Keywords
OCT Eyestar - Anterion - swept source (SS) OCT - ocular biometry
Schlüsselwörter
Eyestar - Anterion - Augenbiometrie - Swept Source (SS)
Background
Although the prediction of refractive outcomes after cataract surgery has consistently
improved over time, the most accurate formula for these predictions remains a subject
of ongoing debate among researchers and practitioners in the field [1]. The foundation for calculations made using these formulas is the precise measurement
of ocular biometric data, the quality of which directly impacts postoperative refractive
outcomes.
Eyestar 900 (ES) and Anterion (AN) are two advanced swept-source optical coherence
tomography (SS-OCT) devices designed to provide high-resolution imaging and accurate
biometric data.
This retrospective analysis of study data compares their agreement in measuring key
ocular parameters, including anterior chamber depth (ACD), axial length (AL), central
corneal thickness (CCT), corneal curvature along the flat (K1) and steep (K2) meridians,
steep meridian axis (AX), and white-to-white corneal diameter (WTW). This evaluation
will help to determine their reliability and potential clinical implications.
Materials and Methods
This study aimed to evaluate the agreement between two SS-OCT devices, the Eyestar
900 (Haag-Streit, Köniz, Switzerland) and the Anterion (Heidelberg Engineering, Heidelberg,
Germany), in measuring key ocular parameters of the anterior and posterior segments.
The study involved human participants who provided written informed consent prior
to inclusion. The study was reviewed by the appropriate ethics committee and conformed
to the tenets of the Declaration of Helsinki, Good Clinical Practice, ISO 14 155,
as well as all applicable Swiss legal regulations.
Population
Participants were recruited from a secondary referral centre, with inclusion criteria
encompassing individuals of legal age, irrespective of sex, ethnicity, or corneal
status. Individuals from vulnerable populations, those unable to provide consent,
those unable to maintain fixation during the examination, individuals with active
ocular inflammation or infection, and those with a tear film break-up time of less
than 5 seconds were excluded.
The study included participants with diverse corneal and endothelial irregularities,
such as keratoconus, or with a history of corneal allogenic intrastromal ring segments
(CAIRS), Descemetʼs membrane endothelial keratoplasty (DMEK), laser-assisted in situ
keratomileusis (LASIK), or radial keratotomy (RK), as well as those with lenticular
irregularities, including patients with cataracts or intraocular lenses (IOLs).
Study procedure
To ensure that the study equipment was at operating temperature, it was powered on
at least 30 minutes ahead of the scheduled data collection. All measurements were
conducted in a single session under identical conditions after informed consent was
obtained. Participants were first measured on the Anterion and then alternated between
devices until three measurements per eye were taken on each device, the order of which
was not randomised. The data from the three measurements were averaged for each parameter
in each eye of each participant.
All measurements were performed by trained operators. In cases where the initial measurement
failed to meet the deviceʼs quality control algorithm, the procedure was repeated
once. If the device provided a numerical value, it was included in the analysis. No
follow-up visits were conducted.
Statistical analysis
To evaluate the agreement between the Eyestar 900 and Anterion, we used Bland-Altman
analysis with 95% limits of agreement (LoA) and calculated the intraclass correlation
coefficient (ICC) [2]. The statistical analysis focused on identifying systematic differences and the
degree of correlation between the measurements from the two devices.
To address potential intra-subject biases, it is standard practice to include only
one eye per participant in statistical analyses involving ocular data due to the natural
correlation between eyes. In our study, we have included both eyes (OU) as they differed
significantly in status–for example, one eye being preoperative and the other postoperative–as
this introduces substantial differences between them. Recognising that some degree
of intra-subject correlation still exists, we compared the right (OD) and left (OS)
eyes for each parameter on the Eyestar data to gain an understanding of how the eye
status affects correlation.
Calculations were performed in RStudio 2023.12.1 Build 402 (Posit Software, PBC, Boston,
MA, USA) using the packages BlandAltmanLeh, ggplot2, and irr.
Results
A total of 86 eyes from 43 patients were included in the analysis. The participants
can be classified into two groups based on their history of cataract surgery: those
who had not undergone cataract surgery (phakic) and those who had received surgical
treatment for cataracts (pseudophakic). [Table 1] summarises the basic characteristics of these two groups of interest.
Table 1 Basic characteristics of study population.
|
Characteristic
|
Phakic
|
Pseudophakic
|
Total
|
|
N eyes (participants)
|
64 (26)
|
22 (17)
|
86 (43)
|
|
Mean age ± SD [years]
|
61.6 ± 14.8
|
62.6 ± 11.7
|
61.0 ± 15.3
|
|
Age range [years]
|
23 – 85
|
36 – 78
|
23 – 85
|
Across the sample, 20 participants presented with corneal irregularities, amongst
which there were 3 CAIRS, 5 keratoplasties (DMEK/DSAEK/PKP), 4 LASIK, 1 RE-PRK LASEK,
1 Fuchsʼ dystrophy, and 1 RK. Additionally, 20 participants presented with preoperative
cataracts of varying degrees. Finally, 3 participants had no cornea or lens status
that was notable in the context of this study. Due to the limited number of observations
for each corneal condition, it was not feasible to conduct subgroup analyses.
[Table 2] presents the mean for each parameter, accompanied by the standard deviation and
number of observations (n). In conducting these calculations, all values exported
by the devices were included, irrespective of the outcome of the deviceʼs quality
control algorithm. For the listed parameters, where the number of observations (n)
was less than 86, the respective device did not return a value for a number of participants
(86 – n).
Table 2 Means and mean differences of biometric parameters.
|
Parameter
[unit]
|
Eyestar 900 OU
Mean ± SD (n)
|
Anterion OU
Mean ± SD (n)
|
Mean Difference
(ICC)
|
|
ACD [mm]
|
3.4 ± 0.6 (71)
|
3.6 ± 0.8 (85)
|
0.080 (> 0.926)
|
|
AL [mm]
|
23.8 ± 1.3 (79)
|
23.8 ± 1.3 (83)
|
0.015 (> 0.990)
|
|
CCT [µm]
|
546.9 ± 42.6 (82)
|
543.8 ± 46.7 (86)
|
0.914 (> 0.921)
|
|
Flat K(1) [D]
|
42.9 ± 1.7 (83)
|
42.8 ± 3.1 (86)
|
− 0.024 (> 0.904)
|
|
Steep K(2) [D]
|
43.9 ± 1.9 (83)
|
44.2 ± 2.6 (86)
|
− 0.210 (> 0.902)
|
|
WTW [mm]
|
12.1 ± 0.4 (85)
|
11.8 ± 0.4 (85)
|
0.215 (> 0.810)
|
|
Steep axis [°]
|
102.5 ± 54.5 (83)
|
95.7 ± 45.1 (86)
|
5.968 (< 0.732)
|
The mean differences between the ES and the AN (ES-AN) are presented in [Table 2] along with the calculated ICC. Excellent agreement was observed for ACD, AL, CCT,
K1, K2, and WTW ([Fig. 1]
a–f).
Fig. 1 Anterior chamber depth (ACD), axial length (AL), central corneal thickness (CCT),
corneal curvature along the flat (K1) and steep (K2) meridians, and white-to-white
corneal diameter (WTW).
However, the corneal axis only exhibited moderate agreement (ICC < 0.732), with a
mean difference of 5.97° and a notable standard deviation of 58.6°. [Fig. 2] displays the Bland-Altman plot for the AX measurement.
Fig. 2 Axis of the steep meridian (AX).
To estimate the level of correlation between both eyes of a participant, the ICC was
calculated for all parameters between OD and OS of the ES as shown in [Table 3].
Table 3 Mean differences and ICC of biometric parameters between OD and OS measured on the
Eyestar 900.
|
Parameter [unit]
|
Mean Difference
|OD-OS|
|
ICC
|
|
ACD [mm]
|
0.31 ± 0.56
|
0.068 < ICC < 0.672
|
|
AL [mm]
|
0.24 ± 0.34
|
0.928 < ICC < 0.979
|
|
CCT [µm]
|
19.95 ± 31.63
|
0.365 < ICC < 0.771
|
|
Flat K(1) [D]
|
0.55 ± 0.60
|
0.793 < ICC < 0.938
|
|
Steep K(2) [D]
|
0.59 ± 0.52
|
0.814 < ICC < 0.944
|
|
WTW [mm]
|
0.10 ± 0.10
|
0.874 < ICC < 0.967
|
|
Steep axis [°]
|
38.32 ± 41.48
|
0.212 < ICC < 0.573
|
There is excellent agreement between OD and OS for AL, good agreement for WTW and
K2, moderate agreement for K1, and poor agreement for CCT, AX, and ACD.
Discussion
A notable limitation of this study is the heterogeneity of the study population. A
diverse range of patients was included, encompassing those with various corneal irregularities
and both phakic and pseudophakic eyes. This approach was taken to reflect the wide
spectrum of patients who may present for IOL implantation and to understand the performance
of the devices across a broad clinical scenario. While this heterogeneity may have
introduced additional variability, it reflects the diverse patient population encountered
in clinical settings.
Despite this limitation, our findings are consistent with those reported by Lender
et al. [3] and Sorkin et al. [4], both of whom found good agreement between the ES and the AN. Lender et al.ʼs study
included cataract surgery candidates over the age of 40 undergoing standard biometric
measurements prior to surgery, excluding participants whose ocular measurements had
a standard deviation exceeding 0.05 mm or D, and stratified participants based on
astigmatism levels using a cutoff of 1.00 D. In contrast, Sorkin et al. collected
data from the medical records of patients older than 18 years who underwent evaluation
for cataract surgery and excluded patients with prior cataract surgery or significant
corneal irregularities, whereas our study intentionally included such cases.
In our study, the ES and AN demonstrated substantial agreement for all parameters
except for the axis of the corneal meridian (AX). A mean difference of 5.97° suggests
that the devices provide similar results on average; however, the large standard deviation
indicates significant variability. The ICC reflects the consistency or reliability
of the measurements; here, a moderate ICC value indicates a fair degree of agreement
between the two methods, though not perfect. The combination of a moderate ICC and
high variability raises questions about the interchangeability of these devices, at
least for this parameter. A potential explanation for this discrepancy could be the
relatively small sample size combined with the inclusion of patients from different
subgroups, which may have increased measurement variability. The lack of subgroup
analyses likely prevented us from identifying specific factors contributing to this
variability and the inclusion of eyes with corneal
irregularities could have contributed to the discrepancies observed between the devices,
limiting the generalisability of our findings. Though generally, the axis is poorly
defined in eyes with low astigmatism (|K2-K1|), and for an astigmatism of 0 D, the
axis is arbitrary, making differences in corneal axis significant only for higher
levels of astigmatism.
Interestingly, despite their exclusion criteria and stratification based on astigmatism,
Lender et al. [3] also identified weak correlations in AX measurements between devices, similar to
our findings. Future studies with larger, more homogeneous subgroups are needed to
elucidate the performance of these devices in specific patient populations. Enhancing
the accuracy of AX measurements is crucial for the successful implantation of toric
lenses, and ongoing development in this area is needed.
As shown in [Table 2], the AN exhibited a higher success rate for measurements, as evidenced by the number
of observations for each parameter, whereas the ES data exhibited less deviation when
analysed in terms of standard deviation. This finding contrasts with that reported
by Sorkin et al. [4]. This discrepancy might be attributed to the different confidence thresholds of
the devices for the data obtained. Similar to the study by Sorkin et al., our analysis
of unsuccessful measurements revealed no distinctive characteristics or comorbidities.
Overall, we observed a high level of concordance between the two devices, consistent
with results reported in the literature. Moreover, other studies have found positive
correlations between the ES and various ocular biometry devices [5], [6], [7], [8], [9], further supporting its reliability across different clinical settings.
Supplements
Tables for averaged AN and ES data:
