Comparing ocular biometry and autorefraction measurements from the Myopia Master with the IOLMaster 700 and the Huvitz HRK‐8000A autorefractor

To compare axial length (AL) and corneal radius (CR) measured with the Oculus Myopia Master and the Zeiss IOLMaster 700, and cycloplegic refractive error measured with the Myopia Master and the Huvitz Auto Ref/Keratometer (HRK‐8000A).


INTRODUC TION
Current clinical advice advocates full correction of myopia, not un-, under-or over-corrected. 1,2 Accurate and repeatable cycloplegic refraction and ocular biometry measurements are therefore key parameters for research and for improving clinical decision-making for patients with myopia. Two of the most important biometry variables that are key for clinical decision-making are axial length (AL) and corneal radius (CR), 3 with AL combined with objective cycloplegic refractive error being the primary outcome variables when assessing myopia risk of onset, progression and treatment efficacy. 4 CR is essential for assessing whether the refractive error is axial, refractive or mixed axial/refractive in nature. [5][6][7] Commercially available ocular biometers use different optical principles for measuring ocular parameters. When comparing the risk of onset, progression and treatment efficacy between clinical practices, published epidemiological, experimental and treatment studies, it is relevant to consider the limits of agreement (LoA) between different instruments for the same parameter. The utility of the LoA and what might be deemed acceptable and interchangeable may differ in clinical versus research settings.
The Oculus Myopia Master (Oculus Optikgeräte GmbH, myopi a-master.com) combines measurements of CR, AL and autorefraction in one acquisition, and has been reported to be an efficient tool for screening children with ametropia. 8 The instrument software allows for comparison of individual measurements with ethnicity-, sex-and age-matched normative growth curves [9][10][11] for easier oninstrument interpretation of the data and to estimate the risk of myopia onset and progression. It is not known to what degree the Myopia Master is interchangeable with other conventional devices in research and clinical practice. The Zeiss IOLMaster 700 (Carl Zeiss Meditec AG, zeiss.com) is an ocular biometer that is widely used in research and clinical practice, with reportedly high repeatability and reliability in children 12 and adults, 13 and good inter-instrument agreement with ocular biometers that measure AL and CR based on the same 14 and different optical principles. [13][14][15][16] The Huvitz Auto Ref/ Keratometer (HRK-8000A, Huvitz Co. Ltd., huvitz.com) is widely used for autorefraction [17][18][19] with reportedly high repeatability. 20 The aim of this study was to assess the LoA between the Myopia Master and the IOLMaster 700 for measurements of CR and AL, and between the Myopia Master and the HRK-8000A for measurements of cycloplegic refractive errors. The 95% LoA between the Myopia Master and the IOLMaster 700 for AL and CR, and between the Myopia Master and the HRK-8000A for cycloplegic refractive error, were larger than deemed acceptable, implying that measurements obtained by the instruments are not directly interchangeable for these three parameters.

Participants and measurements
Seventy-four healthy adults (16 male) aged 19-41 years were recruited from the National Centre for Optics, Vision and Eye Care at the University of South-Eastern Norway. Data were collected on the same day in September 2021. The following ocular biometry parameters were measured with both the Myopia Master and the IOLMaster 700: AL (in mm), the CR in the flatter meridian (CR flat ) in mm, the CR in the steeper meridian (CR steep ) in mm and the mean CR (CR mean ) in mm. Objective cycloplegic refractive error and CR were each measured with the Myopia Master and the HRK-8000A autorefractor.
Ocular biometry and autorefraction were performed at least 30 min after the instillation of one (for blue or green irides) or two (for brown irides) drops of cyclopentolate hydrochloride 1% (Minims single dose; Bausch.co.uk) in each eye. All eyes were assessed for residual accommodation and pupil dilation prior to the measurements. An eye was deemed to have sufficient cycloplegia when the pupil no longer reacted to light and the monocular accommodation amplitude was <2 D (measured with an RAF rule). In one participant, a third drop was instilled with an additional waiting time of 20 min before measurements were taken.
All participants were examined with all three instruments within a maximum time of 24 min. The measurements were obtained in automatic capture mode in accordance with manufacturers' recommendations, and were deemed acceptable when they satisfied the recommended quality criteria for each individual device. The participants positioned their head against the forehead and chin rests and were instructed to fixate at the built-in fixation target. Trained optometrists operated the instruments after explaining the measurement procedure to each participant. The instrument calibrations were checked prior to data collection using the manufacturers' calibration protocols. None of the participants reported any previous refractive or corneal surgery. In nine participants, measurements from only one eye satisfied the quality criteria either due to a poor tear film during measurement or insufficient corneal reflections causing unreliable measurements in the other eye.
The study followed the tenets of the Declaration of Helsinki and was approved by the Regional Committee for Medical and Health Research Ethics (Southern Norway Regional Health Authority). Each participant signed a written informed consent form before data collection after an explanation of purpose, procedures and possible consequences of the study.

Instruments
The Myopia Master uses non-contact partial coherence interferometry (PCI) with an 880 nm light source to measure AL. AL is automatically measured six times, and the reported AL is the mean of all measurements with feasible signal-to-noise ratio (SNR) peaks. The repeatability of the AL measurements is 0.03 mm according to the manufacturer. 21 CR is derived from the non-linearity of a reflected test image (equally spaced points and a ring) projected onto the cornea and captured by a digital camera sensor. For autorefraction, a digital camera measure the deviation of an infrared light source (880 nm) as it was reflected from the retina, from which the degree of ametropia (in dioptres, D) can be derived. The autorefraction values reported by the instrument are the mean of three measurements obtained with a pupil diameter of 2.0 mm.
The principles of operation of the IOLMaster 700 have been described previously. 13 In short, the IOLMaster 700 uses swept-source optical coherence tomography (SS-OCT) scans with a 1055 nm light source to measure central corneal thickness, anterior chamber depth, lens thickness and AL. CR is measured using reflected light spots on the corneal surface, and the instrument calculates the final mean CR from three average measurements, each of which consists of five single measurements. 15 The instrument reports the mean AL of three scans in each of the six meridians. 15 When the standard deviations between the measurements were too high -indicative of a poor-quality measurement -the instrument displayed a warning message and the measurements were repeated. The instrument has been reported to have high measurement repeatability and reproducibility. 12,13 The Huvitz HRK-8000A uses a Hartmann-Shack wavefront sensor and an associated wavefront analysis algorithm to measure and calculate the refractive error. There is an incorporated automatic three-axis eye tracker which optimally repositions the sensor to compensate for the participant's eye movements between measurements. The autorefraction values reported by the instrument are the mean of five measurements obtained with a minimum pupil diameter of 2.0 mm. The instrument has been reported to have high measurement repeatability. 20

Data analysis
In the analysis, data are presented with the following precision: 0.001 mm for AL, 0.01 mm for CR and 0.25 D for refractive error (sphere and cylinder power). Spherical equivalent refractive error (SER) was calculated as sphere + ½ cylinder specified for a 12 mm vertex distance. The cylinder power and axis measurements were converted into Jackson crosscylinder (J 0 and J 45 ) vector components. 22 The CR mean was defined as the arithmetic mean of CR steep and CR flat . All statistical analyses were performed using R statistical software (v. 4.2.1., r-proje ct.org). 23,24 The distributions of the data were checked using histograms and Q-Q plots. Bland-Altman plots were used to assess the agreement of ocular biometry measurements between the Myopia Master and the IOLMaster 700, and in SER and CR between the Myopia Master and the Huvitz HRK-8000A. To estimate the mean difference (bias) and the 95% limits of agreement (LoA), the paired differences of the measurements from each instrument were analysed with a mixed-effects regression model using the packages lme4 25 and SimplyAgree 26 in R. The participant ID was included as a random effect to account for the inclusion of both eyes from each participant. The 95% confidence intervals (95% CI) of the LoA  were estimated using a bootstrap procedure as described by Parker et al. 27 The paired differences were considered statistically significant when the 95% CI of the mean differences excluded zero, and the p-value was <0.05. Acceptable LoA were defined as ±0.05 mm (≈±0.12 D) for AL, ±0.06 mm (≈±0.25 D) for CR and ±0.40 D for SER, based on reported repeatability and LoA between different instruments (Appendix S1) and as reviewed by Brennan   The required sample size was estimated using the method proposed by Lu et al. 29 For a mean AL difference of 0.001 mm, a standard deviation of AL differences of 0.4 mm and a delta of 0.1 mm, the required sample size was 109 at a 5% level with a power of 80%. Delta was defined as the acceptable agreement limit, see details given in the methods section.
A comparison of the measurements from all three instruments is presented in Table 1. Figure 1 shows Bland-Altman plots of the paired differences against the average for AL and CR measured with the Myopia Master and the IOLMaster 700 and HRK-8000A (CR only). CR mean , measured with the Myopia Master, was significantly flatter than that obtained with the IOLMaster 700 and the HRK-8000A (both p < 0.001). The 95% LoA for CR mean was −0.020 to 0.090 mm and −0.074 to 0.121 mm, respectively. The difference in CR mean between the Myopia Master and the IOLMaster 700 was within ±0.06 mm in 82.6% (CI 75.4, 88.1) of the eyes and within ±0.12 mm in 99.3% (CI 95.0, 99.9). There was no difference in mean AL measurements between the Myopia Master and the IOLMaster 700. The 95% LoA for AL was −0.097 to 0.089 mm. The difference was within ±0.05 mm in 84.9% (CI 77.9, 89.9) and within ±0.10 mm in 92.8% (CI 87.1, 96.1). The agreement between the instruments is summarised in Table 2.
When comparing measurements of SER obtained with the Myopia Master and the HRK-8000A, the SER measured with the Myopia Master was on average 0.19 D (CI −0.25, −0.13) more negative. In 66.9% (CI 58.7, 74.2) of the eyes, the difference between the instruments was within ±0.25 D, while the difference was within ±0.40 D in 82.0% (CI 74.7, 87.5). For J 0 and J 45 , there were no mean differences between the Myopia Master and the HRK-8000A; however, the 95% LoA were wider than for SER (Table 2). A difference less than or equal to ±0.40 D between the measurements was observed in 64.5% (CI 56.2, 72.0) of the eyes for J 0 and in 72.5% (CI 64.4, 79.3) for J 45 . Figure 2 shows Bland-Altman plots of the paired differences against the average refractive errors for the two instruments.

DISCUSSION
In this study, the LoA and the corresponding CIs indicated that the Myopia Master and the IOLMaster 700 should not be used interchangeably for measurements of AL and CR. Likewise, the Myopia Master and the HRK-8000A should not be used interchangeably for cycloplegic autorefraction and CR measurements.
The observed mean (SD) AL difference of −0.004 (0.047) mm is comparable with reported inter-instrument agreement studies comparing other PCI-based instruments with the IOLMaster 700, which is based on SS-OCT. 12,13,15,30 The 95% LoA were, however, wider in the present study. The LoA between the instruments exceeded that deemed to be acceptable for CR, AL and refractive error (see Data analysis section). For AL, 15% of the eyes examined had a larger difference in measurements than the acceptable LoA criterion (±0.05 mm). Even with a more liberal criterion (±0.1 mm), 7% of the eyes still had measurements that were not comparable. Measurements of AL were more sensitive to change than those of refractive error, with ±0.05 mm change in AL corresponding to ±0.12 D change in SER, 28 and measurements of AL are the preferred metric for tracking myopia progression. 13 The width of the 95% LoA for AL may reflect the differences in optical principles that the measurements were based on (here PCI vs. SS-OCT) in the two instruments. SS-OCT has higher scan resolution and higher T A B L E 2 Agreement between ocular biometry measurements from the Myopia Master and the IOLMaster 700 (first four rows) and in refractive error measurements and mean corneal radius from the Myopia Master and the HRK-8000A (last four rows).

Parameters
Mean difference (SD) 95% CI of the difference accuracy than PCI and facilitates, for example, the ability to measure the AL along six different axes and ensures that the AL is measured along the optical path through the fovea. 15 Both instruments claim to measure the AL from the anterior cornea to the retinal pigment epithelium, but the IOLMaster 700 uses light of a longer wavelength (1055 nm) than the Myopia Master (880 nm). The longer wavelength reportedly reduces light scattering of ocular structures 31 and increases tissue penetration depth. 31,32 The different wavelengths did result in a slightly longer median AL for the IOLMaster 700, but no difference in mean AL between the instruments. Previous investigations that compared the IOLMaster 700 and IOLMaster 500 reported similar 95% LoA to that in this study, but with no difference in CR mean nor any difference in the flat or steep corneal meridians. 12,13 Here, the CR measured with the Myopia Master was significantly flatter than with the IOLMaster 700 (Tables 1 and 2), and 17% and 0.7% of eyes had a difference >±0.06 mm and >±0.12 mm, which corresponds to 0.25 and 0.50 D, respectively. The agreement between instruments was poorer for the steep meridian than for the flat meridian (Table 2, Figure 1).
In a review of the repeatability of autorefraction, the mean estimated 95% LoA for repeatability of cycloplegic SER was ±0.41 D. 28 Although SER is an important clinical measurement, most autorefractors do not have comparable accuracy to that of ocular biometers, except perhaps the HRK-8000A with a reported repeatability of ±0.056 D (95% CI 0.052, 0.061). 20 Compared with the HRK-8000A, the Myopia Master measured on average 0.19 D more minus SER, but no differences on average for J 0 and J 45 (Table 2, Figure 2). This was within an acceptable LoA for SER (±0.40 D). However, the proportion of differences within ±0.40 D in SER, J 0 and J 45 between the autorefractor measurements were 82%, 65% and 72%, respectively. This, together with the 95% LoA (−0.86 to 0.47 D, −1.13 to 1.01 D and −1.47 to 1.36 D, Table 2) and their outer confidence limits ( Table 2), underline that the differences in autorefractor measurements with the two instruments cannot be ignored. The finding of a more myopic SER with the Myopia Master compared with the HRK-8000A (both instruments are monocular-closed-view devices) is in line with that reported for the Shin-Nippon SRW-5000 binocular open-view autorefractor in comparison with a monocular-closed-view aberrometer. 33 In contrast, in a comparison of the SRW-5000, which is based on the image size principle such as the Myopia Master, with a binocular open-view aberrometer (COAS-HD VR), a more myopic SER was reported for the wavefront sensor-based device. 34 It was argued that this was because the measurements by the wavefront-sensor-based device cover a larger area of the pupil. 34 This issue is circumvented in the HRK-8000A by automatic repositioning of its sensor within a central 2-mm zone to ensure that only the central area of the pupil is measured.
A key strength of the study was that cycloplegic measurements were carried out in the same session for both autorefraction and ocular biometry, limiting any influence of accommodation. A limitation is that the repeatability data of the Myopia Master were not collected, and are currently lacking beyond those reported for AL by the manufacturer (±0.03 mm). In terms of the SER, a limitation is that the measurements were carried out with a step size of 0.25 D F I G U R E 2 Bland-Altman plots for comparisons between the Myopia Master and the Huvitz HRK-8000A measurements of cycloplegic spherical equivalent refraction (SER), J 0 and J 45 vector components in adults. The mean difference is shown as a solid line, while 95% limits of agreement and the corresponding confidence intervals are shown by the dashed and dotted lines, respectively. Myopic eyes = orange squares, emmetropic eyes = grey circles, hyperopic eyes = blue triangles. instead of 0.01 D. Another possible limitation is that there were different operators on each of the three instruments, although the instruments' automatic capture modes were enabled and should have limited any between-operator variation. The IOLMaster 700 has been reported to be operator independent, 13,30 but we did not assess to what degree this may have affected the results obtained from the Myopia Master. Finally, our findings are based on measurements of adults aged 19-41 years and may not be applicable to other age groups, notably children.