ORIGINAL ARTICLE

Corneal Asymmetry Analysis by Pentacam

Scheimpflug Tomography for Keratoconus
Diagnosis
Jonatán D. Galletti, MD; Pablo R. Ruiseñor Vázquez, MD; Natalia Minguez, MD;
Marianella Delrivo, MD; Fernando Fuentes Bonthoux, MD; Tomás Pförtner, PhD;
Jeremías G. Galletti, MD, PhD

ABSTRACT

PURPOSE: To evaluate intereye corneal asymmetry

in Pentacam (Oculus Optikgeräte GmbH, Wetzlar, Ger-
D etection of early keratoconus signs constitutes a key
aspect of preoperative screening for corneal refrac-
tive surgery and, in addition to rigorous biomicros-
copy examination, there are numerous technologies to assist
the clinician in this task. Most diagnostic options focus on the
many) indices as a diagnostic method between normal
patients and patients with keratoconus. typical ectatic corneal changes, such as abnormal anterior and
posterior curvature,1 localized thinning,2 focally decreased
METHODS: A retrospective, observational case series of epithelial thickness,3 and biomechanical instability.4,5 But
177 healthy, 44 indeterminate, and 121 patients with because none of these methods provide absolute sensitivity
keratoconus classified by Pentacam ectasia detection and specificity, especially when dealing with truly subclini-
indices, randomized to analysis and validation datasets. cal cases (eyes with unremarkable slit-lamp and topography
Intereye asymmetry in 20 Scheimpflug tomography
corneal descriptors was calculated and compared to findings that develop ectatic changes in time), there is still a
develop diagnostic models. need for improvement of current diagnostic models.
Asymmetrical progression is a well-characterized feature
RESULTS: Intereye asymmetry was not correlated with of keratoconus.6,7 Previous studies have explored in depth
anisometropia in healthy controls but was correlated the intereye difference in corneal shape in patients with kera-
with the ectasia grade of the worse eye in patients with toconus,7-10 but to the best of our knowledge, there are no
keratoconus. Patients with keratoconus had significantly
greater intereye asymmetry in all descriptors except for established diagnostic models for clinical use based on this
relational thickness indices. Intereye asymmetry in front specific feature of the disease. Because normal corneas are
elevation at the thinnest corneal location afforded the markedly symmetric,11-14 the possibility of detecting early
single highest diagnostic performance (71% sensitiv- ectatic changes by evaluating intereye differences is attrac-
ity and 85% specificity), whereas the best multivariate tive. In addition, most corneal descriptors show significant
model combining intereye asymmetry in anterior and
posterior keratometry, corneal thickness, and front and overlap between healthy and subclinical keratoconic eyes,1,15
back elevation at the thinnest point provided 65% sen- hence the prospect of increasing sensitivity by also evaluating
sitivity and 97% specificity. Multivariate models upheld their asymmetry.
their performance in the validation dataset. Most (more The Pentacam HR Scheimpflug tomography system (Ocu-
than 90%) indeterminate patients, according to conven- lus Optikgeräte GmbH, Wetzlar, Germany) has gained wide
tional Pentacam analysis, showed within-normal-range
corneal asymmetry. acceptance in clinical practice for keratoconus screening be-
cause it provides several corneal descriptors and diagnostic
CONCLUSIONS: Healthy corneas are markedly sym-
metric irrespective of anisometropia, but corneal asym- From ECOS (Clinical Ocular Studies) Laboratory, Buenos Aires, Argentina
metry analysis does not provide sufficient sensitivity to (JDG, PRRV, NM, MD, FFB, TP, JGG); the Department of Ophthalmology,
be used alone for detecting keratoconus. However, its Hospital de Clínicas José de San Martín, University of Buenos Aires, Argentina
remarkable specificity suggests that it could be used (JDG, PRRV, NM); and the Institute of Experimental Medicine, National
combined with conventional single cornea Pentacam Academy of Medicine/CONICET, Buenos Aires, Argentina (JGG).
analysis to reduce the false-positive rate or in dubious Submitted: July 27, 2014; Accepted: November 19, 2014
cases.
The authors have no financial or proprietary interest in the materials pre-
sented herein.
[J Refract Surg. 2015;31(2):116-123.]
Correspondence: Jeremías G. Galletti, MD, PhD, ECOS (Clinical Ocular
Studies) Laboratory, (1119) Pueyrredón 1716 7 B, Buenos Aires, Argentina.
E-mail: jeremiasg@gmx.net
doi:10.3928/1081597X-20150122-07

116 Copyright © SLACK Incorporated

 Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

models that have been previously described.2,15-21 We thinnest corneal thickness, corneal thickness progres-
have recently shown that although testing with this de- sion indices (minimum, average, and maximum), an-
vice can diagnose many eyes with subclinical kerato- terior and posterior elevation at thinnest corneal loca-
conus, a significant fraction of cases could still go un- tion, Ambrosio’s relational thickness indices (average
detected and the false-positive rate could be an issue.15 and maximum), and the normalized indices deviation
Remarkably, corneal asymmetry is not considered by of normality of the front elevation, deviation of nor-
the Pentacam software for keratoconus detection or, to mality of the back elevation, deviation of normality of
the best of our knowledge, by any other Scheimpflug corneal thickness progression, deviation of normality
device approved for clinical use. of corneal thinnest point, deviation of normality of re-
In this study, we evaluated corneal asymmetry in lational thickness, and overall deviation of normality.
healthy patients and patients with keratoconus to ex- The methodology has been disclosed by the manufac-
plore its applicability to disease detection in the clini- turer for some but not all of these indices.2,19,22,23
cal setting. We separately evaluated the sensitivity and Patients were classified solely by their Pentacam
specificity of single parameter cutoff values for corneal findings into one of three categories. The most sensi-
asymmetry, and looked at how combinations of these tive ectasia detection indices according to previous re-
new corneal descriptors compared to the Pentacam’s ports were considered.2,15 Healthy controls consisted
already established corneal indices. of patients with unremarkable Scheimpflug tomogra-
phy in both eyes, defined as showing normal values
PATIENTS AND METHODS (less than 1.6) in the standardized indices (back eleva-
This study was a retrospective, observational case tion, corneal thickness progression, relational thick-
series. The research protocol followed the tenets of ness, and overall indices) and for the Ambrosio’s maxi-
the Declaration of Helsinki and was approved by the mum relational thickness index (339 or greater).
Hospital de Clínicas ethics committee. The records of Patients with at least one abnormal value in any of
patients who had been referred for spectacle or con- the aforementioned indices (2.6 or greater for the stan-
tact lens prescription or keratoconus diagnosis and dardized index or maximum relational thickness less
examined at ECOS Laboratory between March and No- than 339) were diagnosed as having keratoconus.
vember 2013 were reviewed. Exclusion criteria were Patients with suspicious values in one or both eyes
the following: previous eye surgery or trauma, corneal for any of the standardized indices (1.6 or greater and
scarring, any eye disease other than keratoconus, and less than 2.6) were included in a separate, indetermi-
chronic use of topical medication. Patients were asked nate group. It should be noted that the latter group
to cease contact lens wear at least 3 weeks before ex- does not correspond to the keratoconus suspect cat-
amination. egory usually found in the literature, but to a relatively
All patients were examined at ECOS Laboratory by common situation when screening patients with Pen-
three trained ophthalmologists (JDG, PPRV, and MD). tacam HR Scheimpflug tomography.15 In a previous
Each patient underwent slit-lamp examination, Pen- study, we showed that approximately 20% of normal
tacam HR Scheimpflug tomography, and Placido disk eyes and 20% of subclinical keratoconic corneas have
topography and aberrometry (iTrace, software version suspicious values (1.6 or greater and less than 2.6) for
4.2.1; Tracey Technologies, Houston, TX). Patients the Pentacam’s standardized indices.15 These eyes can-
were told to blink before the examinations and only not be considered keratoconic due to insufficient find-
reliable studies were included. Topographic examina- ings, but the possibility of subclinical ectasia cannot be
tions with artifacts or irregularities were discarded, ruled out. We did not include the indeterminate group
whereas for the Pentacam, only acquisitions with ac- in the analysis and only examined the performance of
ceptable quality (as defined by the manufacturer) were the diagnostic models in this sample for exploratory
included (25 images in 2 seconds). One examination purposes.
per eye was analyzed and data of both eyes were re- Two datasets (analysis and validation) with two
corded. Average corneal power and higher-order aber- groups each were compiled by random 2:1 allocation
rations of the corneal 5-mm central surface were pro- of healthy patients and patients with keratoconus. The
vided by the iTrace software, and refractive spherical Keratoconus Severity Score was used for keratoconus
equivalent and cylinder. Anisometropia was defined grading,24 which is based on average corneal power and
as an intereye difference 1 diopter (D) or greater in corneal higher-order aberrations obtained from Placi-
either variable. The following Pentacam descriptors do topography. The Keratoconus Severity Score scale
were analyzed: flat, steep, and mean anterior keratom- includes scores of 0 (unaffected, normal topography), 1
etry, mean posterior keratometry, central, apex, and (unaffected, atypical topography), 2 (suspect), 3 (mild

Journal of Refractive Surgery • Vol. 31, No. 2, 2015 117

Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

keratoconus), 4 (moderate keratoconus), and 5 (severe this article). Intereye asymmetry for each of the ana-
keratoconus).24 lyzed variables is summarized in Table B (available
Intereye asymmetry of each parameter was calcu- in the online version of this article) and depicted for
lated as the absolute value of the difference between some descriptors in Figure 1. Compared with control
fellow eyes to ease interpretation. The normality of the patients, median values for intereye asymmetry were
data was assessed by the Kolmogorov–Smirnov test significantly greater (P < .001) in patients with kera-
and then either parametric or non-parametric com- toconus for every variable except for standardized re-
parison tests were performed accordingly. Receiver lational thickness and maximum relational thickness.
operating characteristic curves were used to calculate
sensitivity, specificity, and area under the curve of Anisometropia and Corneal Intereye Asymmetry
each parameter. Optimal cutoff points were derived In the control group, 42 (37%) patients (17 males,
from the receiver-operating characteristic curves as 41%) had spherical equivalent anisometropia of 1
the value closest to the perfect classification point.4 D or greater and 32 (28%) patients (16 [50%] males)
A composite asymmetry score was created by exam- had cylindrical anisometropia of 1 D or greater. When
ining five corneal descriptors (anterior and posterior comparing the spherical equivalent of patients with
keratometry, thinnest corneal thickness, and front and anisometropia and those without, median intereye
back elevation at thinnest corneal location) according difference was significantly greater only in the overall
to their optimal cutoff points, as previously described. standardized deviation index (0.15 vs 0.29, P = .002).
To calculate the composite score, which ranged from When comparing the cylinder of patients with aniso-
0 to 5, the number of positive asymmetry descriptors metropia and those without, median intereye differ-
was counted for each patient. In addition, logistic re- ence was significantly greater only in steepest anterior
gression with all variables entered at once was used to keratometry (0.3 vs 0.5, P = .04). In this group, spheri-
obtain the best combination of these five descriptors. cal equivalent anisometropia was only significantly
The resulting logistic function was applied with a 0.5 correlated with intereye asymmetry in average corne-
cutoff: logit = 0.210 3 mean anterior keratometry 1 al thickness progression (r = 0.205, P = .028) and the
5.546 3 mean posterior keratometry 1 0.015 3 thin- overall standardized deviation index (r = 0.218, P =
nest corneal thickness 1 0.539 3 front elevation at .02), whereas cylindrical anisometropia was only sig-
thinnest location 1 0.116 3 back elevation at thinnest nificantly correlated with intereye asymmetry in steep
location – 2.541. anterior keratometry (r = 0.290, P = .002).
All data were recorded and sorted using Micro-
soft Excel 2010 (Microsoft Corporation, Redmond, Keratoconus Grade and Corneal Intereye
WA). Statistical analysis was performed with Prism 5 Asymmetry
(GraphPad Software, La Jolla, CA) and SPSS version In the keratoconus group, the steepest (of both eyes)
17 software (SPSS, Inc., Chicago, IL). A P value of less mean anterior keratometry was positively correlated
than .05 was considered statistically significant. Data with the intereye asymmetry in all variables but refrac-
are shown as mean ± standard deviation unless other- tive cylinder, back elevation at the thinnest corneal
wise stated. point, and relational thickness indices. The highest
Keratoconus Severity Score of both eyes was also posi-
RESULTS tively correlated with the intereye asymmetry in all
Demographics and Corneal Intereye Asymmetry variables but relational thickness indices.
The analysis dataset consisted of a control group of
115 patients (mean age: 32 ± 8 years, 59 [51%] female) Diagnostic Performance of Intereye Asymmetry
and a keratoconus group of 75 patients (mean age: 32 The diagnostic performance of intereye asymme-
± 10 years, 37 [49%] female), whereas the validation try in each variable for differentiating between con-
dataset included a control group of 62 patients (mean trol patients and patients with keratoconus is shown
age: 33 ± 10 years, [56%] female) and a keratoconus in Table 2. The best variables describing anterior and
group of 46 patients (mean age: 34 ± 11 years, [41%] fe- posterior curvature and the potentially ectatic corneal
male). The indeterminate group comprised 44 patients region (thickness and anterior and posterior elevation
(mean age: 33 ± 8 years, 28 [64%] female). For each at the thinnest corneal location) were evaluated togeth-
group, corneal descriptors per eye are summarized in er as multivariate intereye asymmetry. The combined
Table 1 and the proportion of patients with abnormal score of intereye asymmetry in mean anterior (positive
ectasia detection indices in neither, one, or both eyes if 0.3 D or greater) and posterior (positive if 0.1 D or
is shown in Table A (available in the online version of greater) keratometry, and corneal thickness (positive if

118 Copyright © SLACK Incorporated

 Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

12 µm or greater) and front (positive if 2 µm or

45.0 ± 1.7
-6.5 ± 0.3

44 (100%)
Indeterminate (n = 44)

528 ± 30
525 ± 29
5±2

8±4
greater) and back (positive if 5 µm or greater)

OS

–
–
–
–
–
elevation at the thinnest corneal point had an
area under the receiver-operating character-
istic curve of 0.85 (95% confidence interval:

45.0 ± 1.6
-6.6 ± 0.3
532 ± 30
529 ± 29

43 (98%)
1 (2%)
0.80 to 0.90): a value of 2 or greater positive

5±2

8±4
OD

–
–
–
–
criteria cutoff provided 0.82 sensitivity and
0.70 specificity, whereas a value of 3 or greater
positive criteria cutoff afforded 0.67 sensitivity

Corneal Descriptors in Control, Keratoconus, and Indeterminate Groupsa

46.5 ± 4.6
and 0.92 specificity. A logistic regression ap-

-6.7 ± 0.8
483 ± 45
464 ± 53

19 (41%)
Keratoconus (n = 46)

17 ± 16

37 ± 34

7 (15%)

8 (17%)

6 (13%)
4 (9%)

2 (4%)
proach combining the intereye asymmetry in

OS
the aforementioned five variables had 0.65 sen-
sitivity and 0.97 specificity, with an area un-

Validation Dataset (n = 108)

der the receiver-operating characteristic curve

46.1 ± 2.8
-6.6 ± 0.6
489 ± 38
468 ± 43

19 (41%)
17 ± 18

38 ± 36

5 (11%)
9 (20%)
8 (17%)
3 (7%)
2 (4%)
of 0.86 (95% confidence interval: 0.80 to 0.92).

OD
Intereye Asymmetry in the Validation
Dataset and the Indeterminate Group

43.8 ± 1.6
-6.3 ± 0.3
In the validation dataset, the combined

545 ± 33
542 ± 33

61 (98%)

1 (2%)
2±2

4±3
score of intereye asymmetry with a value of 3
OS
Control (n = 62)

–
–
–
or greater positive criteria cutoff provided 0.67
sensitivity and 0.89 specificity, whereas the
logistic regression function yielded 0.72 sensi- -6.2 ± 0.25
43.7 ± 1.6

544 ± 32
541 ± 32

61 (98%)

1 (2%)
2±1

4±3
tivity and 0.94 specificity. Five patients (8%)
OD

–
–
–
in the control group had an intereye asymme-
TABLE 1

try score of 3 and 2 patients (3%) had a score

of 5. The combined intereye asymmetry score
46.5 ± 3.7
-6.8 ± 0.8
493 ± 35
479 ± 40

41 (55%)

13 (17%)
distribution in the indeterminate group was as
12 ± 12

27 ± 26
Keratoconus (n = 75)

4 (5%)
6 (8%)

7 (9%)
4 (5%)
OS

follows: 14 (32%) patients with a score of 0,

15 (34%) with a score of 1, 10 (22%) with a
score of 2, 3 (7%) with a score of 3, and 2 (5%)
46.3 ± 2.43
Analysis Dataset (n = 190)

with a score of 4. The logistic regression func-

OD = right eye; OS = left eye; K = keratometry; KSS = Keratoconus Severity Score

-6.7 ± 0.5
495 ± 32
481 ± 39

41 (55%)

14 (19%)
13 ± 13

27 ± 26

5 (7%)
6 (8%)

5 (6%)
4 (5%)
tion classified 3 (7%) patients as abnormal.
OD

The distribution of intereye asymmetry scores

in both datasets combined is represented in
Figure 2 and summarized in Table 3.
44.1 ± 1.4
-6.3 ± 0.2

113 (98%)
535 ± 30
532 ± 29

1 (1%)
1 (1%)
2±2

3±3

DISCUSSION
Control (n = 115)
OS

–
–
–

In this study, we observed little corneal

asymmetry in normal patients, in agreement
with previous reports.9,12,13 Remarkably, mean
115 (100%)
44.1 ± 1.4
-6.3 ± 0.2

Data are given as mean ± standard deviation.

538 ± 30
535 ± 29
2±2

3±4

intereye differences for most corneal descrip-

OD

–
–
–
–
–

tors were not greater in anisometropic eyes in

the control group, suggesting that both corneas
in normal patients tend to develop symmetri-
Thinnest corneal thickness

cally despite differences in total refraction.

Central corneal thickness

Mean spherical anisometropia in the control

group (1.09 ± 1.34 D) was greater than in other
Front elevation at

Back elevation at
Mean posterior K
Mean anterior K

reports (0.63 ± 1.02 D),9,10 and yet we could

thinnest point

thinnest point
Parameter

only detect a correlation between this vari-

able and the intereye difference for average
KSS 0
KSS 1
KSS 2
KSS 3
KSS 4
KSS 5

corneal thickness progression and the overall

a

standardized deviation index in this group.

Journal of Refractive Surgery • Vol. 31, No. 2, 2015 119

Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

Figure 1. Intereye corneal asymmetry in

control patients and patients with kerato-
conus. Whisker-box plots of intereye asym-
metry of mean anterior keratometry (Mean
ant. K), mean posterior keratometry (Mean
post. K), thinnest pachymetry (Thinnest
pach.), and front and back elevation at
thinnest corneal location in control patients
(n =177, green boxes) and patients with
keratoconus (n = 121, red boxes) . *** =
statistically significant difference (P < .001)
in median values between groups according
to Mann–Whitney U test.

TABLE 2
Diagnostic Performance of Intereye Asymmetry in
Individual Corneal Descriptors for Keratoconus
Parameter AROC (95% CI) Cutoff Sensitivity (%) Specificity (%)
Flat anterior keratometry 0.73 (0.66 to 0.81) ≥ 0.4 66.7 66.1
Steep anterior keratometry 0.77 (0.69 to 0.85) ≥ 0.8 62.5 88.4
Mean anterior keratometry 0.74 (0.66 to 0.83) ≥ 0.3 66.7 78.6
Mean posterior keratometry 0.77 (0.70 to 0.85) ≥ 0.1 61.1 79.5
Central corneal thickness 0.75 (0.68 to 0.82) ≥ 10 68.1 76.8
Apex corneal thickness 0.76 (0.69 to 0.83) ≥9 73.6 70.5
Thinnest corneal thickness 0.78 (0.71 to 0.85) ≥ 12 65.3 79.5
Minimum pachymetric progression index 0.76 (0.68 to 0.84) ≥ 0.10 68.1 76.8
Average pachymetric progression index 0.78 (0.70 to 0.85) ≥ 0.08 66.7 77.7
Maximum pachymetric progression index 0.76 (0.68 to 0.84) ≥ 0.23 61.1 87.5
Front elevation at thinnest point 0.82 (0.75 to 0.89) ≥2 70.8 84.8
Back elevation at thinnest point 0.79 (0.72 to 0.86) ≥5 59.7 92.0
Deviation of front elevation 0.79 (0.72 to 0.87) ≥ 0.93 66.7 87.5
Deviation of back elevation 0.78 (0.70 to 0.86) ≥ 0.79 65.3 89.3
Deviation of pachymetric progression 0.79 (0.71 to 0.86) ≥ 0.49 70.8 75.9
Deviation of corneal thickness 0.82 (0.75 to 0.88) ≥ 0.36 69.4 83.0
Deviation of relational thickness 0.51 (0.42 to 0.60) ≥ 0.40 52.8 48.2
Overall deviation 0.82 (0.76 to 0.89) ≥ 0.47 69.4 87.5
Average ART 0.66 (0.58 to 0.75) ≥ 33 59.7 63.4
Maximum ART 0.53 (0.43 to 0.62) ≥ 66 41.7 69.6
AROC = area under receiver-operating characteristic curve; CI = confidence interval; ART = Ambrosio’s relational thickness

Cylindrical anisometropia in the control group was also relational thickness indices for early ectatic changes.2
greater than in previous studies (0.33 ± 0.44 D)9,10 and Although to a different grade, both eyes are thought to
nonetheless only correlated to intereye difference in be affected in patients with keratoconus.25 Therefore,
steepest anterior keratometry. These findings support an early marker of disease would be expected to be al-
the notion that, in healthy patients, both corneas are tered in both eyes and not show considerable differ-
mirrored in shape to a high extent and that this aspect ence between eyes, in contrast to a late marker, which
could be exploited for early keratoconus diagnosis. would more likely exhibit greater intereye difference
The intereye difference in standardized and maxi- from the change in the worse eye and lack thereof in
mum relational thickness indices was not greater in pa- the better eye. This hypothesis is supported by the
tients with keratoconus than in control patients, a find- proportion of patients with keratoconus with normal
ing that underscores the reported high sensitivity of the values in both eyes for Ambrosio’s relational thickness

120 Copyright © SLACK Incorporated

 Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

Figure 2. Intereye corneal asymmetry score

distribution in control patients, patients with
keratoconus, and indeterminate patients.
Histograms of combined intereye corneal
asymmetry score in control patients (n =
177), patients with keratoconus (n = 121),
and indeterminate patients (n = 44).

max index, which was the lowest among the surveyed TABLE 3
Pentacam metrics in this study (Table A). But despite Summary of Intereye Corneal
the observed corneal asymmetry for most descriptors Asymmetry Scorea
in patients with keratoconus, none of them provided
Positive (+1 point) if
adequate sensitivity and specificity when considered Scoring Criteria Intereye Difference
alone. Our findings are in agreement with Henríquez
Mean anterior keratometry ≥ 0.3 diopters
et al.,10 who reported high specificity but less sensitiv-
ity for these descriptors. Nevertheless, the high speci- Mean posterior keratometry ≥ 0.1 diopters
ficity of some variables could serve for readily detect- Thinnest pachymetry ≥ 12 µm
ing abnormal patients, which can be interpreted from Front elevation at thinnest location ≥ 2 µm
Table B; it is unlikely (less than 5%) that a healthy Front elevation at thinnest location ≥ 5 µm
patient would exhibit a difference greater than 0.6 D in a
Score of 3 is observed in up to 6% to 11% of healthy patients, whereas a
anterior mean keratometry, greater than 0.1 D in poste- score of 4 is found in less than 4% of patients without keratoconus. A score
of 5 should be considered highly abnormal (1% or less of non-keratoconic
rior mean keratometry, greater than 5 µm in posterior patients).
elevation at the thinnest corneal point, or greater than
20 µm in central apex of thinnest corneal thickness.
To improve overall performance, we analyzed com- cannot be generalized to other patient populations
binations of corneal descriptors as diagnostic models. without proper validation. Nevertheless, our findings
Both the combined score and the logistic function suggest that corneal asymmetry analysis should be ap-
described in the Results section showed remarkably plied to indeterminate cases in combination with cor-
high specificity but moderate sensitivity, and the per- neal tomography and/or topography because its low
formance of both models was almost equivalent. How- sensitivity is not sufficient to warrant its use as the sole
ever, the combined score is much simpler and could method for keratoconus detection. In other words, the
be readily assessed by the clinician without additional high specificity of corneal asymmetry implies that an
computation (Table 3). abnormal score should be considered a strong indica-
The false-positive rate of either model was consider- tor of ectatic disease even if other corneal indices are
ably better than what we previously observed for the within the normal range, whereas a normal score does
multivariate metric provided by the Pentacam software not imply the absence of disease.
(overall standardized deviation index), but the sensitiv- Only 2 (3%) of 62 patients who were selected in
ity of the latter was higher.15 Although an actual com- the control group by their unremarkable single cornea
parison in the same patient sample between Placido analysis with Pentacam Scheimpflug tomography in
topography and Scheimpflug tomography is lacking in both eyes had exceedingly high intereye corneal asym-
the literature, the reported sensitivity in different stud- metry, a proportion consistent with the keratoconus
ies for advanced topographic analysis is better than that prevalence reported elsewhere. Figure A (available
of our intereye corneal asymmetry models.26,27 Of note, in the online version of this article) shows two repre-
the diagnostic performance of these models was evalu- sentative cases from the control group that had three
ated on different samples, and thus these observations or more positive asymmetry criteria and remarkably

Journal of Refractive Surgery • Vol. 31, No. 2, 2015 121

Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

also exhibited abnormal biomechanical profiles when AUTHOR CONTRIBUTIONS

tested with the Ocular Response Analyzer (Reichert, Study concept and design (JDG, FFB, TP, JGG); data collection
Inc., Buffalo, NY), a finding that characterizes early ec- (JDG, PRRV, NM, MD); analysis and interpretation of data (JDG,
tatic changes.4,5 Within the indeterminate group, which PRRV, NM, JGG); drafting of the manuscript (JDG, JGG); critical revi-
consisted of 44 patients who had one or more keratoco- sion of the manuscript (PRRV, NM, MD, FFB, TP)
nus detection indices with a suspicious value but in-
sufficient to be classified as diseased, 3 (7%) to 5 (12%) REFERENCES
patients showed abnormal intereye corneal asymme- 1. Arbelaez MC, Versaci F, Vestri G, Barboni P, Savini G. Use of
a support vector machine for keratoconus and subclinical kera-
try, depending on the diagnostic model employed. toconus detection by topographic and tomographic data. Oph-
This finding exemplifies the conundrum faced by cli- thalmology. 2012;119:2231-2238.
nicians when dealing with real-world patients because 2. Ambrósio R, Caiado ALC, Guerra FP, et al. Novel pachymetric
there is a considerable fraction of patients who exhibit parameters based on corneal tomography for diagnosing kerato-
conus. J Refract Surg. 2011;27:753-758.
inconclusive findings even by Scheimpflug tomogra-
3. Rocha KM, Perez-Straziota CE, Perez-Straziota E, Stulting RD,
phy.15 In this way, corneal asymmetry could serve to Randleman JB. SD-OCT analysis of regional epithelial thick-
identify those patients with ectatic changes from those ness profiles in keratoconus, postoperative corneal ectasia, and
more likely to represent false-positives that arise from normal eyes. J Refract Surg. 2013;29:173-179.
any screening technique with high sensitivity. 4. Galletti JG, Pförtner T, Fuentes Bonthoux F. Improved keratoco-
Figures B-C (available in the online version of this nus detection by ocular response analyzer testing after consid-
eration of corneal thickness as a confounding factor. J Refract
article) show representative cases from the indetermi- Surg. 2012;28:202-208.
nate group that exhibited normal and abnormal cor- 5. Ruiseñor Vázquez PR, Delrivo M, Bonthoux FF, Pförtner T, Gal-
neal asymmetry, respectively, which could represent letti JG. Combining ocular response analyzer metrics for cor-
false-positives and true positives, respectively. Bio- neal biomechanical diagnosis. J Refract Surg. 2013;29:596-602.
mechanical data from these patients were also avail- 6. Li X, Rabinowitz YS, Rasheed K, Yang H. Longitudinal study
of the normal eyes in unilateral keratoconus patients. Ophthal-
able and were consistent with the asymmetry analy- mology. 2004;111:440-446.
sis. These examples lend credit to the potential role
7. Saad A, Guilbert E, Gatinel D. Corneal enantiomorphism in nor-
of corneal asymmetry assessment in early keratoconus mal and keratoconic eyes. J Refract Surg. 2014;30:542-547.
detection but are not in any way conclusive. Therefore, 8. Burns DM, Johnston FM, Frazer DG, Patterson C, Jackson AJ.
the predictive value of the proposed models must be Keratoconus: an analysis of corneal asymmetry. Br J Ophthal-
prospectively validated. mol. 2004;88:1252-1255.
There are limitations to this study. We relied on only 9. Zadnik K, Steger-May K, Fink BA, et al. Between-eye asymme-
try in keratoconus. Cornea. 2002;21:671-679.
one high-quality examination per eye to calculate cor-
neal asymmetry and some studies have suggested that 10. Henriquez MA, Izquierdo L, Mannis MJ. Intereye asymmetry
detected by Scheimpflug imaging in subjects with normal cor-
the Pentacam’s precision can be improved by averaging neas and keratoconus. Cornea. 2013;32:779-782.
repeated measurements.16,17 In addition, the elevation 11. Leung T-W, Lam AK-C, Kee C-S. Corneal shapes of Chinese em-
data obtained by this device do not necessarily match metropes and myopic astigmats aged 10 to 45 years. Optom Vis
that generated by other Scheimpflug devices,28 and Sci. 2013.
therefore the reported diagnostic models might only 12. Dingeldein SA, Klyce SD. The topography of normal corneas.
Arch Ophthalmol. 1989;107:512-518.
apply to the Pentacam. We hypothesize that calculat-
ing intereye differences in measurements might com- 13. Falavarjani KG, Modarres M, Joshaghani M, Azadi P, Afshar
AE, Hodjat P. Interocular differences of the Pentacam measure-
pensate for interdevice differences, but nonetheless ments in normal subjects. Clin Exp Optom. 2010;93:26-30.
the diagnostic value of corneal asymmetry should be 14. Khachikian SS, Belin MW, Ciolino JB. Intrasubject corneal
validated with other Scheimpflug tomography systems. thickness asymmetry. J Refract Surg. 2008;24:606-609.
Our results highlight the potential role of corneal 15. Ruiseñor Vázquez PR, Galletti JD, Mínguez N, et al. Pentacam
asymmetry analysis for keratoconus screening. On the Scheimpflug tomography findings in topographically-normal
patients and subclinical keratoconus cases. Am J Ophthalmol.
one hand, we derived single cutoff values for intereye 2014;158:32-40.
asymmetry in some commonly assessed corneal de-
16. Miranda MA, Radhakrishnan H, O’Donnell C. Repeatability
scriptors, which could be readily applied to clinical of oculus Pentacam metrics derived from corneal topography.
practice and perhaps extended to other Scheimpflug Cornea. 2009;28:657-666.
devices. On the other hand, we showed that either a 17. Miranda MA, Radhakrishnan H, O’Donnell C. Repeatability of
combined score or multivariate analysis of corneal corneal thickness measured using an Oculus Pentacam. Optom
Vis Sci. 2009;86:266-272.
asymmetry provides high specificity, which comple-
18. Savini G, Barboni P, Carbonelli M, Hoffer KJ. Agreement be-
ments the already established keratoconus detection tween Pentacam and videokeratography in corneal power
indices and thus could serve to analyze dubious cases.

122 Copyright © SLACK Incorporated

 Corneal Asymmetry Analysis for Keratoconus Diagnosis/Galletti et al

assessment. J Refract Surg. 2009;25:534-538. 24. McMahon TT, Szczotka-Flynn L, Barr JT, et al. A new meth-
od for grading the severity of keratoconus: the Keratoconus
19. Ambrósio R, Alonso RS, Luz A, Coca Velarde LG. Corneal-
Severity Score (KSS). Cornea. 2006;25:794-800.
thickness spatial profile and corneal-volume distribution: to-
mographic indices to detect keratoconus. J Cataract Refract 25. Wei RH, Zhao SZ, Lim L, Tan DTH. Incidence and characteris-
Surg. 2006;32:1851-1859. tics of unilateral keratoconus classified on corneal topography.
J Refract Surg. 2011;27:745-751.
20. Delrivo M, Ruiseñor Vázquez PR, Galletti JD, et al. Agreement
between Placido topography and Scheimpflug tomography for 26. Buhren J, Kuhne C, Kohnen T. Defining subclinical keratoconus
corneal astigmatism assessment. J Refract Surg. 2014;30:49- using corneal first-surface higher-order aberrations. Am J Oph-
53. thalmol. 2007;143:381-389.
21. Bae GH, Kim JR, Kim CH, Lim DH, Chung ES, Chung T-Y. Cor- 27. Bühren J, Kook D, Yoon G, Kohnen T. Detection of subclinical
neal topographic and tomographic analysis of fellow eyes in keratoconus by using corneal anterior and posterior surface ab-
unilateral keratoconus patients using Pentacam. Am J Ophthal- errations and thickness spatial profiles. Invest Ophthalmol Vis
mol. 2014;157:103-109. Sci. 2010;51:3424-3432.
22. Belin MW, Ambrósio R. Scheimpflug imaging for keratoconus 28. Salouti R, Nowroozzadeh MH, Zamani M, Fard AH, Niknam S.
and ectatic disease. Indian J Ophthalmol. 2013;61:401-406. Comparison of anterior and posterior elevation map measure-
ments between 2 Scheimpflug imaging systems. J Cataract Re-
23. Ambrósio R, Nogueira LP, Caldas DL, et al. Evaluation of corne-
fract Surg. 2009;35:856-862.
al shape and biomechanics before LASIK. Int Ophthalmol Clin.
2011;51:11-38.

Journal of Refractive Surgery • Vol. 31, No. 2, 2015 123

 TABLE A
Pentacam Ectasia Indices in Both Eyesa
Analysis Dataset (n = 190) Validation Dataset (n = 108)
Patients With Patients With
Control Patients Keratoconus Control Patients Keratoconus Indeterminate Group
(n = 115) (n = 75) (n = 62) (n = 46) (n = 44)
Which Eyes Which Eyes Which Eyes Which Eyes Which Eyes
Were Abnormal? Were Abnormal? Were Abnormal? Were Abnormal? Were Abnormal?
Parameter Neither One Both Neither One Both Neither One Both Neither One Both Neither One Both
PPIave 83.5 11.3 5.2 12.0 14.7 73.3 87.1 8.1 4.8 15.2 6.5 78.3 38.6 15.9 45.5
ARTmax 100 – – 5.3 38.7 56.0 100 – – 6.5 21.7 71.7 100 – –
Df 91.3 5.2 3.5 34.7 28.0 37.3 86.9 11.5 1.6 26.1 15.2 58.7 50.0 15.9 34.1
Db 100 – – 45.9 18.9 35.1 100 – – 32.6 8.7 58.7 81.8 13.6 4.5
Dp 100 – – 25.3 21.3 53.3 100 – – 15.2 19.6 65.2 65.9 29.5 4.5
Dt 94.8 1.7 3.5 32.0 22.7 45.3 91.9 3.2 4.8 21.7 23.9 54.3 90.9 4.5 4.5
Da 100 – – 32.9 23.3 43.8 100 – – 19.6 15.2 65.2 100 – –
D 100 – – 9.5 18.9 71.6 100 – – 8.7 13.0 78.3 15.9 38.6 45.5
PPIave = average pachymetric progression; ARTmax = Ambrosio’s maximum relational thickness; Df = standardized front deviation; Db = standardized back devia-
tion; Dp = standardized pachymetric progression deviation; Dt = standardized thickness deviation; Da = standardized relational thickness deviation; D = standard-
ized overall deviation
a
Results are shown in percentage form. Each index was analyzed with the cutoff values described in Methods: PPIave > 1.06, ARTmax < 339 and any of the stan-
dardized D indices > 1.6 were considered abnormal. It should be noted that ARTmax, Db, Dp, Da and D indices were used for classifying patients; therefore the
100% proportion of normal values in both eyes of control and indeterminate patients is meaningless, and it is shown only for consistency.
The Pentacam is manufactured by Oculus Optikgeräte GmbH, Wetzlar, Germany.
 AQ5TABLE B
Intereye Asymmetry in Control Patients and Patients With Keratoconus
Intereye Asymmetry
Control (n = 115) (Percentile) Keratoconus (n = 75) (Percentile)
Parameter 1 5 10 25 50 75 90 95 99 1 5 10 25 50 75 90 95 99
Sphere 0.00 0.09 0.13 0.25 0.62 1.50 3.32 4.06 7.91 0.00 0.13 0.25 0.62 1.25 3.37 5.50 7.83 17.00
Cylinder 0.00 0.00 0.12 0.13 0.50 1.00 1.75 3.66 11.27 0.00 0.12 0.13 0.50 1.63 2.88 5.22 6.80 24.25
Spherical equivalent 0.00 0.06 0.12 0.30 0.69 1.19 2.75 4.37 7.62 0.00 0.06 0.13 0.51 1.50 3.69 8.68 9.98 18.13
Flat anterior keratometry 0.0 0.0 0.1 0.1 0.2 0.4 0.6 0.9 2.5 0.0 0.0 0.1 0.3 0.6 2.3 5.0 7.3 13.2
Steep anterior keratometry 0.0 0.0 0.0 0.1 0.3 0.6 0.9 1.6 2.7 0.0 0.0 0.1 0.3 1.4 3.8 6.6 8.2 13.8
Mean anterior keratometry 0.0 0.0 0.0 0.1 0.2 0.3 0.5 0.6 1.5 0.0 0.0 0.0 0.2 0.6 2.9 5.3 8.1 13.5
Mean posterior keratometry 0.0 0.0 0.0 0.0 0.1 0.1 0.1 0.2 0.3 0.0 0.0 0.0 0.1 0.2 0.6 1.2 1.7 2.5
Central corneal thickness 0 0 1 2 6 9 15 20 49 0 2 3 7 11 20 35 50 96
Apex corneal thickness 0 0 1 3 6 10 15 21 49 1 1 2 7 13 27 46 68 104
Thinnest corneal thickness 0 1 1 2 6 10 17 22 43 0 2 3 8 16 28 49 59 117
Minimum PPI 0.00 0.00 0.01 0.03 0.06 0.10 0.14 0.20 0.43 0.00 0.00 0.02 0.07 0.16 0.51 1.06 1.40 3.02
Average PPI 0.00 0.00 0.01 0.01 0.04 0.07 0.11 0.15 0.37 0.00 0.01 0.02 0.05 0.16 0.66 1.17 1.88 5.19
Maximum PPI 0.00 0.01 0.03 0.06 0.11 0.17 0.26 0.33 0.57 0.00 0.03 0.05 0.12 0.41 0.99 2.15 2.59 8.97
Front elevation at thinnest point 0 0 0 0 1 1 2 2 4 0 0 0 1 5 9 17 25 31
Back elevation at thinnest point 0 0 0 1 1 3 4 5 11 0 0 1 2 9 22 36 55 60
Deviation of front elevation 0.00 0.03 0.05 0.17 0.47 0.74 1.02 1.31 2.70 0.02 0.09 0.10 0.65 1.78 5.74 10.23 11.38 20.36
Deviation of back elevation 0.01 0.02 0.05 0.13 0.34 0.60 0.85 1.36 1.58 0.02 0.06 0.09 0.42 1.36 5.07 8.48 12.33 19.30
Deviation of pachymetric progression 0.01 0.01 0.03 0.11 0.26 0.49 0.79 1.06 2.49 0.01 0.06 0.14 0.34 1.09 4.47 7.95 12.74 35.09
Deviation of corneal thickness 0.01 0.01 0.02 0.07 0.17 0.31 0.46 0.67 1.07 0.00 0.08 0.11 0.26 0.54 1.22 2.03 2.46 4.62
Deviation of relational thickness 0.01 0.05 0.10 0.25 0.43 0.66 0.94 1.28 1.85 0.01 0.04 0.05 0.14 0.42 1.08 1.77 2.56 3.05
Overall deviation 0.00 0.03 0.05 0.09 0.17 0.35 0.54 0.75 1.29 0.02 0.07 0.12 0.31 1.28 3.70 7.16 9.31 11.51
Average ART 0 2 4 10 24 44 77 98 329 1 3 10 20 40 125 220 308 393
Maximum ART 0 5 10 27 47 71 102 139 203 0 5 6 16 47 138 226 279 395
PPI = pachymetric progression index; ART = Ambrosio’s relational thickness
Figure A. Two representative cases from the control group that had three or more positive asymmetry criteria and remarkably also exhibited abnormal
biomechanical profiles when tested with the Ocular Response Analyzer (Reichert, Inc., Buffalo, NY).
Figure B. Two representative cases from the indeterminate group that exhibited normal corneal asymmetry and that probably represent false positives.
Biomechanical testing with the Ocular Response Analyzer (Reichert, Inc., Buffalo, NY) was unremarkable.
Figure C. Two representative cases from the indeterminate group that exhibited abnormal corneal asymmetry and that probably represent true positives.
Biomechanical testing with the Ocular Response Analyzer (Reichert, Inc., Buffalo, NY) produced abnormal profiles.