Straylight, lens yellowing and aberrations of eyes in Type 1 diabetes.

Straylight, lens yellowing and ocular aberrations were assessed in a group of people with type 1 diabetes and in an age matched control group. Most of the former had low levels of neuropathy. Relative to the control group, the type 1 diabetes group demonstrated greater straylight, greater lens yellowing, and differences in some higher-order aberration co-efficients without significant increase in root-mean-square higher-order aberrations. Differences between groups did not increase significantly with age. The results are similar to the findings for ocular biometry reported previously for this group of participants, and suggest that age-related changes in the optics of the eyes of people with well-controlled diabetes need not be accelerated.


Introduction
We have previously reported ocular biometry and amplitude of accommodation of a group of people with type 1 diabetes (DM1) in comparison with controls [1,2]. Generally, the eyes of people with DM1 behaved liked aged versions of eyes of people without diabetes, although there were no accelerated rates of change of parameters with age compared with the controls. We consider that this is because the group has low levels of neuropathy and their diabetes is generally well controlled. Here, we report other parameters related to the optics of the eye in type 1 diabetes: straylight, lens yellowing and aberrations. Retinal image quality depends upon forward light scatter, aberrations and diffraction. Forward light scatter, also referred to as straylight, is responsible for a veiling light over retina and reduction of retinal contrast. Straylight is responsible for the skirt of the point spread function beyond approximately 60 min arc, and affects visual functions such as contrast sensitivity, visual fields and pattern electroretinograms [3]. Both psychophysical and optical methods have been used to measure forward light scatter, with the former being particularly meaningful because of their dependency on participant perception [4]. Straylight increases with age [5,6] and in cataract [7][8][9] and some other pathologies [10][11][12]. Objective corneal scatter have been measured, with two studies [13,14] but not a third [15] finding increases in people with diabetes compared with controls. To the best of our knowledge, influence of diabetes on straylight has not been investigated.
Age-related decrease in ocular transmission at short wavelengths (lens yellowing), as determined by heterochromatic flicker photometry, has been reported to be greater for people with diabetes than for controls [16,17].
It has been found that people with diabetes have greater higher-order root-mean-squared aberrations (HORMS) than controls at fixed pupil sizes [18,19], although it should be appreciated that such differences will be reduced under natural conditions because people with diabetes usually have smaller pupils sizes than those without diabetes [20][21][22][23][24][25]. Calvo-Maroto et al. [26] reported total, corneal and internal higher-order aberrations in small groups of people with D-M1 and DM2 (type 2 diabetes). The aberrations were very high for 5 mm pupils, with total HORMS 0.63 ± 0.23 µm and 0.53 ± 0.25 µm reported for the DM1 and DM2 groups, respectively. However, there were no significant differences between the groups and there was no control group. Additionally, it is not clear that the same reference axis was used for total and corneal aberrations.

Participants
Ethical approval was obtained from the Queensland University of Technology before the commencement of the study, and all participants provided written informed consent. The study adhered to the tenets of the declaration of Helsinki. The full study had 74 participants (mean ± SD, 40 ± 12 years) with DM1 and 64 age-matched control participants (mean ± SD, 43 ± 12 years). Most participants were part of the Longitudinal Assessment of Neuropathy in Diabetes using novel ophthalmic Markers (LANDMark) study [27] at the Institute of Health and Biomedical Innovation, for which the DM1 participants tended to have low levels of neuropathy, retinopathy and nephropathy, and their diabetes was well controlled. Details of clinical tests, selection of eye, and exclusion criteria are given elsewhere [2].
There were different numbers of participants for the tests (Tables 1-3). Because the C-Quant was not available at the start of the study, straylight was not measured for 11 and 7 of the diabetic and non-diabetes participants, respectively, giving 63 and 57 subjects. Due to limited time availability, the Medmont corneal topographer was not used with 39 participants, and a further 3 diabetic and 2 control participants were excluded from analyses because they had poor images. With a further 1 diabetic participant excluded because of a pupil size at the aberrometer < 4.5 mm, aberrations were determined for 46 diabetic and 47 control participants. Thirty-eight diabetic and 45 control participants attempted the lens yellowing experiment, but 8 of the former and 4 of the latter were not able to appreciate the flickering.

Data analysis
Analyses have already been described [2]. They included Kolmogorov-Smirnov and Shapiro-Wilk tests to determine normality of data, the Student unpaired t-test and the chi-square test to determine significance of differences between diabetes and control groups, multiple regression analysis to model influences of age and diabetes duration, linear regression fits of a parameter as a function of age, and ANCOVA analysis to determine the significance of differences in linear regression slopes between diabetes and control groups. Data for all parameters were normally distributed except for straylight and the total and corneal spherical aberrations co-efficients. Residuals of the straylight were normally distributed. For consistency with other parameters, total and corneal spherical aberrations co-efficient data were treated as having normal distributions; the statistical tests used are suffictently robust to account for departures from normality.
The multiple regression analysis was performed for the whole group, with duration for the participants without diabetes given as zero, and for only the participants with diabetes. Gender and/or axial length were included as additional factors where these were significant according to simple linear regression, but they were not significant factors in any of the multiple regressions.

Straylight
Straylight was measured with the C-Quant (Oculus Optikgerte, Wetzler, Germany) that uses a compensation comparison [28][29][30][31][32]. The instrument has a central test field, which is divided into right and left half fields, and an outer ring providing a straylight source at 7 degree eccentricity. When the outer ring flashes on, scatter in the eye causes some of the light to reach the fovea. This light is perceived as a weak flickering in the central test field. In only one of the two half fields, a variable counterphase compensated flicker is introduced. This results in two types of flicker with different depth modulations in the central test field. One flicker is a combination of straylight and compensated light in one half field, and the other flicker is due to straylight in the other half field. The amount of light in the compensated half field can be varied. In a series of trials the participant decides which half field has the stronger flicker. The instrument determines the amount of light at which no flickering is observed in the compensated field, and this is the measure of ocular straylight. After familiarizing participants with the instrument, five measures were taken and averaged.

Lens yellowing
In heterochromatic flicker, two sources of different color illuminate the same area in the visual field alternatively, and the intensity of one of these sources is adjusted until the perception of flicker is eliminated or minimised. The lens is relatively non-absorbing for 550 nm light, so an elevated threshold for a 420 nm stimulus relative to that of a 550 nm stimulus to produce this state is a measure of 420 nm light absorbed by the lens and hence of lens yellowing. Lens yellowing was measured under scotopic conditions. Light of 420 nm and 550 nm was obtained with 430 nm blue and 565 nm green light emitting diodes (LEDs) combined with 420 ± 10 nm and 550 ± 10 nm interference filters, respectively. The LEDs were flashed alternatively at 5 Hz. The light from the sources was combined through a cube beam splitter and subtended 1.0 • . This was placed 7 • nasally from a dim red LED fixation target. Pilot tests determined the optimum flicker rate (5 Hz) and pulse width modulation (effectively luminance variation) per rotation of a knob controlling the blue LED. The pulse width modulation of the green LED was set to 10, and the lens yellowing was specified by blue LED pulse width modulation e.g. pulse width modulation of 10 gave a lens yellowing of 10. It was not possible to measure the light output of the system, and hence the lens yellowing obtained was a relative measure.
At the start of an experiment, each participant was shown the apparatus and a brief description of the experiment was given. The non-tested eye was covered with an eye-patch. Each participant was dark-adapted for 25 minutes. The participant put his/her chin on the chin rest with the eye level with the light sources, and looked at the fixation target while observing the flickering of the LEDs. The participant rotated the control knob to alter the luminance of the blue LED. To avoid Troxler's phenomenon [33], participants were advised to look at the roof of the room for short intervals of time during testing. Participants determining the limits of the range of luminance settings for which no flicker was perceived. The limits were averaged, and the mean of ten such sets was taken as the lens yellowing.

Total, corneal and internal aberrations
Total (ocular) aberrations were obtained from the COAS-HD aberrometer (Wavefront Sciences, Albuquerque, New Mexico). Three readings for a 4.5 mm pupil were taken with reference to the corneal plane and 555 nm wavelength, with aberration co-efficients specified up to the 6th order of Zernike polynomial coefficients in the OSA/ISO referencing system [34]. Along with the root-mean-square higher-order aberration (HORMS) determined from the 3 rd to 6 th aberration orders, we considered spherical aberration (C 0 4 co-efficient), vertical coma (C −1 3 co-efficient), and horizontal coma (C 1 3 co-efficient) because their co-efficients are usually greater than other higher-order aberrations co-efficients. To allow for mirror symmetry when combining right and left eye data, signs of vertical coma co-efficients were changed for left eyes.
Corneal aberrations at 4.5 mm diameter were determined from corneal topography measured with a Medmont E-300 placido disk type anterior corneal topographer. At least two images, having instrument alignment quality specifications ≥ 95%, were used. Raytracing through a model anterior cornea was performed using the Optics Studio 14.2 program (Radiant Zemax, Redmond, USA) with decentrations and tilts of the cornea included in order to reference it to the line-of-sight for aberrometry rather than the keratometric axis. This procedure has been described previously [35], except that this time the cornea was modeled as a "grid sag" surface rather than as a set of Zernike co-efficients. After raytracing, aberration co-efficients were averaged across the readings. As was done for ocular aberrations, signs of vertical coma coefficients were changed for left eyes and HORMS was calculated from co-efficients. Internal aberration co-efficients were determined by subtracting corneal aberration coefficients from the total coefficients, and internal HORMS was calculated.

Results
Characteristics of groups are given in the first few lines of Tables 1-3. There is no significant difference between groups for either spherical equivalent refraction or corrected visual acuity. Table 4 shows multiple regression fits for straylight, lens yellowing and aberrations, both for the whole group and for the diabetes group alone. Figure 1 shows straylight as a function of age for the two groups. For each group there is a significant positive age-related slope (+0.006 and +0.007 log(s)/year for diabetes and control groups, respectively), but the slope difference between groups is not significant. Diabetes duration and age contribute significantly to the multiple regression fits, with diabetes having the greater importance (Table 4). The diabetes group has significantly greater straylight than the control group (mean difference ± 95% CI = +0.12 ± 0.06 log(s), p < 0.001).  Figure 2 (a) shows the log of lens yellowing as a function of age for the two groups. For each group there is a significant positive age-related slope (+0.019 and +0.018 log/year for diabetes and control groups, respectively), but the slope difference between groups is not significant. Diabetes duration and age contribute significantly to the multiple regression fit for the whole group, but diabetes duration is omitted for the diabetes group (Table 4). The diabetes group has significantly greater log lens yellowing than the control group (mean difference ± 95% CI = +0.23 ± 0.15 log, p < 0.001). Figure 3 shows total higher-order aberrations as a function of age for the two groups. Significant age-related slopes occur only for the vertical coma co-efficient (−0.003 and −0.001 µm/year for diabetes and control groups, respectively). The slope difference between groups is not significant for any aberration, although it is close in the case of horizontal coma (F 1,89 = 3.04, p 0.09). Diabetes duration contributes significantly to the multiple regression fits for all four aberration terms, but age contributes significantly only for vertical coma (Table 4). There are significant differences between means of groups for horizontal coma and vertical coma coefficients only (mean difference ± 95% CI = −0.039 ± 0.037 µm (p = 0.04), and −0.040 ± 0.030 µm (p = 0.01), respectively), but only for vertical coma does the difference match the ageing trend (Fig. 3(b)). For corneal higher-order aberrations there are no significant changes with age for any combination of group and aberration, and rates of change with age between groups are not significantly different. Neither diabetes duration nor age contribute significantly to multiple regression fits for any terms (Table 4). There is no significant difference between means of groups for any aberration term.  [17]. Vertical displacement of data between studies should be disregarded.
For internal higher-order there are no significant changes with age for any combination of group and aberration, and rates of change with age between groups are not significantly different. Diabetes duration contributes significantly to multiple regression fits for vertical coma and spherical aberration, but age does not contribute to any of the aberrations (Table 4). There is no significant difference between means of groups for any aberration term.
In the multiple regression fits shown in Table 4, duration of diabetes is a significant contributor to the variation in 7 parameters for the whole group and 6 parameters for the diabetes group. For the whole group, this could happen because it is merely a proxy for having diabetes. To investigate this, diabetes duration was replaced by diabetes status in regressions. For two parameters, total HORMS and internal coma, swapping from diabetes duration to diabetes status means that diabetes is lost as a significant factor. For the other 5 parameters, diabetes is retained as a significant factor, but with some reduction of the fit quality (i.e. R 2 reduced). In the 3 cases where age is a significant factor (straylight, lens yellowing and total vertical coma) there are compensatory increases in its co-efficient of up to 40%. These findings indicate that diabetes duration provides more information than the presence of diabetes alone, and is worthwhile including in regression analyses.

Discussion
Previously, we have reported differences in ocular biometric parameters between a group of DM1 participants and an age-balanced control group [2]. In the further tests reported here, we also find differences such that the eyes of diabetic people looked like aged variations of normal eyes. This occurs for straylight (Fig. 1), lens yellowing (Fig. 2), ocular horizontal coma ( Fig.  3(a)) and ocular vertical coma (Fig. 3(b)). Diabetes duration is a significant predictor for 8 parameters: straylight, lens yellowing, total horizontal coma, total and internal vertical coma, total and internal spherical aberration, and total HORMS. However, as for the biometry, there are no significant group differences in rates of change of parameters with age.
We have no comparisons with previous work for straylight, but can compare our findings with those of two studies for lens yellowing (Fig. 2(b)). Similar to this study, Davies & Morland [17] found a significant effect of age but not of diabetes duration on lens yellowing in their diabetes group. However, contrary to this study, both Davies & Moreland [17] and Lutze & Bresnick [16] found greater rates of increase of lens yellowing with age in diabetes. We believe that the difference in trends is likely due to the mildness of problems in our diabetes group. As mentioned above, the only aberrations with significant differences between the groups are total horizontal coma and total vertical coma. Table 3 indicates that the main components responsible for these are the internal component for horizontal coma and the anterior cornea for vertical coma, but the differences between groups for these components are not significant (p = 0.13 and 0.11, respectively). The significant age-related changes in ocular vertical comas for both groups (Fig. 3(b)) might be related to changes in the pupil centration with age, but previously we did not find significant age-related changes with these for either group [2]. It is generally understood that the aberrations of the eye increase with age, as represented by HORMS, although such changes were not marked in the few studies in which there was control for the refraction distribution [36,37]. If the eyes of diabetic people behave like older eyes, it would be expected that ocular aberrations would be higher in the diabetes group than in the control group with the internal contribution determined by the lens also increasing. Power analysis with = 0.05 and power = 0.80 indicate that a large number of 225 participants/group would be needed to show significance for HORMS.
Shahidi et al. [19] and Valeshabad et al. [18] reported greater total HORMS in diabetic groups than in control groups for 6 mm and 5 mm pupils, respectively. Diabetic type was not specified, but unlike our study all people in the diabetes groups had retinopathy. The smaller pupil size of 4.5 mm was used for this study as it was large enough to assess all but one participant. It is possible that we would have obtained more significant differences at larger pupils. We did not dilate pupils because we considered that there might be interactions between natural pupil size, pupil center and aberrations. In fact the difference in pupil size between the groups is not significant, although there is a tendency for the pupils of eyes of diabetic people to be smaller, and there is no tendency for different pupil centrations (relative to the limbus center) in the groups [2].
The increases in straylight and lens yellowing with diabetes can be attributed in part to the greater lens center thickness that occurs with diabetes as shown in Tables 2 and 3 and found in other studies [2,38,39]. Linear correlations of lens thickness with straylight (R 2 = 0.24) and with log lens yellowing (R 2 = 0.38) are highly significant (p < 0.001 for both). Lens thickness was not included in multiple regressions because of its dependence on age.