






































Hrev_master


[Eye Reports 2011; 1:e9] [page 25]

Diabetes and corneal 
endothelial cell characteristics:
a study based on Eye Bank
data
Geoffrey Brown,1 Eric Siegel,2
Steve Staples,1 Jennifer Doyle,1
Jason Y. Chang3,4
1Arkansas Lions Eye Bank and laboratory;
2Department of Biostatistics;
3Department of Neurobiology and
Developmental Sciences; 4Department 
of Ophthalmology, Jones Eye Institute,
University of Arkansas for Medical
Sciences, Little Rock, AR, USA

Abstract 

The aim of the article is to determine
whether corneal endothelial cell density and
other characteristics, such as cell area, pleo-
morphism and polymegathism, are affected by
diabetes. Corneal endothelial cell density and
other characteristics of donor eyes collected
during 2007 and 2008 in a local Eye Bank were
measured by the HAI Eyebank Specular
Microscope System. Adult donors aged 21 or
older who consented to research were divided
into healthy versus compromised eye-status
groups based on eye disease or past eye sur-
geries. Differences in corneal measures
between diabetic and non-diabetic subjects
were analyzed separately in each group via
Mixed Models ANCOVA, with Diabetes as the
fixed effect, Donor as the random effect, and
Age as the continuous covariate. A total of 253
subjects met study criteria, of which 81 (32%)
had diabetes. In the 180 subjects with healthy
eye status, the medians (ranges) of age were
62 (29-78) years among 52 diabetics (29%),
versus 57 (21-79) years among non-diabetics
(P=0.013). In the 73 subjects with compro-
mised eye status, the medians (ranges) of age
were 70 (32-78) years among 29 diabetics
(40%), versus 70 (29-79) years among non-
diabetics (P=0.77). Between diabetics and
non-diabetics, eye disease and past eye surger-
ies were well-balanced in the compromised
eye-status group, while race and sex were well-
balanced in both eye-status groups. Results
from separate analyses on the two groups indi-
cated that diabetes did not affect corneal cell
density or other corneal-cell characteristics
analyzed. Even though diabetics constituted a
large percentage of the Eye Bank donor popu-
lation, this disease did not have a statistically
significant impact on corneal endothelial cell
density, cell area, pleomorphism or poly-
megathism. 

Introduction

Since the first successful corneal transplant
performed in 1905,1 there is a constant need
for more donor eyes for this procedure. As
such, there were 39,391 corneal transplants
performed in the United States in 2007, and
the number increased to 41,652 in 2008
according to the record provided by the Eye
Bank Association of America (EBAA).2

Corneal endothelial cell density is one of the
most important criteria that determine
whether a donor cornea is eligible for trans-
plant. There is an age-dependent decrease of
corneal endothelial cell density, a phenomenon
that has been documented in people from var-
ious ethnic groups. Examples include those
from the United states,3,4 Japan,4 India,5

Philippine,6 Pakistan,7 Iran,8 and China.9

However, it is important to note that all of the
reports above are data derived from normal,
healthy volunteers. Very often people with a
history of diabetes mellitus were excluded. As
a result, the data presented in those reports do
not necessarily reflect those from donor eyes
in a typical Eye Bank. This is because a typical
Eye Bank does encounter donors with various
health conditions, including diabetes, and can
be very different from those normal, healthy
volunteers reported in those studies.

Exclusion of people with diabetes in those
reports analyzing corneal endothelial cell den-
sity implies that diabetes can alter corneal cell
density. However, the published literature on
this subject contains conflicting results.
Diabetes is reported to increase,10 decrease11-13

or have no effect14-17 on corneal endothelial cell
density. 

In this study, we used data collected from
253 research-consenting adults who donated
their corneas to a local Eye Bank during 2007
and 2008 to examine whether there was evi-
dence suggesting that diabetes might adverse-
ly affect corneal endothelial qualities such as
endothelial cell density. To do this, we classi-
fied donors into two groups, a healthy group
and a group with compromised eye status,
based on history of eye disease or past eye sur-
geries. We then compared the age-adjusted
impact of diabetes on four corneal cell param-
eters: cell density, cell area, pleomorphism,
and polymegathism in each group. The per-
centage of hexagonal cells was used as an indi-
cator of pleomorphism and the coefficient of
variation in cell size was used as an indicator
of polymegathism.

Materials and Methods

Data of this report were extracted from the
Arkansas Lions Eye Bank and laboratory (the

Eye Bank) housed at the University of
Arkansas for Medical Sciences, which routine-
ly obtained informed consent from individual
donors for use of their eyes in research. The
consent included the authorization to share
personal health information recorded in the
Medical/Social History Questionnaire for
research. This Eye Bank received donor eyes
from 154 individuals (113 male, 41 female) in
2007 and 156 individuals (91 male, 65 female)
in 2008. For comparison, the total number of
donors in 2004, 2005 and 2006 was 181, 183
and 148, respectively. Among donors in 2007
and 2008, 275 adults aged 21 years or older
allowed the Eye Bank to use their data for
research. Specimens from this group were
excluded from further processing if they were
of poor conditions, if the donors had positive
serology for viruses such as HIV or hepatitis, or
if the donors had a history of other medical
conditions that were deemed to pose a risk to
corneal-transplant recipients. A total of 501
corneas from 253 donors were eventually
examined on the specular microscope, from
which this report was generated. 

The HAI Eyebank Specular Microscope
System (Hightech American Industrial

Eye Reports 2011; volume 1:e9

Correspondence: Jason Y. Chang, Department of
Neurobiology and Developmental Sciences, Slot
510, University of Arkansas for Medical Sciences
4301 W. Markham St., Little Rock, AR 72205, USA.
Tel. +1.501.686.7025 - Fax: +1.501.686-6382. 
E-mail: changjasony@uams.edu

Key words: corneal endothelial cells, diabetes,
Eye Bank.

Acknowledgements: this work was partially sup-
ported by funds from Research to Prevent
Blindness. We greatly appreciate Eye Bank
donors who not only donated their corneas but
also allowed us to conduct this research. The
thoughtful critiques and suggestions provided by
Dr. Michael Brown are highly appreciated.

Contributions: GB, SS, JD, data collection and
assistance in manuscript writing; ES, data analy-
sis and assistance in manuscript writing; JYC,
organizer of the project, data analysis and the
main writer of the manuscript.

Conflict of interest: the authors report no con-
flicts of interest. 

Received for publication: 14 June 2011.
Accepted for publication: 12 September 2011.

This work is licensed under a Creative Commons
Attribution NonCommercial 3.0 License (CC BY-
NC 3.0).

©Copyright G. Brown et al., 2011
Licensee PAGEPress, Italy
Eye Reports 2011; 1:e9
doi:10.4081/eye.2011.e9

Non
-co

mmerc
ial

 us
e o

nly



[page 26] [Eye Reports 2011; 1:e9]

Laboratories, Inc., Lexington, MA, USA) was
used to document corneal characteristics.
Approximately 100 cells of each cornea were
analyzed, and the computer automatically cal-
culated corneal endothelial cell density, poly-
megathism and pleomorphism for each speci-
men. The mean cell area (in μm2/cell) of each
specimen was calculated as 1,000,000 divided
by its corneal-cell density (in cells/mm2).

According to the Eye Bank record, 73 (29%)
of the 253 study subjects had various types of
eye diseases (e.g., cataract, glaucoma, macular
degeneration) or eye operations (e.g., intra-
ocular lens, trauma repair, shunt, corrective
surgery). In the past, people with these condi-
tions were excluded from studies analyzing the
effect of diabetes on corneal endothelial cell
density because these conditions could poten-
tially affect one or more of the parameter stud-
ied.12-15,17,18 In our study, a preliminary analysis
was conducted for confounding of eye disease
and eye surgeries with diabetes and age.
Because of the results, subjects were classified
as having compromised eye status if they had
eye disease or a history of eye surgeries, ver-
sus having healthy eye status if they had nei-
ther, and the two eye-status groups were ana-
lyzed separately in parallel analyses, as
described in the next section.  

Statistical analyses 
In the preliminary analysis for confounding,

the associations of age with eye disease and
eye surgeries were tested using the Wilcoxon
rank sum (WRS) test, while the associations of
diabetes with eye disease and eye surgeries
were tested using the Pearson chi-square test;
the magnitude of all associations were quanti-
fied using Spearman’s rank-correlation coeffi-
cient. Within each eye-status group, donor
ages were summarized by diabetes status
(yes/no) as the median (range), and tested for
imbalance via WRS test, while race and sex
were summarized by diabetes status as propor-
tions, and tested for imbalance via Fisher’s
exact test. Within the compromised eye-status
group, the incidence of eye disease, intra-ocu-
lar lens (IOL) surgeries, and non-IOL surger-

ies were summarized by diabetes status as pro-
portions, and tested for imbalance using
Fisher’s exact test. Corneal-cell measures were
summarized by eye status and decade of age as
means and standard deviations (SDs). Left
eyes were compared to right eyes via paired t-
test for possible differences in corneal meas-
ures. The impact of diabetes on corneal meas-
ures was conducted in separate, parallel analy-
ses on the two eye-status groups using mixed-
models analysis of covariance (ANCOVA),19

with Diabetes as the fixed effect, Donor as the
random effect, and Age as the continuous
covariate. The validity of ANCOVA models were
pre-tested via mixed-models regression tech-
niques, specifically by testing for significant
age-by-diabetes interactions. Following the
practice of Ray and Rosner,21 inter-eye correla-
tions (IECs) were estimated as intra-class cor-
relations (ICCs) from the mixed-models vari-
ance components, using the following formula: 

where s2
b denotes the variance between

donors, s2
w denotes the variance within

donors, and their sum denotes the total vari-
ance. To help interpret the diabetes effects bio-
logically, predicted reference means were cal-
culated for a reference age of 60 years in the
healthy eye-status group and 70 years in the
compromised eye-status group; both reference
ages were chosen to be near the medians of
their groups. Finally, the question was investi-
gated of whether the age trends, age-adjusted
diabetes effects, or total variances and IECs
differed significantly between subjects with
healthy versus compromised eye status. To do
this, the data for each corneal-cell measure
from the two groups were combined into a sin-
gle mixed-models analysis that contained an
eye-status main effect plus eye-status interac-
tions with age and diabetes. Eye-status differ-
ences in the age trends and diabetes effects
were tested for significance using the Type-III

F-statistics for the eye-status interactions with
age and diabetes, respectively. Eye-status dif-
ferences in the total variances and IECs were
tested for significance via likelihood-ratio chi-
square test of the full model (having different
variance components between groups) against
a reduced model (having equal variance com-
ponents between groups). An alpha=0.05 sig-
nificance level was used for all statistical tests. 

Results

Donor characteristics
Among a total of 310 donors in 2007 and

2008, a total of 275 adults 21 or older (134 in
2007 and 141 in 2008) gave informed consent
to participate in research. Of these, 22 had
their specimens excluded from processing
because of poor condition, positive viral serol-
ogy, or medical history for conditions that put a
transplant recipient at risk. Of the remaining
253 subjects, 73 (29%) were classified as hav-
ing compromised eye status (eye disease or a
history of eye surgeries), while 180 (71%)
were classified as having healthy eye status
(no eye disease and no history of eye surger-
ies). A total of 81 subjects in the study (32%)
had diabetes. 

The preliminary analysis for confounding
showed (i) that eye disease was associated
strongly with age (rank correlation 0.40; WRS
P<0.0001) and mildly with diabetes (rank cor-
relation 0.12; chi-square P=0.06), (ii) that eye-
surgery history was associated strongly with
age (rank correlation 0.38; WRS P<0.0001)
and moderately with diabetes (rank correla-
tion 0.16; chi-square P=0.01), and (iii) that
eye-surgery history had high association with
eye disease (rank correlation 0.72; chi-square
P<0.0001). Because of the evident confound-
ing, we chose to study the two eye-status
groups in separate, parallel analyses. 

Table 1 gives breakdowns of donor charac-
teristics by diabetes status within the two eye-
status groups. In adults with healthy eye sta-
tus, the median age was 62 years for diabetics

Article

Table 1. Donor characteristics by diabetes status.

In adults with In adults with 
healthy eye status compromised eye status Overall

Diabetic Nondiabetic P* Diabetic Nondiabetic P*

Donor number 52 128 --- 29 44 --- 253
Median age (range) 62 (29-78) 57 (21-79) 0.013 70 (32-78) 70 (29-79) 0.77 62 (21-79)
Non-Caucasian 5 (10%) 6 (5%) 0.30 1 (3%) 2 (5%) 1.00 14 (6%)
Female 13 (25%) 42 (33%) 0.37 12 (41%) 17 (39%) 1.00 84 (33%)
Any eye disease 0 (0%) 0 (0%) --- 28 (97%) 40 (91%) 0.64 68 (28%)
Any non-IOL surgery 0 (0%) 0 (0%) --- 5 (17%) 3 (7%) 0.25 8 (3%)
Any IOL surgery 0 (0%) 0 (0%) --- 21 (72%) 25 (57%) 0.22 46 (18%)
*P values are via Wilcoxon rank-sum test on Age, and via Fisher’s exact test on all other donor characteristics.

Non
-co

mmerc
ial

 us
e o

nly



[Eye Reports 2011; 1:e9] [page 27]

versus 57 years for non-diabetics (P=0.013);
however, the age distributions still overlapped
substantially, with ranges of 29-78 years for
diabetics versus 21-79 years for non-diabetics.
In adults with compromised eye status, the
median age was 70 years for diabetics versus
70 years for non-diabetics (P=0.77). Race and
sex showed no significant diabetes imbalance
in either eye-status group, and the incidences
of eye diseases and past eye surgeries showed
no significant diabetes imbalance in the com-
promised eye-status group. 

Corneal endothelial-cell measures
of each eye status group

Left eyes were compared to right eyes via
paired t-test for significant differences in
corneal epithelial-cell measures; none were

found (all P >0.50). Table 2 shows descriptive-
ly how corneal cell density (cells/mm2), cell
area (μm2/cell), pleomorphism (% of cells that
are hexagonal), and polymegathism (coeffi-
cient of variation in cell size) varies with age
and eye status. Both eye-status groups show
similarly decreasing trends with age in cell
density, similarly increasing trends with age in
cell area, and similarly unclear trends with age
in pleomorphism and polymegathism. 

Age-adjusted diabetes effect in
adult donors with healthy eye status

The corneal-cell measures of adult donors
with healthy eye status were analyzed for age-
adjusted diabetes effects using mixed-models
regression techniques. None of the age trends
had a significant quadratic component, indi-

cating that it was reasonable to fit all corneal
measures with linear trends. All age-by-dia-
betes interactions were non-significant, indi-
cating for each measure (i) that the linear
trends had the same slope for diabetics versus
non-diabetics, and therefore (ii) that mixed-
models ANCOVA was the appropriate analysis
model. Table 3 shows the results of the mixed-
models ANCOVAs on subjects with healthy eye
status. Inter-eye correlations ranged from a
low of 26.4% for pleomorphism to a high of
69.7% for cell density. The age effect was high-
ly significant (P<.0001) with positive slope for
cell area, highly significant (P<.0001) with
negative slope for cell density, but insignifi-
cant for pleomorphism (P=0.63) and poly-
megathism (P=0.24). The age-adjusted dia-
betes effects were statistically insignificant

Article

Table 2. Corneal cell measures versus age group and eye status.

Age group Eye status # of eyes* Cell density° Cell area# Pleomorphism§ Polymegathism$

21-30 Healthy 30 3019 (258) 333.7 (29.0) 54.47 (7.14) 19.47 (3.31)
Compromised 2 2918 (74) 342.9 (8.7) 54.50 (0.71) 19.00 (1.41)

31-40 Healthy 24 3072 (451) 330.7 (37.8) 55.58 (6.90) 18.50 (3.05)
Compromised 3 2979 (105) 336.0 (11.6) 60.67 (2.52) 16.67 (0.58)

41-50 Healthy 56 2662 (293) 380.2 (42.1) 52.46 (7.19) 19.19 (4.84)
Compromised 4 2605 (123) 384.5 (18.9) 50.50 (3.87) 16.00 (1.63)

51-60 Healthy 94 2706 (319) 375.8 (56.0) 52.91 (7.08) 20.05 (5.68)
Compromised 18 2791 (279) 361.6 (34.8) 53.33 (6.52) 21.11 (3.20)

61-70 Healthy 106 2627 (230) 383.6 (34.7) 52.52 (6.10) 19.83 (3.21)
Compromised 44 2392 (453) 440.9 (131.) 50.55 (7.24) 21.59 (4.03)

71-80 Healthy 50 2530 (404) 406.9 (76.9) 54.32 (6.76) 20.08 (2.58)
Compromised 70 2413 (452) 429.7 (87.3) 51.30 (6.87) 21.51 (3.59)

Average of all groups 2639 (389) 388.8 (72.4) 52.76 (6.86) 20.12 (4.12)
Data from both eyes of the same individual were used to calculate the mean (SD) of the group. *Measurements were available on a total of 501 eyes from 253 subjects. °Cell density, average density of cells in the
corneal specimen, in cells/mm2. #Cell area, average area of cells in the corneal specimen, in μm2/cell. §Pleomorphism, percentage of cells in the specimen that have a hexagonal shape. $Polymegathism, Coefficient of
variation of cell size in the corneal specimen, in percentage units.

Table 3. Mixed-models ANCOVAs for diabetes effect adjusting for age, conducted on adult donors with healthy eye status.

Corneal cell measure Total variance Effect name Estimate±SE° DF t P
(units) (IEC)* statistic

Cell area 2512.1 Age# +1.36±0.24 179 +5.61 <.0001
(μm2/cell) (62.1%) Diabetes§ –7.64±7.60 179 –1.01 0.32

(reference means)$ (384.46±4.26)^
(376.82±6.26)** - - -

Cell density 101899 Age# –9.71±1.58 179 –6.16 <.0001
(cells/mm2) (69.7%) Diabetes§ +38.8±49.6 179 +0.78 0.43

(reference means)$ (2650.3±27.7)^
(2689.0±40.8)** - - -

Pleomorphism (%) 46.355 Age# –0.014±0.029 178 –0.48 0.63
(26.4%) Diabetes§ –0.402±0.913 178 –0.44 0.66

(reference means)$ (53.287±0.512)^
(52.885±0.751)** - - -

Polymegathism (%) 17.601 Age# +0.022±0.019 179 +1.18 0.24
(37.2%) Diabetes§ +0.103±0.586 179 +0.18 0.86

(reference means)$ (19.766±0.328)^
(19.869±0.482)** - - -

*Inter-eye correlation between left and right eyeballs from the same donor; °Standard Error of the estimate. #Slope of the linear age trend, in units/year, for subjects in the healthy eye-status group. §Age-adjusted differ-
ence, diabetics minus non-diabetics, for subjects in the healthy eye-status group. $Reference means are the predicted means of 60-year-old subjects in the healthy eye-status group, where ^ denotes the predicted mean
of the non-diabetic 60-year-old and * denotes the predicted mean of the diabetic 60-year-old. Reference means double as Y-intercepts in the regression equations Y = Reference + Slope x (Age – 60), where Y is the
corneal-cell measure, Age is the subject’s age in years, Slope is the slope of the linear age trend, and Reference is the reference mean denoted by ^ and **for non-diabetics and diabetics, respectively.

Non
-co

mmerc
ial

 us
e o

nly



[page 28] [Eye Reports 2011; 1:e9]

(lowest P=0.32, cell area) on all four cell meas-
ures. Table 3 also shows the ANCOVA models’
predicted reference means for two 60-year-old
individuals with healthy eye status, one with
diabetes and one without, in order to facilitate
biological interpretation of the diabetes
effects. Computations from the values in Table
3 show that the reference diabetic had 2.0%
lower cell area, 1.5% higher cell density, 0.7%
lower pleomorphism, and 0.5% higher poly-
megathism than did the reference non-diabet-
ic. It should be noted that the reference means
in Table 3 double as Y-intercepts in the regres-
sion equations Y = Reference + Slope x (Age -
60), where Y is the predicted mean corneal-cell
measure, Age is the subject’s age in years,
Slope is the slope of the linear age trend, and
Reference is the appropriate reference mean
for diabetics or non-diabetics with healthy eye
status. Graphic illustration of the results for
cell density is shown in Figure 1.

Age-adjusted diabetes effect in
adult donors with compromised
eye status

The corneal-cell measures of adult donors
with compromised eye status were likewise
analyzed for age-adjusted diabetes effects
using mixed-models regression techniques.
Again, none of the age trends had a significant
quadratic component, indicating that it was
reasonable to fit all corneal measures with lin-
ear trends. And again, all age-by-diabetes
interactions were non-significant, indicating
for each measure (i) that the linear trends had
the same slope for diabetics versus non-diabet-
ics, and therefore (ii) that mixed-models
ANCOVA was the appropriate analysis model.
Table 4 shows the results of the mixed-models
ANCOVAs on subjects with compromised eye
status. Inter-eye correlations ranged from a
low of 24.9% for cell area to a high of 51.7% for
cell density. The age effect was statistically sig-
nificant (P=0.0075) with positive slope for cell
area, highly significant (P=0.0017) with nega-
tive slope for cell density, insignificant
(P=0.10) for pleomorphism, and statistically
significant (P=0.010) with positive slope for
polymegathism. The age-adjusted diabetes
effects were statistically insignificant (lowest
P=0.19, pleomorphism) on all four cell meas-
ures. Table 4 also shows the ANCOVA models’
predicted reference means for two 70-year-old
individuals with compromised eye status, one
with diabetes and one without, in order to
facilitate biological interpretation of the dia-
betes effects. Computations from the values in
Table 4 show that, compared to the reference
non-diabetic, the reference diabetic had 2.1%
higher cell area, 3.6% lower cell density, 3.4%
lower pleomorphism, and 3.0% higher poly-
megathism. Here, the reference means in
Table 4 double as Y-intercepts in the regres-

Article

Figure 1. Corneal endothelial cell density as a function of age in donors with healthy eye
status. Vertical axis shows cell density in cells/mm2, while horizontal axis shows age in
years. Solid black circles (dashed black line) represent observed values (predicted means)
of cell density in diabetics, while open gray circles (solid line) represent observed values
(predicted means) of cell density in non-diabetics. The regression equation for diabetics
(dashed black line) is: Cell Density (in cells/mm2) = 2689.0-9.71 x (Age – 60), where Age
is in years. The regression equation for non-diabetics (solid gray line) is: Cell Density (in
cells/mm2) = 2650.3-9.71 x (Age – 60), where Age is in years. 

Figure 2. Corneal endothelial cell density as a function of age in donors with compromised
eye status. Vertical axis shows cell density in cells/mm2, while horizontal axis shows age in
years. Solid black circles (dashed black line) represent observed values (predicted means)
of cell density in diabetics, while open gray circles (solid line) represent observed values
(predicted means) of cell density in non-diabetics. The regression equation for diabetics
(dashed black line) is: Cell Density (in cells/mm2) = 2392.1-14.05 x (Age – 70), where Age
is in years. The regression equation for non-diabetics (solid gray line) is: Cell Density (in
cells/mm2) = 2482.3-14.05 x (Age – 70), where Age is in years.

Non
-co

mmerc
ial

 us
e o

nly



[Eye Reports 2011; 1:e9] [page 29]

sion equations Y = Reference + Slope x (Age -
70), where Y is the predicted mean corneal-cell
measure, Age is the subject’s age in years,
Slope is the slope of the linear age trend, and
Reference is the appropriate reference mean
for diabetics or non-diabetics with compro-
mised eye status. Graphic illustration of the
results for corneal cell density is shown in
Figure 2.

Differences between ANCOVA-
model results for subjects with
healthy versus compromised eye
status

Although the focus of this report is on the
age-adjusted diabetes effect, it is reasonable to
ask whether subjects having compromised eye
status (Table 4) differed significantly from
subjects having healthy eye status (Table 3)
with respect to the age trend or age-adjusted
diabetes effect in any corneal-cell measure. We
investigated this question by combining the
data for each corneal-cell measure from the
two groups into a single mixed-models analy-
sis, as described in the Statistical Analysis sec-
tion. For polymegathism, the difference in age
trends between Tables 3 and 4 was almost sig-
nificant (interaction P=0.07), but for the other
three measures, the differences in age trends
between tables was not significant (lowest
interaction P=0.18). And none of the four
corneal-cell measures showed a statistically
significant difference in their diabetes effect
between the two tables (lowest interaction
P=0.21). 

However, likelihood-ratio chi-square tests
showed that, when compared to subjects with
healthy eye status, those with compromised
eye status had significantly larger between-

donor and within-donor variances for both cell
area and cell density. This finding reinforces
the appropriateness of the decision to conduct
separate analyses on subjects with healthy ver-
sus compromised eye status. For cell area,
between-donor variance increased to 1.5-fold
and within-donor variance increased to 7.5-
fold, such that total variance increased to 3.8-
fold (9565.0 divided by 2512.1) while IEC fell by
37.2 percentage points (62.1% versus 24.9%)
from Table 3 to Table 4 (c2=150.9, DF=2;
P<0.0001). For cell density, a similar analysis
indicates that between-donor variance
increased to 1.3-fold and within-donor vari-
ance increased to 2.8-fold, such that total vari-
ance increased to 1.8-fold (180,845 divided by
101,899) while IEC fell by 18.0 percentage
points (69.7% versus 51.7%) from Table 3 to
Table 4 (c2=36.05, DF=2; P<0.0001). On the
other hand, the total variance and IEC showed
very little change between tables for pleomor-
phism (chi-square=0.14, DF=2; P=0.93), and a
suggestive, but insignificant change between
tables for polymegathism (chi-square=4.81,
DF=2; P=0.090). 

Discussion

Corneal endothelial cell density from Eye
Bank donors was previously examined by the
Cornea Donor Study using 1101 qualified
donor cornea collected from January 10, 2000
to August 2, 2002.22 These investigators
noticed a non-linear age-dependent drop of cell
density until age 60. Beyond this point, the cell
density changed little. It should be noted that
only corneas qualified for that particular study

were used in that analysis, which meant that
all corneas analyzed in that report had a cell
density between 2300-3300 cells/mm2.22 In con-
trast, our data included 253 research-consent-
ing adult donors during years 2007 and 2008.
Consequently, ours reflected the raw data that
one would encounter in a typical regional Eye
Bank. Results from this study indicated that
there was an age-dependent decrease of
corneal endothelial cell density in both eye-sta-
tus groups (Tables 3 and 4). It should be noted
that, among those donors at >70 years old,
~73% (88 out of 120 eyes) met the standard
(≥2,300 cells/mm2) set by the Cornea Donor
Study. Thus, age alone should not be used as a
criterion for donor cornea. In practice, our Eye
Bank uses 2000 cells/mm2 as the cutoff criteri-
on to determine whether the cell density of a
donor cornea is qualified for transplant.

According to data provided by the NIH (year
2007), 10.7% (23.5 million) of the population
in the United States age 20 years or older have
diabetes. The prevalence of diabetes increased
to 23.1% (12.2 million) of the population age
60 years or older.23 Our data from 2007 and
2008 indicated that ~32% of our adult donors
had diabetes, which was significantly higher
than the 18% reported in the Cornea Donor
Study that involved 1,101 corneas.22

Importantly, ~38% of donors among those 60
years or older in our study population had dia-
betes. This was much higher than the above-
mentioned 23.1% in the United State. Based
on the report from the Arkansas Department of
Health and Human Services,24 an estimated
233,255 adult residents in Arkansas had dia-
betes in 2005. The cost of hospital charges was
~$87 million for Arkansas in 2005, which
included 651 lower extremity amputations and

Article

Table 4. Mixed-models ANCOVA of diabetes effect adjusting for age, conducted on adult donors with compromised eye status.

Corneal cell measure Total variance Effect name Estimate±SE° DF t P
(units) (IEC)* statistic

Cell area 9565.0 Age# +2.50±0.91 68 +2.76 0.0075
(μm2/cell) (24.9%) Diabetes§ +8.71±18.72 68 +0.47 0.64

(reference means)$ (422.65±12.19)^
(431.36±14.55)** - - -

Cell density 180845 Age# –14.05±4.29 68 –3.27 0.0017
(cells/mm2) (51.7%) Diabetes§ –90.2±89.4 68 –1.01 0.32

(reference means)$ (2482.3±58.2)^
(2392.1±69.5)** - - -

Pleomorphism (%) 46.786 Age# –0.108±0.065 68 –1.66 0.10
(31.3%) Diabetes§ –1.773±1.341 68 –1.32 0.19

(reference means)$ (52.068±0.874)^
(50.295±1.043)** - - -

Polymegathism (%) 13.240 Age# +0.094±0.036 68 +2.63 0.010
(42.9%) Diabetes§ +0.644±0.743 68 +0.87 0.39

(reference means)$ (21.182±0.484)^
(21.827±0.578)** - - -

*Inter-eye correlation between left and right eyeballs from the same donor; °Standard Error of the estimate. #Slope of the linear age trend, in units/year, for subjects in the compromised eye-status group. §Age-adjusted dif-
ference, diabetics minus non-diabetics, for subjects in the compromised eye-status group. $Reference means are the predicted means of 70-year-old subjects in the compromised eye-status group, where ^ denotes the pre-
dicted mean of the non-diabetic 70-year-old and ** denotes the predicted mean of the diabetic 70-year-old. Reference means double as Y-intercepts in the regression equations Y = Reference + Slope x (Age – 70), where
Y is the corneal-cell measure, Age is the subject’s age in years, Slope is the slope of the linear age trend, and Reference is the reference mean denoted by ^ and ** for non-diabetics and diabetics, respectively.

Non
-co

mmerc
ial

 us
e o

nly



[page 30] [Eye Reports 2011; 1:e9]

1565 hospitalizations for ketoacidosis.
This high prevalence of diabetic donors

raised an important question, i.e., whether this
disease can affect corneal endothelial cell
quality, including cell density, thus compromis-
ing the number of qualifying corneas. This
issue becomes even more important when we
consider that the number of qualified corneas
received in each Eye Bank may decrease
because more and more refractory corneal pro-
cedures are being currently performed.25 The
effect of diabetes on corneal endothelial cells
was indeed a subject of discussion in several
reports. Based on a comparison of 30 diabetic
patients and 30 non-diabetic subjects using
contact specular microscope, Siribunkum, et
al. reported that the diabetics had statistically
significant increase in corneal endothelial cell
density and decrease in mean cell area.10 A
comparison of 158 type II diabetic patients and
165 control subjects led Mathew, et al. conclud-
ed that type II diabetes caused an increase
(~5%) of corneal endothelial cell density.26

The finding of an increased endothelial cell
density was not supported by other reports. For
example, Lee et al. conducted an age-adjusted
comparison of 200 diabetics to 100 normal sub-
jects, and found that that diabetics had statis-
tically significantly less (~ 5%) cell density as
compared to controls (diabetes: 2577±27, con-
trol: 2700±39). They also found that diabetics
had more irregular shaped cells and more vari-
ation in cell sizes.12 Modis, et al.27 recently
reported that there was a statistically signifi-
cant decrease (~3%) in endothelial cell densi-
ty in type I diabetes relative to healthy subjects
(2428±219 versus 2495±191). However, they
did not find a statistically significant differ-
ence in endothelial cell density between type II
diabetes and age-matched controls. An earlier
study by Inoue et al. using a multivariate
regression analysis comparing 99 diabetics
and 97 control subjects also found that diabet-
ics had lower endothelial cell density.11

In contrast to those reports above, other
studies found diabetes did not cause statistical
differences in corneal endothelial cell density.
Larsson, et al. analyzed 49 type I diabetics, 60
type II diabetics and their respective age-
matched controls and found there was no dif-
ference in endothelial cell density between
diabetics and control subjects.15 Pardos and
Krachmer studied 52 diabetic patients with
long-term (>14 years) proliferative retinopa-
thy also found no evidence that diabetes
caused statistically significant difference in
endothelial cell density when compared to 32
control subjects.14 Finally, Inoue, et al. exam-
ined a group of 1394 cataract patients before
their surgery procedures and found that the
presence of type 2 diabetes had no effect on
endothelial cell density in this particular group
of patients.18 Results from this study indicated
that diabetes did not cause statistically differ-

ent endothelial cell density (Tables 3,4; Figures
1,2), cell area, pleomorphism and polymegath-
ism (Tables 3,4), regardless of eye status. 

Given our findings that corneas from diabet-
ic donors are comparable to those from non-
diabetic donors and are appropriate for cornea
transplant, an important issue is whether
corneas from diabetic donors perform as well
as those from non-diabetic donors after the
transplant surgery. This issue is important
because a recent report by Mathew, et al. indi-
cated that surgeries that affect cornea, such as
manual small incision cataract surgery, could
cause a decrease in corneal endothelial cell
density, and this decrease appeared to be more
pronounced in type II diabetic patients.26

Based on this finding, there is a possibility
that corneas from diabetic donors may behave
differently compared to those from non-diabet-
ic donors. We currently do not have data to
address this issue and will investigate this in
future studies.

This current study was unique in two ways.
While data from live volunteers reported in
those above-mentioned reports were valid,
they did not necessarily represent the popula-
tion encountered in an Eye Bank. In contrast,
this study was unique in that it represented
the actual data from a regional Eye Bank
where donor corneas were collected and
shipped out routinely for corneal transplants.
Second, while other studies were performed in
subjects without eye diseases and eye surger-
ies, we performed two parallel analyses on
those with and those without compromised eye
conditions. The same conclusion was achieved
in both analyses.

There were certainly some limitations to
this study. First, it represented a set of 2-year
data from an Arkansas Eye Bank. This could be
used to compare data from Eye Banks in other
regions (e.g., Alaska, Maine or Florida) or
other countries (e.g., France, Guatemala or
Australia) but it was not necessarily a fair rep-
resentation of what they would encounter in
those Eye Banks. Second, this study used a
closed pool of donors. Instead of a continuous
recruitment of volunteers to reach a certain
target number (e.g., 100 diabetics and 100 con-
trols), we used data from donors within a 2-
year period and had no control over how many
subjects we would have in each category. Also,
there was a screening process of the potential
donors before we collected their corneas.
Those did not meet our criteria were rejected,
thus their data were not used for this analysis.
Furthermore, we were limited to use only data
from those donors who consented to our
research. In this regard, we were fortunate to
have ~94% of adult donors who gave informed
consent to use their data for research. 

An additional limitation we encountered in
carrying out this project was that we obtained
the information regarding diabetes from the

Medical/Social History Questionnaire, which
gave us a positive identification of diabetic
donors but was limited such that it was with-
out detailed disease history for each individ-
ual. We do not have a detailed record of type I
versus type II diabetes for each donor. As a
result, we grouped both types together in the
analysis. Consequently, we could not conclude
whether type I diabetes was significantly dif-
ferent from type II diabetes for those parame-
ters we evaluated in this study. In a report ana-
lyzing corneal endothelial cells in patients with
type I and type II diabetes, Modis, et al. con-
cluded that type I diabetic corneas were more
susceptible to environmental changes than
type II corneas.27 Based on their findings, there
is a possibility that corneas from type I and
type II diabetic donors may perform differently
after the transplant. This issue warrants fur-
ther investigation in the future.

In conclusion, we performed a study on data
collected from the Arkansas Lions Eye Bank and
laboratory during 2007 and 2008. In findings
similar to those of other studies, we observed
among subjects with healthy eye status (Table
3) an age-dependent decrease in corneal
endothelial cell density and an age-dependent
increase in cell area, along with statistically
insignificant age trends in percentage of hexag-
onal cells and coefficient of variation in cell
size. Among subjects with compromised eye sta-
tus (Table 4), we observed statistically signifi-
cant age-dependent decrease in cell density as
well as increase in cell area and variation in cell
size. The age trend for percentage of hexagonal
cells was not significant. Importantly, subjects
with compromised eye status had significantly
larger variance components for cell density and
cell area; for both measures, the difference was
especially pronounced for the component of
variance between eyeballs within the same
donor. This finding suggests that previous stud-
ies12-15,17,18 were prudent to have excluded sub-
jects with eye disease or past eye surgeries. Also
importantly, there was a large percentage of
donors who had diabetes: 32% overall, 29%
among subjects with healthy eye status and 40%
among subjects with compromised eye status.
Differences in corneal measures between dia-
betic and non-diabetic subjects were analyzed
in both eye-status groups via separate, parallel
mixed-models ANCOVAs. Results indicated that,
when adjusted for age, diabetes did not have a
statistically significant impact on corneal
endothelial cell density, cell area, pleomorphism
and polymegathism, either in subjects with
healthy eye status, or in subjects whose status
was compromised by eye disease or history of
eye surgeries. Based on the parameters ana-
lyzed in this study, corneas from diabetic donors
were comparable to those from non-diabetic
donors for cornea transplants. Future analysis is
required to determine whether the outcomes of
corneal transplant are comparable from corneas

Article

Non
-co

mmerc
ial

 us
e o

nly



[Eye Reports 2011; 1:e9] [page 31]

obtained from diabetic and non-diabetic donors,
and whether corneas from type I and type II dia-
betic donors perform equally well after the
transplant.

References

1. Moffatt SL, Cartwright VA, Stumpf TH.
Centennial review of corneal transplanta-
tion. Clin Experiment Ophthalmol 2005;33:
642-57.

2. EBAA. Eye Bank Association of America
Press Release (April 24, 2009).
http://www.restoresight.org/files/2008pres
srelease_statreport.pdf Accessed: 22
September 2009.

3. Yee RW, Matsuda M, Schultz RO,
Edelhauser HF. Changes in the normal
corneal endothelial cellular pattern as a
function of age. Curr Eye Res 1985;4:671-8.

4. Matsuda M, Yee RW, Edelhauser HF.
Comparison of the corneal endothelium in
an American and a Japanese population.
Arch Ophthalmol 1985;103:68-70.

5. Rao SK, Ranjan Sen P, Fogla R, et al.
Corneal endothelial cell density and mor-
phology in normal Indian eyes. Cornea
2000;19:820-3.

6. Padilla MD, Sibayan SA, Gonzales CS.
Corneal endothelial cell density and mor-
phology in normal Filipino eyes. Cornea
2004;23:129-35.

7. Ashraf KM, Saeed MU, Zia R. Corneal
endothelial cell density in a normal
Pakistani population. Eye (Lond)
2006;20:116-8.

8. Hashemian MN, Moghimi S, Fard MA, et
al. Corneal endothelial cell density and

morphology in normal Iranian eyes. BMC
Ophthalmol 2006;6:9.

9. Yunliang S, Yuqiang H, Ying-Peng L, et al.
Corneal endothelial cell density and mor-
phology in healthy Chinese eyes. Cornea
2007;26:130-2.

10. Siribunkum J, Kosrirukvongs P, Singa -
lavanija A. Corneal abnormalities in dia-
betes. J Med Assoc Thai 2001;84:1075-83.

11. Inoue K, Kato S, Inoue Y, et al. The corneal
endothelium and thickness in type II dia-
betes mellitus. Jpn J Ophthalmol 2002;46:
65-9.

12. Lee JS, Oum BS, Choi HY, et al.
Differences in corneal thickness and
corneal endothelium related to duration in
diabetes. Eye (Lond) 2006;20:315-8.

13. Roszkowska AM, Tringali CG, Colosi P, et
al. Corneal endothelium evaluation in type
I and type II diabetes mellitus.
Ophthalmologica 1999;213:258-61.

14. Pardos GJ, Krachmer JH. Comparison of
endothelial cell density in diabetics and a
control population. Am J Ophthalmol 1980;
90:172-4.

15. Larsson LI, Bourne WM, Pach JM,
Brubaker RF. Structure and function of the
corneal endothelium in diabetes mellitus
type I and type II. Arch Ophthalmol 1996;
114:9-14.

16. Matsuda M, Ohguro N, Ishimoto I, Fukuda
M. Relationship of corneal endothelial
morphology to diabetic retinopathy, dura-
tion of diabetes and glycemic control. Jpn
J Ophthalmol 1990;34:53-6.

17. Quadrado MJ, Popper M, Morgado AM, et
al. Diabetes and corneal cell densities in
humans by in vivo confocal microscopy.
Cornea 2006;25:761-8.

18. Inoue K, Tokuda Y, Inoue Y, et al. Corneal

endothelial cell morphology in patients
undergoing cataract surgery. Cornea
2002;21:360-3.

19. Littell RC, ed. SAS System for Mixed
Models. 4th ed. Analysis of Covariance.
Cary, NC: SAS Institute; 1996. pp. 171-227.

20. Ray WA, O'Day DM. Statistical analysis of
multi-eye data in ophthalmic research.
Invest Ophthalmol Vis Sci 1985;26:1186-8.

21. Rosner B. Statistical methods in ophthal-
mology: an adjustment for the intraclass
correlation between eyes. Biometrics
1982;38:105-14.

22. Sugar A, Gal RL, Beck W, et al. Baseline
donor characteristics in the Cornea Donor
Study. Cornea 2005;24:389-96.

23. National Diabetes Statistics, 2007 fact
sheet. National Institute of Diabetes and
Digestive and Kidney Diseases. U.S.
Department of Health and Human
Services, National Institutes of Health.
http://diabetes.niddk.nih.gov/dm/pubs/sta-
tistics/ Accessed: 22 September 2009.

24. Seaton D. Diabetes: The burden of dia-
betes in the Natural State. 2007.
http://www.healthy.arkansas.gov/programs
Services/epidemiology/ChronicDisease/Do
cuments/publications/diabetes_report200
7.pdf Accessed: 22 September 2009.

25. Chu W. The past twenty-five years in eye
banking. Cornea 2000;19:754-65.

26. Mathew PT, David S, Thomas N.
Endothelial cell loss and central corneal
thickness in patients with and without
diabetes after manual small incision
cataract surgery. Cornea 2011;30:424-8.

27. Modis L Jr, Szalai E, Kertesz K, et al.
Evaluation of the corneal endothelium in
patients with diabetes mellitus type I and
II. Histol Histopathol 2010;25:1531-7.

Article

Non
-co

mmerc
ial

 us
e o

nly




