cornea - confocal microscopy - Langerhans cells - fibers - age
córnea - microscopia confocal - células de Langerhans - fibras - idade
Corneal innervation, especially in the sub-basal epithelial plexus, is an important
factor in the reflex arc and is involved in corneal protection and corneal epithelium
maintenance[1 ],[2 ]. While the plexus sub-basal epithelial has been studied using electron and light
microscopy, these studies are hampered by the fact that corneal nerve fibers degenerate
within 14 hours after death[1 ]. The emergence of confocal microscopy adapted for in vivo use has allowed the study
of all layers of the cornea, including its innervation and its relationship with Langerhans
cells (LCs).
LCs are part of a main histocompatibility complex class II (MHCII) and work to transport
tissue to capture, process, and present antigens[3 ]. As cells mature, they begin to have less capacity to capture antigens and then
turn into a strong stimulator of T lymphocytes. In general, the density changes and
number of LCs can help us understand the immune mechanisms of the cornea[4 ] as they are closely related to the sub-basal epithelial plexus. They increase in
number and density in inverse relation to the number and density of the fiber plexus
in the presence of corneal infection (keratitis)[5 ].
More recently, the assessment of LCs and unmyelinated fibers in the cornea has become
the subject of study for other medical specialties, particularly neurology. Corneal
confocal microscopy (CCM) is a noninvasive method for diagnosing and monitoring systemic
diseases such as diabetic neuropathy[6 ]; fewer nerve bundles correlate with a loss of corneal sensitivity and neuropathy
severity in type 1 diabetic patients. An early diagnosis of peripheral neuropathy
is made by observing a reduction of sub-basal epithelial plexus fibers, which occurs
even before a reduction in corneal sensitivity[7 ]. Improved blood glucose levels in type 1 diabetic patients undergoing pancreas transplantation
was shown to recover a number of nerve fibers.
Fabry disease also evolves with peripheral neuropathy and has been studied with CCM.
It is characterized by a reduction in the number and the density of nerve fibers in
the plexus sub-epithelial basement[8 ]. CCCM has also been employed in diagnosing peripheral neuropathy in patients undergoing
chemotherapy as some of these drugs can dose-dependently induce peripheral neuropathy[9 ].
Reference levels are largely based on North American samples; it is unclear what Brazilian
population values are. This study aims to evaluate the morphological characteristics
of the sub-basal plexus with in vivo cornealCCM in healthy Brazilian individuals to
provide reference values that may be used for comparison and future research.
METHOD
Design and study population
This was an observational, cross-sectional descriptive study conducted with a convenience
sample of healthy individuals referred from the Hospital Universitário Antonio Pedro
(HUAP), Universidade Federal Fluminense (UFF). We included all individuals who signed
the informed consent form, between 21 and 70 years of age, and of both sexes. We excluded
subjects who had a history of trauma or corneal diseases and/or peripheral nerve disorders,
had previous eye surgery, had a history of contact lens use, had more than 6 months
of exposure to a potentially toxic drug to nerves and/or the cornea, or had cognitive
and/or sensory deficits that would prevent them from performing the exam.
Data collection by in vivo CCM
The confocal microscope used in the study was an HRT II coupled to the cornea module
from Heidelberg Engineering, Germany. The objective lenses were prepared with one
drop of Vidisic® gel (Bausch Lomb), and another drop was instilled into the lower
fornix of the patient’s conjunctiva. Anestalcon® (proparacaine HCl 0.5%, Alcon) anesthetic
drops were instilled into the patient’s eye prior to initiation of the study.
The patient was positioned correctly and comfortably on the device and was oriented
to fix their look for the examination. The control of the central perpendicular alignment
of the cornea was obtained by observing the red reflex laser corneal shown on the
handset screen. By identifying the well-focused epithelium in the corneal apex, images
were captured at 0 microns depth for the first picture. The focus deepened, and sequential
images were recorded to the level of the posterior stroma. Images were acquired in
sequence and separated at a distance of approximately 1 micron and a fixed field of
0.16 mm2 . The same procedure was performed on both eyes.
The examination in both eyes lasted about 10 minutes. After capturing the images in
both eyes, the Tomocap was discarded, and another drop of gel was instilled in the
eye. The individual was then released and received guidance to instill lubricating
eye drops during the day of the exam (offered to the individual).
The images obtained during the examination were 384 × 384 pixels over an area of 400
μm2 and were stored in the computer coupled to the confocal system. The 5 best pictures
of the sub-basal plexus epithelial of each eye were selected for a total of 10 images
per individual. The blinded images were separately evaluated by three observers: an
ophthalmologist (G.D.) and two neurologists (C.P. and O.J.M.N.). The following parameters
previously established were evaluated: number and fiber density (defined as # of fibers
× 0.16), number and LC density, degree of tortuosity (no fibers, grade 1 for parallel
and slightly crooked fibers; grade 2 for twisted, non-parallel fibers; and grade 3
for tortuous fibers without direction), and thickness (categorized as thick, medium,
or thin fibers) ([Figure 1 ]).
Figure 1 Confocal microscope images of corneas showing (A) thick, parallel fibers; (B) parallel
fibers of medium thickness; and (C) thin non-parallel, unorganized fibers.
Statistical analysis
All statistical procedures were performed using the computer programs Microsoft Excel© and SPSS (Statistical Package for Social Science) version 16.0 and statistical programming
language R. Analyses were performed according to the following steps:
Description of variables and study population: a descriptive analysis of the population
was performed by calculating the percentage distribution of the categories when the
variables were categorical and the mean and standard deviation (SD) when variables
were continuous or discrete. The following categorical variables were included: sex,
fiber tortuosity (parallel, not parallel, not parallel - same direction, not parallel
- without direction), and thickness (none, fine, medium, large). All analyses were
performed for the entire population and stratified by sex.
Intra and inter-rater reliability: Fifty-five healthy subjects were included. For
each individual, 10 pictures were taken (5 of the each eye), totaling 550 individual
photos, which were evaluated by 3 observers. For the degree of reliability of categorical
variables, simple agreement was calculated (percentage of concordant numbers), as
were Cohen’s kappa coefficients and their respective 95% confidence intervals (CIs)
and weighted kappas. The criterion of Landis and Koch[20 ] was used to interpret agreement. Inter-observer agreement was verified in accordance
for the right eye, left eye, and all photos. For the degree of reliability of continuous
or discrete variables, an evaluation of intra-observer variability through the intra-class
correlation coefficient (ICC) with respective 95%CIs, and inter-observer analysis
was performed by comparison analysis means (t test for normally distributed variables
and Mann-Whitney U or Kruskal-Wallis H tests for those variables that did not follow
normal assumptions), analysis of variance (ANOVA), and correlation of variables between
observers (Pearson correlation for variables with normal distributions and Spearman
for those that were non-parametric).
Descriptive analysis of the variables: In this stage, due to the variables identified
in CCM to be evaluated by the three observers, the measurements were aggregated with
the purpose of obtaining a diagnosis for each individual (consensus among observers).
So for the number of fibers and LCs, the measures were grouped using the “average”
of observers 1 and 3, with observer 2 excluded due to low correlation with the others
(see reliability analysis results). Categorical variables (tortuosity and thickness)
were grouped using the measure of the “mode.” As the concordance rate was not 100%
between observers, all were included with the third observer’s evaluation used in
non-concordant cases. The variables of the study were explored and evaluated by age
and sex. For continuous variables, mean comparisons were performed with the normality
of those tested with the Kolmogorov-Smirnov and Shapiro tests. The difference of means
was evaluated using the Mann-Whitney U or Kruskal-Wallis H tests for variables with
normal and non-parametric distributions, respectively. The relationships between age
variables and CCM parameters were calculated by Pearson correlation (normal distributions)
and Spearman (nonparametric variables). In relation to qualitative variables (categorical
variables), the percentage differences were evaluated using chi-square statistics.
Fisher’s exact test was considered when the expected frequencies were below 5%. The
influences of age and sex on CCM parameters were also investigated using simple linear
regression and multiple (continuous and normally distributed variables) or generalized
linear models (discrete or binary). Bivariate and multivariate analyzes were performed.
The model fit was performed using ANOVA (nested models) and the Akaike information
criterion (AIC, not nested models). In all analyses, p < 0.05 was considered significant.
RESULTS
Of the 55 healthy individuals evaluated, there were between 5-8 individuals per sex
per age category. Within the age categories, there was no difference between the average
age of the males or females ([Figure 2 ]). In general, the average age was 44.9 ± 13.2, number of fibers 5.35 ± 1.36, fiber
density 33.4 ± 8.5 fibers per field, and LCs 5.13 ± 8.10 ([Table 1 ]). Considering the morphological fiber characteristics, 45% had parallel fibers (tortuosity),
and 63.6% had average thickness. No statistically significant difference was observed
between sexes for the assessed variables including age, age category, and CCM parameters.
Figure 2 Average age within each age category by sex.
Table 1
Population characteristics - Corneal confocal microscopy.
Age
Male (n = 25)
Female (n = 30)
Total (n = 55)
p-value
46.2 ± 13.5
43.9 ± 13.1
44.9 ± 13.2
0.54
Age Category
21-30 years
5 (20%)
6 (20%)
11 (20%)
31-40 years
5 (20%)
6 (20%)
11 (20%)
41-50 years
5 (20%)
8 (27%)
13 (24%)
51-60 years
5 (20%)
5 (16.5%)
10 (18%)
61-70 years
5 (20%)
5 (16.5%)
10 (18%)
Total
25 (100%)
30 (100%)
55 (100%)
CCM1
# of fibers
5.52 ± 1.2
4.20 ± 1.5
5.35 ± 1.4
0.37
Density of fibers
34.5 ± 7.6
7.68 ± 8.6
33.4 ± 8.5
0.37
Langerhans cells
7.00 ± 10.9
3.57 ± 4.2
5.13 ± 8.1
0.49
Width
Absence
0 (0%)
2 (6.7%)
2 (3.6%)
Thin
6 (24%)
10 (33.3%)
16 (29.1%)
Medium
18 (72%)
17 (56.7%)
35 (63.6%)
Large
1 (4%)
1 (3.3%)
2 (3.6%)
Total
25 (100%)
30 (100%)
55 (100%)
Tortuosity
Absence
0 (0%)
2 (6.7%)
2 (3.6%)
Parallel
11 (44%)
14 (46.7%)
25 (45.5%)
Non-parallel – same direction
13 (52%)
12 (40%)
25 (45.5%)
Non-parallel – no direction
1 (4%)
2.6 (7%)
3 (5.5%)
Total
25 (100%)
30 (100%)
55 (100%)
Inter- and intra-rater reliability
Reliability between and within observers was calculated using Cohen’s kappa, concordance
correlation coefficient, and ICC. The results from the analysis of reliability revealed
that observer 2 had low correlation with observers 1 and 3 and needed to be retrained
on the scoring process. Data from all three observers is provided in the supplementary
materials; however, due to the questionable validity of observer 2, we have excluded
them from the results presented below (see [Table 2 ] and [Figure 3 ]).
Table 2
Data from all three observers.
Observer 1
Observer 2
Observer 3
Right eye
Left eye
Total
Right eye
Left eye
Total
Right eye
Left eye
Total
N
%
N
%
N
%
N
%
N
%
N
%
N
%
N
%
N
%
Tortuosity*
Absence
4
2
8
3
12
2
5
2
29
10
34
6
0
0
6
2
6
1
Parallel
140
54
128
44
268
49
73
28
101
35
174
32
72
28
107
37
179
32
Not parallel, but in the same direction
104
40
123
42
227
41
132
51
108
37
240
44
157
60
155
53
312
57
Same direction
12
45
31
11
43
8
50
19
52
18
102
18
31
12
22
8
53
10
Total
260
100
290
100
550
100
260
100
290
100
550
100
260
100
290
100
550
100
Thickness*
Absence
4
2
8
3
12
2
5
2
29
10
34
6
1
0
6
2
7
1
Thin
69
26
106
37
175
32
61
24
48
17
109
20
128
49
115
40
243
44
Medium
162
62
165
57
327
60
138
53
129
44
267
48
118
45
130
45
248
45
Large
25
10
11
4
36
6
56
22
84
29
140
26
13
5
39
13
52
10
Total
260
100
290
100
550
100
260
100
290
100
550
100
260
100
290
100
550
100
Right eye
Left eye
Total
Conc. (%)
Kappa (95%CI)
Kappa pond.
Conc. (%)
Kappa (95%CI)
Kappa pond.
Conc. (%)
Kappa (95%CI)
Kappa pond.
Tortuosity
Obs 1 & Obs 2
51%
0.250 0.16-0.33
0.314
46%
0.227 0.15-0.30
0.279
48%
0.240 0.18-0.30
0.296
Obs 1 & Obs 3
86%
0.749 0.67-0.82
0.767
84%
0.750 0.68-0.82
0.776
85%
0.750 0.70-0.80
0.774
Obs 2 & Obs 3
56%
0.307 0.22-0.39
0.346
46%
0.225 0.15-0.30
0.276
51%
0.266 0.20-0.32
0.309
Width
Obs 1 & Obs 2
66%
0.433 0.33-0.53
0.509
61%
0.402 0.32-0.48
0.500
64%
0.419 0.36-0.48
0.510
Obs 1 & Obs 3
90%
0.810 0.75-0.88
0.802
88%
0.790 0.72-0.85
0.814
89%
0.803 0.76-0.85
0.812
Obs 2 & Obs 3
66%
0.429 0.33-0.52
0.480
58%
0.359 0.27-0.44
0.470
62%
0.395 0.33-0.45
0.481
Figure 3 Data from all three observers.
Inter- and intra-rater reliability showed “substantial” agreement between observers
1 and 3 for tortuosity (kappa = 0.75, 95%CI 0.70-0.80) and thickness (kappa = 0.80,
95%CI 0.76-0.85). Observations for the right and left eyes were similar to the total
measurements.
Regarding the continuous variables, the number and density of fibers and LCs were
evaluated with Spearman’s correlation. Once again, the number of fibers was more highly
correlated between observers 1 and 3 with values 0.912, 0.904, and 0.913 for total,
right, and left eye measures, respectively. For LCs, the correlation between viewers
1 and 3 reached 0.950 for the right and left eyes.
Considering intra-observer reliability, the 10 pictures for each patient were evaluated
using the ICC correction. The results show low variability within each observer and
for the variables studied. As expected, for the 10 evaluated photos (5 right and 5
left eyes) the pictures of the same subject and object were highly consistent. LCs
had the highest intra-observer reliability with a total ICC = 0.96 and 95%CI 0.93-0.97.
For the number of fibers, the results were similar when separately measuring the right
(ICC = 0.82, 95%CI 0.53-0.91) or left eye (ICC = 0.87, 95%CI 0.71-0.93), with greater
variability for the right eye. For all observers, there was an ICC of 0.85 and, 95%CI
0.65-0.92.
Evaluation of CCM variables
The correlation between the average number of fibers and subject age showed an inverse
relationship between fiber number and density and age in women; that is, there was
a reduction in the number and density of fibers with increasing age for women (p <
0.05, [Figure 4 ]).
Figure 4 Average numbers of fibers by age group and sex.
The relationships between demographic variables, thickness, and tortuosity with the
number of LCs were verified by means of generalized linear models with Poisson model
(used for quantifiable variables). In univariate analysis, there was a statistically
significant relationship between age, sex, and thickness (p < 0.001); therefore, all
these variables were included in the multivariate model. These variables remained
significant in the model (p < 0.001). The results indicate that a 1-year increase
in age corresponded with an average increase of 1.017 (95%CI: 1.008-1.026) in the
number of LCs, adjusting for sex and thickness. Similarly, the average number of LCs
among women was 35% lower than for men, adjusted for the effect of age and thickness
(odds ratio [OR] = 0.645, 95%CI: 0.500-0.831). Moreover, the average number of LCs
among those with fibers with medium or large thickness was 1.13 times higher compared
with those of thin fibers by controlling the effects of age and sex (OR = 2.13, 95%CI:
1.531-2.978). However, there were two outliers that should be considered in these
results ([Figure 5 ]).
Figure 5 (A) Average numbers of Langerhans cells by age group and sex. Age categories 31-40
and 61-70 are represented twice to demonstrate the results excluding the outliers.
Black and gray represent males and females, respectively. (B) Box plots showing numbers
of Langerhans cells by age group and sex without the outliers.
No statistically significant percentage difference was noted between demographic variables
and fiber characteristics (tortuosity and thickness) ([Figures 6 ] and [7 ]).
Figure 6 Tortuosity by age group, sex, and type. The numbers within the bars represent the
number of patients with the specified shaded type.
Figure 7 Fiber width by age group, sex, and type. The numbers within the bars represent the
numbers of patients with the specified shaded type.
Discussion
We analyzed 550 images from 55 healthy volunteers and found an inverse relationship
between fiber number and age, with a decrease of 0.148 fibers for each additional
year of age. This reduction was higher in females. We also observed that older individuals
had more tortuous and thinner nerve fibers. This study is the first of its kind in
a Brazilian population.
Other studies evaluating the density of fibers per field are shown in [Table 3 ]
[4 ],[10 ],[11 ],[12 ]. Only studies using the same methodology employed here were included. Our population
had a greater mean number of fibers per field than the other studies.
Table 3
Comparable studies.
Authors
Country
Age range
Sample size
Density
Standard deviation
Patel et al., 200910
New Zealand
37 ± 10
31 eyes
25.9
7
Niederer et al., 200711
New Zealand
45 ± 17
30 eyes
21.6
5.91
Parissi et al., 201312
Norway
88 ± 15
207 eyes
19
5.1
Niederer et al., 20074
New Zealand
38 ± 16
170 eyes
20.3
6.5
Current Study
Brazil
70 ± 21
110 eyes
33.4
8.5
Regarding the density of nerve fibers of the sub-basal epithelial plexus, there is
variation depending on the type of confocal microscope used. Confocal microscopes
can be laser scanning, scanning slit lamp, or tandem type. The differences between
these microscopes are lighting intensity and type, resolution, and image contrast.
Another issue exists regarding the definition of the method for measuring fiber density
of the sub-basal epithelial plexus. Some authors consider the total length of nerve
fibers visible in a defined area in mm2
[13 ], while other authors[14 ] consider the sum of the number of nerve fibers by field. In our study, we used a
laser scanning microscope type and measured the density of nerve fibers as visible
fibers per field microns per mm2 . Analyses were made only through observation; no automated method was used. A clinical
evaluation performed by a trained specialist is equivalent to results obtained with
automated methods[13 ],[14 ],[15 ]. We cannot definitively say that there were no differences between these methodologies;
however, there was no perceived compromise of the technique regarding our results.
Parissi’s group[12 ] evaluated 207 healthy eyes in subjects aged 88 ± 15 years and also found a negative
correlation between nerve fiber density and age, reporting a decrease of 0.25% to
30% per year, regardless of sex, eye studied, or methodology used to delineate the
nerve fibers. In accordance with earlier findings[16 ],[17 ], an inverse relationship was found between nerve fiber density and age. In our work,
similar to the findings of Niederer[4 ], this inverse correlation was more significant in women. For each 1-year increase
in age, there is a reduction in the number of fibers (-0.148). Dehaghani’s group[18 ] carried out a 3-year longitudinal study and reported that a reduction of 0.05 mm/mm2 in the total length of the nerve fiber per field per year increased with age, but
they did not observe differences between males and females.
As seen in [Figure 2 ], the average number of fibers by age group and sex may be misleading. We observed
a greater average number of fibers in the 20-30-year-old female group and then from
the age of 30 onwards the average number reduces to 5 and remains consistent across
until age 70. This initial elevation gives rise to an annual decrease of 0.148; however,
this may actually represent a shorter time-span, possibly decreasing from adolescence
and plateauing after 30. This difference may disappear with a greater number of participants;
therefore, the study should be repeated with a larger sample. It should be noted though
that this trend of decreasing fibers in women over time has been reported by several
other groups[4 ],[18 ]; thus, the current literature lends itself to validating our findings rather than
representing an artifact from a limited sample size.
Regarding nerve fiber tortuosity, our results as well as the others[14 ] showed that most of the fibers were oriented in a direction perpendicular to the
axis and slightly crooked. Individuals over the age of 30 years were more likely to
have no parallel fibers, fewer perpendicular fibers, and more itytortuosity.
In our study of LCs, we found an age-dependent increase in the number. In individuals
younger than 30 years, the prevalence of the cells was lower for females. The average
number of cells for women was 35% lower than for men adjusted for the effect of age
and sex. Our findings are consistent with other works evaluating LCs[19 ].
Due to the two patients in our study who had an extremely high number of LCs for unknown
reasons, the variation we saw was far greater than in other studies[5 ], 3.0 ± 0.4 per field compared to our 5.1 ± 8.10 cells/mm2 . This variation between the two studies is as expected for the technique. However,
the reason for increased LCs in these two patients, who were reportedly healthy, is
unclear. A study by Cruzat and colleagues[5 ] demonstrated a correlation between plexus epithelial sub-basal zone, or the peripheral
nervous system, and the cornea immune response. Assessing patients with infectious
corneal keratitis, the authors reported a decrease in the number of nerve fibers and
hence an increase in LC density in the acute phase of the disease.
The limitations of this study are clearly the small sample size, which is a restricted
representation of the general Brazilian population. Nevertheless, we feel that offering
a sample of each age group adds to the current literature and provides the first work
in this area in Brazil.
