Article Text

Download PDFPDF

Original article
Inner retinal visual dysfunction is a sensitive marker of non-proliferative diabetic retinopathy
  1. Gregory R Jackson1,
  2. Ingrid U Scott1,
  3. David A Quillen1,
  4. Laura E Walter1,
  5. Thomas W Gardner2
  1. 1Penn State Hershey Eye Center, Penn State College of Medicine, Hershey, Pennsylvania, USA
  2. 2Kellogg Eye Institute, University of Michigan, Ann Arbor, Michigan, USA
  1. Correspondence to Dr Gregory R Jackson, 500 University Drive, HU19, Penn State College of Medicine, Hershey, PA 17033, USA; gjackson{at}


Aims To determine the effect of diabetes on inner and outer retinal function in persons with diabetes and no clinically detectable retinopathy or with non-proliferative diabetic retinopathy (NPDR).

Methods Visual function was assessed in 18 adults with normal retinal health, 23 adults with diabetes and 35 adults with NPDR and normal visual acuity. Contrast sensitivity and frequency doubling technology (FDT) sensitivity were used to assess ganglion cell function. Acuity, dark adaptation, light-adapted visual sensitivity and dark-adapted visual sensitivity were measured to evaluate cone and rod photoreceptor visual function. The presence and severity of diabetic retinopathy was determined by grading of 7-field stereoscopic fundus photographs using the Early Treatment Diabetic Retinopathy Study grading system.

Results Participants with NPDR exhibited impairment of all measured visual functions in comparison with the normal participants. Inner retinal function measured by FDT perimetry was the most impaired visual function for patients with NPDR, with 83% of patients exhibiting clinically significant impairment. Rod photoreceptor function was grossly impaired, with almost half of the patients with NPDR exhibiting significantly impaired dark-adapted visual sensitivity.

Conclusion Both inner retinal and outer retinal functions exhibited impairment related to NPDR. FDT perimetry and other visual function tests reveal an expanded range of diabetes induced retinal damage even in patients with good visual acuity.

  • Psychophysics
  • diagnostic tests/investigation

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Diabetic retinopathy (DR) is commonly viewed as a microvascular complication of diabetes mellitus which is defined by the presence of vascular lesions visible on the fundus on clinical examination. In addition to vascular dysfunction, DR causes neurodegeneration of the retina.1–3 Previous reports have shown that patients with diabetes and mild non-proliferative diabetic retinopathy (NPDR) exhibit a wide variety of visual function abnormalities, including loss of contrast sensitivity and impaired visual fields, even in the absence of clinically evident vascular lesions or macular oedema.4–11 These observations suggest that diabetes may directly insult the neural retina and that visual function may be a sensitive functional biomarker of the diabetes-induced damage. The mechanisms responsible for these visual function alterations are complex and incompletely understood. Histopathological and morphometric studies of streptozotocin rat and Ins2Akita mouse models of diabetes found that experimentally induced diabetes causes ganglion cell dysfunction and death.12 13 Diabetes causes apoptosis of ganglion cells in the human retina as observed in both histopathological and imaging studies.14–17 Selective loss of short-wavelength cone photoreceptors has also been reported in psychophysical studies of patients with diabetes and NPDR.18 However, to our knowledge, no human anatomical studies have compared relative rates of ganglion cell loss with photoreceptor loss.

Based on these findings in human and rodent eyes, we tested whether patients with diabetes and patients with NPDR exhibit greater impairment of inner retinal visual function compared with photoreceptor-mediated visual function. To characterise inner retinal function, contrast sensitivity and frequency doubling technology (FDT) perimetry were used. Ganglion cell function is thought to be primarily responsible for contrast sensitivity and FDT perimetry. Light-adapted visual sensitivity (standard white-on-white perimetry), dark-adapted visual sensitivity (absolute threshold) and dark adaptation were used to assess outer retinal (photoreceptor) function. Cone photoreceptor light sensitivity is primarily responsible for light-adapted visual sensitivity, whereas rod photoreceptor function is responsible for dark-adapted visual sensitivity and dark adaptation performance. Although the entire visual system may contribute to a psychophysical response, the relative magnitude of impairment across the chosen tests provides insight into the earliest or most severely affected cell type.

The visual function tests evaluated in this study were chosen because previous studies found that they are at least moderately impaired in patients with diabetes and/or patients with mild to moderate NPDR (reviewed by Jackson and Barber19). Quantification of inner retina and outer retina impairment in the same well-defined cohort may provide new insight into the cellular pathogenesis of DR, and an enhanced understanding of visual loss in diabetes and diabetic retinopathy may guide future investigations into the mechanisms of vision loss in diabetes. Identification of a sensitive marker to early visual dysfunction may serve as an outcome measure in reversibility experiments or proof of concept interventional studies.

Materials and methods

Three groups of patients were recruited: healthy adults (normal group), participants with type 1 or type 2 diabetes without DR (diabetes group), and participants with NPDR (NPDR group). The inclusion criteria for the normal group were: (1) ≥18 years old; (2) best corrected distance visual acuity of 20/25 or better; and (3) an Early Treatment Diabetic Retinopathy Study (ETDRS) DR severity level of grade 10—no detectable retinopathy. The diabetes group met the following inclusion criteria: (1) ≥18 years old; (2) best corrected distance electronic ETDRS (E-ETDRS) acuity of 20/25 or better; and (3) ETDRS DR severity level of grade 10—no detectable retinopathy. The inclusion criteria for the NPDR group were: (1) ≥18 years old; (2) best corrected distance E-ETDRS acuity of 20/40 or better; and (3) ETDRS DR severity level of grade 20—microaneurysms only, 35—mild NPDR, 43—moderate NPDR, 47—moderately severe NPDR, or 53—very severe NPDR. Exclusion criteria for all groups were: (1) any eye disease, other than DR or minimal cataract; (2) ocular trauma; and (3) neurological conditions that can impair vision. If both eyes met the normal group eligibility criteria, the participant was included in the normal group. Participants with diabetes and absent DR in both eyes were included in the diabetes group. If either eye met the criteria for the NPDR, the participant was included in the NPDR group. The test eye was the eye with NPDR if only one eye had NPDR. If both eyes had NPDR, the test eye was the eye with better acuity. If both eyes had identical acuity, the right eye was chosen as the test eye.

The protocol was approved by the Penn State Hershey Institutional Review Board, and adhered to the tenets of the Declaration of Helsinki. Written informed consent was obtained from the subjects before participation in the study. Prior to visual function measurements, participants underwent the ETDRS refraction protocol to determine the best optical correction for the test distance. The following measurements were made on the first visit: acuity, contrast sensitivity, FDT perimetry and photopic visual field. Visual acuity was measured with the electronic visual acuity (EVA) tester (JAEB Center, Tampa, Florida, USA) using the E-ETDRS protocol. The EVA tester is calibrated for each visual acuity measurement so that the mean luminance of the monitor is in the specified range of 85–105 cd/m2. Contrast sensitivity was assessed with the Pelli-Robson contrast sensitivity chart with a luminance of 100 cd/m2 (Haag-Streit USA, Mason, Ohio, USA). The Matrix perimeter (Carl Zeiss Meditec, Dublin, California, USA) was used to measure the FDT 24-2 visual field. The Matrix stimulus is a 0.25 cycles per degree sinusoidal grating which is phase reversed at 18 Hz. The grating appears to have twice as many alternating light and dark bars than are actually present. The minimum contrast threshold of the 5.0°-diameter stimulus is measured at each of the 55 test locations. The frequency doubling illusion on which FDT perimetry is based is thought to arise in the magnocellular and be ganglion cell dependent.20 21 A 700-series Humphrey Field Analyser (Carl Zeiss Meditec) was used to measure the photopic 24-2 visual field. A blood sample was obtained from each participant to measure non-fasting blood glucose level and glycosylated haemoglobin (HbA1c).

At the second visit, within 3 weeks of the first visit, dark adaptation and scotopic visual field sensitivity were measured. Participants' eyes were dilated using 1% tropicamide and 2.5% phenylephrine hydrochloride. Rod-mediated dark adaptation was measured for a 2° circular test spot located 5° superior to the fovea using the AdaptRx, a computer automated dark adaptometer (Apeliotus Vision Science, Hershey, Pennsylvania, USA). Dark adaptation measures the sensitivity recovery of the rod photoreceptors in the dark following exposure to a moderate intensity 2-ms flash (5.8×104 scotopic cd/m2 sec, equivalent to an 80% bleaching level). After bleaching, sensitivity recovery was measured for up to 20 min. The speed of dark adaptation was characterised by the rod intercept derived from the slope of the second component of rod-mediated dark adaptation.22 The dark-adapted visual field was measured using a modified Octopus 101 (Haag-Streit USA). After the patient adapted in the dark room for 40 min, an 80-point, 48° diameter visual field was performed using the normal test strategy and a 1.7° diameter, 500-nm test target. Thus, dark adaptation measured the speed of rod photoreceptor sensitivity recovery and the dark-adapted visual field measured the absolute threshold of rod photoreceptors. Seven-field stereo colour photographs were taken in both eyes using a Topcon TRC 50-EX fundus camera (Topcon USA, Paramus, New Jersey, USA). The photographs were graded by GRJ in a masked fashion using the ETDRS grading system, and TWG provided adjudication for borderline cases.

To compare the performance on each of the visual function tests among the three groups of participants, Kruskal–Wallis test was performed. The Wilcoxon test was performed for post-hoc multiple comparisons. For each visual function test, diagnostic test sensitivity was calculated. Sensitivity is the percentage of NPDR patients that fall outside the normal reference range (mean±2SD) of the normal group. Tests exhibiting high sensitivity indicate a large performance difference between the groups, and low sensitivity indicates a small difference. This method provides an indication of clinical meaningfulness by ranking relative impairment among tests with disparate measurement units.


The sample consisted of 18 normal participants, 23 diabetic participants without DR and 35 participants with NPDR. The participants' characteristics are listed in table 1. The primary finding is that multiple aspects of visual function may be grossly impaired in patients with NPDR and good acuity. This finding is illustrated by the participant's results shown in figure 1. This participant is a 57-year-old woman with 21 years of type 2 diabetes mellitus and moderate NPDR. The participant has 20/25 best corrected visual acuity and she exhibits gross impairment of dark adaptation, FDT field and moderate impairment of the light-adapted visual field.

Table 1

Participants' characteristics

Figure 1

A patient with 21 years of type 2 diabetes mellitus, 20/25 visual acuity and moderate non-proliferative diabetic retinopathy (NPDR) exhibits gross visual function abnormalities (A). Dark adaptation (B) was grossly impaired compared with the typical normal adult. The frequency doubling technology field (C) exhibited quite deep deficits in sensitivity that were exhibited to a lesser but significant extent on the light-adapted visual field (D).

The NPDR group exhibited a mean visual acuity loss of about five letters compared with the normal and diabetes groups (p=0.001) (table 2). Because of the restricted acuity inclusion criteria, 17% of the NPDR group exhibited abnormal ETDRS visual acuity (table 3). Contrast sensitivity differed between the three groups (p<0.001). The NPDR group exhibited a mean 2.6 letter decrease compared with the normal group (p<0.001), and similar acuity compared with the diabetes group (p=0.35). Twenty-six per cent of patients with NPDR exhibited abnormal contrast sensitivity.

Table 2

Comparison of visual function outcomes

Table 3

Percentage of participants falling outside the normal reference range

The Matrix FDT visual field discriminated among the normal, diabetes and NPDR groups (p<0.0001). Sensitivity to the central 5° diameter target, which is centred on the fovea, was reduced 7.1 dB in the NPDR group compared with the normal group (p<0.0001). The impairment was substantial as evidenced by a sensitivity of 83%; that is, 29 of 35 patients with NPDR fell outside the normal reference range. The diabetes group exhibited a mean 2.9 dB loss compared with the normal group at the central test location (p=0.0002). To examine whether the Matrix FDT field was diffusely depressed in NPDR patients, the mean sensitivity of all 55 test locations was calculated for each participant. The NPDR group exhibited a mean sensitivity loss of 3.93 dB compared with the normal group (p=0.0002) and the diabetes group was indistinguishable from the normal group (mean loss 1.25 dB; p=0.10). Fewer (46%) NPDR patients fell out of the normal reference range. FDT sensitivity loss did not vary as a function of eccentricity among the three groups (p=0.11), and there was no discernible characteristic pattern to the impairment outside of the foveal location.

For the light-adapted visual fields, the mean foveal sensitivities of the three groups were similar (p=0.11). The mean field sensitivity was reduced in the NPDR group by 2.2 dB compared with the control group (p<0.0001) and about 1.3 dB compared with the diabetes group (p=0.02) (table 2). Thirty-four per cent of the NPDR group had mean sensitivity values outside the normal reference range. The photopic visual field did not vary as a function of eccentricity among the three groups (p=0.82).

The difference in dark adaptation speed between the three groups was marginally significant (p=0.07). The NPDR group's rod intercept was, on average, 1 minute slower than the normal and diabetes groups. Twenty-five per cent of the NPDR group exhibited clinically significant dark adaptation impairment. The dark-adapted visual field was depressed 2.9 dB in the NPDR group compared with the normal and diabetes groups (p=0.0001). The diagnostic sensitivity for the dark-adapted visual fields was 47%. Scotopic sensitivity did not vary by eccentricity among the three groups (p=0.66). These results indicate that a subset of participants with NPDR had impaired rod photoreceptor dysfunction as part of their NPDR.

The diagnostic test sensitivities listed in table 3 suggest that for the NPDR group impairment of inner retinal function as assessed by the Matrix FDT is greater than the impairment of outer photoreceptor function as assessed by the rod-mediated scotopic visual field or the cone-mediated photopic visual field. To calculate the relative effect sizes across these disparate measurements, impairment values were calculated for each test and matched Wilcoxon signed rank tests were used for statistical comparison between the tests. The average participant with NPDR exhibited 18% greater impairment on the Matrix foveal threshold compared with mean scotopic visual field (p=0.04). Impairment of inner retinal function assessed by the Matrix was 55% greater than impairment detected by the photopic visual field (p<0.001). Rod-mediated function exhibited 41% greater impairment compared with cone-mediated function (p<0.001). Thus inner retinal function was most impaired followed by rod function and cone function respectively.

To address the question whether blood glucose control could account for the visual function test results, the diabetes group and NPDR group were combined and HbA1c was correlated with each of the visual function test results. Worse blood glucose control (higher HbA1c levels) correlated (Spearman r=0.40, p=0.006) with worse visual acuity. Lower FDT foveal sensitivity was marginally related to higher HbA1c levels (r=0.24, p=0.06). Lower mean scotopic visual field sensitivity was marginally related to higher HbA1c levels (r=0.25; p=0.06). No other visual function parameters were related to HbA1c (all p≥0.29). Blood glucose levels were measured during each visit. The visual acuity was worse in patients with higher blood glucose levels (r=0.30, p=0.03), and lower contrast sensitivity was correlated with higher blood glucose levels (r=0.26, p=0.02). The foveal threshold of the FDT was marginally correlated with higher blood glucose levels (r=0.25; p=0.06). All other visual function results were uncorrelated with blood glucose levels at the time of testing (all p ≥0.50).


The purpose of this study was to determine the effects of diabetes on visual function impairment in persons with diabetes and mild to moderate NPDR. Unique to this study is the investigation of multiple aspects of inner and outer retinal function within the same cohort of patients. Inner retinal function as assessed by the Matrix FDT perimeter was the most impaired by NPDR. Outer retinal function was also impacted by NPDR. Rod-mediated visual sensitivity was the second most impaired visual function followed by cone-mediated visual sensitivity. The severity of visual dysfunction associated with inner retinal and photoreceptor function is notable given the good acuity and mild disease severity of the participants. These findings support the hypothesis that diabetes exerts an early impact on the inner retina, although outer retinal function is also significantly altered in a subset of patients.

Almost every aspect of visual function has been demonstrated to be impaired in diabetes and NPDR (reviewed by Jackson and Barber19). Parravaneo et al23 examined FDT in patients with DM without DR. The patients with DM exhibited a mean deviation reduction of 2.5 dB very similar to the 2.9 dB reduction reported here. The FDT abnormalities were found in only patients with HbA1c values >7%, a finding that we were unable to replicate. Parikh et al24 evaluated the use of the FDT perimeter (predecessor to the Matrix FDT) as a screening tool for severe to very severe NPDR, PDR and clinically significant macular oedema. Parikh et al found that the FDT exhibited a very high diagnostic test sensitivity of 91% for severe NPDR and PDR, but a modest sensitivity of 50% for clinically significant macular oedema. These studies and our own data indicate that impaired FDT sensitivity is a prominent feature of visual dysfunction associated with DR. Consistent with our findings, patients with DM and DR exhibit reduced rod visual sensitivity.4 5 25–27 It is notable that NPDR has quite large adverse effects on both inner retinal function and rod function before moderate visual acuity loss occurs.

The mechanisms responsible for these visual function deficits are unclear but imaging studies in humans with minimal retinal retinopathy show nerve fibre layer and inner nuclear layer thinning15 16 23 28 and histopathological studies in humans and animals show increased apoptosis and loss of ganglion cells12–14 and microglial cell activation indicative of degeneration in the inner retina.29 Similar findings have been confirmed in short-term studies of diabetic rodents (reviewed by Antonetti et al30). Inner retinal dysfunction was not correlated to short-term measurements of blood glucose control, which suggest that chronic metabolic alterations may account for the observed changes in retinal function. The linkage between visual dysfunction and chronic insults such as inflammation or long-term stability of blood glucose control should be examined by including biomarkers for oxidative stress and other systemic markers of diabetes disease severity. Reversibility of these visual deficits should be investigated through short-term manipulation of blood sugar and insulin to determine whether hyperglycaemia and/or abnormal insulin signalling contributes to visual dysfunction. Studies evaluating visual function in this population should incorporate measurements of vascular function beyond standard fundus imaging. Assessment of capillary dropout, blood flow and vessel diameter may provide insight to the mechanisms underlying visual dysfunction. Thus, measuring visual function beyond visual acuity and in combination with advanced imaging may be useful to elucidate the mechanisms underlying visual function loss caused by diabetes. These tools may also provide novel methods for the evaluation of treatments aimed at prevention of vision loss during the early stages of DR.



  • Funding This study was supported by the JDRF Diabetic Retinopathy Center at the Penn State College of Medicine, Pennsylvania, USA. DAQ is the George and Barbara Blankenship Professor.

  • Competing interests GRJ is an employee and investor in Apeliotus Vision Science, the manufacturer of the AdaptRx. The authors are solely responsible for the content of the manuscript.

  • Patient consent Obtained.

  • Ethics approval Ethics approval was provided by Penn State Hershey Institutional Review Board.

  • Provenance and peer review Not commissioned; externally peer reviewed.