SUBJECTS AND METHODS

Informed consent was obtained from all subjects after the nature of the study had been fully explained. The tenets of the Declaration of Helsinki were followed and the study gained approval from the Flinders University ethics committee. Inclusion criteria were age 15 years or older, and normal healthy eyes with a VA better than 0.1 logMAR (6/7.5 Snellen equivalent) for the control group, or keratoconus (for the keratoconus group), or having undergone PK for keratoconus with an uncomplicated postoperative course (for example, no rejection, cataract development, etc) of at least 12 months (for the PK group). Keratoconus was defined by the staging system of Krumeich et al as being at least stage 1 (eccentric corneal steepening, induced myopia/astigmatism <5D, corneal radii <48D, Vogt’s striae, no scars).³¹ Subjects with probable mild keratoconus as suggested by a history of keratoconus or having had PK for keratoconus in the other eye, as well as abnormal retinoscopy but not meeting the criteria for stage 1, were included as a separate forme fruste keratoconus group.

Exclusion criteria were any ocular pathology (other than keratoconus for the keratoconus group) or abnormality such as amblyopia and strabismus, any previous ocular surgery (other than penetrating keratoplasty for the PK group), any neurological problem, systemic disease, or taking of any medication which may affect contrast sensitivity, inability to speak English sufficiently to be instructed to perform the tests, insufficient mental ability to perform the tests, and physical disability which would make it arduous to perform the tests (for example, wheelchair bound). Subjects with keratoconus and PK for keratoconus were drawn from the anterior segment clinic of the Department of Ophthalmology at Flinders Medical Centre (FMC) on a consecutive attendance basis. Control subjects were drawn from medical students and staff of FMC.

Data were collected on the visual performance of 11 eyes of 11 keratoconus subjects, six eyes of six forme fruste keratoconus subjects, 21 eyes of 21 PK subjects, and 24 eyes of 24 normal subjects. The subjects had a mean age of 36.4 (SD 9.8) years. Data were collected in a single session. All subjects were refracted and optimally corrected before data collection. For the subjects who normally wore contact lenses (keratoconus 8/11, PK 7/21, normal 2/24) refraction and optimal correction took place over their contact lenses. The measurements taken were logMAR high contrast visual acuity, LCVA, Pelli-Robson contrast sensitivity (PRCS), LCVA with glare, and PRCS with glare.

Both VA and LCVA were measured on a computerised monitor based system.³² The program utilises the psychophysical “staircase” method, using a forced choice protocol, to determine the acuity end point, which is taken as the average of 13 staircase reversals. This offers excellent reliability and validity and is free from learning effects. The program uses the same 5×4 letters used in Bailey-Lovie logMAR charts and the results are given in logMAR.³² Low contrast visual acuity is analogous to VA testing except that the target optotypes are reduced in contrast. Twenty five per cent (Weber) contrast optotypes were utilised since this has been reported to be the most suitable contrast level for detecting visual loss in early cataracts under glare conditions and thus may be suitable for keratoconus subjects.³³ Testing was conducted at 3.0 metres and the monitor had a maximum luminance of 185 cd/m². Low contrast visual acuity was also measured under glare conditions (LCVAglare).

There is no accepted standardised method for glare testing. In this study, LCVA and Pelli-Robson contrast sensitivity were measured with and without the presence of a light source directed at the subject. The glare source consisted of two projection lamps placed either side of the test chart and monitor. The baseline room illuminance was 80 lux. The luminance of the projector sources was 1800 cd/m². Natural pupils were used and care taken to ensure neither occlusion of the glare source nor macular photostress occurred.³⁴

The Pelli-Robson contrast sensitivity (PRCS) chart was used according to previously published methodology.^35,36 This test was chosen from among other available clinical charts for its superior validity and reliability^36,37 especially compared to sinusoidal grating charts such as the Vistech and FACT.³⁸ The use of a letter reading task for both CS and LCVA testing also allowed easy comparison of the two techniques. The Pelli-Robson chart was positioned 3 metres from the subject and had a luminance of 100 cd/m². The test was scored by the modified method of letter by letter scoring,³⁹ and results were given in units of log contrast sensitivity (logCS). Contrast sensitivity was also measured under the same glare conditions (PRCSglare) as LCVAglare.

The results are reported as raw measures and three derived measures of contrast and glare loss. Taking the simple arithmetic difference between glare and no glare, or high and low contrast demonstrates the effect of contrast or glare independent of baseline VA or CS. This facilitates comparison of groups of disease versus no disease where matching for VA is not possible. Low contrast loss (LCL) is defined as the difference between the high contrast VA and LCVA: LCL = LCVA − VA.^40,41 Low contrast visual acuity glare loss (GL_LVCA) is defined as the difference between the LCVAglare and LCVA: GL_LCVA = LCVAglare − LCVA.⁴⁰ Pelli-Robson contrast sensitivity glare loss (GL_PRCS) is defined as the difference between PRCS and PRCSglare: GL_PRCS = PRCS − PRCSglare.⁴⁰

One way analyses of variance (ANOVA) with Sheffé post hoc significance testing were used to compare groups. All statistical analyses were performed on the SPSS software package v10.0 (SPSS Inc, Chicago, IL, USA).

RESULTS

The results of visual performance of the normal control group and the study groups are listed in table 1. The three groups were similar for age (F_2,53 = 3.32; p >0.05). The normal controls performed better on all raw vision measures than both the keratoconus and PK groups. Similarly, the PK group performed better than the keratoconus group. However, as the groups were not matched for VA (F_2,48 = 30.60; p<0.001), they were also compared using three difference measures. Only two significantly different results were identified. The keratoconus group had a greater low contrast loss than the normal group (0.15 (0.12) v 0.05 (0.04), F_2,48 = 6.16; p<0.01, post hoc Sheffé p<0.05)—that is, the measurement of LCVA identified loss of vision in the keratoconus group not predicted by VA testing. Conversely, the keratoconus group had a smaller glare loss with the Pelli-Robson contrast sensitivity chart than the normal group (−0.16 (0.36) v 0.07 (0.11), F_2,49 = 4.23; p<0.05, post hoc Sheffé p<0.05)—that is, the vision in the normal group worsened under glare but improved in the keratoconus group. On all three difference measures the PK group was indistinguishable from the normal and keratoconus groups. Low contrast visual acuity and CS testing did identify visual loss in keratoconus and PK not appreciated with VA testing, but glare testing did not provide any such useful information.

To examine the influence of disease severity on vision loss in the keratoconus group, a forme fruste keratoconus group was also tested. This group was age similar (33.0 (13.0) years; p>0.05) and was also indistinguishable from normals on all measures of vision (VA −0.15 (0.04) p>0.05; LCVA −0.06 (0.08) p>0.05; LCVAglare −0.03 (0.09) p>0.05; PRCS 1.65 (0.12) p>0.05; PRCSglare 1.65 (0.15) p>0.05; LCL 0.10 (0.05) p>0.05; GL_LCVA 0.03 (0.09) p>0.05; GL_PRCS 0.00 (0.13) p>0.05).

DISCUSSION

The normal control group performed significantly better on all raw measures of vision compared to the keratoconus and PK groups. However, this chiefly reflects that the normal and study groups were not matched for VA. When adjusting for VA differences by using the difference measures low contrast loss and glare loss, the normal group was only significantly better on LCL (logMAR drop between VA and LCVA) and only compared to the keratoconus group (normal 0.05 (0.04), keratoconus 0.15 (0.12); p<0.05). The PK group was indistinguishable from both normal and keratoconus groups on these three difference measures. This suggests that VA is a very effective measure for distinguishing between normal, keratoconus, and PK subjects, but that CS and LCVA do provide some additional information.

Previous studies on keratoconus subjects have found reduced LCVA compared to normal subjects, but none of these studies controlled for VA by looking at LCL.^{9,14,19,23–25} However, deriving LCL from published VA and LCVA data suggests there is slightly greater LCL in keratoconus subjects compared to normals (at 10% but not 30% contrast),¹⁴ slightly more LCL in keratoconus subjects with apical corneal scarring compared to keratoconus subjects without scarring,^23,24 and greater LCL in more severe keratoconus.²⁵ While these reports support our findings, these differences were not tested statistically. Similarly, previous studies where CS was measured in keratoconus found reduced CS compared to normals, in some cases these losses were shown to be independent of VA.^11–22 Two of these studies measured PRCS in keratoconus.^21,22 Rose et al found smaller losses of PRCS than we did, which may have been because they reported binocular measurements, or because their keratoconus group were less impaired than ours, which is possible since they gave no definition of keratoconus, and all subjects were able to habitually wear their optimal correction.²¹ Conversely, Brahma et al’s group had much worse PRCS and VA than ours, but this was because their group were all presenting for surgery and were therefore unable to achieve satisfactory visual correction.²² This conclusion is supported by their VA and PRCS data for the other eye, which were better than our keratoconus data, although these data were a mix of 12 keratoconus and six PK eyes.²² Another possible reason for differences between studies is a difference in the relative proportions of those wearing RGP contact lenses, as these have been shown to improve contrast sensitivity in keratoconus compared to spectacle wearers.^13,42

Low contrast visual acuity data in PK subjects has previously been shown to include subnormal values in one series,¹⁹ and to be indistinguishable from keratoconus eyes in another.²⁸ Subnormal CS after PK for keratoconus has also been shown previously.^{15–17,19,26} Two longitudinal studies of keratoconus subjects before and after PK both found improved CS after PK.^22,27 Several cross sectional studies comparing PK to keratoconus eyes support this finding.^{16,17,26,29,43} Our data consolidate these findings, in that subjects with fully healed PK have CS better than keratoconus subjects, but not as good as normals, and LCVA losses which do not exceed normals. Those studies where PK subjects have been shown to have greater deficits in CS than keratoconus subjects were either in the early postoperative period,^27,28 or during a rejection episode.²⁹

The two glare loss measures did not yield any further vision losses in the keratoconus and PK groups, not identified with LCVA or PRCS testing alone. For GL_LCVA, all groups were similar, but for GL_PRCS, the normal group (0.07 (0.11)) was significantly worse than the keratoconus group (−0.16 (0.36); F_2,49 = 4.23; p<0.05, post hoc Sheffé p<0.05). While the keratoconus group did better than the normal group in terms of the change in CS under glare conditions the raw PRCS under glare scores were still worse in the keratoconus group compared to the normals (1.41 (0.21) v 1.73 (0.11); F_2,49 = 17.09; p<0.001, post hoc Sheffé p<0.001). That glare testing is of no value in either keratoconus or PK reflects that the main mechanism of visual loss in keratoconus and PK is optical aberration rather than light scatter.¹⁹ Our same experimental set-up finds large glare losses in cataract since the main mechanism of optical disturbance from cataract is forward light scatter.⁴⁰ If the main problem is aberrations, visual performance must be pupil dependent.⁴⁴ Therefore, the use of natural pupils, which will constrict in the presence of a bright glare source, reduces the chance of finding decreased visual performance under glare and, in this study, probably explains the significant increase in PRCS in the presence of glare for the keratoconus group. This raises the possibility that glare testing with a less bright glare source may be more useful in keratoconus. Alternatively, instead of testing vision under glare, perhaps low illumination testing, such as with the Smith Kettlewell Institute Low Luminance (SKILL) Card,⁴⁵ which could allow use of maximum natural pupil size and hence sample more aberrated optics, may be more appropriate in keratoconus.

The only longitudinal study of keratoconus subjects before and after PK (n = 18), in which glare testing was done (using the Brightness Acuity Tester (BAT) with VA and PRCS charts), found glare losses for keratoconus but not for PK.²² The reason for the difference to our results may be that the BAT glare source is not as bright as our projector glare source, thus may cause less pupillary constriction. Two small cross sectional studies (n = 6 and 7) also showed a trend toward CS deficit under glare for keratoconus subjects compared with normal subjects.^16,17 Conversely, two small studies (n = 4 and 5) found glare losses in PK greater than those seen in keratoconus.^15,16 These small studies may have included keratoconus cases with apical corneal scarring, which has been shown to cause forward light scatter and measurable glare loss.^24,30 Similarly, the PK series may have included cases with graft oedema, perhaps early in the postoperative recovery, which could also cause glare loss. Therefore, while glare testing is not useful for measuring vision in all cases of keratoconus or PK, it may still be useful in the presence of apical corneal scarring, or graft oedema—for example, in the first month after surgery²⁷ or after a rejection episode.²⁹

Defining the point along a spectrum from normal to disease state at which the transition from one entity to the other occurs is not a difficulty confined to keratoconus.⁴⁶ However, for keratoconus this difficulty has led to debate over the existence of unilateral keratoconus,⁴⁷ issues for study design,⁴⁸ and difficulty determining patient suitability for laser refractive surgery.⁴⁹ For this study we defined keratoconus according to an accepted, but conservative system.³¹ To illustrate that the inclusion of more mild cases of keratoconus could confound the results, we also tested a group of six forme fruste keratoconus subjects. This group were indistinguishable from the normal group on all vision measures. If we had included these subjects in the keratoconus group then the results of this study would have been different. This highlights the problem of sampling disease severity for cross sectional design studies. This problem could be avoided by using a longitudinal outcome study design. Indeed, the clinical impression that PK patients see better than keratoconus patients arises from comparing pre-PK keratoconus patients who want visual improvement to their post-PK state. Such pre-PK keratoconus subjects often proceed to surgery because they are unable to tolerate optimum refractive correction, so it is possible that this clinical impression is exaggerated by undercorrection of refractive error. Thus it is important in longitudinal study execution to ensure vision testing is conducted with optimal refraction in place.

While PK provides vision superior to keratoconus, it does not achieve levels of vision as good as normals. The pursuit of improvement in the surgical management of keratoconus requires vision testing to establish these improvements. Based on our findings, studies of treatments for keratoconus should include a measure of contrast vision, either PRCS or LCVA.