Visual Outcomes Following the Use of

Intravitreal Steroids in the Treatment of

Postoperative Endophthalmitis
Gaurav K. Shah, MD, Joshua D. Stein, MS, Sanjay Sharma, MD, MSc (Epid), Arunan Sivalingam, MD,
William E. Benson, MD, Carl D. Regillo, MD, Gary C. Brown, MD, William Tasman, MD

Objective: To compare visual outcomes between cases of acute postoperative endophthalmitis that did or
did not receive intravitreal steroids.
Design: Retrospective nonrandomized comparative trial.
Participants: Fifty-seven patients with postoperative endophthalmitis.
Intervention: Thirty-one patients with postoperative endophthalmitis resulting from cataract extraction
received both intravitreal antibiotics and steroids, whereas the remaining 26 received only intravitreal antibiotics.
Main Outcome Measures: Improvement in visual acuity.
Results: Multivariate logistic regression was used to analyze the variables that potentially influence a
three-line visual acuity improvement. The mean baseline visual acuities of both groups were comparable. The use
of intravitreal steroids reduced the probability of developing a three-line improvement in visual acuity (odds ratio
[OR] ⫽ 0.287; 95% confidence interval [CI] [0.072– 0.852]). On the basis of logistic regression analysis using our
multivariate model, gender, baseline visual acuity, and pars plana vitrectomy were not significantly associated
with visual outcome differences between the two groups.
Conclusions: Patients who received intravitreal steroids had a significantly reduced likelihood of obtaining
a three-line improvement in visual acuity. At a minimum our study provides no support for their use and,
therefore, steroids may not be efficacious for acute endophthalmitis related to cataract extraction.
Ophthalmology 2000;107:486 – 489 © 2000 by the American Academy of Ophthalmology.

Endophthalmitis can occur after any intraocular procedure,
including cataract, glaucoma, corneal, and retinal surgery.
Although endophthalmitis is an infrequent complication of
intraocular surgery, it may have a profound effect on visual
function.1
Traditionally, vitrectomy and intravitreal antibiotics have
been used for the treatment of postoperative endophthalmi-
tis. Recently, however, the Endophthalmitis Vitrectomy
Study (EVS) demonstrated the therapeutic efficacy of vit-
rectomy in selected cases of postoperative endophthalmitis,
Originally received: May 25, 1999.
Accepted: November 12, 1999. Manuscript no. 99271.
Retina Service, Wills Eye Hospital, Philadelphia, Pennsylvania.
Supported by the Heed Ophthalmic Foundation, Cleveland, Ohio (GKS),
the Vitreo-Retinal Research Fund, Philadelphia, Pennsylvania (GKS, SS),
and by an unrestricted grant from Research to Prevent Blindness, New
York, New York (SS), and Thomas Duane Student Research Fellowship,
Philadelphia, PA (JS).
Presented in part at the Annual Meeting of the Association for Research in
Vision and Ophthalmology, Fort Lauderdale, Florida, May 1999, and
Retina Society Meeting, Maui, Hawaii, December 1999.
Dr. Shah is now affiliated with Barnes Retina Consultants, Washington
University, St. Louis, Missouri, and Dr. Sharma is affiliated with the
Departments of Ophthalmology and Epidemiology at Queen’s University,
Kingston, Ontario, Canada.
Reprint requests to Gaurav K. Shah, Barnes Retina Institute, Suite 17413,
St. Louis, MO 63110.
while also demonstrating that vitrectomy was not necessary
in the management of most cases of endophthalmitis.2 All
patients in the EVS received both intravitreal antibiotics and
oral and topical steroids, and no patient received intravitreal
steroids.
Although strong scientific evidence exists regarding the
efficacy of antibiotics for the treatment of endophthalmitis,
little evidence supports the decision to add intravitreal ste-
roids in these cases.3–7,8 –12 Nevertheless, many surgeons
routinely administer them.
Our study was designed to evaluate the efficacy of in-
travitreal steroids as an adjunct to antibiotic treatment of
acute postoperative endophthalmitis after cataract surgery.
Methods
The computerized database of all consecutive patients diagnosed
with acute postoperative endophthalmitis after cataract surgery
seen in the Retina Service of Wills Eye Hospital over a 1-year
period was obtained. Most, but not all, patients had postoperative
pain, decreased vision, vitreitis, and hypopyon after surgery. We
included patients whose ocular symptoms of pain and decreased
vision were of less than 4 weeks’ duration after cataract extraction.
Patients were excluded if additional procedures (corneal or glau-
coma surgery) were performed at the time of the initial surgery.
Furthermore, patients with chronic endophthalmitis or inflamma-
tions of other origins, such as retained lens fragments, were ex-
cluded from the study.

486 © 2000 by the American Academy of Ophthalmology ISSN 0161-6420/00/$–see front matter
Published by Elsevier Science Inc. PII S0161-6420(99)00139-6
 Shah et al 䡠 Use of Intravitreal Steroids in Endophthalmitis

Table 1. Clinical Symptoms in Each Group Compared Table 2. Clinical Signs in Each Group Compared
with the EVS with the EVS

Percent Percent Percent Percent

in in Percent in in in Percent in
Group 1 Group 2 EVS Visual Group 1 Group 2 EVS
(n ⴝ 26) (n ⴝ 31) (n ⴝ 420) Acuity (n ⴝ 26) (n ⴝ 31) (n ⴝ 420)
Pain 58 67 74 LP 20 25 26
(n ⫽ 15) (n ⫽ 21) (n ⫽ 310) (n ⫽ 5) (n ⫽ 8) (n ⫽ 109)
Redness 76 60 82 HM 45 38 44
(n ⫽ 19) (n ⫽ 19) (n ⫽ 344) (n ⫽ 12) (n ⫽ 12) (n ⫽ 185)
Decreased vision 98 94 94 5/200-CF 15 25 16
(n ⫽ 25) (n ⫽ 29) (n ⫽ 395) (n ⫽ 4) (n ⫽ 8) (n ⫽ 67)
Hypopyon 88 84 86 ⬎5/200 20 12 14
(n ⫽ 23) (n ⫽ 26) (n ⫽ 361) (n ⫽ 5) (n ⫽ 4) (n ⫽ 59)

EVS ⫽ Endophthalmitis Vitrectomy Study. EVS ⫽ Endophthalmitis Vitrectomy Study; CF ⫽ counting fingers; HM ⫽
hand motion; LP ⫽ light perception.

Results
All patients were treated with intravitreal amikacin (400 ␮g/0.1
ml) and vancomycin (1 mg/0.1 ml), either at the time of vitrectomy
Fifty-seven patients with acute endophthalmitis after cataract sur-
or at the time of vitreous tap. No patient received systemic anti-
gery met our various criteria. Twenty-six patients (46%) did not
biotics. Assignment of intravitreal steroids (dexamethasone, 400
receive intravitreal steroids (group 1), whereas the remaining 31
␮g/0.1 ml) was nonrandom in nature and depended on physician (54%) did (group 2). Fourteen of 26 (55%) patients from group 1
preference. All patients treated with intravitreal steroids received underwent pars plana vitrectomy, whereas the remaining 12 (45%)
only one injection, and none had multiple injections. None of the patients underwent tap and injection. Eighteen of 31 (61%) pa-
patients in our study received oral steroids. On the Retina Service, tients from group 2 underwent vitrectomy, whereas the remaining
some physicians use steroids in some cases and not in others. 13 (39%) underwent vitreous tap and therapeutic injection. These
However, two physicians always use intravitreal steroids and two differences were not statistically significant (P ⬎ 0.05). The mean
never do. There is no reason to believe that those who used steroids age in group 1 was 64 (range, 56 – 84) and the mean age in group
had more or less severe cases than those who did not. Cases are 2 was 62 (range, 55– 89). No significant proportional gender dif-
referred on emergency bases and are cared for by the physician on ferences were found between the two groups. The presenting signs
call. and visual acuities in both groups were virtually identical to each
Patients who did not receive vitrectomy had a vitreous tap other and to the patients in the EVS (Tables 1 and 2).
performed through the pars plana with a 22-, 25-, or 27-gauge Follow-up in both groups ranged from 2 months to 2 years,
needle. In patients who underwent vitrectomy, mechanized vitre- with a mean follow-up of 8.5 months in group 1 and 7.5 months in
ous biopsy was obtained by means of 20-gauge pars plana scle- group 2. The initial visual acuities were light perception (20/50) in
rotomy and vitreous cutter/aspirator attached to a syringe. Approx- group 1 and light perception (20/40) in group 2, with mean acuity
imately 0.1 to 0.3 ml of undiluted fluid was removed from the of counting fingers in both groups (P ⬎ 0.05). At each follow-up
vitreous cavity with either vitreous tap or vitrectomy. The undi- interval (1 month, 3 months, and 6 months), the steroid patients
luted 0.1-ml sample and vitrectomy cassette were both submitted had a worse mean visual acuity than the patients who did not
for laboratory analysis. An undiluted aqueous specimen was ob- receive intravitreal steroids. This difference was statistically sig-
tained with a 30-gauge needle inserted through the limbus in all nificant at each interval during the follow-up (Table 3).
cases. Seventeen of 24 (75%) patients who received only intravitreal
All patient’s charts were reviewed for demographic character- antibiotics had a three-line visual improvement by 1 month. Thir-
istics, baseline visual acuity, culture results, vitrectomy or intra- teen of 29 patients (46%) who received both intravitreal antibiotics
vitreal antibiotic injection, antibiotic and steroid injection, post- and steroids had a three-line visual improvement by 1 month. Our
treatment visual acuity, and other pertinent signs of complications
at every follow-up visit. These two groups were analyzed with
respect to the development of visual improvement at follow-up Table 3. Mean Visual Acuity Outcomes between the Two
intervals. A multivariate model using logistic regression deter- Treatment Groups
mined whether gender, age, performance of vitrectomy, or intra-
vitreal steroids were associated with a three-line visual improve- Mean Visual Acuity No Steroids Steroids P value
ment. All visual acuities were based on best-corrected acuities.
These were obtained with a distance Snellen acuity chart. They Baseline CF CF P ⬎ 0.05
(n ⫽ 26) (n ⫽ 31)
were not obtained in a standardized fashion. A three-line visual
1 mo 20/60 20/100 P ⬍ 0.05
improvement was determined to be a “clinically relevant” outcome (n ⫽ 24) (n ⫽ 29)
because this degree of visual improvement is associated with 3 mos 20/50 20/70 P ⬍ 0.05
significant improvement in quality of life (as measured by time- (n ⫽ 22) (n ⫽ 26)
tradeoff technique to quantitate utility states). Statistical signifi- 6 mos 20/50 20/70 P ⬍ 0.05
cance was assumed to have occurred when the chance of type I (n ⫽ 22) (n ⫽ 25)
error was less than 5%. Visual acuities were compared at baseline,
1 month, 3 months, and 6 months by use of a two-tailed Student’s CF ⫽ counting fingers.
t test.

487
 Ophthalmology Volume 107, Number 3, March 2000
Table 4. Visual Acuity Outcomes between the Two Treatment
Groups at 1 Month
Three-Line Less than Three-
Improvement Line Improvement Total
Antibiotics alone 17 7 24
Antibiotics/steroids 13 16 29
Totals 30 23 53
logistic regression revealed that there was a significant negative
association with a three-line visual improvement and the use of
intravitreal steroids (Wald statistic, 4.27; P ⬍ 0.05) (OR ⫽ 0.287;
95% CI [0.072– 0.852]). In our logistic regression model, there
were no other clinical variables, that were significantly associated
with developing a three-line visual improvement at 1 month (Table
4). The 3-month results were similar to those seen at 1 month
(Table 5). No significant differences were found in our two cohorts
with respect to the proportion of patients who were culture nega-
tive and those who were culture positive for Staphylococcus epi-
dermidis (Table 6).
Discussion
In our study, patients received either intravitreal antibiotics
alone or in conjunction with steroids. The treatment modal-
ity was nonrandom in nature and was based on the treatment
preference of the various attending physicians. Patients who
received intravitreal steroids had a significantly reduced
probability of a three-line improvement in visual acuity
developing. In fact, patients who received intravitreal ste-
roids in conjunction with their antibiotics were 3.5 times
less likely to develop a three-line visual improvement com-
pared with those who received only intravitreal antibiotics.
Given the limits of our confidence interval, patients who
receive steroids may have anywhere from a 1.2-fold to a
14-fold reduction in the probability of a three-line visual
improvement. Each group had similar demographic charac-
teristics with comparable baseline visual acuities. Given
that our confidence interval excludes 1, we are able to reject
our null hypothesis (i.e., that there is no difference in the
incidence of a three-line visual improvement at 2 to 3
months between the two cohorts) and conclude that there is,
in fact, a significant difference in the three-line improve-
ment between our two treatment groups.
Our results suggest caution in the use of intraocular
corticosteroids in the treatment of endophthalmitis. Possible
explanations to account for our findings include corticoste-
roid-induced toxicity and/or a blunting of the immune re-
sponse that is necessary to combat bacterial infection. Un-
Table 5. Visual Acuity Outcomes between the Two Treatment
Groups at 3 Months
Three-Line Less Than Three-
Improvement Line Improvement Total
Antibiotics alone 15 7 22
Antibiotics/steroids 12 14 26
Totals 27 21 48
488
Table 6. Culture Results in Each Group
Group 1 Group 2
Culture-negative 11 Culture-negative 9
S. epidermidis 8 S. epidermidis 11
S. aureus 4 S. aureus 6
S. pneumoniae 1 S. pneumoniae 1
S. fecalis 1 S. fecalis 1
Enterococcus 1 Enterococcus 2
Serratia 1
treated, patients with acute postoperative endophthalmitis
can have severe visual loss develop. The EVS has demon-
strated the benefit of vitrectomy in selected cases.2 How-
ever, the use of intravitreal corticosteroids was not ad-
dressed in that trial, and their use remains controversial.
Studies based on animal models of endophthalmitis have
demonstrated clinical improvement with intravitreal ste-
roids added to the antibiotic regimen, as measured by elec-
trophysiologic studies and histopathologic findings. Al-
though studies by Mao and coworkers13 have demonstrated
good treatment outcomes in animals infected with S. aureus
when these animals were treated with intravitreal antibiot-
ics, no advantage was seen when comparing steroids ad-
ministered intravitreally versus systemically in a nonran-
domized retrospective study. A study by Meredith and
colleagues14 using a rabbit model of S. aureus endoph-
thalmitis failed to demonstrate improvement in media clar-
ity of animals treated with intraocular dexamethasone and,
instead, showed an increase in choroidal inflammation re-
sulting in retinal necrosis. Inflammatory scores were in-
creased and corneal opacities were doubled in the group
treated with corticosteroids compared with the eyes without
corticosteroids.
It has been shown by histopathologic analysis of tissues
that infections with cytotoxic bacterial strains lead to sub-
stantial disorganization and lysis of cells located within all
retinal layers. In contrast, noncytotoxic bacterial strains lead
to an infiltration of immune cells in the vitreous while
leaving the retina structurally intact.6,7 Jett et al15 reported
similar results in a study involving two different strains of
Enterococcus fecalis. They noted that there was some ben-
eficial effect from combining intravitreal dexamethasone
and antibiotic therapy only in eyes infected with the less
virulent, nontoxin-producing strains of bacteria. Among
eyes infected with toxin-producing bacteria, intravitreal an-
tibiotic therapy had no beneficial effect, regardless of con-
current administration of intravitreal dexamethasone. Ma-
nipulation of the timing of dexamethasone administration
did not significantly affect the negative treatment outcome
in their study.
Our results suggest a cautionary approach to the use of
intraocular corticosteroids at dosage levels given in this
study. They suggest, as well, that steroids should not be
routinely used as part of the treatment regimen for acute
postsurgical endophthalmitis. A potential explanation for
our results includes the fact that the dose of intraocular
steroids in our patients (400 ␮g) may be too high. The
 Shah et al 䡠 Use of Intravitreal Steroids in Endophthalmitis
intraocular concentration of medication is much higher than
with systemically administered steroids. Although Kwak
and D’Amico9 have demonstrated the lack of toxicity with
500 ␮g of intravitreal dexamethasone, the optimal dose has
yet to be determined.
In the neurologic literature there is a “J”-shaped response
curve with 15 mg/kg of methylprednisolone being insuffi-
cient and 60 mg/kg being too much steroid. The most
adequate recovery of the central nervous system during
trauma was achieved with 30 mg/kg.
The number of patients who underwent vitrectomy was
slightly higher in the nonsteroid group compared with the
steroid group, but this difference did not reach statistical
significance in our model on the basis of logistic regression.
No significant differences were found in the proportion of
culture-negative cases, S. epidermidis, and gram-negative
cases. We did not perform a stratified analysis on various
subgroups defined by infecting organism, given the small
number of patients in each subgroup.
Some potential limitations of our study include the fact
that it was retrospective and nonrandomized. As a result,
both the independent and dependent variables were obtained
in a nonstandardized fashion. Visual acuities were not Early
Treatment Diabetic Retinopathy Study standardized but
were best-corrected Snellen visual acuities. With nonran-
dom allocation, it is possible that the observed differences in
our outcome between the two groups were due to a con-
founding variable. Although this is a possibility, our logistic
regression included recognized variables, which are known
to influence the course of endophthalmitis. These variables
were not significant in our statistical modeling. A final
limitation of our study was the fact that subgroup stratifi-
cation according to infecting organism was not performed,
given the small number of patients in each category. To
have performed these analyses on a limited number of
patients would have yielded statistical results with a very
low power (high type II error) and would have limited
generalizability.
In summary, our data suggest that intravitreal dexameth-
asone, when administered in conjunction with intravitreal
antibiotics for the treatment of acute endophthalmitis after
cataract extraction, resulted in a significant worsening of
visual acuity during a follow-up of at least 6 months com-
pared with eyes that received intravitreal antibiotics alone.
On the basis of our data and on review of the literature, we
believe that the burden of proof is on those who advocate
their use.
References
1. Javitt JC, Street DA, Tielsch JM, et al. National outcomes of
cataract extraction. Retinal detachment and endophthalmitis after
outpatient cataract surgery. Ophthalmology 1994;101:100 – 6.
2. Results of the Endophthalmitis Vitrectomy Study. A random-
ized trial of immediate vitrectomy and of intravenous antibi-
otics for the treatment of postoperative bacterial endoph-
thalmitis. Endophthalmitis Vitrectomy Study Group. Arch
Ophthalmol 1995;113:1479 –96.
3. Aguilar HE, Meredith TA, Drews C, et al. Comparative treat-
ment of experimental Staphylococcus aureus endophthalmitis.
Am J Ophthalmol 1996;121:310 –17.
4. Baum JL, Barza M, Lugar J, Onigman P. The effect of
corticosteroids in the treatment of experimental bacterial en-
dophthalmitis. Am J Ophthalmol 1975;80(3 Pt 2):513–5.
5. Graham RO, Peyman GA. Intravitreal injection of dexameth-
asone. Treatment of experimentally induced endophthalmitis.
Arch Ophthalmol 1974;92:149 –54.
6. Maxwell DP Jr, Brent BD, Diamond JG, Wu L. Effect of intra-
vitreal dexamethasone on ocular histopathology in a rabbit model
of endophthalmitis. Ophthalmology 1991;98:1370 –5.
7. Peyman GA, Herbst R. Bacterial endophthalmitis. Treatment
with intraocular injection of gentamicin and dexamethasone.
Arch Ophthalmol 1974;91:416 – 8.
8. Meredith TA, Aguilar HE, Miller MJ, et al. Comparative
treatment of experimental Staphylococcus epidermidis en-
dophthalmitis. Arch Ophthalmol 1990;108:857– 60.
9. Kwak HW, D’Amico DJ. Evaluation of the retinal toxicity and
pharmacokinetics of dexamethasone after intravitreal injec-
tion. Arch Ophthalmol 1992;110:259 – 66.
10. Stern GA, Engel HM, Driebe WR Jr. The treatment of post-
operative endophthalmitis. Results of differing approaches to
treatment. Ophthalmology 1989;96:62–7.
11. Schulman JA, Peyman GA. Intravitreal corticosteroids as an
adjunct in the treatment of bacterial and fungal endophthalmi-
tis: a review. Retina 1992;12:336 – 40.
12. Coats ML, Peyman GA. Intravitreal corticosteroids in the
treatment of exogenous fungal endophthalmitis. Retina 1992;
12:46 –51.
13. Mao LK, Flynn HW Jr, Miller D, Pflugfelder SC. Endoph-
thalmitis caused by Staphylococcus aureus. Am J Ophthalmol
1993;116:584 –9.
14. Meredith TA, Aguilar HE, Drews C, et al. Intraocular dexa-
methasone produces a harmful effect on treatment of experi-
mental Staphylococcus aureus endophthalmitis. Trans Am
Ophthalmol Soc 1996;94; 241–52; discussion 252–7.
15. Jett BD, Jensen HG, Atkuri RV, Gilmore MS. Evaluation of
therapeutic measures for treating endophthalmitis caused by
isogenic toxin-producing and toxin-nonproducing Enterococ-
cus faecalis strains. Invest Ophthalmol Vis Sci 1995;36:9 –15.
489