Original Research ajog.

org

GYNECOLOGY
Reducing surgical site infections after hysterectomy:
metronidazole plus cefazolin compared with
cephalosporin alone
Sara R. Till, MD, MPH; Daniel M. Morgan, MD; Ali A. Bazzi, MD; Mark D. Pearlman, MD;
Zaid Abdelsattar, MD, MSc; Darrell A. Campbell, MD; Shitanshu Uppal, MBBS

BACKGROUND: Organisms that are isolated from vaginal cuff
infections and pelvic abscesses after hysterectomy frequently include
anaerobic vaginal flora. Metronidazole has outstanding coverage against
nearly all anaerobic species, which is superior to both cefazolin and
second-generation cephalosporins. Cefazolin plus metronidazole has been
demonstrated to reduce infectious morbidity compared with either cefa-
zolin or second-generation cephalosporins in other clean-contaminated
procedures, which include both as colorectal surgery and cesarean
delivery.
OBJECTIVE: The purpose of this study was to evaluate whether the
combination of cefazolin plus metronidazole before hysterectomy was
more effective in the prevention of surgical site infection than existing
recommendations of cefazolin or second-generation cephalosporin.
STUDY DESIGN: This was a retrospective cohort study of patients in
the Michigan Surgical Quality Collaborative from July 2012 through
February 2015. The primary outcome was surgical site infection. Patients
who were >18 years old and who underwent abdominal, vaginal, lapa-
roscopic, or robotic hysterectomy for benign or malignant indications were
included if they received 1 of the following prophylactic antibiotic regi-
mens: cefazolin, second-generation cephalosporin, or cefazolin plus
metronidazole. Multivariate logistic regression modeling was performed to
evaluate the independent effect of an antibiotic regimen, and propensity
score matching was used to validate the findings.
RESULTS: The study included 18,255 hysterectomies. The overall rate
of surgical site infection was 1.8% (n¼329). The unadjusted rate of
surgical site infection was 1.8% (n¼267) for cefazolin, 2.1% (n¼49) for
second-generation cephalosporin, and 1.4% (n¼13) for cefazolin plus
metronidazole. After adjustment for differences in patient and operative
factors among the antibiotic cohorts, compared with cefazolin plus
metronidazole, we found the risk of surgical site infection was significantly
higher for patients who received cefazolin (odds ratio, 2.30; 95% confi-
dence interval, 1.06e4.99) or second-generation cephalosporin (odds
ratio, 2.31; 95% confidence interval, 1.21e4.41).
CONCLUSION: In this large cohort, the use of prophylactic cefazolin
plus metronidazole resulted in lower surgical site infection rates
after hysterectomy compared with cefazolin or second-generation
cephalosporin.
Key words: cephalosporin, hysterectomy, metronidazole, surgical site
infection

S urgical site infections after hyster-

ectomy have decreased nearly 10-
fold in the last 20 years.1,2 Routine use
of preoperative antibiotics has played a
significant role in this decline. The
American College of Obstetricians and
Gynecologists recommends the use
of cefazolin, cefoxitin, or ampicillin/
sulbactam as single agents before hys-
terectomy.3 Recent data has shown that
first-generation cephalosporins, specif-
ically cefazolin, are the most widely
used preoperative antibiotic regimen.4
However, organisms isolated from
vaginal cuff infections most often
Cite this article as: Till SR, Morgan DM, Bazzi AA, et al.
Reducing surgical site infections after hysterectomy:
metronidazole plus cefazolin compared with cephalo-
sporin alone. Am J Obstet Gynecol 2017;217:187.e1-11.
0002-9378/$36.00
ª 2017 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.ajog.2017.03.019
include anaerobic vaginal flora, the
most common being Bacteroides spp,
Prevotella spp, Peptostreptococcus spp,
and Gardnerella spp.5 Some providers
have begun to use second-generation
cephalosporins in an effort to improve
anaerobic coverage. However, Bacter-
oides fragilis groups have demonstrated
significant levels of resistance to
second-generation antibiotics such as
cefoxitin, whereas resistance rarely is
seen with metronidazole.6,7 Addressing
the inadequacy of anaerobic coverage
might provide an opportunity to
further reduce the rate of surgical site
infections after hysterectomy.
This study was designed to deter-
mine whether the combination of
cefazolin plus metronidazole before
hysterectomy resulted in different rates
of postoperative surgical site infection
compared with existing recommenda-
tions of cefazolin or second-generation
cephalosporin.
Materials and Methods
Patients who underwent hysterectomy
from July 2012 to February 2015 in the
Michigan Surgical Quality Collabora-
tive were included in this retrospective
cohort study. The Collaborative is fun-
ded by the Blue Cross and Blue Shield
of Michigan/Blue Care Network, and
includes patients from all insurance
payers. At each of the 73 participating
hospitals, a trained dedicated nurse
abstractor collects patient characteris-
tics, intraoperative processes of care,
and 30-day postoperative outcomes for
hysterectomy cases. To ensure complete
capture of the data, nurse abstractors
make phone calls to the patients to
verify whether they were admitted in a
hospital other than where the index
surgery was performed. The standard-
ized data collection method is validated
routinely through scheduled site visits,
conference calls, and internal audits.
Detailed methods of the registry’s

AUGUST 2017 American Journal of Obstetrics & Gynecology 187.e1

Original Research GYNECOLOGY ajog.org

data collection have been described

FIGURE 1
previously.2,4
Sample selection flow diagram
Patients were included in the study if
they were >18 years of age, had under-
gone abdominal, vaginal, laparoscopic, or
robotic hysterectomy, and received 1 of
the following prophylactic antibiotic
regimens: first-generation cephalosporin,
second-generation cephalosporin, or
first-generation cephalosporin plus
metronidazole. In our cohort, the only
first-generation cephalosporin used was
cefazolin; therefore, our 3 comparison
groups were cefazolin, second-generation
cephalosporins, and cefazolin plus
metronidazole. Of note, this analysis
focused on the cephamycin subgroup
of second-generation cephalosporins,
including cefoxitin and cefotetan. Cases
with gynecologic malignancy and those
for benign indications were included in
the study. Cases with no recorded anti-
biotic information, with the use of beta-
lactam alternative antibiotics, and the
receipt of antibiotics not recommended
by the American College of Obstetricians
and Gynecologists or Surgical Care
Improvement Project were not included
in this analysis, based on our previous
study that reported higher surgical site
infections in these groups.4 In addition,
cases with missing surgical site infection
information were excluded from the The American Congress of Obstetricians and Gynecologists/Surgical Care Improvement Project
analysis. guidelines include (1) cefazolin, cefoxitin, cefotetan, cefuroxime, or ampicillin-sulbactam, (2) clin-
This study met the criteria for damycin plus gentamicin or quinolone or aztreonam, or (3) metronidazole plus gentamicin or
“exempt” status by the University of quinolone.
Michigan Institutional Review Board- ACOG, American Congress of Obstetricians and Gynecologists; SCIP, Surgical Care Improvement Project.
Medical (HUM00073978). Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017.
From the patients included in the
analysis, we abstracted demographic
information and medical comorbidities having chewed tobacco within the past times were reported in hours from the
that included age, body mass index year). Patients with a final diagnosis start of the surgery (incision) to the
(BMI; kilograms/square meters), cova- coded 179e184 based on the primary closing of the skin incision.
riates that are associated with perfor- International Classification of Diseases, Surgical site infections were defined
mance status that included an American 9th edition, were defined as having by the Centers for Disease Control and
Association of Anesthesiology (ASA) the diagnosis of gynecologic cancer. All Prevention criteria. In brief, the defini-
classification score (defined as a dichot- other International Classification of tion includes infections that occur
omous variable as ASA class <3 or not), Diseases, 9th edition, diagnoses were within 30 days of surgery, and comprises
and preoperative medical history that defined as benign final disease. superficial surgical site infection that
included diabetes mellitus (defined as Approach to hysterectomy was cate- involves only skin and the subcutaneous
requiring oral hypoglycemic agents and/ gorized as open (all abdominal hyster- tissue of the incision, deep incisional
or insulin), hypertension (defined as ectomy cases and cases converted from surgical site infection that involves the
documentation in preoperative evalua- laparoscopic or robotic cases) or mini- fascial and muscle layers, and organ-
tion or if receiving antihypertensive mally invasive, which included laparo- space surgical site infections that
medications), and tobacco use (defined scopic, robotic, and vaginal (including include “any part of the body deeper
as smoking cigarettes, cigars, or pipe or laparoscopic-assisted) cases. Operative than fascial/muscle layers that is opened

187.e2 American Journal of Obstetrics & Gynecology AUGUST 2017

ajog.org GYNECOLOGY Original Research

or manipulated during the operative

TABLE 1
procedure.”8 In this study, deep and or-
Predictors of surgical site infection (unadjusted)
gan- space surgical site infections were
both considered “deep surgical site Any surgical site infection, n (%)
infections” because the fascia and muscle
Variable Yes (n¼329) No (n¼17,926) P value
layers of the vaginal cuff are contiguous
with the organ space. The primary Demographics
outcome of the study was a composite Age  65 y
outcome of any surgical site infection. Yes 35 (1.8) 1,941 (98.2) .913
The term any surgical site infection in- No 294 (1.8) 15,985 (98.2)
dicates that there was either a superficial
or deep surgical site infection. Superficial Non-white race
and deep infections were not analyzed Yes 92 (2.1) 4,298 (97.9) .094
separately in this analysis because there No 237 (1.7) 13,628 (98.3)
were not enough deep surgical site Body mass index, kg/m2
infection cases to power this analysis.
For all included patients, descriptive 30, obese 203 (2.3) 8,496 (97.7) <.001
and comparative statistics of de- <30, not obese 126 (1.3) 9,430 (98.7)
mographics, comorbidities, operative Medical comorbidities
details, and postoperative surgical site American Society of Anesthesiologists
infections were analyzed. For bivariate class 3
analyses, chi-square analysis or Fisher’s
Yes 104 (2.7) 3,723 (97.3) <.001
exact test were used. For continuous
variables, parametric 1-way analysis of No 225 (1.6) 14,203 (98.4)
variance or nonparametric Wilcoxon Dependent functional status
Mann-Whitney tests were used to assess Yes 3 (3.5) 82 (96.5) .23
significance in the bivariate relationship.
No 326 (1.8) 17,844 (98.2)
To ascertain the independent effect of
antibiotic categories included in the Diabetes mellitus
analysis, we constructed multivariate Present 44 (2.9) 1,485 (97.1) .001
logistic regression models. Variables Absent 285 (1.7) 16,441 (98.3)
were retained in the final model if they
Tobacco use in past year
were related to the outcome in a clini-
cally plausible manner or if they were Yes 92 (2.2) 4,072 (97.8) .025
significant at a level of 0.1 in the bivariate No 237 (1.7) 13,854 (98.3)
analysis. Hypertension
For all logistic regression models, to
Present 128 (2.3) 5,343 (97.7) <.001
account for violations in model as-
sumptions because of nonindependence Absent 201 (1.6) 12,583 (98.4)
of observations within clusters of data Cardiac disease
(hospital level), we used Huber-Eicker- Present 34 (2.2) 1,484 (97.8) .181
White robust standard errors. These
Absent 295 (1.8) 16,442 (98.2)
robust standard errors and the hospital-
level clustering allowed the model to Preoperative indication cancer
better reflect the collected data charac- Yes 54 (4.25) 1,218 (95.8) <.001
teristics. Results of the logistic regression No 275 (1.6) 16,708 (98.4)
models were confirmed with the use
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017. (continued)
of propensity score matching, which
allows for comparisons between patients
who received a different treatment
(antibiotic type) but who had a similar selection bias, confounding, and differ- metronidazole vs second-generation
surgical risk profile. The propensity ences between treatment groups in cephalosporins, and (3) cefazolin vs
score is the probability from 0e1 of each observational studies. Three separate second-generation cephalosporins. The
patient who received a particular anti- propensity score-matching analyses were propensity scoreematching approach
biotic regimen, given a set of known performed: (1) cefazolin plus metroni- was accomplished in 2 steps. First, the
variables, and is used to reduce potential dazole vs cefazolin, (2) cefazolin plus probability of receiving an antibiotic

AUGUST 2017 American Journal of Obstetrics & Gynecology 187.e3

Original Research
TABLE 1
Predictors of surgical site infection (unadjusted) (continued)
Variable
Perioperative variables
Surgical approach
Open
a cases) for cefazolin, 2.1% (n¼49 cases)
Minimally invasive
Estimated blood loss, mL
250
<250
Surgical time, h
3
<3
Bowel surgery
Yes
No
Lymph node dissection
Yes
No
a
Laparoscopic, vaginal, and robotic hysterectomy.
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017.
regimen was modeled with the use of
multiple logistic regressions. From this
regression, the propensity score for
receiving this regimen was computed for
each patient. Second, antibiotic group
patients were matched 1:2 to patients
from comparative groups based on a k-
nearest neighbor match algorithm with
common support restriction, with cali-
pers set at 0.005.
We used STATA software (version 14.0
SE for Macintosh; StataCorp LP, College
Station, TX) for all analyses.
Results
A total of 22,992 patients who had un-
dergone hysterectomy were available in
the database during the study period.
Patients with missing information
regarding the antibiotic that had been
administered (n¼418; 1.8%) and surgi-
cal site infection (n¼29; 0.1%) were
excluded. An additional 375 patients
(1.3%) were excluded because of pre-
operative sepsis (n¼12), open wound
before hysterectomy (n¼56), emergent
surgery (n¼57), and preexisting infected
187.e4 American Journal of Obstetrics & Gynecology AUGUST 2017
GYNECOLOGY ajog.org
Cefazolin alone was used for anti-
biotic prophylaxis in 82% of cases
(n¼14,971). Second-generation cepha-
Any surgical site infection, n (%) losporin was used in 13% of cases
(n¼2365). Cefazolin plus metronidazole
Yes (n¼329) No (n¼17,926) P value
was used in 5% of cases (n¼919). The
overall rate of infection was 1.8%
(n¼329 cases). The unadjusted rate of
167 (3.3) 4,903 (96.7) <.001 surgical site infection was 1.8% (n¼267
162 (1.2) 13,023 (98.8)
for second-generation cephalosporin,
and 1.4% (n¼13 cases) for cefazolin
124 (3.2) 3,802 (96.8) <.001 plus metronidazole. As highlighted in
205 (1.4) 14,124 (98.6) Table 2, patients who received cefazolin
plus metronidazole were significantly
more likely to have risk factors predictive
87 (2.8) 3,043 (97.2) <.001 of surgical site infection than patients
242 (1.6) 14,883 (98.4) who received either cefazolin or second-
generation cephalosporins.
Given the substantial differences be-
15 (12.3) 107 (87.6) <.001
tween the antibiotic cohorts with regard
314 (1.7) 17,819 (98.3) to risk factors for surgical site infection,
we compared rates of surgical site
50 (3.9) 1,216 (96.1) <.001 infection using multivariate logistic
279 (1.6) 16,710 (98.4)
regression analysis. After controlling for
patient factors (BMI, ASA category,
hypertension, diabetes mellitus, and
smoking status), treatment factors (sur-
gical time, surgical route, estimated
or dirty wound (n¼250). Patients who blood loss, bowel surgery, and lymph
received an antibiotic regimen other node dissection), and disease factors
than 1 of the 3 prophylactic antibiotic (malignancy status), we found that the
regimens that currently were the focus of risk of surgical site infection was signif-
this study were excluded (n¼3915). A icantly higher for patients who received
total of 18,255 patients (79.4%) were cefazolin (odds ratio [OR], 2.30; 95%
included in the analysis. Details of confidence interval [CI], 1.06e4.99) or
patient selection are highlighted in second-generation cephalosporin (OR,
Figure 1. 2.31; 95% CI, 1.21e4.41) compared
Table 1 highlights the demographic with those who received cefazolin plus
and operative information of the cohort metronidazole (Table 3).
by the presence or absence of surgical site The rate of surgical site infection was
infection. Patients who experienced 3.9% among patients with malignancy,
surgical site infection were significantly compared with 1.6% among patients
more likely to have BMI >30 kg/m2, ASA with benign disease. Subgroup analysis
class >3, tobacco use in past year, was performed to assess the effect of
hypertension, diabetes mellitus, cancer, antibiotic regimens among patients with
operative time of >3 hours, laparotomy, malignancy, given the significantly
estimated blood loss >250 mL, bowel higher baseline risk for surgical site
surgery, or lymph node dissection. The infection. Compared with cefazolin plus
factors in the cohort that were predictive metronidazole, the adjusted risk for
of surgical site infection (P<.1) or that surgical site infection was significantly
had clinical significance were then higher for patients who received cefa-
adjusted with the use of logistic regres- zolin (OR, 2.98; 95% CI, 1.51e7.53) or
sion models and balanced in propensity second-generation cephalosporin (OR,
score matching to compare the 3 anti- 2.25; 95% CI, 1.14e6.89) among
biotic groups. patients with malignancy. Among
ajog.org GYNECOLOGY Original Research

TABLE 2
Baseline comparison of characteristics among antibiotic groups
Cefazolin Second-generation Cefazolin þ metronidazole
Variable (n¼14,971), n (%) (n¼2365), n (%) (n¼19), n (%) P value
Demographics
Age  65 y 1481 (9.9) 338 (14.3) 157 (17.1) <.001
Non-white race 3680 (24.6) 446 (18.9) 264 (28.7) <.001
Body mass index  30 kg/m 2
7170 (47.9) 1068 (45.2) 461 (50.2) .014
Medical comorbidities
American Society of Anesthesiologists Class 3 3034 (20.3) 479 (20.3) 314 (34.2) <.001
Functional status dependent 61 (0.4) 14 (0.6) 10 (1.1) .008
Diabetes mellitus 1200 (8.0) 217 (9.2) 112 (12.2) <.001
Tobacco use in past year 3475 (32.3) 480 (20.3) 209 (22.7) .007
Hypertension 4365 (29.2) 745 (31.5) 361 (39.3) <.001
Cardiac disease 1176 (7.9) 228 (9.6) 114 (12.4) <.001
Preoperative indication cancer 872 (5.8) 230 (9.7) 170 (18.5) <.001
Perioperative variables
Surgical approach, open 3892 (26.0) 804 (34.0) 374 (40.7) <.001
Estimated blood loss,  250 mL 3098 (20.7) 501 (21.2) 327 (35.6) <.001
Surgical time  3 h 2538 (17.0) 404 (17.1) 188 (20.5) .024
Bowel surgery 71 (0.5) 24 (1.0) 27 (2.9) <.001
Lymph node dissection 813 (5.4) 219 (9.3) 234 (25.5) <.001
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017.

patients with benign disease, the
adjusted risk for surgical site infection
was higher for patients who received
cefazolin (OR, 1.7; 95% CI, 0.55e5.42)
or second-generation cephalosporin
(OR, 1.8; 95% CI, 0.68e5.11), but this
difference did not reach statistical sig-
nificance. Subgroup analysis was per-
formed to assess the effect among
patients with benign disease who had
open procedures. In this population, the
adjusted risk for surgical site infection
was significantly higher for patients who
received cefazolin (OR, 3.94; 95% CI,
1.21e12.7) compared with cefazolin
plus metronidazole. Patients who
received second-generation cephalo-
sporin also had a higher risk for infection
compared with cefazolin plus metroni-
dazole (OR, 3.32; 95% CI, 0.84e13.16),
but this difference did not reach statis-
tical significance.
Figure 2 highlights the results of the
predictive analysis based on the logistic
regression results to quantify the impact
of changing antibiotic regimen on the
rates of surgical site infection. We esti-
mate that the addition of metronidazole
to cefazolin would lead to an absolute
risk reduction of 0.8%. In other words,
the number that was needed to treat to
prevent 1 surgical site infection is
roughly 125 hysterectomy cases for the
entire cohort. However, the number that
was needed to treat was 14 for cases with
malignancy and 13 for open abdominal
benign hysterectomy. Similarly, use
of cefazolin plus metronidazole rather
than second-generation cephalosporins
would lead to an absolute risk reduction
of 0.83%. Results of the logistic regres-
sion models were confirmed with the use
of propensity score matching. Matched
cohorts were balanced with regard to all
predictive variables. Details of 2 pro-
pensity score matching analyses for the
matched and unmatched cohorts are
available in Table 4. Matched cefazolin
and cefazolin plus metronidazole co-
horts had surgical site infection rates of
1.5% and 0.8%, respectively (P¼.008).
Matched second-generation cephalo-
sporin and cefazolin plus metronidazole
cohorts had surgical site infection rates
of 2.9% and 1.2%, respectively (P¼.059).
In addition, we performed a propensity
score matching analysis that compared
cefazolin to second-generation cephalo-
sporin cohorts. Although the regression
analysis indicated similar adjusted risk
for surgical site infection for both cefa-
zolin and second-generation cephalo-
sporins, the addition of propensity score
matching allowed for a direct compari-
son between the 2 regimens. As
demonstrated in the Supplementary
Table, matched cefazolin and second-
generation cephalosporin cohorts had
surgical site infection rates of 2.9% and
2.4%, respectively (P¼.536).
Comment
In this retrospective cohort study, cefa-
zolin plus metronidazole was associated
with lower risk of posthysterectomy

AUGUST 2017 American Journal of Obstetrics & Gynecology 187.e5

Original Research GYNECOLOGY ajog.org

TABLE 3
Logistic regression model: independent predictors of surgical site infection
All Surgical Site Infection
Variable, adjusted for use in logistic
regression model Unadjusted odds ratio Adjusted odds ratio 95% Confidence interval P value
Antibiotic category
Cefazolin þ metronidazole Reference Reference Reference Reference
Cefazolin 1.3 2.30 1.06e4.99 .035
Second-generation cephalosporin 1.5 2.31 1.21e4.41 .011
Body mass index, kg/m2
<30, not obese Reference Reference Reference Reference
30, obese 1.79 1.51 1.19e1.91 .001
American Society of Anesthesiologists class
<3 Reference Reference Reference Reference
3 1.76 1.18 0.9e1.54 .223
Diabetes mellitus
Absent Reference Reference Reference Reference
Present 1.71 1.18 0.83e1.69 .353
Tobacco use in past year
No Reference Reference Reference Reference
Yes 1.32 1.46 1.14e1.87 .003
Hypertension
Absent Reference Reference Reference Reference
Present 1.5 1.04 0.81e1.34 .223
Final disease
Benign Reference Reference Reference Reference
Malignant 2.58 1.89 1.28e2.79 .001
Surgical approach
Minimally invasivea Reference Reference Reference Reference
Open 2.74 2.25 1.74e2.9 <.001
Estimated blood loss, mL
<250 Reference Reference Reference Reference
250 2.25 1.28 0.98e1.67 .076
Surgical time, h
<3 Reference Reference Reference Reference
3 1.76 1.43 1.09e1.86 .009
Bowel surgery
No Reference Reference Reference Reference
Yes 7.95 1.86 0.7e4.92 .211
Lymph node dissection
No Reference Reference Reference Reference
Yes 2.46 1.32 0.88e1.97 .182
a
Laparoscopic, vaginal, and robotic hysterectomy.
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017.

187.e6 American Journal of Obstetrics & Gynecology AUGUST 2017

ajog.org
FIGURE 2
Predicted reduction in the
absolute risk of surgical site
infection based on regression
model
The asterisk indicates that the data were
adjusted for body mass index, American Society
of Anesthesiologists category, hypertension,
diabetes mellitus, smoking status, surgical time,
surgical route, estimated blood loss, bowel
surgery, lymph node dissection, and malignancy
status. Data are given as mean and standard
error.
SSI, surgical site infection.
Till et al. Metronidazole and surgical site infection. Am J
Obstet Gynecol 2017.
surgical site infection compared with
cefazolin or second-generation cephalo-
sporin. Despite the fact that patients who
received cefazolin plus metronidazole in
this cohort had a significantly higher risk
for infectious morbidity at baseline,
the risk-adjusted rates of surgical site
infections (both in logistic regression
and propensity matched cohorts) were
lower. Of note, the rate of surgical site
infection in this study was slightly lower
than the 2.1%2 and 2.06%4 found in 2
other studies that examined surgical site
infection among patients from the
Michigan Surgical Quality Collaboration
cohort. This is due to the fact that this
analysis excluded patients who received
non-beta lactam antibiotics and those
who received antibiotics not in accor-
dance with guidelines of the College or
Surgical Care Improvement Project.
Both of these subgroups have been
shown to have a higher baseline rate of
surgical site infections.3
In our subgroup analysis, the use of
cefazolin plus metronidazole resulted in
a significant reduction in surgical site
GYNECOLOGY
infections among patients with malig-
nancy and among those who underwent
open procedures for benign indications.
There was a trend toward a reduction in
infection rate among all patients with
benign indications, but this did not
reach statistical significance. We hy-
pothesize that the baseline risk of surgi-
cal site infection is lower in patients who
undergo laparoscopic hysterectomy for
benign indications, thereby increasing
the likelihood of type II error (failure to
reject a null hypothesis).
We hypothesize that the significant
reduction in infection rate with use of
cefazolin plus metronidazole demon-
strated in our study is due to the fact that
(1) metronidazole offers improved
anaerobic antibacterial coverage and (2)
first-generation cephalosporins (cefazo-
lin in particular) generally are consid-
ered to have better Gram-positive
coverage than second-generation ceph-
alosporins. The most common organ-
isms that are isolated from vaginal cuff
infections are anaerobic vaginal flora.
Bacterial vaginosis, an overgrowth of
anaerobic vaginal flora, is highly preva-
lent, with estimates that range from
19.8e29.2%.5,9-11 However, bacterial
vaginosis is often asymptomatic and/or
untested before hysterectomy.12 The
presence of bacterial vaginosis at the
time of hysterectomy is a risk factor for
postoperative infection.9,10 Treatment
for bacterial vaginosis before hysterec-
tomy has been associated with a
decreased risk for vaginal cuff infec-
tion.13 Some studies have suggested
universal preoperative testing for bacte-
rial vaginosis with a plan-to-treat those
who are affected before surgery. A deci-
sion model analysis predicted that uni-
versal preoperative administration of
metronidazole would result in lower cuff
infection rate and lower cost compared
with preoperative screening for bacterial
vaginosis.14 Some institutions have
shifted to using second-generation
cephalosporin in an effort to improve
anaerobic coverage. However, resistance
to metronidazole is quite rare among
B fragilis groups compared with 15-25%
resistance to second-generation cepha-
losporins. In summary, cefazolin plus
metronidazole offers more complete
AUGUST 2017 American Journal of Obstetrics & Gynecology
Original Research
coverage for both skin and vaginal flora
than either cefazolin alone or second-
generation cephalosporins.
Our study differs from existing
literature in that we evaluated the com-
bination of cefazolin plus metronidazole
for hysterectomy antibiotic prophylaxis.
Several studies have demonstrated
no difference in infectious morbidity
with the addition of metronidazole to
second-generation cephalosporins. One
small randomized controlled trial
(n¼226) found no difference in post-
operative fever among patients who un-
derwent vaginal hysterectomy who were
assigned randomly to ceftriaxone (third-
generation cephalosporin) vs cefuroxime
(second-generation cephalosporin) plus
metronidazole. There were no cases of
vaginal cuff or deep pelvic infection in
either group, thus this study was likely
underpowered to detect a difference in
this outcome.15 A large nonrandomized
prospective cohort study demonstrated
no difference in rate of postoperative
infection between patients who received
cefuroxime alone vs cefuroxime plus
metronidazole.16 Of note, the infection
rate in this study was quite high across all
routes of hysterectomy, ranging from
3.9% of vaginal cases to 6.3% of abdom-
inal cases. Because our study controlled
for hospital-level effects, we may have
accounted more adequately for disparities
because of differing rates of postsurgical
infection in individual hospitals.
Cefazolin plus metronidazole has
been demonstrated to reduce infectious
morbidity compared with cefazolin or
second-generation cephalosporins in
other clean-contaminated procedures.
One randomized controlled trial
demonstrated significantly lower infec-
tion rate after cesarean delivery among
patients who received cefazolin plus
metronidazole compared with those
who received cefazolin alone.17 Two
large retrospective cohort studies in
colorectal surgery have demonstrated
a 2-fold increase in postoperative
infection for patients who received
second-generation cephalosporin alone
compared with patients who received
cefazolin plus metronidazole.18,19
Metronidazole has very low associated
risks. There are case reports of allergic
187.e7
Original Research GYNECOLOGY ajog.org

TABLE 4
Propensity score matching
Unmatched, mean % Matched, mean %
Variable (standard deviation) (standard deviation)
Cefazolin þ Cefazolin þ
Cefazolin vs cefazolin þ Cefazolin metronidazole Cefazolin metronidazole
metronidazole cohort (n¼14,971) (n¼919) P value (n¼1453) (n¼853) P value
Demographics
Age 65 y 1481 (10) 157 (17) <.001 179 (24) 128 (22) .611
Non-white race 3680 (25) 264 (29) .005 254 (33) 204 (36) .397
Body mass index, 30 kg/m 2
7170 (48) 461 (50) .181 423 (56) 308 (54) .506
Medical comorbidities
American Society of Anesthesiologists 3034 (20) 314 (34) <.001 375 (49) 275 (48) .642
class  3
Dependent functional status 61 (0) 10 (1) .003 11 (1) 6 (1) .726
Diabetes mellitus 1200 (8) 112 (12) <.001 148 (19) 101 (18) .398
Tobacco use in past year 3475 (23) 209 (23) .743 187 (25) 137 (24) .780
Hypertension 4365 (29) 361 (39) <.001 392 (52) 285 (50) .521
Cardiac disease 1176 (8) 114 (12) <.001 159 (21) 101 (18) .136
Perioperative variables
Open surgical approach 3892 (26) 374 (41) <.001 365 (48) 301 (53) .099
Malignant final disease 1228 (8) 253 (28) <.001 288 (38) 212 (37) .752
Estimated blood loss 250 mL 3098 (21) 327 (36) <.001 313 (41) 263 (46) .082
Surgical time 3 h 2538 (17) 188 (20) .006 213 (28) 166 (29) .694
Bowel surgery 71 (0) 27 (3) <.001 21 (3) 17 (3) .822
Lymph node dissection 813 (5) 234 (25) <.001 239 (31) 192 (34) .416
Outcome: surgical site infection 267 (1.8) 13 (1.4) .409 22 (1.5) 7 (0.8) .008
Second-generation Cefazolin þ Second-generation Cefazolin þ
Second-generation cephalosporins cephalosporins metronidazole cephalosporins metronidazole
vs cefazolin þ metronidazole cohort (n¼2365) (n¼919) (n¼1010) (n¼761)
Demographics
Age 65 y 338 (14) 157 (17) .045 118 (21) 97 (20) .499
Non-white race 446 (19) 264 (29) <.001 161 (29) 163 (33) .172
Body mass index 30 kg/m2 1068 (45) 461 (50) .01 314 (57) 267 (54) .379
Medical comorbidities
American Society of Anesthesiologists 479 (20) 314 (34) <.001 221 (40) 217 (44) .19
class 3
Dependent functional status 14 (1) 10 (1) .134 6 (1) 5 (1) .909
Diabetes mellitus 217 (9) 112 (12) .01 93 (17) 79 (16) .722
Tobacco use in past year 480 (20) 209 (23) .122 135 (25) 118 (24) .847
Hypertension 745 (32) 361 (39) <.001 279 (51) 241 (49) .596
Cardiac disease 228 (10) 114 (12) .02 93 (17) 79 (16) .722
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017. (continued)

and hypersensitivity reactions, but the effects are headache and nausea, but the general anesthesia is considered mini-
rate has not been defined because of low impact of single prophylactic dosing mal. Metronidazole is commonly used to
incidence. The most common adverse before abdominal procedure with treat Clostridium difficile infection, and

187.e8 American Journal of Obstetrics & Gynecology AUGUST 2017

ajog.org GYNECOLOGY Original Research

TABLE 4
Propensity score matching (continued)
Second-generation Cefazolin þ Second-generation Cefazolin þ
Second-generation cephalosporins cephalosporins metronidazole cephalosporins metronidazole
vs cefazolin þ metronidazole cohort (n¼2365) (n¼919) (n¼1010) (n¼761)
Perioperative variables
Open surgical approach 804 (34) 374 (41) <.001 255 (46) 249 (51) .161
Malignant final disease 345 (15) 253 (28) <.001 178 (32) 159 (32) .995
Estimated blood loss  250 mL 501 (21) 327 (36) <.001 201 (37) 205 (42) .086
Surgical time 3 h 404 (17) 188 (20) .024 164 (30) 133 (27) .33
Bowel surgery 24 (1) 27 (3) <.001 17 (3) 13 (3) .670
Lymph node dissection 9 (0.3) 234 (25) <.001 15 (0.4) 18 (0.4) .269
Outcome: surgical site infection 49 (2.1) 13 (1.4) .214 16 (2.9) 6 (1.2) .059
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017.

risk for the precipitation of superinfec-
tion is thought to be very low. Resistance
to metronidazole is quite low, as dis-
cussed earlier. The most likely adverse
outcomes are related to process and
systematic error, specifically that the
addition increases risk for medication
error or delay in optimal administration
timing of prophylactic antibiotics.
Overall, the risk benefit ratio favors
additional of metronidazole given
potential risk reduction of surgical site
infection.
Cefazolin plus metronidazole might
be a more cost-effective and convenient
option compared with second-
generation cephalosporin alone. At the
University of Michigan Health System,
the cost for 2 g of cefazolin (intravenous)
is $1.56; medication needs to be redosed
at 4-hours operative time. The cost for
500 mg of metronidazole (intravenous)
is $1.03; medication needs to be redosed
at 8 hours operative time. The cost for 2 g
of cefoxitin (intravenous) is $6.52;
medication needs to be redosed at
2 hours operative time. For hysterec-
tomies with operative time of >2 hours,
cefazolin plus metronidazole could
reduce the number of cases in which
redosing is either missed or not done in a
timely fashion. In addition, because of
frequent redosing, the cost of using
cefoxitin is 5 times higher than the cost
of cefazolin plus metronidazole.
The strengths of this study include a
large multiinstitutional cohort and a
well-validated database. Generalizability
could be limited, given that the database
includes hospitals only within the state
of Michigan. However, the database
draws from a wide range of hospitals and
populations (ie, community and aca-
demic settings, large and small hospital
systems, and both urban and rural areas)
across the socioeconomic spectrum.
Limitations include the retrospective,
nonrandomized design, which inher-
ently limits our ability to control for
unknown confounders. A randomized
controlled trial would allow for a more
definitive determination of the effect
of cefazolin plus metronidazole for hys-
terectomy prophylaxis. However, the
expected effect size, based on this retro-
spective study, would necessitate a very
large study cohort, which would be time-
intensive and cost-prohibitive. Patients
who received metronidazole in this
cohort were more likely to have baseline
risk factors for postoperative infection
than patients who received cephalospo-
rins alone; the nature of this database
does not allow us to infer the reason that
a surgeon chose to add metronidazole
for a particular patient. Specifically, a
higher proportion of patients with
malignancy as the preoperative indica-
tion for surgery received cefazolin plus
metronidazole compared with their
benign counterparts. Of note, we limited
our analysis to prophylactic antibiotic
administration and did not account for
additional antibiotics that may have been
administered later.
In summary, cefazolin plus metroni-
dazole appears to reduce the risk for
surgical site infection after hysterectomy
compared with cefazolin or second-
generation cephalosporin alone. Imple-
mentation of this regimen has the
potential to decrease substantially infec-
tious morbidity that is associated with
hysterectomy, particularly among patient
populations at higher risk for surgical site
infection, such as those with malignancy
and those who undergo open procedures
for benign indications. n
Acknowledgment
The authors thank Sarah Block for assistance in
preparing the manuscript; Ms Block is employed
by the University of Michigan Department of
Obstetrics and Gynecology (no funding or salary
support was provided as compensation for this
contribution).
References
1. Mittendorf R, Aronson MP, Berry RE, et al.
Avoiding serious infections associated with
abdominal hysterectomy: a meta-analysis of
antibiotic prophylaxis. Am J Obstet Gynecol
1993;169:1119-24.
2. Morgan DM, Swenson CW, Streifel KM, et al.
Surgical site infection following hysterectomy:
adjusted rankings in a regional collaborative. Am
J Obstet Gynecol 2016;214:259.e1-8.

AUGUST 2017 American Journal of Obstetrics & Gynecology 187.e9

Original Research GYNECOLOGY
3. ACOG Committee on Practice Bulletins:
Gynecology. ACOG practice bulletin no.
104: antibiotic prophylaxis for gynecologic
procedures. Obstet Gynecol 2009;113:
1180-9.
4. Uppal S, Harris J, Al-Niaimi A, et al. Prophy-
lactic antibiotic choice and risk of surgical site
infection after hysterectomy. Obstet Gynecol
2016;127:321-9.
5. Soper DE, Bump RC, Hurt WG. Bacterial
vaginosis and trichomoniasis vaginitis are risk
factors for cuff cellulitis after abdominal hyster-
ectomy. Am J Obstet Gynecol 1990;163:
1016-23.
6. Snydman DR, Jacobus NV, McDermott LA,
et al. Update on resistance of Bacteroides fragilis
group and related species with special attention
to carbapenems 2006-2009. Anaerobe
2011;17:147-51.
7. Aldridge KE, Sanders CV. Susceptibility
trending of blood isolates of the Bacteroides
fragilis group over a 12-year period to clinda-
mycin, ampicillinesulbactam, cefoxitin, imipe-
nem, and metronidazole. Anaerobe 2002;8:
301-5.
8. Centers for Disease Control. Procedure-
associated Module SSI (January 2015; modi-
fied 2015). Available at: http://www.cdc.
gov/nhsn/PDFs/pscManual/9pscSSIcurrent.pdf.
Accessed December 7, 2016.
9. Lin L, Song J, Kimber N, et al. The role of
bacterial vaginosis in infection after major gyne-
cologic surgery. Infect Dis Obstet Gynecol
1999;7:169-74.
187.e10 American Journal of Obstetrics & Gynecology AUGUST 2017
10. Larsson PG, Platz-Christensen JJ,
Forsum U, Påhlson C. Clue cells in predicting
infections after abdominal hysterectomy. Obstet
Gynecol 1991;77:450-2.
11. Allsworth JE, Peipert JF. Prevalence of
bacterial vaginosis. Obstet Gynecol 2007;109:
114-20.
12. Klebanoff MA, Schwebke JR, Zhang J,
Nansel TR, Yu K-F, Andrews WW. Vulvovaginal
symptoms in women with bacterial vaginosis.
Obstet Gynecol 2004;104:267-72.
13. Larsson P-G, Carlsson B. Does pre- and
postoperative metronidazole treatment lower
vaginal cuff infection rate after abdominal hys-
terectomy among women with bacterial vagi-
nosis? Infect Dis Obstet Gynecol 2002;10:
133-40.
14. McElligott KA, Havrilesky LJ, Myers ER.
Preoperative screening strategies for bacterial
vaginosis prior to elective hysterectomy: a cost
comparison study. Am J Obstet Gynecol
2011;205:500.e1-7.
15. Brouwer WK, Hoogkamp Korstanje JA.
Single dose ceftriaxone versus single dose
cefuroxime plus metronidazole for preventing
febrile morbidity and urinary tract infection in
vaginal hysterectomy. Eur J Obstet Gynecol
Reprod Biol 1995;61:143-6.
16. Brummer THI, Heikkinen A-M, Jalkanen J,
et al. Antibiotic prophylaxis for hysterectomy, a
prospective cohort study: cefuroxime, metroni-
dazole, or both? BJOG 2013;120:1269-76.
17. Meyer NL, Hosier KV, Scott K,
Lipscomb GH. Cefazolin versus cefazolin plus
ajog.org
metronidazole for antibiotic prophylaxis at ce-
sarean section. South Med J 2003;96:992-5.
18. Deierhoi RJ, Dawes LG, Vick C, Itani KMF,
Hawn MT. Choice of intravenous antibiotic pro-
phylaxis for colorectal surgery does matter. J Am
Coll Surg 2013;217:763-9.
19. Hendren S, Fritze D, Banerjee M, et al.
Antibiotic choice is independently associated
with risk of surgical site infection after colectomy:
a population-based cohort study. Ann Surg
2013;257:469-75.
Author and article information
From the Department of Obstetrics and Gynecology (Drs
Till, Morgan, Pearlman, and Uppal), the Gynecology
Health Services Group (Drs Morgan and Uppal), the
Institute for Healthcare Policy and Innovation (Drs
Abdelsattar and Uppal), and the Department of Surgery
(Dr Campbell), University of Michigan, Ann Arbor, MI; the
Department of Obstetrics and Gynecology, St. John
Hospital and Medical CentereSt. John Providence,
Detroit, MI (Dr Bazzi); and the Department of Surgery,
Mayo Clinic, Rochester, MN (Dr Abdelsattar).
Received Dec. 7, 2016; revised March 11, 2017;
accepted March 21, 2017.
Supported in part (partial salary) by the Michigan
Surgical Quality Collaborative (D.M.M.).
The authors report no conflict of interest.
Presented in abstract form at the American Associa-
tion of Gynecologic Laparoscopists Global Congress,
Orlando, Florida, November 14-18, 2016.
Corresponding author: Shitanshu Uppal, MBBS.
Uppal@med.umich.edu
ajog.org GYNECOLOGY Original Research

SUPPLEMENTARY TABLE
Propensity score matching for cefazolin vs second-generation cephalosporins
Unmatched, mean % Matched, mean %
(standard deviation) (standard deviation)
Second- generation Second- generation
Cefazolin cephalosporins Cefazolin cephalosporins
Variable (n¼14,971) (n¼2365) P value (n¼3743) (n¼2288) P value
Demographics
Age 65 y 1481 (10) 338 (14) <.001 317 (28) 245 (27) .474
Non-white race 3680 (25) 446 (19) <.001 305 (27) 243 (27) .771
Body mass index 30 kg/m 2
7170 (48) 1068 (45) .013 640 (57) 506 (56) .459
Medical comorbidities
American Society of Anesthesiologists 3034 (20) 479 (20) .989 481 (43) 361 (40) .127
class 3
Dependent functional status 61 (0) 14 (1) .204 9 (1) 12 (1) .256
Diabetes mellitus 1200 (8) 217 (9) .056 248 (22) 182 (20) .232
Tobacco use in past year 3475 (23) 480 (20) .002 257 (23) 211 (23) .916
Hypertension 4365 (29) 745 (32) .02 603 (54) 484 (53) .737
Cardiac disease 1176 (8) 228 (10) .003 251 (22) 186 (20) .273
Perioperative variables
Open surgical approach 3892 (26) 804 (34) <.001 499 (45) 429 (47) .259
Malignant final disease 1228 (8) 345 (15) <.001 381 (34) 292 (32) .344
Estimated blood loss 250 mL 3098 (21) 501 (21) .585 384 (34) 292 (32) .283
Surgical time 3 h 2538 (17) 404 (17) .876 344 (31) 275 (30) .789
Bowel surgery 71 (0) 24 (1) .001 25 (2) 20 (2) .954
Lymph node dissection 813 (5) 9 (0.3) <.001 9 (0.3) 9 (0.3) .547
Outcome: surgical site infection 267 (1.8) 49 (2.1) .33 32 (2.9) 22 (2.4) .536
Till et al. Metronidazole and surgical site infection. Am J Obstet Gynecol 2017.

AUGUST 2017 American Journal of Obstetrics & Gynecology 187.e11