Antibiotic Choice for the Prophylaxis of

Post-Cataract Extraction Endophthalmitis

Christopher J. Rudnisky, MD, MPH, Darwin Wan, MD, Ezekiel Weis, MD, MPH
Objective: To determine the 8-year incidence of endophthalmitis after cataract surgery and to determine
which surgical practices were associated with higher rates of endophthalmitis.
Design: Case-control study.
Participants: A total of 75 318 eyes undergoing cataract extractions, performed by 26 different surgeons at
4 public hospitals and 5 nonhospital surgical facilities.
Methods: Cases of endophthalmitis were acquired using a detailed, prospectively designed demographic
database. Controls were tabulated using volume data available from the provincial health care system.
Main Outcome Measures: The primary outcome was the development of endophthalmitis.
Results: A total of 23 cases (13 with culture-positive results) of postoperative endophthalmitis occurred,
yielding an overall 8-year incidence of 0.03%. The incidence of endophthalmitis varied between surgeons from
0% to 0.20%. Two surgeons had higher rates than the rest of the group: 1 high-volume surgeon
(1059.4231.9 mean cases per year) with an incidence of 0.08% (n 7; P 0.004) and 1 low-volume surgeon
(123.544.8 mean cases per year) with an incidence of 0.20% (n 2; P 0.002). On univariate analysis, the
rate of endophthalmitis was not inuenced by the use of intracameral (0.898) or subconjunctival antibiotics
(0.331), whereas the use of moxioxacin was associated with a lower rate of endophthalmitis (P 0.029).
Surgery at 1 private facility (P 0.046) and the use of timolol at the end of the procedure (P 0.007) were
associated with a higher rate of endophthalmitis. Multivariate analysis demonstrated that the odds of
endophthalmitis was lower if a second-generation (P 0.02) or fourth-generation (P 0.008) uoroquinolone
antibiotic was used after surgery. In contrast, the odds of endophthalmitis occurring was higher if timolol (P
0.0002) was used at the end of the procedure or if the surgery was performed at one of the private facilities
(P 0.009).
Conclusions: The rate of endophthalmitis was lower if a uoroquinolone was used after surgery. In contrast,
endophthalmitis was more likely to occur if timolol was used at the end of the procedure or if surgery was
performed at one of the private facilities. Ophthalmology 2014;121:835-841 2014 by the American Academy of
Ophthalmology.
Supplemental material is available at www.aaojournal.org
Endophthalmitis is a visually devastating complication of
cataract surgery. Its incidence ranges from0.028%to 0.345%
1,2
and has varied over time because of changes in surgical tech-
nique,
3
such as the widespread adoption of clear corneal
incisions. However, apart from povidoneeiodine antisepsis,
4
there is no clear consensus regarding prophylaxis. A large,
prospective, randomized, multicenter study initiated by
the European Society of Cataract & Refractive Surgeons
1
found that intracameral cefuroxime reduces the rate of
endophthalmitis; although this was the rst prophylactic
regimen shown to be effective, it has not changed practice
5
because the study had a signicantly higher rate of
endophthalmitis (0.074%) than expected. Fourth-generation
uoroquinolones, which are effective in a laboratory environ-
ment,
6
have not yet been proven to have a clinical prophylactic
benet.
The large volume of cataract surgery performed and
uniform data acquisition in our health region allow for well-
powered comparisons of different prophylactic routines in
patients undergoing phacoemulsication cataract extraction.
The purpose of the present study was to determine the
incidence of postoperative endophthalmitis after cataract
surgery over an 8-year period and to determine which sur-
gical practices, if any, were associated with a higher rate of
endophthalmitis.
Methods
Before initiation of this population-based case-control study,
approval was obtained from the Health Research Ethics Board
at the University of Alberta. A single tertiary retina service practice
provides care for the entire region from a single hospital.
This hospital has used a detailed, prospectively designed de-
mographic database since 2002. This regional retinal service pro-
vides a unique opportunity to capture all endophthalmitis cases
occurring after cataract surgery for many surgeons from multiple
sites. Controls were tabulated using volume data available from the
provincial health care system.
835 2014 by the American Academy of Ophthalmology ISSN 0161-6420/14/$ - see front matter
Published by Elsevier Inc. http://dx.doi.org/10.1016/j.ophtha.2013.08.046
Similar to the Endophthalmitis Vitrectomy Study,
7
the
diagnosis of endophthalmitis was made clinically, and all cases,
including culture-negative endophthalmitis, were considered posi-
tive. Patients were included if they: (1) underwent cataract surgery
between April 1, 2002, and March 31, 2009, in an Edmonton zone
public or private facility, and (2) demonstrated clinical signs and
symptoms of endophthalmitis within 42 days of cataract extraction.
Exclusion criteria included: (1) patients in whom endophthalmitis
developed after intraocular surgery other than cataract extraction,
(2) patients who underwent combined cataract extraction and vit-
rectomy, or (3) patients with chronic endophthalmitis (developing
more than 42 days after cataract extraction).
Surgeon parameters and operative preferences by year were
obtained retrospectively from hospital records and were veried
with each surgeon. This information was then used to populate
covariates for cases and controls. Parameters and preferences
collected included the type of preoperative antibiotics used before
the surgical date, topical antibiotics administered at the facility
before surgery (on the same day), incision type and location, blade
size, type of intraoperative antibiotics and route of delivery (sub-
conjunctival, intracameral, topical), administration of intra-
operative steroid (topical, injected), topical glaucoma agent, and
type of postoperative antibiotics used. According to institution
policies, all patients were prepped using a sterilizing agent, the
most common of which was povidoneeiodine solution; patients
with an iodine allergy were prepped using chlorhexidine. Because
a detailed database of specic preparations was not available for
controls, this covariate was not included in the analysis.
SAS statistical software version 9.2 (SAS Inc., Cary, NC) was
used for all analyses. The association between each covariate and
the development of endophthalmitis was analyzed using the chi-
square test. Logistic regression was used for continuous data.
For the purposes of analysis, uoroquinolones were grouped
together by generation: second-generation uoroquinolones,
ooxacin and ciprooxacin; fourth-generation uoroquinolones,
gatioxacin and moxioxacin. Tests for trend were carried out
using the Mantel-Haenszel chi-square test. Antibiotics were also
analyzed as a binary variable (used versus not used). Surgeon
mean annual volume was categorized according to previously
described groupings: 50 to 250 procedures per year, 251 to 500
procedures per year, 501 to 1000 procedures per year, and more
than 1000 procedures per year.
8
A multivariate logistic regression model was constructed to
determine which variables were most predictive of endoph-
thalmitis. Only covariates with a statistically signicant relation-
ship on univariate analysis (P<0.1) were included. Post hoc
exploratory simple regression analysis was performed to assist in
the interpretation of the analysis of postoperative antibiotic choice.
Power calculations, using an a error of 0.05, were performed
for analyses that had nonsignicant results to determine if the study
was powered to detect a clinically signicant difference in
endophthalmitis rates. For these calculations, a reference rate
indicating a clinically signicant increase in endophthalmitis rates
was set at 0.1%.
Results
During the study period, 75 692 cataract extractions were per-
formed by 29 different surgeons. Covariate information regarding
surgical preferences and technique could not be obtained from
3 surgeons. Their cases, accounting for 0.49% of the data, were
excluded. No cases of endophthalmitis occurred after cataract ex-
tractions by these 3 surgeons.
A total of 75 318 cataract extractions, performed by 26 different
surgeons, were included in the analysis. If patients were equally
split between comparative groups, this study had 88.7% power to
detect a difference in endophthalmitis rates of 0.04%.
Surgery was performed at 4 public hospitals and 5 nonhospital
surgical facilities, where publicly funded care is delivered in a
private setting. Twenty-six cases of postoperative endophthalmitis
occurred, 3 of which were diagnosed more than 42 days after
cataract surgery. These cases of chronic endophthalmitis were
excluded, leaving a total of 23 cases of acute endophthalmitis
occurring after cataract extraction (Table 1, available at
www.aaojournal.org) and yielding an overall 8-year incidence of
0.03%. The incidence ranged from as low as 0.01% in 2004 and
2008 to as high as 0.07% in 2006 (Fig 1). Thirteen cases showed
positive culture results; the overall 8-year incidence of culture-
positive postoperative endophthalmitis was 0.03%.
The incidence of endophthalmitis varied between surgeons
from 0% to 0.20%. Surgeon volume by year varied signicantly
from a low of 4 cases to a high of 1586 cases (Fig 2), but there was
no difference in the rate of endophthalmitis by surgeon mean
annual volume (P 0.507). Two surgeons were found to have
higher rates than the rest of the group: 1 high-volume surgeon
(1059.4231.9 mean cases per year) with an incidence of 0.08%
(n 7; P 0.004) and 1 low-volume surgeon (123.544.8 mean
cases per year) with an incidence of 0.20% (n 2; P 0.002).
Most surgical procedures (n 50 471 [67.0%]) were performed
at a single hospital (Fig 3). There was no difference in the rate of
endophthalmitis when comparing each hospital and nonhospital
surgical facilities grouped together (P 0.769). However,
among the 5 (private) nonhospital surgical facilities, one center
had a statistically signicantly higher rate of endophthalmitis
than the other facilities (0.21%; P 0.046).
Preoperative Antibiotics
A slim majority of surgeons (n 14 [53.8%]) used preoperative
antibiotics on the days leading up to surgery (Fig 4). Of the
surgeons prescribing preoperative antibiotics, 84.6% (n 11)
used a uoroquinolone. In the univariate analysis, neither the
choice to use any preoperative antibiotics (Table 2; P 0.132)
nor any specic antibiotic (P 0.135) affected the rate of
endophthalmitis. Using the reference rate, there was 99.6%
power to detect a 0.1% difference in endophthalmitis rates
between use and nonuse of preoperative antibiotics.
Same-Day Antibiotics
Most surgeons (n 21 [80.8%]) applied antibiotic drops during the
preparation phase on the day of surgery, and all but one (n 20
[95.2%]) used a uoroquinolone (Fig 5). In the univariate analysis,
there was no difference in the rate of endophthalmitis among
different antibiotic choices (P 0.890) or between use and
nonuse (Table 2; P 0.817). Using the reference rate, this
analysis had 88.9% power to detect a difference between use and
nonuse.
Figure 1. Graph showing the incidence rate of endophthalmitis by year.
Ophthalmology Volume 121, Number 4, April 2014
836
Intracameral Antibiotics
Only 6 surgeons (23.1%) used intracameral antibiotics; 5 used
vancomycin and 1 used moxioxacin. Overall, 20.7% of study eyes
(n 15 561) were treated with intracameral antibiotics, for which an
endophthalmitis rate of 0.03% was observed. This was the same as
in eyes not treated with intracameral antibioticsdthere was no sta-
tistically signicant difference between these rates (P 0.898) on
univariate analysis (Table 2). Using the reference rate, this analysis
had 83.9% power to detect a difference between use and nonuse.
Subconjunctival Antibiotics
Subconjunctival antibiotics were used much more commonly, with
11 surgeons (42.3%) using either cefazolin (n 6), gentamicin
(n 4), or cefuroxime (n 1). More than one third of eyes
(35.8%; n 26 971) received an injection of subconjunctival an-
tibiotics at the end of their surgery, but again, there was no sta-
tistically signicant difference in the endophthalmitis rate (P
0.331) in the univariate analysis (Table 2). Even if intracameral and
subconjunctival antibiotics were grouped together and analyzed as
a single variable, there was still no difference in the rate of
endophthalmitis (P 0.209) compared with those that received
neither. Using the reference rate, this analysis had 99.2% power
to detect a difference between use and nonuse.
Intraoperative Steroids
Most surgeons used intraoperative steroids (topical, injected, or
both) as part of the procedure (n 21 [80.8%]). Accordingly,
74.1% of procedures used intraoperative steroids, with 76.2% of
these applying a topical antibiotic and steroid combination drop at
the end of the procedure, 13.7% using subconjunctival dexa-
methasone or betamethasone, and 10.1% using both. In the uni-
variate analysis (Table 2), there was no statistical difference in
endophthalmitis rates between eyes that received subconjunctival
(P 0.259) or topical (P 0.575) steroids. Using the reference
rate, this analysis had 90.6% power to detect a difference
between use and nonuse.
Topical Glaucoma Medications
Twelve surgeons (46.2%) used topical glaucoma medications at the
end of the procedure (n 40 193) to try to prevent a postoperative
intraocular pressure spike. Endophthalmitis developed twice as
frequently in these eyes (0.04%) in comparison with cases in which
Figure 2. Bar graph showing the surgical volume (y-axis) by surgeon (x-axis) and year (z-axis).
Figure 3. Graph showing the number of procedures performed by facility.
NHSF nonhospital surgical facilities; RAH Royal Alexandra Hospital.
Figure 4. Graph showing the number of eyes treated with preoperative
antibiotics.
Rudnisky et al

Prevention of Endophthalmitis after Cataract Surgery
837
topical glaucoma therapy was not used (0.02%; P 0.048). Six
different medications were used: 8.0% received pilocarpine, 39.7%
received brimonidine, 6.2% received Iopidine, 2.6% received
timolol, 16.4% received levobunolol, and 27.2% received betaxolol.
The rate of endophthalmitis varied between 0% (pilocarpine) and
0.20% (timolol), with most being between 0.02% and 0.08%; in the
univariate analysis, the rate of endophthalmitis in timolol-treated
eyes was statistically signicantly higher than in eyes treated with
the other medications (P 0.007). Using the reference rate, there
was 85.9% power to detect a difference between use and nonuse.
Postoperative Antibiotics
All eyes received postoperative antibiotics (Fig 6): 37.5% received
polymyxin B and trimethoprim, 35.1% received moxioxacin,
9.5% received ooxacin, 8.2% received ciprooxacin, 4.3%
received gatioxacin, 4.0% received both moxioxacin and
tobramycin, 1.0% received polymyxin B, and 0.5% received
both moxioxacin and neomycin and polymyxin B (Maxitrol).
Because of these antibiotic combinations and the absence of an
obvious control group (everyone received some form of post-
operative antibiotic), individual indicator variables were created for
each antibiotic. Using all other antibiotics grouped together as
a control in each analysis (Table 2), the rate of endophthalmitis
ranged between 0% (ooxacin) and 0.09% (Polysporin). Only
moxioxacin demonstrated a statistically signicant lower rate of
endophthalmitis (0.01%) in comparison with all others grouped
together (0.04%; P 0.029). Interestingly, if antibiotics were
grouped by class, there was a trend (P 0.04) for the rate of
endophthalmitis to decline as a stronger antibiotic class was used
(Table 3), with the lowest rate of endophthalmitis (0.02%)
observed if a fourth-generation uoroquinolone was used. If
a power calculation were performed to evaluate the strength of the
comparison between the use of a fourth-generation uoroquinolone
versus any other antibiotic, the analysis had 99.5% power to detect
a difference.
Wound Construction
Almost all cases of endophthalmitis (n 22) developed in eyes in
which a temporal incision was used. However, the difference in
endophthalmitis rates between temporal (0.04%) and superior
(0.01%) incisions was not statistically signicant (P 0.06) in the
univariate analysis (Table 2). Regarding wound types, 61.8% of
procedures were clear corneal, 33.4% were limbal, and 4.7%
were a scleral tunnel; there was no statistically signicant
difference in the rate of endophthalmitis between wound types
(P 0.564). The average wound size for endophthalmitis cases
was 2.810.13 mm in comparison with 2.780.14 mm, but this
difference was not statistically signicant (P 0.337).
Antibiotic Combinations
An exploratory analysis was performed to determine if certain
combinations of antibiotic choices improved prophylaxis. To
assess this, surgeon prophylactic regimens were awarded 1 point
for each: if a fourth-generation uoroquinolone was used before
surgery, if either a subconjunctival or intracameral injection was
performed during surgery, and if a fourth-generation uo-
roquinolone was used after surgery. As such, points ranged bet-
ween 0 and 3. Eyes were divided fairly evenly between point
categories: 26.1% had 0 points, 30.4% had 1 point, 17.1% had
2 points, and 26.4% had 3 points. There was a statistically
Table 2. Univariate Analyses of Endophthalmitis Covariates
Variable
Endophthalmitis Rate (%)
P Value* Yes No
Public surgical facility 0.03 0.03 0.769
Preoperative antibiotics 0.02 0.04 0.132
Antibiotic choices
Same day 0.03 0.03 0.817
Intracameral 0.03 0.03 0.898
Subconjunctival 0.02 0.04 0.331
Intraoperative steroids
Subconjunctival
dexamethasone
0.02 0.03 0.259
Topical 0.03 0.03 0.575
Topical glaucoma
medications
0.04 0.02 0.048
Temporal wound
location
0.04 0.01 0.060
Postoperative antibiotic choice
Polymyxin B 0.09 0.03 0.260
Tobramycin 0.04 0.02 0.300
Ooxacin 0 0.03 0.121
Ciprooxacin 0.07 0.03 0.106
Moxioxacin 0.01 0.04 0.029
Gatioxacin 0.06 0.03 0.290
*Chi-square test.
Figure 5. Graph showing the number of eyes treated with same-day an-
tibiotics during the preoperative preparatory phase conducted at the fa-
cility. Note that for tobramycin, n 35.
Figure 6. Graph showing the number of eyes treated with each post-
operative antibiotic choice (dark bars; left y-axis). Light grey diamonds
indicate the rate of endophthalmitis (right y-axis) that was observed for
each antibiotic choice.
Ophthalmology Volume 121, Number 4, April 2014
838
signicant trend for the rate of endophthalmitis to decline as the
point value increased (P 0.04).
Multivariate Exploratory Analysis
Because a single private surgical center, use of timolol, and tem-
poral wound placement were associated with the development of
endophthalmitis on univariate analysis and because moxioxacin
alone a higher antibiotic class, and point score seemed protective,
these variables were included in an exploratory multivariate model.
Timolol was analyzed as a categorical variable with the following
categories: 0 no glaucoma medication used, 1 timolol used,
2 betaxolol or levobunolol used, and 3 any other topical
glaucoma agent. Antibiotic class also was analyzed categorically:
1 polymyxin B, 2 tobramycin, 3 ooxacin or ciprooxacin,
and 4 moxioxacin or gatioxacin.
The variables with the weakest association with endoph-
thalmitis rates were moxioxacin (P 0.916) and antibiotic point
score (P 0.487); these variables were removed from the model as
part of backward, stepwise logistic regression. The nal model
(Table 4) demonstrated that the use of timolol (P 0.0002) or
surgery at one of the private surgical suites (P 0.009) were
associated independently with a higher rate of endophthalmitis,
whereas the use of a second-generation uoroquinolone
(P 0.019) or fourth-generation uoroquinolone (P 0.008) was
independently associated with a lower rate of endophthalmitis.
Discussion
From a total of 75 318 cataract surgeries performed by
26 surgeons at 9 Edmonton-area hospitals or ambulatory
surgical centers, 23 cases of endophthalmitis developed,
yielding an overall incidence of 0.03%. This is on the lower
end of incidences reported in the literature.
1,2,9e11
In a
systematic review of 3 140 650 cataract surgeries, Taban
et al
3
found an aggregate incidence of endophthalmitis
occurring after cataract surgery of 0.128%. Our study
demonstrates that a low endophthalmitis rate can be
achieved in a group of surgeons using heterogeneous
prophylactic regimens.
Multivariate analysis found that the use of timolol was
associated with a higher rate of endophthalmitis. There is
controversy over whether intraocular pressure-lowering
agents can prevent an intraocular pressure spike effec-
tively after surgery.
12e18
Before conducting this study, the
authors would not have hypothesized that there would be
any disadvantage to using topical glaucoma agents at the
end of cataract surgery. Possible explanations for the asso-
ciation between timolol and endophthalmitis could be bottle
tip contamination
19
or transient, postoperative hypotony
with resultant ingress of uid through the wound.
However, only one surgeon used timolol, and it was the
surgeon with the highest rate of endophthalmitis (0.20%).
It is therefore unclear if this signicant association is
because of differences in technique, the use of timolol, a
patient population with a higher baseline risk of
endophthalmitis, or other unmeasured surgeon or patient
confounders. This nding is important because it suggests
that further research into the use of topical b-blockers in
this setting is warranted.
This study also found that surgery at one of the private
surgical centers was associated with a higher rate of
endophthalmitis. However, there was a low volume of
surgery (n 944) performed at this center, which suggests
that chance alone may explain this nding; one fewer case
and there would be no statistically signicant relationship.
However, it is also possible that this center has a higher
rate because it is a low-volume site. The surgeons who
operate there also operate at other sites; their overall rate
was not different in comparison with that of their peers. As
such, it suggests the difference observed could be related to
location-based factors, whether they be surgical assistants,
management procedures, air quality issues, differences in
patient preparatory technique, or unmeasured patient-
specic variables. Although the exact reason is impor-
tant, it cannot be determined with any accuracy from these
data; it is noteworthy that one case occurred in 2005 and
the other in 2007, indicating there was not a so-called mini-
epidemic.
This study suggests that uoroquinolones are a superior
antibiotic prophylactic choice for the prevention of endoph-
thalmitis; multivariate analysis demonstrated that both second-
and fourth-generation uoroquinolones decrease the odds of
endophthalmitis when used after surgery. In vitro and in vivo
animal models
20
have shown that fourth-generation uo-
roquinolones have superior efcacy compared with second-
Table 3. Endophthalmitis Rates by Antibiotic Class
Antibiotic
No. of
Endophthalmitis
Cases
No. of
Control
Participants
Rate
(%)
P
Value
Polymyxin B or
polymyxin
B trimethoprim
1 738 0.14 0.04*
Tobramycin 12 28 234 0.04
Ooxacin or
ciprooxacin
4 13 284 0.03
Moxioxacin or
gatioxacin
6 33 039 0.02
*Mantel-Haenszel chi-square test for trend.
Table 4. Final Multivariate Exploratory Logistic Regression Model
Variable Odds Ratio
95% Wald
Condence
Limits P Value
Private surgical facility 10.57 1.81e61.79 0.009
Glaucoma agent*
Timolol 33.61 5.25e215.28 0.0002
Levobunolol or
betaxolol
3.27 0.91e11.70 0.069
Other topical agent 2.74 0.89e8.47 0.079
Postoperative antibiotic class
y
Tobramycin 0.18 0.02e1.52 0.115
Ooxacin or
ciprooxacin
0.05 0.004e0.61 0.019
Moxioxacin or
gatioxacin
0.04 0.004e0.44 0.008
*Categorical variable; reference category was no glaucoma agent used.
y
Categorical variable; reference category was polymyxin B.
Rudnisky et al

Prevention of Endophthalmitis after Cataract Surgery
839
and third-generation uoroquinolones in terms of potency and
the development of bacterial resistance.
6
Similarly, there were
statistically stronger and slightly lower odds of developing
endophthalmitis when a fourth-generation uoroquinolone
was used (Table 4). Although these ndings are statistically
signicant and were detected within a large overall sample,
this study still reects only 23 cases of endophthalmitis. As
such, these results contribute to, rather than resolve, the
debate over which postoperative antibiotic choice is the best.
Unfortunately, the optimal overall strategy remains contro-
versial. The European Society of Cataract & Refractive Sur-
geons initiated a prospective, randomized, partially masked,
multicenter study in 2003
1
demonstrating that intracameral
cefuroxime is effective in reducing the rate of postoperative
endophthalmitis. Although the injection of intracameral
antibiotics has been supported by other authors,
21
this study
did not nd a protective effect. Indeed, adoption of the
European Society of Cataract & Refractive Surgeons
recommendations has been slow; a subsequent survey
revealed that only 6% of respondents reported injecting
intracameral cefuroxime, and 77% did not plan to start,
5
perhaps in part because there are large studies in which
intracameral antibiotics were not used
11
that have reported
lower endophthalmitis rates (0.04%) than those observed in
the European Society of Cataract & Refractive Surgeons
intracameral antibiotic group (0.07%). Similarly, previous
reports have suggested that subconjunctival antibiotics are
associated with decreased endophthalmitis rates,
22,23
but this
also was not observed in our study, despite a power of 99%.
Surgeon volume was recently reported to be associated
with endophthalmitis. Keay et al
24
found that surgeons with
annual volumes of 50 cases or fewer had a relative risk of
3.8 (95% condence interval, 3.13e4.61) in comparison
with surgeons who perform more than 1000 cases
annually. A recent letter about this study, and the author
response,
25
suggested that the explanation for this gradient
may be a choice by high-volume surgeons for more inten-
sive prophylactic strategies, such as intracameral antibiotics.
In our study, one of the highest rates of endophthalmitis was
experienced by a surgeon who had the third highest volume
of 26 surgeons (mean, 1059 cases/year). The discrepancy
between our ndings and those of Keay et al, whose
study was very large, may be because the variables in
endophthalmitis studies could be surrogates for something
else more specic. Intuitively, patient ora, postoperative
behavior, and activities, which are unmeasurable, or other
unmeasured surgeon preferences, such as wound construc-
tion technique, could inuence infection risk. Similarly,
strategies such as patching after surgery or other unmea-
sured intraoperative details also could affect the rate of
endophthalmitis. Clearly, if an answer is to be found, the net
must be cast wider than the usual focus on costly prophy-
lactic antibiotics and their differences.
There are a number of limitations of this study. Endoph-
thalmitis studies are often criticized for missing cases, and
although it is possible that a case could have been missed if a
patient sought care at another hospital in Alberta, it is un-
likely given that the retina service for the region (Edmonton
Health Zone) operates out of only 1 hospital and provides
service to patients in a 2 000 000-km
2
catchment area.
Similarly, the overwhelming majority of patients undergoing
cataract surgery reside in the same catchment area. Another
limitation is the retrospective method by which covariates for
cases and controls were constructed. Recall bias by the
cataract surgeons may have affected the assembly of surgeon
preferences that were used to populate covariate data for
controls. For example, patients who are allergic to penicillin
would not have received subconjunctival cefazolin injections,
and that discrepancy is not captured in this dataset. However,
this weakness has little impact given that there would be
induced bias in favor of the use of antibiotic strategies, and
yet no associations were found and the study was powered
adequately. Another deciency is that most study eyes would
represent the right and left eyes of a single patient. Given the
correlation between eyes, the analyses performed do not ac-
count for intereye correlation, and it is possible that even the
signicant results are actually underpowered. Unfortunately,
the extent to which this issue affects the study cannot be
quantied because of the indirect and retrospective method
used to obtain control data. However, nonsimultaneous
sequential surgery, which almost every eye would have un-
dergone, is performed to mitigate infection risk. As a result,
there may be less correlation between eyes for an outcome
like endophthalmitis than one might normally observe.
In this large, population-based, case-control study, the rate
of endophthalmitis was lower if a uoroquinolone was used
after surgery. In contrast, endophthalmitis was more likely to
occur if timolol was used at the end of the procedure or if
surgery was performed at one of the private facilities.
References
1. ESCRS Endophthalmitis Study Group. Prophylaxis of post-
operative endophthalmitis following cataract surgery: results
of the ESCRS multicenter study and identication of risk
factors. J Cataract Refract Surg 2007;33:97888.
2. Wykoff CC, Parrott MB, Flynn HWJr, et al. Nosocomial acute-
onset postoperative endophthalmitis at a university teaching
hospital (2002e2009). Am J Ophthalmol 2010;150:3928.
3. Taban M, Behrens A, Newcomb RL, et al. Acute endoph-
thalmitis following cataract surgery: a systematic review of the
literature. Arch Ophthalmol 2005;123:61320.
4. Ciulla TA, Starr MB, Masket S. Bacterial endophthalmitis
prophylaxis for cataract surgery: an evidence-based update.
Ophthalmology 2002;109:1324.
5. Chang DF, Braga-Mele R, Mamalis N, et al; ASCRS Cataract
Clinical Committee. Prophylaxis of postoperative endoph-
thalmitis after cataract surgery: results of the 2007 ASCRS
member survey. J Cataract Refract Surg 2007;33:18015.
6. Mather R, Karenchak LM, Romanowski EG, Kowalski RP.
Fourth generation uoroquinolones: new weapons in the
arsenal of ophthalmic antibiotics. Am J Ophthalmol 2002;133:
4636.
7. Endophthalmitis Vitrectomy Study Group. Results of the
Endophthalmitis Vitrectomy Study: a randomized trial of im-
mediate vitrectomy and of intravenous antibiotics for the
treatment of postoperative bacterial endophthalmitis. Arch
Ophthalmol 1995;113:147996.
8. Bell CM, Hatch WV, Cernat G, Urbach DR. Surgeon volumes
and selected patient outcomes in cataract surgery: a population-
based analysis. Ophthalmology 2007;114:40510.
Ophthalmology Volume 121, Number 4, April 2014
840
9. Aaberg TM Jr, Flynn HW Jr, Schiffman J, Newton J. Noso-
comial acute-onset postoperative endophthalmitis survey. A
10-year review of incidence and outcomes. Ophthalmology
1998;105:100410.
10. Freeman EE, Roy-Gagnon MH, Fortin E, et al. Rate of
endophthalmitis after cataract surgery in Quebec, Canada,
1996e2005. Arch Ophthalmol 2010;128:2304.
11. Lloyd JC, Braga-Mele R. Incidence of postoperative endoph-
thalmitis in a high-volume cataract surgicentre in Canada. Can
J Ophthalmol 2009;44:28892.
12. Erdogan H, Ozec AV, Caner C, et al. Effect of latanoprost/
timolol and dorzolamide/timolol on intraocular pressure after
phacoemulsication surgery. Int J Ophthalmol 2011;4:1904.
13. Kir E, Cakmak H, Dayanir V. Medical control of intraocular
pressure with brinzolamide 1% after phacoemulsication. Can
J Ophthalmol 2008;43:55962.
14. Pharmakakis N, Giannopoulos K, Stasinos S, et al. Effect of a
xed brimonidine-timolol combination on intraocular pressure
after phacoemulsication. J Cataract Refract Surg 2011;37:
27983.
15. Borazan M, Karalezli A, Akman A, Akova YA. Effect of
antiglaucoma agents on postoperative intraocular pressure after
cataract surgery with Viscoat. J Cataract Refract Surg 2007;33:
19415.
16. Cetinkaya A, Akman A, Akova YA. Effect of topical brinzo-
lamide 1% and brimonidine 0.2% on intraocular pressure after
phacoemulsication. J Cataract Refract Surg 2004;30:173641.
17. Kasetti SR, Desai SP, Sivakumar S, Sunderraj P. Preventing
intraocular pressure increase after phacoemulsication and the
role of perioperative apraclonidine. J Cataract Refract Surg
2002;28:217780.
18. Katsimpris JM, Siganos D, Konstas AG, et al. Efcacy of
brimonidine 0.2% in controlling acute postoperative intra-
ocular pressure elevation after phacoemulsication. J Cataract
Refract Surg 2003;29:228894.
19. Geyer O, Bottone EJ, Podos SM, et al. Microbial contamina-
tion of medications used to treat glaucoma. Br J Ophthalmol
1995;79:3769.
20. Scoper SV. Review of third-and fourth-generation uo-
roquinolones in ophthalmology: in-vitro and in-vivo efcacy.
Adv Ther 2008;25:97994.
21. Wejde G, Montan P, Lundstrom M, et al. Endophthalmitis
following cataract surgery in Sweden: national prospective
survey 1999e2001. Acta Ophthalmol Scand 2005;83:710.
22. Colleaux KM, Hamilton WK. Effect of prophylactic anti-
biotics and incision type on the incidence of endophthalmitis
after cataract surgery. Can J Ophthalmol 2000;35:3738.
23. Ng JQ, Morlet N, Bulsara MK, Semmens JB. Reducing the
risk for endophthalmitis after cataract surgery: population-
based nested case-control study: Endophthalmitis Population
Study of Western Australia sixth report. J Cataract Refract
Surg 2007;33:26980.
24. Keay L, Gower EW, Cassard SD, et al. Postcataract surgery
endophthalmitis in the United States: analysis of the complete
2003 to 2004 Medicare database of cataract surgeries.
Ophthalmology 2012;119:91422.
25. Naseri A. Surgeon cataract volume and endophthalmitis [let-
ter]. Ophthalmology 2012;119:2415; author reply 2416.
Footnotes and Financial Disclosures
Originally received: May 17, 2013.
Final revision: August 18, 2013.
Accepted: August 22, 2013.
Available online: December 9, 2013. Manuscript no. 2013-792.
Department of Ophthalmology, University of Alberta, Edmonton, Canada.
Presented at: Eye Physicians and Surgeons of Alberta Annual Meeting,
February 2013, Banff, Canada.
Financial Disclosure(s):
The author(s) have no proprietary or commercial interest in any materials
discussed in this article.
Supported by the University of Alberta Endowment Award for Student
Research, Edmonton, Canada. The funding organization had no role in the
design or conduct of this research.
Correspondence:
Christopher J. Rudnisky, MD, MPH, Department of Ophthalmology, Royal
Alexandra Hospital, 10240 Kingsway Avenue, Room 2316, Edmonton,
Alberta T6E2V8, Canada. E-mail: chris@rudnisky.ca.
Rudnisky et al

Prevention of Endophthalmitis after Cataract Surgery
841