Dioptric Differences between Clinically Determined and Metric-Optimised Refractions for Adults with Down Syndrome

Abstract

Purpose

Refractions based on the optimisation of single-value wavefront-derived metrics may help determine appropriate corrections for individuals with Down syndrome where clinical techniques fall short. This study compared dioptric differences between refractions obtained using standard clinical techniques and two metric-optimised methods: visual Strehl ratio (VSX) and pupil fraction tessellated (PFSt), and investigated characteristics that may contribute to the differences between refraction types.

Methods

Thirty adults with Down syndrome (age = 29 ± 10 years) participated. Three refractive corrections (VSX, PFSt and clinical) were determined and converted to vector notation (M, J₀, J₄₅) to calculate the dioptric difference between pairings of each type using a mixed model repeated measures approach. Linear correlations and multivariable regression were performed to examine the relationship between dioptric differences and the following participant characteristics: higher order root mean square (RMS) for a 4mm pupil diameter, spherical equivalent refractive error and Vineland Adaptive Behavior Scales (a measure of developmental ability).

Results

The least squares mean estimates (standard error) of the dioptric differences for each pairing were: VSX vs. PFSt = 0.51D (0.11); VSX vs. clinical = 1.19D (0.11) and PFSt vs. clinical = 1.04D (0.11). There was a statistically significant difference in the dioptric differences between the clinical refraction and each of the metric-optimised refractions (p<0.001). Increased dioptric differences in refraction were correlated with increased higher order RMS (R = 0.64, p<0.001 (VSX vs. clinical) and R = 0.47, p<0.001 (PFSt vs. clinical)) as well as increased myopic spherical equivalent refractive error (R = 0.37, p=0.004 (VSX vs. clinical) and R = 0.51, p<0.001 (PFSt vs. clinical)).

Conclusions

The observed differences in refraction demonstrate that a significant portion of the refractive uncertainty is related to increased higher order aberrations and myopic refractive error. Methodology surrounding clinical techniques and metric-optimisation based on wavefront aberrometry may explain the difference in refractive endpoints.

INTRODUCTION

It is often difficult to refract individuals with Down syndrome using the common clinical technique of subjective refraction. Subjective refraction is an inherently demanding psychophysical task that requires attention to detail and fine discrimination. It is a difficult task even for individuals with excellent optical quality and without intellectual disability. For patients with Down syndrome, it is made even more challenging by the complex refractive errors often present in this population,^1–3 including elevated levels of higher order aberrations.^4,5

Objective refraction methods often utilised in the clinical setting, such as retinoscopy, are a useful alternative to subjective refraction methods for this population, but still rely largely on examiner skill and do not specifically consider the contribution of higher order optical aberrations to the resultant retinal image quality. Thus, refraction methods that utilise wavefront aberration measures of the eye may be better suited for patient populations with substantial higher order aberrations. In order to quantify the optical quality of the eye, various image quality metrics have been defined to mathematically distil wavefront measures down to a single numerical value.⁶ Some of these metrics have been demonstrated to correlate well with visual acuity performance, and thus they serve as valuable predictive tools.^7–11 Refractions can be determined objectively from wavefront aberrations measures of the eye by mathematically combining them with a given refractive correction and calculating the resultant image quality metric value. Thousands of refractions can be tested computationally for a given eye to identify the refraction predicted to provide the best metric value, and thus the optimal visual acuity.^12–14

More recently, objective refraction methods based on measurement of wavefront aberrations and subsequent optimisation of single-value metrics have been explored in a clinical trial of spectacle prescribing for adults with Down syndrome.¹⁵ The purpose of the clinical trial was to compare distance visual acuity outcomes for three different refraction types: two based solely on the objective measurement of wavefront aberrations (one optimising visual Strehl ratio (VSX) and the other pupil fraction tessellated (PFSt)), and a third determined solely from standard clinical techniques. Overall, the study found no difference in visual acuity outcomes across treatment types, although some individual participants did experience measurable improvements in visual acuity with the metric-optimised refractions over clinical refraction.¹⁵

The primary outcome paper for the clinical trial reported visual acuity measurements obtained with each refraction type but did not report a comparison between the three refractions. One might therefore question whether the refractions were similar across treatment types, given the similar acuity outcomes. Alternatively, if the refractions were significantly different despite similar acuity outcomes, this could point to other factors limiting visual acuity in adults with Down syndrome. For example, the majority of participants had significant bilateral refractive error that could have been amblyogenic if left uncorrected during childhood, and thus visual acuity may have been similarly reduced for all refraction types in the clinical trial despite differences in the magnitude of the correction.

This investigation analysed data from the previously reported clinical trial to quantify differences between refraction types in that study’s participants to identify possible explanations for the lack of a difference in visual acuity, as well as to determine the level of refractive uncertainty for this population as indicated by the magnitude of differences between clinical and objective refractions. We hypothesise that various participant characteristics might play a role in clinical refractive uncertainty and thus this analysis explores refractive error magnitude, level of intellectual disability and the magnitude of higher order aberrations as possible sources of differences between refractions. In addition, data from two separate study populations without intellectual disability (adult controls and adults with keratoconus) were included to determine whether refractive uncertainty increases across populations with greater levels of higher order aberrations, or whether the observed differences are unique to participants with Down syndrome.

METHODS

This study was approved by the University of Houston Committee for the Protection of Human Subjects and adhered to the tenets of the Declaration of Helsinki. All testing was conducted at the University of Houston, College of Optometry. Written parental/guardian permission to participate in research was obtained for all participants with written or verbal assent obtained from the participants. The present study reports analysis of data collected in a previously reported clinical trial registered on ClinicalTrials.gov (NCT03367793), in accordance with the National Institutes of Health policy.

Participants

Thirty adult participants with Down syndrome were randomised in the clinical trial. The average age of the participants was 29 years (±10 years) with an age range of 18–52 years. Inclusion criteria included a previous diagnosis of Down syndrome and an ability to sit for study measures. Exclusion criteria included nystagmus, significant corneal or lenticular opacities and/or a clinical diagnosis of keratoconus. There was no minimum level of refractive error required for participation in the study, nor was previous spectacle wear a cause for exclusion. Twenty-one of the participants were habitual spectacle wearers; the other nine participants presented to the study unaided. At the baseline examination, a parent or legal guardian completed the Vineland Adaptive Behavior Scales (Pearson Education, pearsonassessments.com/) to assess the developmental ability of each participant. This comprehensive assessment includes items related to communication, daily living skills, socialisation and motor skills.

Study Measures

All participants received a complete eye examination performed by a single paediatric optometrist with over 30 years of experience providing care to individuals with differing abilities, such as patients with Down syndrome. During the examination, the optometrist determined a best clinical refraction using measurement techniques commonly performed during a comprehensive eye examination, as described below. Wavefront aberrations from the 2^nd through the 10^th radial order were measured at the end of the examination by a separate examiner using the COAS-HD wavefront aberrometer (Johnson & Johnson Vision, jnjvisionpro.com/products). All wavefront measures were collected 30 minutes post-instillation of tropicamide 1% and phenylephrine 2.5%. Measurements were taken until three to five good quality measurements per eye were obtained. Habitual pupil diameter was determined from infrared photorefraction measures (PowerRef 3, Plusoptix, plusoptix.com/en-gb/products) prior to dilation in dim illumination and used to re-size the wavefront measurements to a diameter specific to each eye for determination of two metric-optimised refractions as described below. In addition, wavefront measures were secondarily re-sized to a common 4mm pupil diameter in order to compare higher-order root mean squared (RMS) aberrations across eyes.

Refraction Methods

Best Clinical Refraction

The expert examiner determined a clinical refraction based on information from the following common clinical techniques: static retinoscopy, Grand Seiko WAM-5500 autorefraction (Visionix, shop-usa.visionix.com/), subjective refraction and consideration of the spectacle power of any presenting correction, as well as the visual acuity with that correction. The examiner was permitted to obtain clinical measures both pre- and post-dilation with 1% tropicamide and 2.5% phenylephrine. Both letters and shapes available in the M&S electronic visual acuity system (M&S Technologies, Inc., mstech-eyes.com/) were used throughout the clinical examination to guide determination of the clinical refraction but clinical trial visual acuity outcomes were obtained using letter by letter scoring on a Bailey-Lovie style chart. The protocol used for determining visual acuity through each refraction type (the two metric-optimised refractions and the clinical refraction) has been described previously.¹⁶

While distance autorefraction and retinoscopy were obtained for all study participants, subjective refraction was only obtained from those individuals who were developmentally able to provide subjective input. One form of subjective refraction that was frequently utilised for the sphere power determination was a monocular red-green bichrome (duochrome) target whereby the examiner listened to how easily the participant read the letters on the red versus the green backgrounds, rather than asking the participant to compare the clarity of the letters. A final clinical prescription was left up to the examiner’s discretion based on synthesis of all of the findings from the clinical tests. The examiner did not have access to the metric-optimised refractions described below.

Metric-Optimised Refractions

Images captured from the COAS-HD were processed and Zernike coefficients from the 2^nd to 10^th order calculated over the eye’s habitual pupil size. An average of the Zernike coefficients was calculated for each eye from the three to five good quality images obtained from the COAS-HD using the custom software program Spectacle Sweep (University of Houston College of Optometry Core Programming Module, Houston, Texas) written with MATLAB (MathWorks, mathworks.com/products/matlab.html). Spectacle Sweep then mathematically applied >20,000 refractions to the average Zernike coefficients of each eye of each participant in sphero-cylindrical combinations ranging from at least ±3DS in 0.25DS steps in either direction from the participant’s habitual sphere power and from 0.00 to at least −4.00DC in −0.25DC steps (with greater values explored for participants with higher habitual cylinder power) over the entire range of cylindrical axes in 1° steps. For each of these sphero-cylindrical combinations, the residual wavefront error for the mathematical representation of the combination of the spectacle lens and eye was calculated along with the resultant values for two image quality metrics: Visual Strehl Ratio in the spatial domain (VSX) and Pupil Fraction Tesselated (PFSt).⁶ The final objective metric-optimised refractions were the sphero-cylindrical combination that resulted in the best value for each of VSX and PFSt, respectively.

Data Analysis

Vector Notation Conversion

From the refractive methods described above, each participant was prescribed three different pairs of glasses: best clinical refraction, VSX-optimised refraction and PFSt-optimised refraction. Given the difficulty making direct comparisons between sphero-cylindrical refractions, each refraction was converted to vector notation for analysis using the following components: M (spherical equivalent), J₀ (astigmatic component of the horizontal and vertical meridians) and J₄₅ (astigmatic component of the oblique meridians).¹⁷ Conversion to vector notation allows for scalar mathematical methods (e.g., addition, subtraction, averaging) to be applied to the data as all portions of the sphero-cylindrical notation have been converted to the same unit (D).^17,18 The standard convention was followed where a positive value for J₀ and J₄₅ corresponds to the 180 and 45 meridians, respectively, and a negative value for J₀ and J₄₅ corresponds to the 90 and 135 meridians, respectively.

Dioptric Difference Comparison

Once all refractions were converted to vector notation, the dioptric difference between refractions was calculated as the square root of the sum of the squared differences between refraction types.^18,19 The formula for this calculation as applied to two refractions (ref1 and ref2) was as follows:

$$\mathit{\text{Dioptric}}\mathit{\text{Difference}}=\sqrt{{\left(M_{\mathit{\text{ref}}1}-M_{\mathit{\text{ref}}2}\right)}^2+{\left(J_{0,\mathit{\text{ref}}1}-J_{0,\mathit{\text{ref}}2}\right)}^2+{\left(J_{45,\mathit{\text{ref}}1}-J_{45,\mathit{\text{ref}}2}\right)}^2}$$

An example comparison for a single eye is shown in Table 1 where the sphero-cylindrical and vector notation are listed for each of the three refraction types and the dioptric difference between all pairings of refraction types calculated. The data for the single eye presented in Table 1 illustrates the calculations performed for each eye individually to determine its appropriate vector variables. Vector component data for the median of all eyes analysed in this study are presented in the results section (Table 2).

Table 1.

Example refraction comparison for a single eye.

Vector	VSX +1.50 – 1.75 × 154	PFSt +1.75 – 3.00 × 152	Clinical +2.25 – 2.00 × 160
M	+0.63	+0.25	+1.25
J0	+0.54	+0.84	+0.77
J45	−0.69	−1.24	−0.64

Calculated Dioptric Differences
VSX vs PFSt = 0.73D	VSX vs Clinical = 0.67D	PFSt vs Clinical = 1.12D

First, each refraction type is converted from sphero-cylindrical to vector notation. Second, the dioptric difference is calculated for each pair of refraction types. J₀, astigmatic component in the horizontal and vertical meridians; J₄₅, astigmatic component in the oblique meridians; M, spherical equivalent refractive error; PFSt, Pupil Fraction Tesselated refraction; VSX, Visual Strehl Ratio.

Table 2.

Median (first and third quartiles) for each vector component in dioptres by refraction type.

Eye	Vector	VSX	PFSt	Clinical
OD (n=30)	M	+0.13 (−4.13, +1.47)	+0.25 (−3.88, +2.00)	+0.25 (−2.66, +1.47)
	J0	+0.14 (−0.39, +0.56)	+0.16 (−0.37, +0.59)	+0.26 (−0.27, +0.78)
	J45	−0.50 (−0.92, +0.07)	−0.57 (−0.97, +0.02)	−0.31 (−0.73, 0)
OS (n=30)	M	+0.25 (−3.53, +1.41)	+0.06 (−3.28, +1.75)	−0.06 (−2.41, +1.34)
	J0	−0.12 (−0.58, +0.68)	−0.10 (−0.46, +0.66)	+0.20 (−0.25, +0.47)
	J45	+0.45 (+0.19, +1.02)	+0.40 (+0.09, +0.97)	+0.33 (+0.04, +0.67)

J₀, astigmatic component in the horizontal and vertical meridians; J₄₅, astigmatic component in the oblique meridians; M, spherical equivalent refractive error; OD, right eye; OS, left eye: PFSt, Pupil Fraction Tesselated refraction; VSX, Visual Strehl Ratio.

Higher-order RMS

Higher-order RMS was calculated on the Zernike coefficients representing the average of three to five measures on each eye (re-sized to a 4mm pupil diameter) by taking the square root of the sum of the squares of all terms in the 3^rd – 10^th orders.

Vineland Score Calculation

The Vineland sub-domains were tallied, converted to v-scores, and combined to determine a single overall adaptive behaviour score for each participant. The adaptive behaviour score was then converted to a standard score based on the published norms in the instrument’s manual.

Comparison with Other Study Populations

Dioptric differences between clinical and VSX-optimised refractions for the participants with Down syndrome were also compared to differences between refractions for two samples obtained in separate studies. These samples included 40 eyes from 20 controls with normal best corrected visual acuity (mean acuity −0.07 ± 0.05 logMAR) aged 21 – 43 years¹³ and 21 eyes from 12 individuals with keratoconus aged 21 – 65 years.²⁰ A clinical manifest refraction was performed for each of these groups followed by VSX-optimised refraction of dilated wavefront aberration measures, similar to the procedure described for participants with Down syndrome in the present study. The clinical manifest refraction for the control and keratoconus groups was not performed by the same examiner as the Down syndrome group. However, the manifest refraction was performed by one of two experienced clinicians for each group (three clinicians in total) in a manner consistent with clinical convention where maximum plus to best visual acuity was used as the subjective endpoint. For participants with normal best corrected visual acuity, subjective refraction was determined through a non-dilated pupil with 0.25D sphere and cylinder adjustments until a final endpoint of maximum plus that retained best visual acuity was achieved (a hyperfocal refraction).¹³ For participants with keratoconus, subjective refraction was also determined through a non-dilated pupil. Cylinder adjustments were conducted with the ±0.25D Jackson Cross Cylinder while spherical adjustments were adapted to a step size based on the subject’s ability to detect a difference between the offered choices.²⁰ In addition to comparing dioptric differences in refraction across study groups, correlations between the dioptric difference in refraction and both higher-order RMS over a 4mm pupil and spherical equivalent refractive error were explored.

Statistical Approach

Statistical analyses were performed using SAS (PROC GLIMMIX; SAS Institute, sas.com/en_us/home.html). A mixed model repeated measures approach was used to identify significant differences between refraction types while clustering eyes within participants. As a descriptive exploration, linear correlations (all eyes pooled together) between the magnitude of dioptric differences and the following participant characteristics were performed: higher-order RMS for a 4mm pupil, spherical equivalent refractive error as determined by the clinical refraction and Vineland Adaptive Behavior Scales standard score. Lastly, multivariable regression via a mixed linear regression model was performed to estimate the effect that each of the independent variables collectively had on differences between refraction type while also accounting for the clustering of eyes within participants.

RESULTS

Participants with Down syndrome in the study had the following range of refractive error, as determined by the clinical refraction for all eyes: sphere power −14.75 to +6.50DS, cylinder power 0 to −8.25DC and spherical equivalent power −15.75 to +5.12DS. The median M, J₀, and J₄₅ components for the right and left eyes for each of the three refraction types are listed in Table 2. Plots of the individual participant data are shown in Appendix 1.

Dioptric Differences between Refraction Types

The smallest dioptric difference between refraction types was for the comparison between VSX and PFSt metric-optimised refractions (least squares mean estimate (standard error) = 0.51D (0.11)). Greater differences were observed between the clinical refractions and each of the metric-optimised refractions: least squares mean estimate (standard error) = 1.10D (0.11) for VSX versus Clinical and 1.04D (0.11) for PFSt versus Clinical. The dioptric difference between the clinical and metric-optimised refraction comparisons was significantly greater (p<0.001) than the difference between the two metric-optimised refractions. The boxplots in Figure 1A show the median and interquartile range of differences between refraction types for individuals with Down syndrome. Given that the aim of the clinical trial was to compare metric-optimised refractions to clinical refraction and given the relatively small mean difference between the metric-optimised refractions (0.51D), the remainder of the analysis was focused on factors associated with the comparisons between clinical and metric-optimised refractions rather than the metric-optimised refractions to each other.

Figure 1.

Dioptric differences were significantly greater between clinical and metric-optimised refraction types (p<0.001 for both Visual Strehl Ratio (VSX) versus Clinical and Pupil Fraction Tesselated (PFSt) versus Clinical) in individuals with Down syndrome as shown in Figure 1A. Figure 1B shows a comparison of the magnitude of the dioptric differences between VSX-optimised and clinical refractions for the three different populations of study participants.

Figure 1B describes the median difference between VSX versus Clinical for the individuals with Down syndrome shown in Figure 1A (different scale) to typical controls corrected to 0.00LogMAR or better¹³ and individuals with keratoconus.²⁰ The median difference for participants with Down syndrome (0.97D) was larger than the difference between these two refraction types for control subjects with normal best corrected acuity (0.35D). However, it was smaller than the median difference between these two refraction types for participants with keratoconus (2.77D), suggesting greatest refractive uncertainty for individuals with keratoconus.

Relationship between Dioptric Differences and Aberrations

To determine whether the dioptric difference between refraction types was associated with the level of higher order aberrations in an eye, the dioptric difference was compared to higher order RMS (Figures 2A and 2B). For both refraction pairs tested in individuals with Down syndrome, the dioptric difference and higher order RMS were significantly (p<0.001), positively correlated, meaning that eyes with greater higher order RMS had larger differences between clinical and metric-optimised refractions. This strong, positive correlation remains when data from the controls and individuals with keratoconus are included (p<0.001) (Figure 2C). Visual inspection of the data again indicates that individuals with Down syndrome fall in between these two other groups in regard to higher order RMS versus dioptric differences. It also highlights that the two individuals with Down syndrome with larger amounts of higher order RMS are not “outliers” but merely fall along the spectrum when including all three populations.

Figure 2.

Dioptric differences were positively correlated with higher-order root mean square (RMS) calculated over a 4mm pupil for both the Visual Strehl Ratio (VSX) versus Clinical (2A) pair (p<0.001) and Pupil Fraction Tesselated (PFSt) versus Clinical (2B) pair (p<0.001) in individuals with Down syndrome. Figure 2C shows the positive correlation between higher-order RMS and dioptric difference of VSX and clinical refractions for three study populations with varying levels of higher-order RMS combined (p<0.001).

Relationship between Dioptric Differences and Refractive Error

To determine whether the dioptric difference between refraction types was associated with the amount and type of refractive error, the dioptric difference was compared to the spherical equivalent refractive error (as determined by the clinical refraction) (Figures 3A and 3B). For both refraction pairs tested in individuals with Down syndrome, dioptric difference and spherical equivalent refractive error were significantly (p=0.004 for VSX versus clinical; p<0.001 for PFSt versus clinical), negatively correlated, meaning that eyes with higher amounts of myopic refractive error had larger differences between clinical and metric-optimised refractions. This negative correlation was maintained when combining all three study populations together (p<0.001) (Figure 3C).

Figure 3.

Dioptric differences were negatively correlated with spherical equivalent refractive error (determined from the clinical refraction) for both the Visual Strehl Ratio (VSX) versus Clinical (Figure 3A) pair (p=0.004) and Pupil Fraction Tesselated (PFSt) versus Clinical (Figure 3B) pair (p<0.001) in individuals with Down syndrome. Figure 3C shows the correlation between the spherical equivalent refractive error and dioptric difference for VSX and clinical refraction in a combined sample from three study populations (p<0.001).

Relationship between Dioptric Differences and Vineland Adaptive Behavior Scales

To determine whether the dioptric difference between refraction types was associated with performance on the Vineland Adaptive Behavior Scales, the dioptric difference was compared to the standard score from the Vineland assessment (Figures 4A and 4B). There was no significant correlation between the Vineland standard score and the dioptric difference in refraction for either pair. As the other two study populations did not include individuals with intellectual disabilities, the Vineland assessment was not completed and there were no data to compare to the current study’s participants.

Figure 4.

There was no significant correlation between the Vineland standard score and dioptric difference in refraction for either pair in individuals with Down syndrome. PFSt, Pupil Fraction Tesselated refraction; VSX, Visual Strehl Ratio.

Multivariable Regression Analysis

For individuals with Down syndrome, all three characteristics [higher order RMS, spherical equivalent and Vineland score] showed a significant relationship with the magnitude of the refraction differences for the VSX and Clinical pair (Table 3 multivariable results). The Vineland Adaptive Behavior score was still non-significant for the PFSt and Clinical pair of refractions (Table 4).

Table 3.

Univariable and multivariable regression for the dioptric difference between the Visual Strehl Ratio (VSX) and Clinical refractions in individuals with Down syndrome.

VSX versus Clinical
	Univariable		Multivariable
Variable	Point Estimate (Standard Error)	P−value	Point Estimate (Standard Error)	P−value
Higher Order RMS	7.43 (1.20)	<0.001	7.14 (1.04)	<0.001
Spherical Equivalent	−0.069 (0.024)	0.08	−0.063 (0.017)	0.001
Vineland	−0.011 (0.007)	0.12	−0.010 (0.005)	0.03

RMS, root mean squared higher-order aberrations. Vineland refers to the Vineland Adaptive Behavior score.

Table 4.

Univariable and multivariable regression for the dioptric difference between the Pupil Fraction Tesselated (PFSt) and Clinical refractions in individuals with Down syndrome.

PFSt versus Clinical
	Univariable		Multivariable
Variable	Point Estimate (Standard Error)	P−value	Point Estimate (Standard Error)	P−value
Higher Order RMS	5.17 (1.25)	<0.001	4.57 (1.02)	<0.001
Spherical Equivalent	−0.087 (0.019)	<0.001	−0.083 (0.017)	<0.001
Vineland	−0.005 (0.007)	0.48	−0.004 (0.004)	0.43

RMS, root mean squared higher-order aberrations. Vineland refers to the Vineland Adaptive Behavior score.

DISCUSSION

This study compared the dioptric differences between three different types of refractions: clinical refraction, VSX metric-optimised refraction, and PFSt metric-optimised refraction. The two metric-optimised refractions were the most similar, differing by approximately 0.50D on average, while the average difference between the clinical refraction and either metric-optimised refraction was over 1D. There are several methodological and participant characteristics that may impact these differences between refraction types. Understanding these characteristics and what makes refraction endpoints less certain is important, particularly when examining populations who may not be as readily able to participate in traditional clinical techniques.

Methodological Considerations

To provide more context for this discussion, it is important to note that while not termed an “objective” refraction, the clinical refraction is based on methods considered objective in the clinical examination. Retinoscopy is treated as an objective technique as it does not require input from the patient to determine endpoints; a trained clinician interprets the light exiting the pupil as it is reflected off of the retina (termed “reflex”). However, there is a subjective component on the part of the examiner as to what lenses neutralise the observed reflex, and this subjectivity may be further influenced by the quality of the reflex. For example, a reflex sometimes exhibits “scissoring” where the paraxial rays represented by the peripheral portion of the reflex show a different motion to the central rays/centre portion of the reflex.²¹ In these cases, it is more difficult to determine an endpoint compared with a reflex that is crisp with distinct, uniform motion.

Retinoscopists are trained to determine neutrality for the centre portion of the reflex (corresponding to the centre of the pupil) particularly in cases of scissoring.²² Peripheral rays are ignored as they contradict the movement of the central rays. The expert examiner determining clinical refractions in the present study concentrated on the centre of the reflex when determining retinoscopy endpoints. This strategy is inherently different from that of the metric-optimised refraction. With metric-optimisation, wavefront error is assessed over the entire pupil, and this wavefront is used to calculate the image quality metric (either VSX or PFSt). In addition to assessing wavefront error, the VSX also incorporates a neural contrast sensitivity weighting function of a normal eye, which is capable of relatively ranking corrections for highly-aberrated eyes.²³

While endpoints for retinoscopy can be thought of as pupil location dependent (e.g., the clinician emphasises the reflex seen in the centre of the pupil), PFSt differs in that equal weight is given to all locations within the pupil that meet a certain criterion. In the calculation of PFSt, the entire pupil is thought of in terms of sub-apertures and criteria are established for what qualifies as a “good” versus “bad” sub-aperture. The area of good sub-apertures is divided by the total area of the pupil to compute the pupil fraction and provide the PFSt value.⁶ The location of these sub-apertures does not affect the metric and may include both central and peripheral portions of the pupil. In this manner, retinoscopy and PFSt-optimised refractions have an inherent difference. Differences with respect to the location of the optical information emphasised may also occur between retinoscopy and VSX, although the direct comparison is less clear.

Another potential source of differences between metric-optimised and clinical refractions could be the pupil diameters for which the refractions were determined. Metric-optimised refractions were re-sized to the habitual pupil diameter measured prior to dilation under dim illumination. This same dim illumination was used when the expert examiner performed non-cycloplegic retinoscopy and subjective refraction (when possible). However, the examiner also used retinoscopy and autorefraction findings under cycloplegia when considering the final clinical refraction, and thus it is not possible to state that only one single pupil diameter was used during the determination of clinical refraction. Despite this, we do not believe that differences in pupil diameter would account for the dioptric differences observed between clinical and metric-optimised refractions in this study. A previous study in our laboratory determined metric-optimised refractions for both a 4mm and 6mm pupil diameter for adult participants with Down syndrome and found that the dioptric differences between refractions were on average only 0.50 D across pupil diameters.⁵ This suggests that even if there is some variability in refraction related to pupil diameter, it does not account for the large dioptric differences between metric-optimised and clinical refractions observed here (Figure 1).

From examiner notes about the quality of each eye’s reflex, only 16% of the eyes in this study were noted to have “good”, “great” or “crisp” reflexes. The quality of the retinal reflex is likely related to the amount of aberration in the eye, such as is seen in eyes with keratoconus, a disease resulting in increased higher order aberrations which can be identified by assessment of the retinoscopic reflex.²⁴ In cases where scissoring is more prominent due to higher order aberrations, the endpoints from retinoscopy become less certain and may be more likely to vary from the endpoints found with metric-optimised refractions. This trend is reflected in the data from this study; clinical refractions differed more from metric-optimised refractions in eyes with increased amounts of higher order aberration.

It is important to note that the clinical refraction for individuals with Down syndrome was not determined solely by retinoscopy. For those who could participate in subjective refraction, endpoints were also determined by the fluency with which letters were read through different sphero-cylindrical combinations. However, even when the participants with Down syndrome were able to complete a subjective refraction, it is likely that these endpoints were more variable in individuals with elevated aberrations compared to normally developed individuals with lesser higher order aberrations. Keratoconic patients from the Collaborative Longitudinal Evaluation of Keratoconus (CLEK) study exhibited greater variability in the repeatability of subjective refraction compared to myopic individuals without keratoconus,¹⁸ and our data demonstrate a strong correlation between the differences in clinical and objective refraction with increasing levels of high-order aberrations (Figure 2). However, the reader is cautioned against deriving a predictive relationship between the amount of increasing HOA and a fixed dioptric difference value. This is because, at a minimum, the clinical refraction depends on an interaction between residual lower order aberration terms (which are not a part of the figure) and higher order aberrations. The clinical refraction that results from these interactions would then influence the dioptric difference from metric-derived refractions. In addition, higher order RMS in Figure 2 was calculated for a 4mm pupil, and the dioptric difference was calculated for the habitual pupil diameter, which was close to 4mm on average for both eyes. Therefore, these data should be interpreted that as the eye’s aberrations increase (in particular, those in the higher orders that cannot be corrected with spectacles and have a tendency to influence the residual lower order aberrations of sphere and cylinder) then the uncertainty in the refraction also increases.

Another trend seen in the data is the higher amount of oblique astigmatism found with metric-optimised refractions versus larger amounts of with- and against-the-rule astigmatism for clinical refractions (Table 2). It is possible that the wavefront measurements are better able to capture the oblique data of the eye compared to clinical methods, and this may also be related to the larger sampling of the pupil. However, it is also possible that an explanation for oblique versus with/against-the-rule astigmatism comes from the way that retinoscopy is performed. Clinicians are often instructed to start with the beam from the retinoscope oriented vertically, aligned with the 90° meridian of the eye and observe for change in skew as the beam is rotated towards the 180° meridian.²¹ The expert examiner in the current study adopted this classic methodology, starting with a quick sweep of the vertical and horizontal meridians before returning to the vertical meridian to “find” the axis of the eye. By starting retinoscopy with these 90 and 180 meridia, clinicians may inherently and subconsciously tend towards these major meridia when determining the axis during performing retinoscopy.

When considering prescribing patterns globally, clinicians are often cautioned against prescribing large amounts of cylinder power at oblique axes for fear of creating spatial distortion during binocular viewing that may impede activities of daily living, including locomotion.^25,26 One guideline, particularly when prescribing for adult patients, is to bias the axis towards the 90 or 180 meridians to reduce this distortion, limiting axis changes to less than 20° and altering the cylinder power accordingly to minimise residual astigmatism.²⁷ While these recommendations are found in the literature, they are less likely to impact the results reported in the current study. The expert examiner determined the cylinder axis by using the data point that instilled the most confidence. For example, keeping an oblique axis on those with definitive retinoscopy endpoints, or for individuals with strong subjective responses during axis determination. There was no intentional alteration of the findings to prescribe an axis closer to the 90 or 180 meridians for any of the clinical refractions determined here.

A last methodological difference to consider when discussing the difference between clinical and metric-optimised refractions is the binocular nature of the human visual system. When calculating VSX and PFSt for a participant’s right eye, there was no consideration by the software of the image quality metrics in the left eye. However, when prescribing clinically, binocular vision status may play a role in determining the final refraction. The cohort for this study often had poor binocularity due to strabismus and/or reduced visual acuity such that the examiner did not emphasise binocular balancing but instead tried to maximise visual acuity for each eye individually. However, there was still some consideration of binocularity, especially regarding the symmetry of astigmatism. There is a tendency for the two eyes to have mirror astigmatism (i.e., addition of the two axes equals 180°).²⁸ In cases where an individual had a strong preference and/or a crisper reflex for a particular axis in one eye, then the examiner would account for this when determining the astigmatic axis of the fellow eye, particularly if the findings from that eye were less conclusive.

Participant Considerations

There was a significant relationship (Figure 3) between the spherical equivalent refractive error and dioptric differences between metric-optimised and clinical refractions, with larger differences for participants with high levels of myopia. These differences were likely also based on the considerations previously discussed. The wavefront measures may be leveraging different information for eyes with higher amounts of myopia by sampling over the full extent of the pupil diameter, which is not the case for retinoscopy or autorefraction. There is also the retinoscopy technique to consider. For example in higher myopes, a closer working distance may be used on occasion to analyse the dimmer retinoscopy reflex better. While the expert examiner would estimate the correction factor needed for the closer working distance, there is more dioptric variability with a shorter working distance, and this may have affected the accuracy of the spherical measurement in these cases. Further, when considering how subjective refraction may be impacted by higher amounts of myopia, there are minification effects from higher powered minus lenses that are absent with lens powers closer to zero. Nevertheless, it should also be noted that while the refractions differed more in eyes with greater myopic refractive error, there was no consistent shift of the clinical refractions being more myopic than the metric-optimised findings, or vice versa, as seen in the M-values shown in Table 2.

While higher order aberrations were significantly correlated with dioptric differences, some variation was observed. For example, some individuals with RMS of 0.2 μm had < a 0.5D dioptric difference between results, while others with the same RMS had > 2D dioptric difference (Figure 2C). One potential explanation for this variability is the inherent aggregate nature of RMS. It could be that specific aberrations affect this dioptric difference more than others, or that the interactions between aberrations are affecting the resultant image quality in different ways.²⁹ For example, individuals with more favourable interactions of higher order aberrations may have better overall image quality, creating greater refractive certainty, whereas individuals with unfavourable interactions could have poorer image quality and greater refractive uncertainty.

The Vineland Adaptive Behavior Scales standard score had little predictive value on the dioptric differences between the refraction pairings. A weak relationship was observed between VSX and clinical refractions in the multivariable analysis (Table 3), with a higher standard score (representative of higher developmental functioning) being associated with smaller dioptric differences. However, given the small size of the effect, it seems unlikely that this is a meaningful factor when considering the dioptric differences observed.

It is important to note that while unilateral amblyopia was an exclusion criterion, neither bilateral amblyopia nor the amount of refractive error were reasons for exclusion. Patients with Down syndrome have higher rates of reduced vision compared with age-matched, normally developing children,^30,31 and the previously noted higher rates of significant refractive error and higher order aberrations are present from an early age,^4,32 likely resulting in the development of bilateral amblyopia. Thus, the visual acuity measured from the adults in the current study may well have been impacted by bilateral amblyopia, resulting in no clinically significant difference in the resulting visual acuity measured with the three types of refraction, even though there was a significant difference between the metric-optimised and the clinical refraction.

This lack of treatment effect could be a result of residual or previously untreated amblyopia where the “just noticeable difference” (JND) for the fine discrimination task of optotype acuity was greater in this population. Thus, the expected improvements in visual acuity with changes in refraction were not observed. Normally developed adults with residual amblyopia completing an orientation discrimination task had a JND 2 to 3x higher in their amblyopic eye when compared with the fellow eye when determining the tilt of short line (<4°) stimuli at various orientations.³³

When analysing data from the current study in the context of control eyes and those with keratoconus, it is evident that the refractive uncertainty that exists in individuals with Down syndrome lies between these two other groups. Additionally, while individuals with keratoconus may suffer from larger refractive uncertainty based on their the optics of their eyes, these patients are able to communicate more effectively with their eyecare practitioner as to the lenses they prefer. This occurs less frequently in individuals with Down syndrome and thus clinicians may be uncertain whether the lenses prescribed are optimal. These same practitioners may not have the time or expertise necessary to arrive at the same conclusions as our expert examiner, and with the uncertainty that clearly exists when prescribing for this population, may end up with a final prescription that is inadequate for the patient’s refractive error.

In summary, this study identified that the largest dioptric differences between refraction types occurred between metric-optimised refractions and clinical refractions. These differences were largest for individuals with increased levels of higher order aberrations, although overall spherical equivalent, as determined by the clinical refraction, also played a significant role in the refractive uncertainty in this population.

Key Points:

• Clinical refraction was significantly different from both metric-optimised refractions.

• Participants with greater amounts of higher order aberrations had larger differences between refraction types.

• Participants with greater amounts of myopic spherical equivalent refractive error had greater difference between refraction types.

Appendix 1.

Shown here are plots of the differences between the spherical equivalent (M) (A&B), J₀ (C&D), and J₄₅ (E&F) values for clinical versus metric-optimised refractions as a function of the clinical findings. The solid lines indicate the mean differences and the dashed lines represent the 95% limits of agreement (1.96 * standard deviation of the differences). PFSt, Pupil Fraction Tesselated refraction; VSX, Visual Strehl Ratio.