjustment using logistic regression for the potential risk factors listed in Table 1 did not alter the odds ratio for infectious morbidity; nor did controlling for study centre. Women with febrile morbidity stayed 5.6 days compared with 4.4 days in afebrile women. The mean hospital stay for subjects with a pelvic abscess was 12 days vs 5 days for patients without abscess (difference −7 days, 95% CI −11 to −4 days.) Pelvic abscess occurred in 4/576 women not receiving prophylactic antibiotics versus 3/993 who did (RR 2.3, 95% CI 0.6 to 9.2); and in 2/215 women with bacterial vaginosis versus 4/1265 without (RR 2.9, 95% CI 0.5 to 16). (Vaginal smears were not available on all patients.)
Table 1. Baseline and surgical characteristics of study population. Values are presented as mean (SD) or %.
Povidone–iodine gel (n= 780)
Control (n= 790)
ASA = American Society of Anesthesiology.
TAH ± BSO: total abdominal hysterectomy with or without salpingo-oophorectomy.
Radical TAHBSO, or with node dissection/omentectomy
Table 2. Primary and secondary outcome measures. Values are presented as n (%) or mean [SD].
Povidone–iodine gel (n= 780)
Control (n= 790)
Unadjusted risk ratio
95% Confidence interval
Length of stay greater than five days with febrile morbidity; febrile morbidity with antibiotic treatment; re-admission to hospital for an infection within 30 days; outpatient or office visit for infection within 30 days.
Febrile morbidity: temperature ≥38°C on two days, excluding the first 24 post-operative hours.
In an analysis stratified by the presence or absence of bacterial vaginosis, women with bacterial vaginosis (n= 215) treated with povidone–iodine gel had a relative risk of infectious morbidity of 0.7 (95% CI 0.4 to 1.3). compared with the control group. In women without bacterial vaginosis, the relative risk in gel compared with control women was 1.0 (95% CI 0.8 to 1.3). In women with bacterial vaginosis, the relative risk for any pelvic infection (cellulitis, abscess, haematoma or ‘urinary tract infection’ with fever) in the gel group compared with the control group was 0.4 (95% CI 0.1 to 1.9); in women without bacterial vaginosis, the relative risk for a pelvic infection was 1.1 (95% CI 0.6 to 2.2). No severe allergic reactions to povidone–iodine occurred.
We found a relative risk of 0.9 for infectious morbidity in the group treated with povidone–iodine gel, a decrease which was not statistically significant. This composite outcome measure, which included prolonged stay or antibiotic treatment in patients with febrile morbidity, or readmission or repeat visits to physicians for infection, has the strength of clinical relevance and high sensitivity for the consequences of post-operative infectious morbidity requiring additional health care. Nevertheless, it is non-specific, capturing not only infections related to vaginal contamination of the operative site but all other important infectious morbidity. Therefore, the power of this study to detect a difference due to improved vaginal antisepsis is low.
Because febrile morbidity is an objective, uniformly documented and clinically relevant index of the overall infection rate,12 this measure of infection permitted us to control for potential diagnostic bias if the surgeon ordering antibiotic therapy or deciding on hospital discharge recalled the patient's study group. The relative risk of febrile morbidity in the gel group was somewhat (14%), but not significantly, decreased.
The absence of pelvic abscess in the gel group compared with seven in the controls was striking and significant (P < 0.01). Because pelvic abscess occurred in about 1% of women in the control group, the potential benefit of using povidone–iodine gel is limited to a very small proportion of patients who experience high morbidity. We were unable to identify a subgroup of patients at higher risk for vaginal abscess who might be selectively targeted for povidone–iodine gel use, as none of the potential risk factors examined including obesity, bacterial vaginosis and antibiotic prophylaxis, were significant predictors of pelvic abscess. Although the difference in pelvic abscess was striking and biologically consistent with improved vaginal antisepsis, the overall impact of povidone–iodine gel vaginal antisepsis on infectious morbidity was disappointingly small. There is an urgent need for innovative research into improving vaginal antisepsis for hysterectomy.
Other participants in the study group were: P. Bernstein, M. Cheng (Toronto, Ont.); J. Blanchet, C. Bouchard (Quebec City, Que.); B. Fallis (Orillia, Ont.); G. Gill (Halifax, N.S.); G. Lefebvre, C. Ling (Ottawa, Ont.); O. Pinsonneault (Sherbrooke, Que.); C. Roye (Brantford, Ont.).
The authors would like to thank L. Yetisir, statistician; M. Cheng-Fitzpatrick, study co-ordinator; W. Woods and D. Edgar, microbiology; M.Tran and E. Moen, data managers, and the patients, gynaecologists and research nurses at all study sites.
This randomised controlled trial was funded by the Medical Research Council of Canada. Purdue Frederick, Pickering, Ontario, Canada supplied the Betadine gel and control samples and contributed to a research nurse training meeting.