To characterize defects in color vision in patients with choroideremia.
Prospective cohort study.
Thirty patients with choroideremia (41 eyes) and 10 age-matched male controls (19 eyes) with visual acuity of ≥6/36 attending outpatient clinics in Oxford Eye Hospital underwent color vision testing with the Farnsworth-Munsell 100 hue test, visual acuity testing, and autofluorescence imaging. To exclude changes caused by degeneration of the fovea, a subgroup of 14 patients with a visual acuity ≥6/6 was analyzed. Calculated color vision total error scores were compared between the groups and related to a range of factors using a random-effects model.
Mean color vision total error scores were 120 (95% confidence interval [CI] 92, 156) in the ≥6/6 choroideremia group, 206 (95% CI 161, 266) in the <6/6 visual acuity choroideremia group, and 47 (95% CI 32, 69) in the control group. Covariate analysis showed a significant difference in color vision total error score between the groups ( P < .001 between each group).
Patients with choroideremia have a functional defect in color vision compared with age-matched controls. The color vision defect deteriorates as the degeneration encroaches on the fovea. The presence of an early functional defect in color vision provides a useful biomarker against which to assess successful gene transfer in gene therapy trials.
Choroideremia is a progressive X-linked inherited retinal degeneration affecting the retinal pigment epithelium (RPE), choroid, and outer retina. Early symptoms include nyctalopia, reduction in peripheral vision, and loss of visual acuity (VA), resulting in legal blindness around the third to fourth decade. Female carriers present with variable retinal involvement, often in the absence of symptoms.
The disease is caused by mutation of the CHM gene, which codes for Rab escort protein-1 (REP1), which has a key role in protein prenylation. This is an important intermediate step for the protein interactions that govern cell signaling and intracellular regulatory pathways. In particular, the prenylation of the Rab27a protein is reduced by the lack of REP1. Rab27a is a retina specific protein, thought to affect melanosome transport and localization in the RPE and choroid. A layer-specific chm knockout mouse model demonstrates degeneration may occur in the RPE and, to a lesser extent, photoreceptors in the absence of REP1. REP1 is expressed ubiquitously throughout the retina, including the cone photoreceptors (MacDonald IM, et al. IOVS 2005; 46: ARVO E-Abstract 540).
Since loss of the RPE and choroid will invariably lead to secondary death of the overlying photoreceptors, it can be difficult to isolate independent functional defects in these cells. Reports in younger choroideremia patients and animal models suggest degeneration may occur in the choroid, RPE, and photoreceptors independently. Functional impairment of the rod photoreceptors has been identified in the chm knockout mouse. Choroideremia patients can maintain excellent foveal VA until very late in the disease process ; however, this does not exclude the presence of a color vision defect. Since the choroideremia gene is also expressed in cones and these cells are dependent on surrounding retinal cells that also express REP1 as part of the cone visual cycle, it would not be unexpected to find subtle defects in color vision in these patients before the onset of degeneration.
Patients were assessed as part of an ongoing choroideremia gene therapy clinical trial, approved by the National Research Ethics Service Committee for West London & GTAC ( ClinicalTrials.gov identifier NCT01461213 ) and conducted in accordance with the Declaration of Helsinki. Patients with a clinical diagnosis of choroideremia, referred for consideration of inclusion in the gene therapy project, attending outpatient clinics at Oxford Eye Hospital were included in the analysis. The ethics approval covers necessary screening undertaken to assess suitability for gene therapy, to which all patients consented. The diagnosis of choroideremia was confirmed by genetic testing, which was approved by the National Research Ethics Service Committee for Essex 2 (Reference 08/H0302/96), and all patients were consented.
Color vision testing took place at the initial assessment visit under a fixed lighting booth with Standard Illuminant C (correlated color temperature 6774K). VA measurements were undertaken for distance (6 m). Individualized microperimetry (Maia; CenterVue, Padova, Italy) was undertaken to quantify central visual field loss by using a staircase threshold testing algorithm centered around foveal fixation. A new 100 hue set with undisturbed color caps was used (manufactured by DG Color, Wiltshire, UK) and testing was conducted in accordance with previously published guidelines. Briefly, this involved presenting the patient with 4 boxes of colored caps in succession to each eye separately. The caps at each end were fixed and the patient was asked to arrange the remaining color caps to resemble a smooth progression from one end to the other. The order of the caps was recorded on an Excel spreadsheet (Microsoft, Redmond, Washington, USA), adapted from the manufacturer’s results recording software. An error score was calculated for each cap and summed to calculate the total error score (color vision total error score), Autofluorescence (both normalized and nonnormalized images for a 30 degree and 55 degree field) images were collected for each patient. The normalization process integrated in the Heidelberg image capture software was used to enhance the contrast of the images obtained.
Ten age- and sex-matched control subjects were also tested. In order to differentiate the effects of reduced visual field on the ability to carry out the 100 hue test from functional loss, the control subjects repeated the test with a simulated field constriction using restricted-view spectacles and obscuring part of the colored caps. This was done to mimic a visual field defect similar to that experienced by patients with choroideremia, which can reduce the visibility of the caps. This allows differentiation of a defect due to function or visual field. This is a way of answering the concerns raised by previous work with regard to a color vision defect being due to a field defect rather than degeneration.
In order to exclude the obvious effects of foveal degeneration on impaired color discrimination, only patients with normal foveal appearance as shown by the presence of hypofluorescence from the macular pigment surrounded by autofluorescence were included ( Supplemental Figure 1 , available at AJO.com ). The position of the fovea is visible on the AF image in Supplemental Figure 1 owing to the presence of hypofluorescence surrounded by autofluorescence. This indicates the presence of macular pigment, which binds to foveal cones. Patients were divided into 2 subgroups: a group with preserved VA (≥6/6) and a group with reduced VA (<6/6). All patients had a VA of 6/36 or better to ensure 100 hue testing could be accurately undertaken.
Forty-one eyes from 25 patients with choroideremia were included. Nine patients were tested in 1 eye only because their VA and autofluorescence did not meet the inclusion criteria of VA sufficient to perform the test accurately (6/36 or better) and a normal-looking fovea on imaging. Eleven patients had vision in both eyes that fit into a single subgroup. Five patients had 1 eye in each subgroup. In addition, color vision testing was conducted in 19 eyes from 10 age-matched normal male subjects.
The analyses were performed using Stata version 13 (StataCorp LP, College Station, Texas, USA).
These data have a hierarchical structure of eyes within subjects. The unit of analysis was an eye. All analyses of these data needed to accommodate the dependence of measurements on eyes within subjects. We used a random-effects model, which takes into account the hierarchical structure of the data. The factor describing the group to which an eye belongs (VA <6/6, VA ≥6/6, or control) was a fixed effect in the analysis of variance model, with age added as a covariate. The Shapiro-Wilk test was used to test for normality within each group (defined by VA and eye) to ensure the independence of observations. Data for the continuous measurements were transformed by taking logarithms before analysis if the distribution was markedly skewed.
Age-adjusted means and 95% confidence intervals were calculated for color vision total error score, and for the red, blue, and green error scores. The statistical significance of pairwise comparisons between the 2 groups defined by VA and the controls was reported using a 2-sided test without correction for multiple comparisons. The test statistic was the z-statistic, derived from the coefficients of the terms in the random-effects model.
Reduced-field color vision total error score was measured for a sample of controls. Analyses of these data needed to accommodate the dependence of repeated measurements on the same control subjects. The difference between reduced and full field was analyzed using a random-effects model of the variable color vision total error score. The factor describing whether the measurement was a full-field or a reduced-field assessment was included in the model as a fixed effect. The test statistic was the z-statistic, derived from the coefficients of the term for this factor in the random-effects model.
In order to assess the impact on color vision of encroachment of the degeneration onto the fovea, a factor with 3 levels was derived from the shortest distance to the edge of the healthy retinal island as shown in supplemental Figure 2 . The 3 levels were <400 μm, 400–800 μm, and >800 μm, identified using autofluorescence on the Heidelberg scanning laser ophthalmoscope (Heidelberg Engineering GmbH, Heidelberg, Germany). Analysis of color vision total error score for the patient group was analyzed using a random-effects model with this 3-level factor as a fixed effect in the model. All pairwise comparisons between mean color vision total error score for the 3 levels were calculated with P values. The color vision total error score is log transformed for analysis. Back-transforming the pairwise mean differences results in a ratio of mean color vision total error score between classes, and the P value refers to a 2-sided test using the z-statistic derived from the coefficients of the terms for this factor, of whether this ratio is significantly different from 1.
To investigate the association of cone function in general, microperimetry threshold was compared with color vision total error score after adjustment for age and VA. Microperimetry was added to the random-effects model as a covariate. A similar analysis was carried out for area of remaining retinal tissue. Analysis of variance (ANOVA) was used to investigate the effect of genetic mutation on color vision total error scores.
The order of the caps was recorded and partial error scores calculated at each position ( Figure 1 ).
The total error score was calculated from the partial error scores. Age adjustment was made by including the covariate age in the analysis to control for the change in test performance due to age. Descriptive statistics are shown in Tables 1 and 2 .
|Choroideremia Subgroup VA <6/6, n = 19||Choroideremia Subgroup VA ≥6/6, n = 22||Control Group, n = 19||VA ≥6/6 vs VA <6/6||Control vs VA ≥6/6||Control vs VA <6/6|
|Mean (95% CI)||Mean (95% CI)||Mean (95% CI)||P Value||P Value||P Value|
|Distance VA range||6/7.5–6/36||6/4–6/6||6/4–6/6|
|Near VA range||N4.5–N9||N4.5||N4.5–N5|
|Color vision total error score||206.4 (160.5, 265.6)||120.2 (92.3, 156.3)||47.3 (32.2, 69.4)||.001||<.0001||<.0001|
|Microperimetry threshold (dB)||13.43 (11.14, 15.72)||18.65 (16.25, 21.05)||<.0001|
|Shortest distance from foveal center to edge of retinal island (μm)||359.2 (272.6, 473.4)||694.4 (521.7, 925.2)||<.0001|
|Area of central retinal island (mm 2 )||8.09 (5.64, 11.61)||8.15 (5.64, 11.79)||.95|
|Red error score (caps 1–5)||14.16 (9.39, 18.88)||8.63 (3.68, 13.57)||2.03 (0.67, 3.39)||.09||.03||<.0001|
|Green error score (caps 36–40)||16.05 (13.06, 19.03)||7.42 (4.33, 10.51)||4.99 (3.05, 6.92)||.002||.22||<.0001|
|Blue error score (caps 55–59)||18.59 (14.00, 23.19)||8.24 (3.51, 12.98)||4.66 (2.60, 6.73)||<.0001||.26||<.0001|
A significant difference in color vision total error score between each of the subgroups of choroideremia eyes (mean error score of 206 in the group with VA <6/6; mean error score of 120 in the group with VA ≥6/6) and the control eyes (mean error score of 47) was seen ( P < .0001). The choroideremia eyes with reduced VA (<6/6) had a significantly higher color vision total error score than the choroideremia eyes with preserved VA (≥6/6, P = .001).
Having identified the presence of a baseline color vision defect in choroideremia patients, even before the onset of foveal degeneration, we were keen to identify if any one cone class was preferentially affected.
When the means of these scores were plotted, however, no identifiable major axis was visible as shown in Figure 2 . Hence the observations supported a general reduction in color vision across the spectrum, which deteriorates further as VA falls later in the disease (≤6/6).
In order to explore this further, the relationship between the 100 hue and CIE diagram was assessed, because groups of caps can be analyzed in subgroups to represent pure red, green, and blue responses. This likely represents maximal stimulation of the L, M, and S cones, respectively, determined by the wavelength represented by these colored caps. At all 3 points, the choroideremia eyes with reduced VA show a significant reduction in color vision compared with controls ( P < .0001). The choroideremia eyes with preserved VA had reduced sensitivity for red ( P = .03) but not for green or blue compared with controls ( P > .20).
Choroideremia is also associated with a reduction in the visual field. In order to rule out reduced field of vision as a cause for the color vision defects seen, the normal control subjects repeated 100 hue color vision testing wearing simulation spectacles with a markedly reduced field of view. In addition, half of each cap was covered over to reduce the amount of color visible. The simulated field reduction in the control subjects was able to assess the role of visual field reduction in the measurement of color vision loss. The difference between the full and reduced field for color vision total error score was analyzed using a random-effects model, and on the log transformed scale was estimated as 0.28 (95% CI -0.93, 0.66, P = .14). This indicated that the color vision results obtained with a reduced field were not significantly worse than testing in full visual field conditions.
To determine if the observed color vision defect is truly functional rather than occurring secondary to retinal degeneration, we compared color vision scores to other parameters of retinal health. Microperimetry was undertaken and central threshold recorded as a measure of foveal function. Further analyses of the color vision total error score were carried out with age and VA as covariate factors. Microperimetry threshold and area of remaining retina were included in this model as separate possible factors. Area of remaining retinal island tissue was not a statistically significant covariate; however, microperimetry threshold was a statistically significant factor relating to the color vision total error score ( P = .02).
In order to relate the results of the analysis of color vision total error score to encroachment of the degeneration onto the fovea, the distance data were split into 3 groups. Using adaptive optics in healthy human subjects, the pit diameter has been measured at 1.85 ± 0.23 mm (radius 925 μm). Therefore 3 levels for distance from the foveal center were chosen: 400 μm to represent a region well within the foveal pit, 400–800 μm to represent a region involving the foveal pit in most patients, and >800 μm to represent a region completely outside the foveal pit in most individuals. These measurements therefore provided a proxy for the likelihood of the fovea being affected directly by the degeneration. The 3-level covariates describing the shortest distance to the edge of the retinal island are reported in Table 3 . The covariates age and VA were included in the model. Age was not a statistically significant factor for explaining color vision total error scores, but VA was significant in the mathematical model ( P = .024). The mean color vision total error score ratio between classes, for all pairwise comparisons, is reported in Table 4 , with the 95% CI and P value for the comparison of the ratio with 1.
|Class||VA <6/6||VA ≥6/6|
|Mean Color Vision Total Error Score||95% CI||Mean Color Vision Total Error Score||95% CI|
|Comparison Between Classes||Mean Color Vision Total Error Score Ratio Between Classes (95% Confidence Interval), P Value|
|Distance <400 μm compared with distance 400–800 μm||1.27 (0.89, 1.81), P = .19|
|Distance <400 μm compared with distance >800 μm||1.50 (0.95, 2.35), P = .08|
|Distance 400–800 μm compared with distance >800 μm||1.18 (0.56, 1.28), P = .43|
There is an increase in the variability of the color vision scores when the remaining central island of healthy retina falls below 20 mm 2 or the edge of the island of healthy retina comes closer than 400 μm to the fovea, as shown in Figure 3 . This may be an early signal of functional cone loss.