Impact of age on measurement variability for axial length in myopic children

ABSTRACT

Clinical relevance

Axial length is a primary outcome in the management of progressive myopia. However, young children may have difficulty fixating during these measurements compared to older children, which can result in higher measurement variability. This may affect perceived axial length progression, leading to inappropriate management.

Background

This study assessed the impact of patient age on measurement variability for axial length measurements taken with the IOLMaster 700 and IOLMaster 500 in myopic children.

Methods

A retrospective review of records was undertaken at a university optometry clinic. Five axial length measurements captured at the same visit were collected with the IOLMaster 700 and IOLMaster 500 for myopic patients ≤16 years. The within-subject standard deviation and R² were calculated for each instrument to examine the effects of age on instrument variability.

Results

Data was collected for 51 patients (30 female and 21 male), and the mean age was 10.98 ± 2.77 years. Mean axial length measured with the IOLMaster 700 was longer compared to the IOLMaster 500 (difference −0.02 ± 0.02 mm; p < 0.001). There was no effect of age on within-person variability for the measurement of axial lengths with either instrument, with R² values of 0.021 (p = 0.305) and 0.13 (p = 0.420) for the IOLMaster 700 and IOLMaster 500, respectively. The within-subject variability of axial measurements with the IOLMaster 700 was significantly lower than that with the IOLMaster 500 (p < 0.001).

Conclusion

Measurement variability of axial length measurements with the IOLMaster 700 and IOLMaster 500 was not dependent on age. However, axial length measurements captured with the IOLMaster 700 were significantly longer and less variable than those with the IOLMaster 500. Eye health care practitioners should be aware of the differences between the two instruments and refrain from using them interchangeably, especially for myopia control where small changes in axial length can affect patient management.

KEYWORDS:

• Age
• axial length
• IOLMaster
• myopia
• variability

Introduction

Axial length is traditionally used in the calculation of refractive outcomes for cataract surgery, but it is becoming the preferred measurement in the management of progressive myopia.^1,2 It is defined as the distance from the anterior corneal surface to either the inner limiting membrane or retinal pigment epithelium of the retina.³ Axial elongation underlies myopia development and progression, with longer axial lengths associated with increased risk of pathological ocular complications and irreversible visual impairment.^4,5 As such, eye health care practitioners use axial length not only as an objective means of monitoring myopia progression but also to assess the risk of onset for pre-myopes.⁶

Different biometry devices utilise various technologies, each with their own advantages and disadvantages. Early ocular biometry devices utilised A-scan ultrasound technology, but current biometers utilise non-contact optical methods. While ultrasound biometry is better able to penetrate dense cataracts, it requires direct applanation to the eye of the patient.⁷ Therefore, non-contact optical biometers provide key advantages in comfort and ease of use when imaging children in the context of managing progressive myopia.

Axial length captured with the IOLMaster 500 (Carl Zeiss Meditec AG, Jena, Germany) has been used as the primary outcome measure in many clinical trials and for the clinical management of progressive myopia. It is based on partial coherence inferometry (PCI) also known as time-domain optical coherence tomography (OCT) with a resolution of 12 µm.⁸ Infrared light is reflected from each interface of the eye where there is a change in refractive index. This light is then measured with the aid of a moving reference mirror from which geometric distances are calculated.^8,9

In comparison, the newer IOLMaster 700 (Carl Zeiss Meditec AG, Jena, Germany) uses swept-source OCT technology which has high axial resolution of 5 µm.¹⁰ This involves time-coding the wavenumber by using a tuneable wavelength laser source to sequentially scan the eye. The light reflected from the eye is mixed with the light reflected from a stationary reference mirror to acquire an interferometric signal from which measurements are extracted.¹¹

The main difference between the two technologies is that time-domain OCT requires physical moving and positioning of the reference mirror whereas swept-source OCT utilises a stationary reference mirror, resulting in significantly different capture times.^11,12 Therefore, the speed and sensitivity at which the IOLMaster 500 captures measurements are limited by the speed and precision at which the reference mirror can be moved and positioned. Various ocular biometry measures such as corneal curvature, anterior chamber depth and axial length must be taken sequentially.

In comparison, the IOLMaster 700 simultaneously acquires multiple radial biometry measurements along the visual axis while taking only half the time of the IOLMaster 500.¹² This acquisition time difference is more likely to have an impact during myopia management which requires repeat measurements over time. In addition, younger children are more likely to have more difficulty keeping steady fixation during measurements compared to adults resulting in higher within-person variability.¹³ Even very small errors in axial length could inappropriately influence myopia management.²

The UNSW Optometry Clinic obtained the IOLMaster 700 in mid-2018, in addition to the existing IOLMaster 500. Therefore, the clinical protocol was updated to measure the axial length with both instruments while transitioning between the two instruments. Previous studies examining the agreement between the IOLMaster 700 and IOLMaster 500 have found close agreement in axial length measurements between the two instruments when measured in both adults and children (refer to Table 1). However, the impact of age on axial length within-person variability has not yet been assessed.

Table 1. A summary of previous studies comparing IOLMaster 700 and IOLMaster 500.

Paper	Country	n	Optical media	Mean age (range) in years	Mean AL for IOL Master 500 (range) in mm	Mean AL for IOL Master 700 (range) in mm	Difference of means (700–500) (p-value) in mm	95% CI for mean difference	LoA in mm
Current study	Australia	51 eyes, 51 children	Clear lenses	10.98 ± 2.77 (4.85 to 16.20)	24.55 ± 1.04	24.57 ± 1.05	0.02 (p < 0.001)		−0.025 to + 0.061
Ye et al.^Citation14	China	250 eyes, 125 children	Clear lenses	Approximately 8.29	23.68 ± 1.26	23.70 ± 1.25	0.01 (not reported)	-	−0.07 to + 0.09
Huang et al.^Citation15	China	100 eyes, 100 children	Clear lenses	10.37 ± 1.81 (7–14)	-	24.7 ± 1.07^†	0.025 (not reported) ^†	-	−0.005 to + 0.055
		109 eyes, 109 patients	Cataract	71.07 ± 10.44 (35–92)	-	23.6 ± 1.73^†	0.01 (not reported) ^†	-	−0.03 to + 0.049
		92 eyes, 27 adults	Clear lenses	28.90 ± 6.40 (20–48)	-	24.1 ± 1.85^†	0.019 (not reported) ^†	-	−0.021 to 0.06
Bullimore et al.^Citation16	USA	100 eyes, 100 participants	49 clear lenses	33.7 ± 9.2 (22–58)	24.21 ± 1.24	24.25 ± 1.24	+0.04±0.01 (<0.001)	+0.03 to + 0.04	+0.01 to + 0.06
			51 cataracts	62.3 ± 8.1 (47–85)	23.99 ± 1.28	24.01 ± 1.29	+0.03 ± 0.02 (<0.001)	+0.02 to + 0.03	−0.01 to + 0.06
Cho et al.^Citation17	South Korea	137 eyes, 78 participants	Cataract	66.2 ± 10.1 (33–86)	23.79 ± 1.67 (21.41 to 31.53)	23.82 ± 1.67 (21.42 to 31.53)	+0.039 ± 0.038 (not reported)	-	−0.041 to 0.106
Akman et al.^Citation18	Turkey	175 eyes, 101 participants	Cataract	68.32 ± 12.71 (24–81)	23.627 ± 1.25 (20.92 to 28.333)	23.622 ± 1.32 (20.91 to 29.69)	−0.005±0.02 (0.008)	−0.008 to −0.002	−0.045 to 0.035
Kunert et al.^Citation19	Germany	120 eyes, 120 participants	Cataract	68.2 ± 9.89 (37–89)	23.651 ± 1.822 (20.53 to 34.54)	23.655 ± 1.831 (20.49 to 34.56)	+0.004±0.025 (0.093)	−0.001 to 0.009	−0.045 to 0.053
Srivannaboon et al.^Citation12	Thailand	100 eyes, 100 participants	Cataract	64.53 ± 11.20	23.48 ± 1.17 (21.21 to 28.36)	23.50 ± 1.18 (21.21 to 28.40)	+0.017±0.024 (0.89)	-	−0.03 to 0.06

The present study aimed to first confirm the agreement between the axial length measurements taken with IOLMaster 700 and IOLMaster 500 in children as found in other studies, and then assess the impact of age on the within-person variability in axial length measurements taken with the instruments in myopic children from a range of ethnicities.

Methods

Study design

This was a retrospective record review conducted at the UNSW Optometry Clinic in Sydney, Australia. This study followed the tenets of the Declaration of Helsinki, and approval was obtained from the institutional Human Research Ethics Committee before study commencement. Participant parents/guardians gave written consent for the use of de-identified data for research. Clinical records were de-identified by a third-party outside the research team, who already had access to clinical records, and then transcribed by the research team.

Participants

Records of patients who attended the UNSW Optometry Clinic (University of New South Wales, Sydney, Australia) between September 2018 and September 2019 were reviewed, and those who fulfilled the inclusion criteria were included in the data analysis.

Inclusion criteria comprised of children ≤16 years of age at the time of measurement, cycloplegic refraction of ≤−0.50DS, no prior use of myopia control including atropine and orthokeratology, no strabismus or amblyopia, no known ocular or systemic pathology and valid ocular biometry measurements taken from both the IOLMaster 700 and IOLMaster 500.

Data collated

Axial length measurements

Axial length measurements were collated from de-identified clinical records from the UNSW Optometry Clinic in which multiple clinicians practice. As part of clinic protocol, all axial length measurements were taken before cycloplegic refraction.

For the IOLMaster 700 (Carl Zeiss Meditec AG, Jena, Germany), each scan yields axial length values for six meridians which are each comprised of an average of three B-scans. These six meridional values were averaged for statistical evaluation. The instrument will alert the operator to suboptimal results if the standard deviation for axial length measurements is greater than 0.027 mm.¹⁰ These suboptimal measurements were excluded. In addition, each IOLMaster 700 scan produces an OCT image of the fovea which indicates accurate patient fixation. Measurements were excluded if foveal images indicated poor fixation.

For the IOLMaster 500 (Carl Zeiss Meditec AG, Jena, Germany), the manufacturer does not recommend the use of measurements with a signal-to-noise (SNR) ratio of less than two and so these measurements were excluded. Single measurements that deviated more than ±0.50 mm from the average of the other four measurements were also excluded, as these were attributed to the loss of fixation during the scan. Therefore, the first five consecutive acceptable measurements were obtained for each eye and included for statistical analysis.

Other data

In addition to axial length measurements, other data collated included date of birth, gender and cycloplegic autorefraction. Patients were also required to fill out a questionnaire, which collected information about their ethnicity.

Sample size calculations

Sample size was calculated for precision between the two instruments, using the formula:

$S_W\frac{z}{\sqrt{2n\left(n^{\prime }-1\right)}}=\mathit{C}{S}_W$

where $S_W$ is the within-subject standard deviation, z is the critical z for the α level, C is the relative margin of error, n is the sample size and n’ is the number of repeated measurements.²⁰ As $S_W$ is on both sides of the formula, it is cancelled out. For α of 0.05 and a 10% relative margin of error, a sample size of 48 participants was required for a 10% level of confidence and 5 measurements with each instrument.²⁰ Five repeated measurements were used as this resulted in more conservative calculation of sample size.

Statistical analysis

Data analysis was performed using StataCorp 2019 (Stata Statistical Software Release 16; College Station, TX: StataCorp LLC) and RStudio 2023 (Version 4.3.0; Boston, MA: RStudio Team). Data from the right eye only was included in the statistical analysis unless it did not meet the inclusion criteria, in which case data from the left eye was used.

Agreement between the IOLMaster 500 and IOLMaster 700

Shapiro-Wilk test was used to determine the normality of data, and a paired t-test was performed to compare measurements between eyes and measurements between the IOLMaster 700 and IOLMaster 500.

Bland-Altmann analysis was used to compare the agreement between axial length measurements taken with the IOLMaster 700 and IOLMaster 500. Limits of Agreement (LoA) were calculated using:

$\mathit{LoA}=\mathit{mean}\mathit{of}\mathit{differences}\pm 1.96\times \mathit{standard}\mathit{deviations}\mathit{of}\mathit{differences}$

The 95% confidence intervals for LoA were calculated using exact methods, considering the LoA as a pair.^21,22 Intraclass correlation coefficient (two-way mixed, absolute agreement) was calculated to compare the measurements obtained from each instrument.²³

Within-person variability

Within-person variability and R² was calculated for both the IOLMaster 700 and IOLMaster 500 to examine the effects of age on instrument variability. Linear mixed models with and without instrument variance were compared using one-way ANOVA to evaluate differences in within-subject variability for the two instruments. This method was also used to examine the effect of gender and ethnicity on within-person variability for the two instruments.

Results

Fifty one eyes from 30 female and 21 male patients were included for analysis (49 right eyes and 2 left eyes). The mean ± standard deviation age of the sample was 10.98 ± 2.77 years, range 4.85 to 16.20 years. Mean spherical equivalent refraction were −3.16 ± 1.51D (range of −0.75 to −6.38 D). The majority of children were East Asian (including South-East and North East Asian, n = 18) or Caucasian (n = 16). The rest of the participants were from Southern and Central Asia (n = 11) or other (n = 6).

Mean ± standard deviation axial lengths obtained with IOLMaster 700 and IOLMaster500 were 24.57 ± 1.05 mm (range 22.11–26.86 mm) and 24.55 ± 1.04 mm (range 22.10–26.85 mm).

Axial length measurements were highly correlated in both instruments with an intraclass correlation coefficient of 1.000 (95% CI: 0.998–1.000). The IOLMaster 700 measured significantly longer axial lengths than the IOLMaster 500 with a bias of +0.02 mm (95% CI: +0.012 to +0.024 mm) (P < 0.001) (Bland-Altmann graph shown in Figure 1). Limits of agreement (with 95% confidence intervals) were −0.025 (−0.036 to −0.018) to + 0.061 (0.054 to 0.072) mm.

Figure 1. Bland-Altmann plot for axial length measured with the IOLMaster 700 compared to the IOLMaster 500. The 95% limits of agreement are indicated by the dotted lines, while the bold line indicates the mean difference. 95% CI for the limits of agreement are indicated by the error bars.

There was no effect of age on within-person variability for the measurement of axial lengths with either instrument, with R² of 0.13 (p = 0.420) and 0.02 (p = 0.305) for the IOLMaster 700 and 500, respectively (Figure 2). The within-person variability of axial measurements with the IOLMaster 700 was significantly lower compared to the IOLMaster 500 with mean standard deviation of 0.005 (95% CI: 0.004 to 0.006) and 0.039 (95% CI: 0.032 to 0.048).

Figure 2. Standard deviations of within-person measurements for the IOLMaster700 (indicated in closed circles) and IOLMaster 500 (indicated in open circles) plot against patient age. Linear trend lines are indicated for the IOLMaster 700 (long dashes) and the IOLMaster 500 (short dashes).

There was no significant difference in the variability between East Asian or Caucasian ethnicities for the IOLMaster 700 (p = 0.053) or the IOLMaster 500 (p = 0.057). Other ethnicities were not analysed due to a small sample size. There was a statistically significant difference in the variability of axial length measurements between female and male groups for both instruments. For the IOLMaster 700, females were more variable than males (p = 0.006) and for the IOLMaster 500, males were more variable compared to females (p < 0.001).

Examination of Figure 2 revealed six outliers with high within-subject standard deviation (>0.05) for the IOLMaster 500 for which there is no consistent trend. From these outliers, five were female and one was male. Average age of these participants was 11.50 years (range 8.97 to 16.20 years), and average spherical equivalent refraction was −4.48 D (range −2.13 to −6.38 D). These participants did not have corresponding high within-subject standard deviation for the IOLMaster 700 with mean standard deviation of 0.0028.

Discussion

The primary aim of this study was to confirm the agreement between the IOLMaster 700 and 500 for axial length measurements as found in previous studies (refer to Table 1). The acquisition time of the IOLMaster 500 is twice as long as the IOLMaster 700,¹² so it was hypothesised that younger children may have more difficulty keeping steady fixation. This could result in higher within-person variability and less accurate axial length measurements. Therefore, the secondary aim was to examine the effect of age upon the within-person variability.

Axial length measurements captured with the IOLMaster 700 were statistically longer compared to those captured with the IOLMaster 500 by + 0.02 ± 0.02 mm (p < 0.001). This is consistent with previous studies conducted in both children and adults, with and without media opacities (refer to Table 1). All but one study¹⁸ reports slightly longer axial length measurements with the IOLMaster 700, ranging from + 0.004 to +0.04 mm.^{12,15–17,19}

The two studies which have compared the IOLMaster 700 and 500 in Chinese children have also found axial lengths measured with the IOLMaster 700 to be slightly longer compared to those measured with the IOLMaster 500, with difference of means of 0.01 and 0.025 mm, respectively.^14,15 In agreement, this study found the IOLMaster 700 to measure axial length slighter longer than the IOLMaster 500 in a multi-ethnic population.

The IOLMaster 700 also measures axial length slightly longer compared to other instruments, albeit with smaller mean differences. This includes the IOLMaster V3 (mean difference 0.009 mm),²⁴ MyopiaMaster (mean difference 0.02 mm)¹⁴ and the Pentacam (mean difference 0.01 mm).²⁵ Only the MYAH (Topcon EU, Visia Imaging, Japan) measured a slightly longer axial length than the IOLMaster 700 with a mean difference of 0.01 mm.²⁵

Previous studies, especially those comparing the IOLMaster 700 and 500, have concluded that the mean differences found is clinically insignificant but have done so within the context of calculating intra-ocular lens power for the purpose of cataract surgery. A measurement error of 0.02 mm in axial length results in a postoperative refractive error of only 0.07D.²⁶ However, the criteria for clinical significance in myopia management is different; a difference of 0.02 mm between instruments as found in the present study represents 20% of axial length growth per year for a 10-year-old child with axial length at the 50th percentile.²⁷ As such, this measurement difference between the two instruments could give false positive changes in axial length if instruments are used interchangeably for the purpose of myopia management, potentially resulting in inappropriate management strategies or conclusions. As a result, eye health care practitioners should be aware of the differences between the two instruments and refrain from using them interchangeably.

The second study aim was to examine the effect of age on within person variability, as younger patients are more likely to have worse fixation compared to adults. Investigations into binocular fixation behaviour in children 4–15 years of age have found fixation improved with increasing age, as did time between intruding eye movements.¹³ The number of intruding saccades away from fixation also reduced with age. However, this study found no effect of age on within-person variability for both the IOLMaster 700 and IOLMaster 500, with R² failing to reach significance for both instruments. Therefore, practitioners can be confident that within-person variability for axial length measurements for both the IOLMaster 700 and 500 will not vary according to the age of the patient.

Gender and ethnicity were also examined as potential factors affecting the variability of the instruments. No effect of ethnicity was found. Axial length measurements for females were more variable compared to males in the IOLMaster 700, but the opposite was found for the IOLMaster 500. This may be a result of the small sample size and warrants further investigation with a larger sample size.

For comparison of within-person variability between the two instruments, it was found that axial length measurements were significantly less variable when measured with the IOLMaster 700 compared to the IOLMaster 500. Each capture with the IOLMaster 700 produces 6 measurements, whereas each capture with the IOLMaster 500 produces a single measurement. As such, to obtain multiple measurements with the IOLMaster 500, clinicians are required to perform multiple captures which increases capture time and variability. A within-person variability of 0.0049 mm in the IOLMaster 700 and 0.028 mm in the IOLMaster 500 (p < 0.001) was found.

Previous studies which perform multiple consecutive axial length measurements with the IOLMaster 700 have found intra-observer repeatability of 0.02 mm, which is more similar to 0.028 mm obtained with the IOLMaster 500 in the present study.¹⁵ A lower within-person variability is particularly useful in the context of myopia management, which requires the detection of small changes in axial length over 3–6 months.²⁸ Therefore, the ability of the IOLMaster 700 to obtain multiple measurements within a single capture may be a key advantage over the IOLMaster 500 in the context of myopia control.

The study population was reflective of typical myopia control patients in Australia but as such were mostly limited to those with Asian or Caucasian ethnicities. In addition, they were from a limited range of refractive errors, as the focus in this study was on myopic children. The study design is a retrospective chart review within a university optometry clinic, and so there is a possibility that some measurements were taken by optometry students who are less experienced. However, the study design included quality review of each axial length measurement which was included for comparison. Therefore, potential effect of the instrument experience of the user should be minimised.

This study is also limited to comparing only two instruments when there are many more in the market which utilise different underlying technologies and may be more affected by patient age and fixation. Future studies investigating inter-visit reliability to help determine the repeatability of the IOLMaster 700 between visits is required.

Conclusion

The within-person variability for axial length measurements with the IOLMaster 700 and IOLMaster 500 is not dependent on age in myopic children. However, axial length measurements captured with the IOLMaster 700 were significantly less variable compared to the IOLMaster 500, suggesting the IOLMaster 700 is preferable for myopia management which requires repeat axial length measurements over short periods of time. In addition, axial length measurements with the IOLMaster 700 are longer compared to the IOLMaster 500. Therefore, eye health care practitioners should be wary of the small but significant difference between the two instruments and refrain from using them interchangeably.

Acknowledgements

The authors thank the UNSW Optometry Clinic for its assistance and Dr Nancy Briggs for her contribution towards the statistical analysis used in this study. There are no financial disclosures.

Disclosure statement

No potential conflict of interest was reported by the author(s).