Accepted Manuscript

Prospective, Randomized Clinical Trial of Povidone-Iodine 1.25% Solution versus

Topical Antibiotics for Treatment of Bacterial Keratitis
Sherwin J. Isenberg, Leonard Apt, Mario Valenton, Savitri Sharma, Prashant Garg,
Philip A. Thomas, Pragya Parmar, Jayaraman Kaliamurthy, Johann M. Reyes, Daniel
Ong, Peter D. Christenson, Madeline Del Signore, Gary N. Holland
PII:

S0002-9394(16)30516-5

DOI:

10.1016/j.ajo.2016.10.004

Reference:

AJOPHT 9926

To appear in:

American Journal of Ophthalmology

Received Date: 28 December 2015

Revised Date:

12 October 2016

Accepted Date: 12 October 2016

Please cite this article as: Isenberg SJ, Apt L, Valenton M, Sharma S, Garg P, Thomas PA, Parmar
P, Kaliamurthy J, Reyes JM, Ong D, Christenson PD, Del Signore M, Holland GN, Prospective,
Randomized Clinical Trial of Povidone-Iodine 1.25% Solution versus Topical Antibiotics for Treatment of
Bacterial Keratitis, American Journal of Ophthalmology (2016), doi: 10.1016/j.ajo.2016.10.004.
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

ABSTRACT

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

Purpose: To compare povidone-iodine 1.25% ophthalmic solution to topical

antibiotics for treatment of bacterial keratitis in areas of the world where use of
effective topical antibiotics may not be an option.
Study Design: Randomized, controlled, investigator-masked clinical trial.
Methods: We randomized 172 individuals with bacterial keratitis to topical
treatment with povidone-iodine or antibiotics (neomycin-polymyxin B-gramicidin
in the Philippines; ciprofloxacin 0.3% in India). Using survival analysis, we
compared intervals from start of treatment to presumed cure (primary outcome
measure, defined as a closed epithelial defect without associated inflammatory
signs) and to recovering (residual epithelial defect <1mm2 with only minimal
inflammation).
Results: Median interval to presumed cure in the Philippines was 7 days for
povidone-iodine and 7 days for neomycin-polymyxin B-gramicidin (95%
confidence interval [CI] for difference in median interval, -9.5 to 0.7 days] and in
India was 12 days for povidone-iodine and 17 days for ciprofloxacin (95% CI, 35.2 to 3.2 days). Hazard ratio (HR) for presumed cure among those treated
with povidone-iodine (vs. antibiotics) was 1.46 in the Philippines (95% CI, 0.902.36; p=0.13) and 1.70 in India (95% CI, 0.73-3.94; p=0.22). Comparisons of
intervals to recovering and HR for recovering also revealed no significant
differences between treatment groups in either country.
Conclusions: There is no significant difference between the effect of topical
povidone-iodine 1.25% and topical antibiotics commonly available in the
developing world for treatment of bacterial keratitis. Povidone-iodine 1.25%,
which is widely available and inexpensive, can be considered for treatment of
bacterial keratitis when antibiotic treatment is not practical.

ACCEPTED MANUSCRIPT

RI
PT

Isenberg, et al. Treatment of Bacterial Keratitis 1.

SC

Prospective, Randomized Clinical Trial

of Povidone-Iodine 1.25% Solution versus
Topical Antibiotics for Treatment of Bacterial Keratitis

M
AN
U

Sherwin J. Isenberg, Leonard Apt, Mario Valenton, Savitri Sharma,

Prashant Garg, Philip A. Thomas, Pragya Parmar, Jayaraman Kaliamurthy,
Johann M. Reyes, Daniel Ong, Peter D. Christenson, Madeline Del Signore,
and Gary N. Holland

Short title: Trial of Povidone-iodine Treatment for Bacterial Keratitis

TE
D

Supplemental material available at AJO.com

AC
C

EP

From the Center To Prevent Childhood Blindness (SJI, LA) and the Ocular Inflammatory
Disease Center (GNH), UCLA Stein Eye Institute and the Departments of
Ophthalmology (SJI, LA, GNH), and Medicine (PDC), David Geffen School of Medicine
at UCLA, Los Angeles, CA; the Los Angeles Biomedical Research Institute at HarborUCLA Medical Center, Torrance, CA (SJI, PDC, MD); the Philippine General Hospital,
Manila, Philippines (MV, JMR, DO); the L.V. Prasad Eye Institute, Hyderabad, India
(SS, PG); and the Joseph Eye Hospital, Tiruchirapalli, India (PAT, PP, JK).

Corresponding author: Gary N. Holland, M.D., UCLA Stein Eye Institute, 100 Stein
Plaza, UCLA, Los Angeles, CA 90095. Telephone: 310-825-5440. E-mail:
uveitis@jsei.ucla.edu

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 2.

TE
D

M
AN
U

SC

RI
PT

INTRODUCTION
Infectious keratitis is an important cause of blindness in the developing world,
where a number of factors, including malnutrition with vitamin A deficiency, substantially
increase the risk of corneal infection. In India, for example, corneal disease, much of
which is infectious keratitis, is one of the top 10 causes of visual impairment, regardless
of gender, age, or socioeconomic factors.1 The problem is especially severe in children;
corneal scarring has traditionally been considered to be the most common reason for
avoidable blindness in children worldwide.2,3
For antimicrobial agents to reduce the burden of infectious keratitis in resourcepoor areas of the world, they must meet certain criteria, including effectiveness against
a broad spectrum of organisms, a favorable safety profile, ease of preparation, and
minimal expense. Studies suggest that povidone-iodine meets these criteria. With rare
exception, it is effective against all bacteria, viruses, and fungi in vitro, given sufficient
contact time; true bacterial resistance to povidone-iodine probably does not exist.4 With
regard to safety, povidone-iodine has been associated with few allergic reactions.
In previous studies, povidone-iodine ophthalmic solution was shown to be
effective for prophylaxis against infection when used during preoperative preparation or
as a postoperative antimicrobial treatment.5,6 A study in Kenya showed that povidoneiodine ophthalmic solution was an effective prophylactic agent for prevention of
ophthalmia neonatorum.7 In a randomized, controlled trial to investigate its use for
treatment of infectious conjunctivitis, povidone-iodine ophthalmic solution appeared to
be as effective as antibiotics for bacterial cases and was more effective against
chlamydial.8
In this study, we investigated whether povidone-iodine 1.25% ophthalmic solution
could be used to treat bacterial keratitis, by comparing its efficacy to topical antibiotics in
prospective, randomized clinical trials conducted in the Philippines and in India. If
shown to be effective, it potentially could decrease the incidence of corneal scarring and
vision loss in many parts of the world.

AC
C

EP

METHODS
This study was approved by the Human Subjects Protection Committee of the
Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center and by the
institutional review boards at each participating hospital prior to initiation of the study.
Subjects or their parents or guardians provided written informed consent in their native
languages. Assent was also obtained from subjects 18 years old and younger. The
clinical trial conformed to the tenents of the Declaration of Helsinki, and was registered
at clinicaltrials.gov (registration number: NCT00386958; URL:
https://register.clinicaltrials.gov).
Sample size calculations were based on a pilot study of individuals in Manila
meeting inclusion criteria for the current study and treated with neomycin-polymixin Bgramicidin; median time to achieving criteria of the primary outcome measure for the
proposed study was 7 days. It was assumed that the median time for ciprofloxacin
would be identical. Calculations planned for separate comparisons of the primary
outcome measure between investigative and control subgroups in both countries, for a
total of four study arms. It was determined that 40 study participants per arm (160 total

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 3.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

study participants) would provide at least 80% power to detect differences in 10-day
cumulative probabilities of reaching the study end-point of at least 0.32 in each country,
corresponding to a hazard ratio of 2.7, at a 0.05 level of significance.
Patient Populations
Recruitment of study participants began in November 2002; all follow-up
examinations were complete in January 2006. Inclusion criteria were age 1 month;
keratitis duration (as reported by the patient or family) 14 days; no prior treatment of
the keratitis; and an epithelial defect 2-8mm in greatest dimension that did not extend to
the limbus. The presence of bacteria had to be confirmed by culture or by microscopic
examination of corneal scrapings using gram stain within 24 hours of study entry for
continued participation. Exclusion criteria were treatment with topical or systemic
corticosteroid or immunosuppressive drugs within 14 days of study entry; corneal
perforation or impending perforation (based on the judgment of the treating
ophthalmologist); presence of dacryocystitis, neurotrophic keratopathy, exposure
keratitis, or keratitis sicca; human immunodeficiency virus infection; blindness of the
unaffected eye; and allergy or intolerance to povidone-iodine, iodine, or the study
antibiotics (by study site).
Study Procedures and Data Collection
To maximize uniformity between study sites, all investigators met in Manila,
Philippines, to agree upon standardized examination techniques, culture techniques,
and data reporting. Standardized dosing regimens and follow-up schedules (described
in detail in supplemental text available at www.ajo.com) were selected by consensus,
based on standards of care at each of the study centers. The study sites were visited
by the Principle Investigator (SJI) during recruitment to insure consistency between
sites.
Eyes with infectious keratitis were randomized to receive either topical povidoneiodine 1.25% or the control topical antibiotic, based on the study site, as described
below. Stratified randomization by involved eye was performed, based on location and
age group (<18 years of age vs. 18 years of age), utilizing lists generated before study
initiation. Corneal foreign bodies were removed before treatment.
Povidone-iodine 1.25% solution (Purdue Frederick, Norwalk CT) was prepared
by dilution, using Balanced Salt Solution (Alcon Laboratories Inc., Fort Worth TX).
Control antibiotics were based on standards of care at each site at the time of the study.
In the Philippines, neomycin-polymixin B-gramicidin ophthalmic solution (Bausch &
Lomb Pharmaceuticals, Tampa, FL) was used; at the two Indian sites (Hyderabad and
Tiruchirapalli), ciprofloxacin 0.3% ophthalmic solution (Ciprodac, Cadila
Pharmaceuticals Ltd., Ahmedabad, India) was used. The only other eye medication
that was permitted was atropine ophthalmic solution (Akorn, Inc., Buffalo Grove, IL),
administered to the affected eye(s) twice daily. Concentrations were based on age, as
follows: 0.25% solution for ages <1 year; 0.5% solution for ages 1-3 years; and 1% for
ages 3 years and older.
All patients were hospitalized for at least 7 days, as was customary at the study
sites during the period of the study. To insure compliance, all medications were
administered by study nurses. Each application of drug was recorded, and records
were subsequently reviewed by study personnel. Physicians and microbiologists were
masked to treatment assignment during the course of the study; because study nurses

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 4.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

applied medications well before the daily eye examinations, physicians had no
opportunity to see the transient brown color imparted by povidone-iodine.
At daily eye examinations the following variables were assessed: visual acuity;
pain; associated ocular surface and adnexal inflammatory signs (discharge, chemosis,
redness, eyelid swelling); corneal signs (epithelial defect, size of stromal infiltrate, tissue
loss, tissue firmness, keratic precipitates); and anterior chamber cells. Standardized
scores were used for each variable (see Supplemental Table 1, available at
www.ajo.com). A daily status assessment was made on the basis of examination
findings; study eyes were categorized as failure, worsening, persistence,
improving, or presumed cure. In post-hoc analyses, some patients were reassigned
a score of recovering at last examination, as described below. Criteria for each
category are outlined in the supplemental material (see www.ajo.com).
Within the first 48 hours, if there was either deterioration in all study variables or
appearance of a descemetocele, then treatment was changed and the patient was
removed from the study, as outlined in our dosing and follow-up schedules (see
supplemental text available at www.ajo.com). After 48 hours, treatment was changed
and the patient removed from the study when a status of failure was assigned. If the
status of a study eye was not improved or better by day 10 of treatment, the patient was
also removed from the study. Therapy after removal from the study was based on best
medical judgment. Change in epithelial defect size alone was never used to make
treatment decisions.
Data Analysis and Statistical Techniques
Data from the Philippines and from India were considered separately. Intercenter differences between the two study sites in India were not considered, as both are
located in the southern part of the country and were believed to see a similar spectrum
of disease. In the primary analysis, we used the Kaplan-Meier technique to compare
times to presumed cure between study participants receiving povidone-iodine and those
receiving the control antibiotic in each country. The Kaplan-Meier technique was also
used to estimate the probability of presumed cure for each subgroup at 10 days of
treatment, the time at which study participants without improvement were censored.
Several secondary comparisons were performed. Because some patients left the
hospital and were lost to follow-up before the primary study end-point was achieved,
use of the post-hoc outcome measure recovering allowed us to compare patients at
discharge for a beneficial treatment effect, based on criteria that are often used clinically
for planning hospital discharge (see supplemental material at www.ajo.com). The same
Kaplan-Meier analyses described above were repeated using recovering at the outcome
variable.
Despite potential problems associated with comparing proportions between
groups when there is differential case follow-up,9,10 we felt that it was acceptable to
compare last examinations as an interval analysis, because most patients last
examinations occurred at a fairly uniform time point in the course of treatment: at
discharge, between days 7 and 10 of treatment. In one comparison, we assumed that
patients lost to follow-up never achieved the primary outcome, and the proportion of
study participants who had achieved the status of presumed cure were compared
between treatment arms, using the Fisher exact test. In a similar interval comparison,

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 5.

SC

RI
PT

the proportions of study participants who had achieved the status of recovering or
presumed cure by last examination were compared between treatment arms.
In a risk factor analysis using Cox proportional hazards models, we investigated
the effect of the following host and disease factors on the primary outcome measure
(status of presumed cure): sex, age, laterality, visual acuity, gram stain characteristic,
ulcer size, ulcer depth, and inflammation score. Those factors that were significantly
associated with the primary outcome variable in univariable comparisons were then
included in multivariable regressions to determine the influence of those factors on the
relationship between povidone-iodine and presumed cure, and to determine the
independence of each factors effect. Subgroup analyses were also performed in which
the relationship between povidone-iodine and presumed cure was determined for study
participants grouped on the basis of ulcer size. Hazard ratios >1 favored treatment with
povidone-iodine.

AC
C

EP

TE
D

M
AN
U

RESULTS
A total of 413 individuals were screened for the study from 2002 through 2006,
172 of whom were randomized to be treated with either povidone-iodine (40 individuals
in the Philippines, 38 in India) or the control antibiotic (49 in the Philippines; 45 in India).
The Supplemental Figure (available at www.ajo.com) provides detailed information
about screening, enrollment, and treatment groups in a Consolidated Standards for
Reporting Trials (CONSORT) flow diagram. The reasons that 241 screened individuals
did not reach randomization (125 individuals in the Philippines; 116 in India) include the
following: lack of organisms seen on initial corneal scraping (n=159, 66%); ulcer size
(n=23, 9.5%); duration of keratitis (n=15, 6.3%); corneal or scleral perforation (n=23,
9.5%); or a combination of these reasons (n=21, 8.7%). Of those with multiple reasons
for exclusion, six also had evidence of fungal infection (hyphae observed on initial
corneal scraping). Demographic, medical, and ophthalmic data for study participants
are included in Table 1. In the Philippines, 11 study participants were less than 18
years of age; in India, five study participants were less than 18 years of age.
Supplemental Table 2 lists the causal bacteria for groups based on treatment
assignment and study site. The three most common bacteria isolated from study eyes in
the Philippines were various Moraxella species (n=44, 49%), Pseudomonas aeruginosa
(n=12, 13%) and Staphylococcus aureus (n=10, 11%), while in India, they were
Streptococcus pneumoniae (n=18, 21%), Pseudomonas aeruginosa (n=17, 20%), and
Staphylococcus epidermidis (n=13, 16%). There were significantly more Moraxella spp.
infections in the Philippines (44 [49%] of 89 cases) than in India (2 [2.4%] of 83 cases,
p<0.0001).
Median corneal ulcer size was 2.75 mm2. For study purposes, we considered
ulcers larger than the median area to be large. There were significantly more large
ulcers at presentation among study participants in India (n=55, 66%) than among study
participants in the Philippines (n=32, 36%; p<0.0001). In India, the mean ulcer size at
discharge was 3.71.4mm2 for study participants treated with povidone-iodine and
3.81.5mm2 for controls.
Seven study participants were not followed until hospital discharge (one in the
Philippines, who was removed from the study when dacryocystitis was diagnosed
(povidone-iodine group) and six in India [five in the povidone iodine group, one in the

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 6.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

ciprofloxacin group], all of whom chose not to continue the study for personal reasons).
There were no statistically significant differences in the prevalence of participants
discontinued from the study before hospital discharge between the four study arms.
Table 2 shows analyses related to outcomes of presumed cure and recovering.
Hazard ratios for each outcome favored povidone-iodine, both in the Philippines and in
India, although differences were not significant. The figure shows a Kaplan-Meier
analysis for presumed cure between treatment arms in each country. In the Philippines,
the 10-day probability of presumed cure for study participants treated with povidoneiodine was 0.91 (95% CI, 0.76 0.98); for study participants treated with neomycinpolymixin B-gramicidin, it was 0.86 (95% CI, 0.73 0.95). In India, the 10-day
probability of presumed cure for participants treated with povidone-iodine was 0.38
(95% CI, 0.21 0.62); for participants treated with ciprofloxacin, it was 0.14 (95% CI,
0.05 0.34). Overall, the presumed cure rate was superior in the Philippines when
compared to India (10-day presumed cure probabilities of 0.88 vs. 0.25, respectively,
p<0.0001).
Table 3 shows the proportion of study participants achieving endpoints during the
course of follow-up (a score of presumed cure or failure), and for those not achieving an
end-point, the scores that had been assigned at hospital discharge or last examination
before loss to follow-up. In a secondary analysis, we assumed that all study
participants who were lost to follow-up or had not achieved the primary study outcome
measure by the time of hospital discharge never achieved cure, while those with scores
of presumed cure did in fact have eradication of infection. In this analysis, 30 (75.0%)
of 40 individuals treated with povidone-iodine and 39 (80%) of 49 individuals treated
with neomycin-polymixin B-gramicidin in the Philippines achieved cure (p=0.62), as
shown in Table 3. In the same analysis, 12 (32%) of 38 individuals treated with
povidone-iodine and 10 (22%) of 45 individuals treated with ciprofloxacin in India
achieved cured (p=0.45). There was also no statistical difference in the proportion of
study participants achieving a score of at least recovering (recovering or presumed
cure) before loss to follow-up or hospital discharge (p=0.41 in the Philippines; p>0.99 in
India). When we compared those with some evidence of improvement before loss to
follow-up or discharge (scores of improved, recovering, or presumed cure) between
treatment groups, there was a significant difference in the Philippines (32 [80%] of study
participants treated with povidone-iodine vs. 47 [96%] of 49 treated with neomycinpolymixin B-gramicidin [p=0.04]), attributable to a higher proportion of treatment failures
in the povidone-iodine treatment group; however, there was no significant difference
between treatment groups in India (p=0.62).
There were a total of 14 treatment failures. In the Philippines, seven (18%) of 40
study participants who received povidone-iodine failed treatment (five with Moraxella
spp. infections, one Streptococcus pneumonia infection, and one with Pseudomonas
aeruginosa infection), while one of 49 who received neomycin-polymixin B-gramicidin
failed treatment (infected with Moraxella sp.). In India, two study participants who
received povidone-iodine failed treatment (one with Pseudomonas aeruginosa infection,
one with Streptococcus pneumoniae infection), while four who received ciprofloxacin
failed treatment (three with Pseudomonas aeruginosa infections; one with
Streptococcus pneumoniae infection).

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 7.

EP

TE
D

M
AN
U

SC

RI
PT

Supplemental Table 3 shows the proportion of eyes with gram-positive bacterial

infections and the proportion of eyes with gram-negative bacterial infections that were
presumed cured or that achieved a score of recovering by last examination. There was
no overall statistically significant difference in treatment effect based on whether
infecting bacteria were gram-positive or gram-negative. In the Philippines, however, the
proportion of gram-negative bacterial infections that achieved scores of recovering or
presumed cure were lower for eyes treated with providone-iodine (15 [65%] of 23
cases) than for eyes treated with neomycin-polymixin B-gramicidin (32 [91%] of 35
cases, p=0.029, Fisher exact test).
Table 4 shows relationships of host and disease factors that might influence the
outcome of presumed cure in each country. In univariate comparisons, eyes with large
ulcers were less likely to achieve a score of presumed cure during the period of study in
both countries. Also, eyes of subjects with visual acuity worse than 0.05 were less likely
achieve scores of presumed cure than those with during the period of study than those
with visual acuity better than 0.30, although the association was borderline in India.
In multivariable regressions, hazard ratios for povidone-iodine treatment were
attenuated slightly after adjusting for ulcer size (1.40 in the Philippines, 1.56 in India),
but not after adjusting for visual acuity (1.46 in the Philippines; 1.89 in India). Eyes with
poorer visual acuity tended to have larger ulcers, and poorer visual acuity lost its
statistical association with lack of presumed cure, after adjustment for ulcer size.
In sub-group analysis, hazard ratios for povidone-iodine treatment in the
Philippines were 1.34 (95% CI: 0.74-2.43) for smaller ulcers and 1.53 (95% CI: 0.653.56) for larger ulcers (p=0.80 for interaction). In India, the hazard ratio for larger ulcers
was 0.73 (95% CI: 0.25-2.15). A hazard ratio for smaller ulcers could not be calculated
in India, because none of the 12 study participants treated with ciprofloxacin had
achieved a score of presumed cure (median 8 days of treatment). Seven (44%) of the
16 small ulcers in India treated with povidone-iodine achieved a score of presumed cure
with a median 10 days of treatment (p=0.01; 0% vs. 44%).
Supplemental Table 4 shows visual acuity at study entry and final study visit for
all subgroups. There were no substantial differences in visual acuity between treatment
groups during the relatively short study interval.
No serious adverse events or side effects were reported from any study site.

AC
C

DISCUSSION
Reduction of blindness from infectious keratitis in the developing world will
depend in part on the early use of effective treatments to reduce the extent of corneal
scarring, yet in many resource-poor areas of the world, antibiotics are either not
available or too expensive to be obtained routinely. Povidone-iodine is a potential
alternative treatment to help achieve this goal. It can be prepared from pre-existing
powders or solutions, which are available worldwide and are inexpensive. In Kenya, a
5ml bottle of the solution costs less than US$0.10 to prepare.7 After penetration,
povidone-iodine affects bacteria by interacting strongly with double bonds of saturated
fatty acids in cell walls and organelle membranes. It also oxidizes amino acids and
nucleotides, causing cell wall lipid membranes to form pores.11
Povidone-iodine has been used successfully as an anti-infective agent in other
ophthalmic situations. It has been shown to decrease microbial contamination of donor

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 8.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

eyes prior to tissue transplantation.12 It was shown to be effective for prophylaxis before
ophthalmic surgery,13 and is now commonly used worldwide as a preoperative
preparative solution. It is also shown to be an effective anti-infective when applied to
the eye following surgery.14 Povidone-iodine reduces bacterial counts when applied to
the conjunctivae of neonates.15 In a study of 3117 newborns in Kenya, it was shown
that povidone-iodine 5% solution was as or more effective than silver nitrate solution or
erythromycin ointment for prophylaxis against ophthalmia neonatorum, while also being,
safe. In a masked and controlled treatment trial involving 464 children in the
Philippines, no differences were found between povidone-iodine and neomycinpolymixin B-gramicidin for treatment of bacterial conjunctivitis and povidone-iodine was
more effective against chlamydial conjunctivitis.8
In contrast, there has been conflicting information about the effect of povidoneiodine on bacterial keratitis, and reported results may have been influenced by
confounding factors. In 1969, Hale reported successful outcomes in four cases of
pseudomonal keratitis treated with topical povidone-iodine in combination with topical
antibiotics.16 In 1985, Schuhman and Vidic found good improvement in 35 of 40
patients with conjunctivitis and keratoconjunctivitis who were treated with povidoneiodine; in some cases, betamethasone was also used.17 In a clinical trial in Nepal, there
was no observed benefit of using povidone-iodine in addition to standard antibiotics,
when compared to treatment with standard antibiotics alone;18 no microbiologic
analyses were performed. In another study, a single application of povidone-iodine 5%
was no more effective than placebo in reducing the frequency of positive cultures
among patients with corneal ulcers;19 approximately half of the study participants were
already using topical antibiotics at enrollment.
Animal studies have also provided conflicting information. In a rabbit model of
staphylococcal keratitis, povidone-iodine was more effective than saline controls, but
less effective than topical ofloxacin.20 In another study, povidone-iodine was more
effective than topical gentamicin.21 In different rabbit model, povidone-iodine was not
effective when a broth of Pseudomonas spp. was placed onto an abraded cornea.22
Because of its potential benefit for treatment of bacterial keratitis, but considering
uncertainty based on these prior studies, a prospective randomized clinical trial of
povidone-iodine vs. topical antibiotics was deemed appropriate.
We chose to use a 1.25% concentration of povidone-iodine to prevent complaints
of stinging, which might reduce compliance, as was observed in the aforementioned
conjunctivitis trial.8 Povidone-iodine has been found to be effective against pathogenic
bacteria, such as Staphylococcus aureus and Pseudomonas spp. in concentrations as
low as 0.1%.23 An in vitro study found povidone-iodine to be effective against Neisseria
gonorrhoeae, Chlamydia trachomatis, and herpes simplex virus, type II in a
concentration of 1.0%.24
The choice of control medications was deliberate. In the Philippines,
neomycin-polymyxin B-gramicidin ophthalmic solution is a common first-line drug for
treatment of bacterial keratitis, as is true in other developing countries. In India,
ciprofloxacin is often used for monotherapy. Ciprofloxacin manufactured in India is
also used in other resource-poor countries, because it is less expensive than
antibiotics available in the West.25

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 9.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

We chose to study the effect of povidone-iodine both in the Philippines and in

India; infectious keratitis is a major health care problem in both countries, and cases are
representative of those seen throughout the developing world. Despite the use of
standard protocols and study definitions at all sites, we recognized that conditions were
likely to vary between countries with regard to the spectrum of causal bacteria and
potential unrecognized host and other confounding factors. Also, there was likely to be
inter-observer variations with regard to interpretation of examination findings. We
therefore prospectively decided to analyze results from each country separately.
As shown in the Figure, the cumulative risk of presumed cure at any given time
point was lower in India than in the Philippines for both arms of the study, but the
relative effect of povidone-iodine vs. controls was similar in both countries. The smaller
proportion of study participants in India scored as presumed cure when compared to
study participants in the Philippines, and larger proportion scored as recovering in the
interval comparison (Table 3), may reflect delayed healing because of the larger ulcer
sizes at presentation in India. Also, it was our impression that ophthalmologists in India
were more willing to discharge patients who only met criteria for a status of improved
than were ophthalmologists in the Philippines.
Although overall trends appeared to favor treatment with povidone-iodine, our
goal was not to show superiority of povidone-iodine over antibiotics. In fact, there may
be specific causes of infection that are better treated with specific antibiotics. In
secondary analyses, there were more treatment failures prior to discharge among study
participants treated with povidone-iodine than with neomycin-polymixin B-gramicidin,
but the association was weak. There was a suggestion that ciprofloxacin was more
effective than povidone-iodine for infections by gram-negative bacteria in India. These
results should be viewed with caution, however, because of the pitfalls associated with
differential loss-to-follow-up.9,10 Overall, we could not identify a statistically significant
difference between povidone-iodine and control antibiotics in this study. Furthermore,
sensitivity testing and the full spectrum of antibiotics used in the West are not availability
in much of the developing world where blindness from corneal infection remains a major
problem.
Although the majority of study participants were adults, our results should be
applicable to children, who are disproportionately affected by infectious keratitis in the
developing world. In our risk factor assessment, age was not associated with the
primary study outcomes. In a different study from our site in Tiruchirapalli, India that
involved 269 patients with corneal ulcers, adults and children had similar rates of culture
positivity, a similar spectrum of bacterial isolates, and similar predisposing factors (most
commonly trauma).25
Our conclusions regarding the utility of povidone-iodine are supported by the fact
that we obtained similar results in two different countries, with different spectrums of
causal bacteria, and that results were consistent across different types of comparisons,
using multiple outcome measures. Nevertheless, there are several limitations to this
study. A large number of screened individuals were rejected, to provide a
homogeneous population for this initial, proof-of-principle study; exclusions may account
for the fact that we saw no perforations, as in other corneal ulcer treatment trials.26
Eligible individuals may not have been representative of the total population with
infectious keratitis. Because of our selected population, and because of variation in

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 10.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

disease factors between different geographic areas, it may not be possible to generalize
our results to all patients with infectious keratitis in all parts of the world. Our study
design was necessarily dictated, in part, by cultural factors and clinical practices at the
study sites. As a result, follow-up was short, often without confirmation of complete
disease resolution. Presumed cure was only a surrogate for confirmation that infection
was eradicated by treatments; however, based on experience at the study sites, it is
believed that infection is eliminated and healing will ensue when the criteria for our
primary outcome measure is achieved. To compensate for the fact that many study
participants were discharged and lost to follow-up before the primary study end-point
was achieved, we performed a post hoc analysis of patients who met criteria for
recovering, which was thought to be a clinically relevant status. Nevertheless, one
must be cautious in the interpretation of post hoc analyses. Visual acuity has been a
major outcome measure in some corneal ulcer treatment trials, but comparisons have
been made at 3 months or later, when complete healing can be expected;26 we
therefore did not compare visual acuity alone between groups as an outcome measure
because of the relatively short duration of our study.
One might argue that our study did not hold povidone-iodine to an appropriate
standard by comparing it to ciprofloxacin. There has been concern over increasing
bacterial resistance to ciprofloxacin.27 A study from Hyderabad, India, found that 21%
of Staphylococcus aureus isolates from patients with infectious keratitis were resistant
to ciprofloxacin.28 In contrast, an Australian study found 96% of bacterial keratitis
isolates were sensitive to ciprofloxacin.29 Another study from Tiruchirapalli, India, found
that 81% of corneal ulcers healed in response to topical ciprofloxacin, and that healing
time was similar to corneal ulcers treated with topical gatifloxacin.30 It is possible that
results would have been better had we treated with a fourth-generation fluoroquinolone,
as in other clinical trials.26 Nevertheless, we believe our choices of control antibiotics
were realistic, considering the options available in resource-poor areas of the world.
Emerging resistance of bacteria to many commercially available antibiotics31,32 in
fact makes use of povidone-iodine more compelling in resource-poor areas of the
world. Newer drugs to which these organisms may be susceptible, including fourthgeneration fluoroquinolones, will not be widely available in these areas.
In conclusion, we found no strong evidence that povidone-iodine is inferior to
antibiotics commonly available in the developing world for treatment of infectious
keratitis caused by a broad range of bacteria. We therefore believe it is reasonable to
consider use of povidone-iodine as an alternative to topical antibiotics in areas of the
developing world where antibiotics are prohibitively expensive or otherwise unavailable
to treat bacterial keratitis. Povidone-iodine is inexpensive, readily available, and easy to
prepare. In addition, povidone-iodine colors the eye brown for about 2 minutes,
confirming proper administration. Use of povidone-iodine may reduce the incidence of
vision loss caused by corneal scarring, which is a particularly important problem among
children.
Despite the encouraging results of this initial study, further investigations are
warranted. Treatment of infectious keratitis with povidone-iodine should be studied in
specific settings within the developing world where it is most likely to be used. Its
efficacy should also be studied among patients with other disease factors, such as
keratomalacia associated with vitamin A deficiency. Additional outcome measures,

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 11.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

including visual acuity and severity of corneal scarring, should be investigated, and
compliance to treatment should be evaluated.

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 12.
ACKNOWLEDGEMENTS

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

A. Funding: This study was supported by Thrasher Research Fund, Salt Lake City, UT
(SJI), The Wendy and Theo Kolokotrones Family Fund, Los Angeles, CA (SJI),
Research to Prevent Blindness, New York, NY (SJI, GNH) and the Skirball Foundation,
New York, NY (GNH).
B. Financial Disclosure: None of the other authors have conflicts of interest with any
aspect of this study. Dr. Holland has served on advisory boards for the following
companies: Genentech, Incorporated (San Francisco, CA, USA); Novartis International
AG (Basel, Switzerland); Santen, Incorporated (Emeryville, CA, USA); and XOMA (US)
LLC (Berkeley, CA, USA). Dr. Holland is recipient of an RPB Physician-Scientist
Award. Funding entities had no role in the conduction or presentation of this study.

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 13.
REFERENCES

6.

7.
8.
9.

10.
11.

12.

13.

14.

15.
16.
17.

RI
PT

SC

5.

M
AN
U

4.

TE
D

3.

EP

2.

Murthy GV, Vashist P, John N, Pokharel G, Ellwein LB. Prevelence and causes
of visual impairment and blindness in older adults in an area of India with a high
cataract surgical rate. Ophthalmic Epidemiol. 2010;17(4):185-195.
Gilbert C. New issues in childhood blindness. Community Eye Health.
2001;14:53-56.
Gogate P, Kalua, K., Courtright, P. Blindness in childhood in developing
countries: time for a reassessment? PLoS Med. 2009;5(12):e1000177.
Houang ET, Gilmore OJ, Reid C, Shaw EJ. Absence of bacterial resistance to
povidone iodine. J Clin Pathol. 1976;29(8):752-755.
Apt L, Isenberg SJ, Yoshimori R, Spierer A. Outpatient topical use of povidoneiodine in preparing the eye for surgery. Ophthalmology. 1989;96(3):289-292.
Isenberg SJ, Apt L, Yoshimori R, Pham C, Lam NK. Efficacy of topical povidoneiodine during the first week after ophthalmic surgery. Am J Ophthalmol.
1997;124(1):31-35.
Isenberg SJ, Apt L, Wood M. A controlled trial of povidone-iodine as prophylaxis
against ophthalmia neonatorum. N Engl J Med. 1995;332(9):562-566.
Isenberg SJ, Apt L, Valenton M, et al. A controlled trial of povidone-iodine to treat
infectious conjunctivitis in children. Am J Ophthalmol. 2002;134(5):681-688.
DiLoreto DA, Jr., Bressler NM, Bressler SB, Schachat AP. Use of best and final
visual acuity outcomes in ophthalmological research. Arch Ophthalmol.
2003;121(11):1586-1590.
Jabs DA. Improving the reporting of clinical case series. Am J Ophthalmol.
2005;139(5):900-905.
Schreier H, Erdos G, Reimer K, Konig B, Konig W, Fleischer W. Molecular
effects of povidone-iodine on relevant microorganisms: an electron-microscopic
and biochemical study. Dermatology. 1997;195 Suppl 2:111-116.
Pels E, Vrensen GF. Microbial decontamination of human donor eyes with
povidone-iodine: penetration, toxicity, and effectiveness. Br J Ophthalmol.
1999;83(9):1019-1026.
Apt L, Isenberg S, Yoshimori R, Paez JH. Chemical preparation of the eye in
ophthalmic surgery. III. Effect of povidone-iodine on the conjunctiva. Arch
Ophthalmol. 1984;102(5):728-729.
Apt L, Isenberg SJ, Yoshimori R, et al. The effect of povidone-iodine solution
applied at the conclusion of ophthalmic surgery. Am J Ophthalmol.
1995;119(6):701-705.
Isenberg SJ, Apt L, Yoshimori R, Leake RD, Rich R. Povidone-iodine for
ophthalmia neonatorum prophylaxis. Am J Ophthalmol. 1994;118(6):701-706.
Hale LM. The treatment of corneal ulcer with povidone-iodine (Betadine). N C
Med J. 1969;30(2):54-56.
Schuhman G, Vidic B. Clinical experience with povidone-iodine eye drops in
patients with conjunctivitis and keratoconjunctivitis. J Hosp Infect. 1985;6 Suppl
A:173-175.

AC
C

1.

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 14.

23.
24.

25.
26.

27.
28.
29.

30.

31.
32.

RI
PT

SC

22.

M
AN
U

21.

TE
D

20.

EP

19.

Katz J, Khatry SK, Thapa MD, et al. A randomised trial of povidone-iodine to

reduce visual impairment from corneal ulcers in rural Nepal. Br J Ophthalmol.
2004;88(12):1487-1492.
Gregori NZ, Schiffman JC, Miller DM, Alfonso EC. Clinical trial of povidone-iodine
(Betadine) versus placebo in the pretreatment of corneal ulcers. Cornea.
2006;25(5):558-563.
Melki SA, Safar A, Yaghouti F, et al. Effect of topical povidone-iodine versus
topical ofloxacin on experimental Staphylococcus keratitis. Graefes Arch Clin Exp
Ophthalmol. 2000;238(5):459-462.
Sharma LG, Goyal JK, Sharma U, Chandak GK, Nepalia LK. Evaluation of
topical povidone-iodine versus gentamycin in staphylococcus coagulase positive
corneal ulcers--an experimental study. Indian J Ophthalmol. 1990;38(1):30-32.
Michalova K, Moyes AL, Cameron S, et al. Povidone-iodine (betadine) in the
treatment of experimental Pseudomonas aeruginosa keratitis. Cornea.
1996;15(5):533-536.
Berkelman RL, Holland BW, Anderson RL. Increased bactericidal activity of dilute
preparations of povidone-iodine solutions. J Clin Microbiol. 1982;15(4):635-639.
Benevento WJ, Murray P, Reed CA, Pepose JS. The sensitivity of Neisseria
gonorrhoeae, Chlamydia trachomatis, and herpes simplex type II to disinfection
with povidone-iodine. Am J Ophthalmol. 1990;109(3):329-333.
Parmar P, Salman A, Kalavathy CM, Kaliamurthy J, Thomas PA, Jesudasan CA.
Microbial keratitis at extremes of age. Cornea. 2006;25(2):153-158.
Srinivasan M, Mascarenhas J, Rajaraman R, et al. Corticosteroids for bacterial
keratitis: the Steroids for Corneal Ulcers Trial (SCUT). Arch Ophthalmol.
2012;130(2):143-150.
Goldstein MH, Kowalski RP, Gordon YJ. Emerging fluoroquinolone resistance in
bacterial keratitis: a 5-year review. Ophthalmology. 1999;106(7):1313-1318.
Sharma V, Sharma S, Garg P, Rao GN. Clinical resistance of Staphylococcus
keratitis to ciprofloxacin monotherapy. Indian J Ophthalmol. 2004;52(4):287-292.
Ly CN, Pham JN, Badenoch PR, et al. Bacteria commonly isolated from keratitis
specimens retain antibiotic susceptibility to fluoroquinolones and gentamicin plus
cephalothin. Clin Experiment Ophthalmol. 2006;34(1):44-50.
Parmar P, Salman A, Kalavathy CM, et al. Comparison of topical gatifloxacin
0.3% and ciprofloxacin 0.3% for the treatment of bacterial keratitis. Am J
Ophthalmol. 2006;141(2):282-286.
Chaurasia S, Ramappa M, Ashar J, Sharma S. Neonatal infectious keratitis.
Cornea. 2014;33(7):673-676.
Hernandez-Camarena JC, Graue-Hernandez EO, Ortiz-Casas M, et al. Trends in
Microbiological and Antibiotic Sensitivity Patterns in Infectious Keratitis: 10-Year
Experience in Mexico City. Cornea. 2015;34(7):778-785.

AC
C

18.

ACCEPTED MANUSCRIPT
Isenberg, et al. Treatment of Bacterial Keratitis 15.
FIGURE LEGEND

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

Kaplan-Meier plot showing cumulative probability of presumed cure for 172 clinical trial
participants with bacterial keratitis, grouped by study site and randomized treatment
assignments. There was no statistically significant difference between treatments at
each site (p=0.13 for the Philippines site; p=0.22 for the India sites).

ACCEPTED MANUSCRIPT

Table 1. Demographic and Clinical Characteristics for 172 Individuals Enrolled in A Randomized Clinical Trial of
Treatment for Bacterial Keratitis with Povidone-Iodine or Topical Antibiotics at Sites in the Philippines and India.
India Sites

Neomycinpolymyxin Bgramicidin
(n=49)

Povidone-iodine
(n=38)

Ciprofloxacin
(n=45)

Male gender (n [percentage])

34 (85%)

39 (80%)

27 (71%)

28 (62%)

Proportion right eyes (n [percentage])

19 (48%)

19 (39%)

20 (53%)

21 (47%)

36 18

46 19

40 18

1 (2%)

5 (13%)

6 (13%)

Other ocularb (eyes [percentage])

1 (3%)

2 (4%)

1 (3%)

5 (11%)

Systemicc (patients [percentage])

2 (4%)

2 (5%)

2 (4%)

Characteristic

Age in years (mean + S.D.)

34 15

EP

SD=standard deviation

AC
C

Corneala (eyes [percentage])

TE
D

Pre-existing disease

SC

Povidone-iodine
(n=40)

M
AN
U

RI
PT

Philippines Site

a Corneal diseases included: preexisting corneal scars (n=6); spheroidal degeneration (n=2); aphakic corneal edema
(n=1), keratoconjunctivitis sicca (n=1); neurotrophic keratitis associated with herpes zoster ophthalmicus (n=1);
corneal graft (n=1).
b Other ocular diseases included: chronic narrow angle glaucoma (n=1); blepharitis/recurrent hordeola (n=4); cataract
(n=2); status post lensectomy, vitrectomy, and retinal detachment repair with scleral buckling following trauma (n=1);
lagophthalmos (n=1).
c Systemic diseases included: diabetes mellitus (n=3); rheumatoid arthritis (n=1), hypertension (n=1); Hansen disease
(n=1).

ACCEPTED MANUSCRIPT

Presumed Curea

RI
PT

Table 2. Outcomes of Presumed Cure and Recovering for Bacterial Keratitis of 172 Eyes Grouped by Study Site
and Randomized Treatment Assignment.
Recoveringb

Philippines Site

India Sites

Philippines Site

India Sites

Povidone-iodine

12

Antibioticd

17

10

-9.5 to 0.7

-35.2 to 3.2

-4.3 to 1.4

-11.7 to 3.2

1.70

1.16

1.20

0.73-3.94

0.72-1.85

0.63-2.2

.22

.54

.59

Povidone-iodine (vs. antibiotic)

HR for event

1.46
0.90-2.36

P valuef

.13

CI=confidence interval

TE
D

95% CI

M
AN
U

95% CIe

SC

Median intervals to eventc (days)

AC
C

EP

a Primary outcome measure; presumed cure was defined as closure of the associated epithelial defect and the absence
of inflammatory signs other than minimal injection.
b Determined in post-hoc analysis, as a means of comparing the speed with which improvement occurred; recovering
defined as the presence of a residual epithelial defect no larger in area than 1mm2, no more than moderate injection,
and no more a score of mild for all other inflammatory signs. Patients were often discharged from the hospital when
the status of recovery was achieved.
c Interval from start of treatment to event.
d Neosporin-polymyxin B-gramicidin at the Philippines site and ciprofloxacin at the India sites.
e 95% CI for the difference in days to event, calculated by subtracting the medial interval for the antibiotic treatment
group from the medial interval for the povidone-iodine treatment group; negative values indicate shorter intervals to
event for the povidone-iodine treatment group.
f Cox proportional hazards survival analysis.

ACCEPTED MANUSCRIPT

Table 3. Status of Bacterial Keratitis at Most Recent Examination for 172 Eyes
Categorized by Study Site and Randomized Treatment Assignment
Philippines Site

India Sites
Povidoneiodine (n=38)

Ciprofloxacin
(n=45)

Presumed curea

30 (75%)

39 (80%)

12 (32%)

10 (22%)

Recoveringb

1 (2.5%)

4 (8%)

7 (18%)

11 (24%)

Improvingc

1 (2.5%)

4 (8%)

8 (21%)

14 (31%)

Persistenced

1 (2.5%)

1 (2%)

3 (8%)

3 (7%)

6 (11%)

2 (4%)

Failuref

SC

M
AN
U

Worsenede

RI
PT

Povidoneiodine (n=40)

Neomycinpolymyxin Bgramicidin
(n=49)

Status

7 (18%)

1 (2%)

2 (5%)

4 (9%)

1 (2%)

Dropped-outg

AC
C

EP

TE
D

a Primary outcome measure; presumed cure was defined as closure of the associated
epithelial defect and the absence of inflammatory signs other than minimal injection.
b Determined in post-hoc analysis, as a means of comparing the speed with which
improvement occurred; recovering defined as the presence of a residual epithelial
defect no larger in area than 1mm2, no more than moderate injection, and no more a
score of mild for all other inflammatory signs. Patients were often discharged from
the hospital when the status of recovery was achieved.
c Improving was defined as a lack of additional melt over two examinations and a
quantitative decrease in severity of the following factors since baseline: infiltrate,
inflammation, and epithelial defect.
d Persistence was defined as all cases with any sign of active keratitis that did not
fulfill criteria for status of recovered, improved, worsened, or failed.
e Worsened was defined as a deterioration in any factor.
f Failure was defined as a deterioration in two or more signs of inflammation or
increase in size or additional melt on two consecutive daily examinations after 48
hours of treatment. Failure required a change in therapy.
g Individual who chose not to participate in the clinical trial after randomization, but
before start of assigned treatment.

ACCEPTED MANUSCRIPT

Table 4. Associations between Host or Disease Factors and Cure of Bacterial Keratitis for 172 Individuals
Treated in A Randomized Clinical Trial and Categorized By Study Site.
Philippines Site

RI
PT

Factor

India Sites

95% CIa

P value

HRa

95% CIa

P value

Gender (male vs. female)

0.91

0.46 1.71

0.77

1.09

0.45 2.62

0.86

Age (>18 yrs vs. 18 yrs)

1.11

0.47 2.59

0.82

0.31

0.07 1.43

0.13

Affected eye (right vs. left)

1.19

0.74 1.90

0.48

0.84

0.34 2.09

0.71

M
AN
U

Visual acuity

SC

HRa

1.00

0.05 to 0.29

1.10

0.59 2.06

0.76

0.53

0.19 1.51

0.23

< 0.05

0.52

0.28 0.96

0.04

0.38

0.13 1.14

0.08

Organism (gram positive vs. negative)

1.23

0.71 2.11

0.46

1.54

0.50 4.70

0.45

Size (> 2.75mm2 vs. 2.75 mm2)

0.33

0.18 0.60

0.0002

0.33

0.11 0.97

0.04

Ulcer depth
0 25%b

EP

1.00

Inflammation score (>3 vs. 3)

0.83

1.00
0.64

1.07

0.24 4.84

0.93

---

---

---

0.55

0.09 3.21

0.51

1.22

0.70 2.12

0.49

0.94

0.37 2.41

0.90

HR=Hazard ratio; CI=confidence interval

a
b
c

1.00

0.38 1.82

AC
C

26 50%
51 100%c

TE
D

0.30 to 1.6b

Hazard ratio and 95% confidence interval determined by Cox regression.

Reference group.
Only three subjects at the Philippines site had ulcer depths >50%.

AC
C

EP

TE
D

M
AN
U

SC

RI
PT

ACCEPTED MANUSCRIPT