In the 21st century we are nearing the time when treatment of ocular disease is becoming a reality. As such, the ability to monitor disease progression and/or disease recovery is as important as the ability to detect disease related ocular change. Minimising measurement variability is intrinsic to monitoring accurately any chosen ocular-visual biomarker that best represents disease progression over time. This is the topic of the paper by Gilmore and co-workers in this issue of BJO (p 1462). These authors have described the measurement variability of the short wavelength (SW) automated perimetry (SWAP) in patients with diabetes. SWAP has been used primarily to detect vision loss and to monitor progressive visual field loss in glaucoma.¹ SWAP is more sensitive in detecting glaucomatous changes than standard white on white (WW) perimetry.² Gilmore and co-workers used the psychophysical frequency of seeing analysis as a measure of within examination variability, where the slope of the frequency of seeing curve represents threshold variability. Earlier Chauhan and co-workers³ integrated a psychophysical frequency of seeing paradigm into WW automated perimetry testing. They found that in patients with glaucoma and those with suspicion of glaucoma the variability in frequency of seeing was not necessarily explained by the response threshold or threshold deviation. Change in the slope of the frequency of seeing curve probably represents spontaneous change in threshold. Gilmore and colleagues find the higher estimate of within examination variability makes SWAP a less valuable biomarker in monitoring progression of certain ocular diseases. The increased sensitivity of SWAP has to be tempered with the higher within test variability of SWAP over standard WW perimetry.

Diabetic retinopathy is the single most common cause of blindness in the developed world.⁴ The prevalence of diabetes is increasing and potential blindness is threatening a rapidly growing number of working age individuals. Currently, there is a one in 300–500 chance for a child developing type 1 diabetes by 18–20 years of age.^5–7 Of these, 98% will show evidence of retinal microvascular changes 15–20 years after diagnosis.⁸

Current treatments for diabetic retinopathy are based on the extent of the retinal vasculopathy. The progression of vascular abnormality involves change from background, to pre-proliferative, to proliferative disease over time. The Diabetic Retinopathy Study (DRS)⁹ and the Early Treatment in Diabetic Retinopathy Study (ETDRS)¹⁰ set the current standard of care. Treatment for proliferative retinopathy involves panretinal laser photocoagulation (PRP) used to ablate the peripheral retina and laser ablative therapy given when high risk proliferative retinopathy develops.

In the future, neuroprotective therapies might conceivably delay onset of proliferative retinal change in diabetes

But, what if the neuronal component of the retina is compromised along with, or even before, the earliest retinal microvascular complications become apparent? This is an area receiving increasing attention and is the focus of collaboration between the Hospital for Sick Children and St Michael’s Hospital (Dr Shelley Boyd) in Toronto. Direct, non-invasive neuroretinal function testing of the human visual system demonstrates functional changes in the neuroretina of individuals with diabetes. Colour processing, in particular the processing of short wavelength stimulus, is abnormal in diabetes.^11–13 Adults with type 1 diabetes show reduced blue-yellow colour vision discrimination before the onset of retinopathy.¹⁴ The deficit in the short wavelength pathway was the focus of the study by Gilmore and colleagues, who tested frequency of seeing (FOS) areas of known decreased SW sensitivity. The importance of this SW sensitivity loss in diabetes is probably linked to the abnormal function of the SW cones. Yamamoto et al^15,16 demonstrated that the short wavelength (S) cones were compromised selectively in adults with type 1 diabetes. These changes were evident with or without evident retinal vasculopathy. A significant (p<0.001) selective reduction in the amplitude of the short wavelength cone response suggests a defect at the level of the S-cone photoreceptor.

Our group and others have found deficits in the integrity of the SW or S-cone pathway in patients with type 1 diabetes, with no evidence of retinopathy, using the colour visual evoked potentials (VEP).^17,18 The SW VEP latencies (time to respond to blue-yellow stimulus) of those with diabetes are delayed when compared with those without diabetes.¹⁷ We investigated the association between glucose control (HbA_1c) and colour vision in preteen (<12.9 years of age) children with type 1 diabetes using the colour VEP. Glucose control was well controlled and did not affect the S-cone pathway in this young group of children with diabetes. However, pubertal status was associated significantly (p = 0.0114) and selectively with S-VEP latency: pubertal children with type 1 diabetes had delayed S-VEP latencies (mean S-VEP latency = 144.3 ms) when compared with the pre-pubertal children with type 1 diabetes (mean S-VEP latency = 134.8 ms).¹⁸

An exciting development is the use of the multifocal electroretinogram (mfERG) to study multiple small regions of the retina individually.^19–27 This enables precise mapping of the neural retina. Multifocal ERG studies of adults with diabetes demonstrate clearly local deficits of retinal function.^19–27 Significant delays in the first order mfERG response, which is predominantly derived from bipolar cells,²⁸ were found.^20,22 Importantly, Han et al^20,21 found localised functional abnormalities that predicted the site of new vasculopathy (microaneurysms or leakage) observed 1 year later on clinical examination and confirmed by 50° fundus photography.

Very recently, high frequency components of mfERG recordings have been found to resemble oscillatory potentials of the full field ERG. Wu and Sutter²⁹ found that multifocal oscillatory potentials (mfOPs) were produced best by a flicker stimulus slowed by the insertion of dark stimulus frames. Bearse and colleagues²⁶ showed that the inner retinal mfOP response, derived from the slow flash mfERG recording, was abnormal in patients with diabetes. In addition response abnormalities were associated with the site of vascular leakage or haemorrhage.

Because diabetes is generally considered a disease of the retinal blood vessels current treatment paradigms are based on the extent of vascular damage and are typically given late in the disease, when significant end organ damage has already occurred. Perhaps the neuronal cells of the retina and visual pathways of the brain are damaged as well as retinal blood vessels in diabetes. In particular, our data show that abnormalities of the koniocellular pathway occur in adolescent children with diabetes in the absence of observable changes in the retinal vasculature. In the future, neuroprotective therapies might conceivably delay onset of proliferative retinal change in diabetes. In streptozotocin induced diabetic rats, neurons and glial cells in the inner plexiform and nuclear layers of the retina undergo apoptosis early in the course of diabetes³⁰ and actually precede the development of microvascular lesions.³¹

For studies investigating treatment paradigms it is essential for the chosen biomarkers to assess accurately ocular function as well as reproducible change. Gilmore and colleagues have set an excellent precedent in investigating variability in responses in paradigms already shown to be sensitive.