Although the alterations in retinal vessels induced by chronic hyperglycaemia are already well known,1 2 the effects on retinal nerve tissue3 4 and on glial cells remain poorly understood. The increase in oxidative stress that occurs during diabetes is likely to lead to retinal ganglion cell (RGC) apoptosis and glial cell impairment and loss.5 Results from several studies imply that neuronal dysfunction can occur prior to overt vasculopathy. Indeed, RGC apoptosis has been noted in animal models as early as 14 weeks, while a manifest vasculopathy is not observed until approximately 6 months after the induction of diabetes.5 6

Morphological changes of retinal nerve tissue have also been reported in human subjects with diabetes without retinopathy, as evidenced by thinning of the retinal nerve fibre layer (RNFL) assessed by means of green filter photographs4 and by scanning laser polarimetry with fixed corneal compensator (GDx NFA).7^–9 However, these results were not confirmed by GDx with variable corneal birefringence compensator (GDx VCC).10

Functional loss has been extensively studied in animal models of diabetes and in human subjects by electrophysiological techniques. An impairment of the innermost retinal layers5 11 12 and of amacrine cells,12^–14 as well as an impairment of the short-, mid- and long-wavelength cone components,15^–18 have been reported in different stages of the pathology.

An early functional impairment of the magnocellular component of ganglion cells (MGC) has also been observed using screening test strategies of first generation frequency doubling technology perimetry (Humphrey FDT), by selective stimulation of the MGC with low spatial frequency doubled stimuli undergoing counterphase flicker at high temporal frequency.19 Recently, second generation FDT (Humphrey Matrix; Welch Allyn, Skaneateles Falls, New York, USA; Carl Zeiss Meditec, Dublin, California, USA) has been released; this technology uses more test locations, smaller stimulus size, improved thresholding strategy, and test patterns equivalent to the Humphrey field analyser patterns, allowing an increased spatial resolution, shorter test duration and improved discriminatory power to detect early visual field defects.20

The aim of this study was to investigate the role of Humphrey Matrix threshold testing in the detection of early functional retinal impairment in the eyes of subjects with type 1 diabetes mellitus (DM1) without any signs of retinal vasculopathy, and to investigate the relationship between both functional and structural retinal parameters and metabolic control.

MATERIALS AND METHODS

Thirty eyes of 30 subjects with DM1 with no sign of retinal vasculopathy, and 30 eyes of 30 age- and sex-matched healthy control subjects without any history of ocular or systemic disease were enrolled in this cross-sectional clinical study. All subjects were examined at the outpatient clinic of the Policlinic of the University of “Tor Vergata”, Rome, Italy.

Inclusion criteria were: age >18 years, clear lens and diagnosis of DM1 at least 1 year earlier. DM1 was defined by clinical features as a history of sudden onset of polyuria and thirst, marked weight loss, blurred vision, paresthesias and age at diagnosis <35 years; laboratory findings included sustained hyperglycaemia, ketoacidosis and ketonuria, serum positivity for anti-islet cell and anti-glutamic acid decarboxylase (GAD) autoantibodies and low plasma C-peptide levels.

Exclusion criteria were: spherical refractive error > ±6 dioptres; astigmatism > ±3 dioptres; active or past retinal pathologies; diagnosis of glaucoma or ocular hypertension (intraocular pressure >22 mmHg); opacities of dioptric media that could bias functional and structural retinal testing; history of ocular surgery; any fluorangiographic sign of retinal vasculopathy due to diabetes; or other local or systemic disease. Subjects with symptoms or signs of cardiovascular disease, vasculopathy, nephropathy, peripheral and autonomic sensory or motor neuropathy, or other metabolic or endocrine disease, were excluded. The presence of a retinopathy was excluded by fluorescein angiography (FA) (TRC-50X; TOPCON Instr. Corp., Tokyo, Japan) performed in a period ranging between time of enrolment and up to 3 months earlier. Medical evaluation, at enrolment, included personal and family history, physical examination and biochemical analyses, such as fasting plasma glucose, lipid profile, serum creatinine, and microalbuminuria and creatinine clearance (24-h urine collection); it also included regular cardiovascular examinations, such as resting electrocardiogram and 24-h blood pressure monitoring, and standard tests for neuropathy, such as biothesiometry and 10 g monofilament.

The percentage of the A_1c form of glycohaemoglobin (HbA_1c) was assessed by high-performance liquid chromatography from blood samples collected during the study visit. Values between 4% and 6.9% were considered normal, and 7% was chosen as the standard cut-off value for defining diabetic subjects with a good glycaemic control.21

All subjects underwent a complete ophthalmological examination including: best-corrected visual acuity (BCVA), Goldmann applanation tonometry, slit-lamp examination and indirect ophthalmoscopy. Functional testing included Humphrey Matrix perimetry and white-on-white Humphrey visual field perimetry (HFA; Carl Zeiss Meditec, Dublin, CA, USA), while RNFL thickness was measured by GDx VCC (software 5.5.0; Zeiss-Humphrey System, Dublin, CA, USA).

The primary objective was to explore the presence of an early reduction of retinal sensitivity in subjects with DM1 investigated by Humphrey Matrix perimetry threshold testing. Secondary objectives included: (1) the proportion of subjects with DM1 showing Humphrey Matrix parameters outside 95% confidence limits based on the control population; (2) the investigation of an early reduction in peripapillary RNFL thickness, measured by GDx VCC; and (3) the investigation of the correlation between functional/structural parameters and metabolic control expressed by HbA_1c.

Humphrey Matrix and HFA

Humphrey Matrix allows retinal testing with high spatial resolution, with 69 stimuli within 30 central degrees.22 Subjects were subjected to two white-on-white 30-2 SITA standard HFA tests.23 All enrolled subjects performed two Humphrey Matrix visual field tests within 2 weeks using the 30-2 threshold program in order to assess the test–retest variability and to rule out a relevant learning curve. The second test was performed at the time of examination, after visual acuity measurement and before any other evaluation. Reliability criteria were defined as less than 33% fixation errors, false-positives and false-negatives. The second test was used for statistical analysis. If the second test was unreliable a third examination was performed and considered for statistical analysis. In the case of a third unreliable test the subject was excluded from the study. Mean deviation (MD) and pattern standard deviation (PSD) values were considered for the analysis. An MD p value <5% was considered as a threshold of abnormality.

Scanning laser polarimetry

RNFL was imaged by GDx VCC, an imaging technology that measures the retardation of reflected light caused by parallel birefringent microtubules of retinal axons. The retardation of reflected light has been proven to be linearly related to RNFL thickness.24 GDx VCC software calculates summary parameters based on eye quadrants, defined as superior (25° to 144°), temporal (355° to 24°), inferior (215° to 334°) and nasal (145° to 214°). GDx VCC parameters considered for this study were: temporal-superior-nasal-inferior-temporal (TSNIT average, superior average, inferior average, TSNIT standard deviation (TSNIT SD) and nerve fibre indicator (NFI). NFI is calculated with a support vector machine algorithm that is based on several RNFL measurements, and a score from 0 to 100 is assigned to each eye: the higher the NFI score, the greater the likelihood that patients have a RNFL defect compatible with a glaucomatous optic neuropathy. Only high-quality images, defined by a well-focused and uniformly illuminated reflectance image, with a centred optic disc and a quality score >8, and without an atypical retardation pattern, were included.

Statistical analysis

Demographic and descriptive data were expressed as mean and SD and 95% CI. Normal distribution of data was assessed by the Shapiro–Wilk test. The right eye was arbitrarily chosen for statistical analyses. Frequencies of categorical variables were compared between groups by chi-squared and Fisher’s exact test as appropriate. Comparisons of continuous variables between groups were performed by independent samples t test and Mann–Whitney U test as appropriate. A p value of <0.05 was considered statistically significant. Differences between controls, DM1 subjects with good glycaemic control and DM1 subjects with poor glycaemic control were performed by Turkey–Kramer HSD test. Moreover, 95% confidence limits were specified for each parameter based on data from control subjects, and the proportion of DM1 subjects showing values outside normal limits was examined (sensitivity).

RESULTS

Among 34 consecutive subjects with DM1 and 32 consecutive healthy subjects considered for the study, 30 DM1 subjects (19 female and 11 male, mean age 36.77 (SD 9.78; 95% CI 33.11 to 40.42) years and 30 healthy subjects (21 female and nine male, mean age 35.87 (SD 8.59; 95% CI 32.66 to 39.08) years were included in the analysis. Three subjects were excluded for the presence of FA signs of diabetic vasculopathy and one for a refractive error > ±6 dioptres in DM1 group and two for a refractive error > ±6 dioptres in the control group. Ninety per cent of subjects performed FA at enrolment; only 10% (three) of subjects were tested not more than 1 month before baseline. Mean duration of DM1 was 12.23 (SD 10.83; range 1–40) years and mean HbA_1c level was 7.38 (SD 1.19; range 4.9–10.3)%. Both DM1 and control subjects presented a BCVA of 0.0 LogMAR. Age and intraocular pressure were similar between groups (table 1).

Enrolled subjects performed two Humphrey Matrix tests. Although an improvement in MD and a decrease in PSD were observed, the differences between the first and the second examination were not statistically significant in either DM1 (MD: −1.58 (SD 2.12) vs −1.10 (SD 2.98) dB, p = 0.28; PSD: 3.14 (SD 1.12) vs 2.88 (SD 0.78) dB, p = 0.09) or control (MD: 1.78 (SD 2.76) vs 1.37(SD 2.11) dB, p = 0.4; PSD: 2.97 (SD 1.04) vs 2.61 (SD 0.45) dB, p = 0.12) subjects.

While at least one of the reliability indices was found to be abnormal at the first Humphrey Matrix test in 10% of DM1 and 13.3% of control subjects, at the second examination all tests presented good reliability and no subject required a third test.

Matrix MD was found to be significantly reduced in DM1 subjects compared with controls (−1.10 (SD 2.98; 95% CI −2.21 to 0.01) vs 1.37 (SD 2.11; 95% CI 0.58 to 2.16) dB, p = 0.0005). In addition, HFA MD was found to be significantly reduced in DM1 subjects compared with controls (−2.22 (SD 2.89; 95% CI −3.30 to 1.15) vs −0.80 (SD 0.36; 95% CI −1.31 to 0.29) dB, p = 0.010). Similarly, HFA PSD was found to be significantly worse in DM1 subjects (2.27 (SD 1.41; 95% CI 1.75 to 2.80) vs 1.57 (SD 0.41; 95% CI 1.41 to 1.72) dB, p = 0.013) (table 2).

Among structural parameters, TSNIT average was found to be significantly reduced in DM1 subjects (53.43 (SD 4.40, 95% CI 51.78 to 55.06) vs 56.41 (SD 4.39, 95% CI 54.77 to 58.05), p = 0.006) (table 3).

Seven DM1 subjects had HbA_1c <7% and 23 subjects HbA_1c ⩾7%. Among functional parameters, Matrix MD was found to be significantly worse in DM1 subjects with HbA_1c ⩾7% compared with controls (−1.35 (SD 3.27; 95% CI −2.76 to 0.07) vs 1.37 (SD 2.11; 95% CI 0.58 to 2.16), p = 0.0004).

HFA MD (−2.30 (SD 1.35; 95% CI −3.69 to −0.91) vs −0.80 (SD 1.35; 95% CI −1.31 to −0.29) dB; p = 0.021) and PSD (2.39 (SD 1.53; 95% CI 1.73 to 3.05) vs 1.57 (SD 0.41; 95% CI 1.41 to 1.72) dB, p = 0.006) were found to be significantly worse in DM1 subjects with HbA_1c ⩾7% compared with controls.

Among structural parameters, TSNIT average (53.12 (SD 4.21; 95% CI 51.29 to 54.94) vs 56.41 (SD 4.39; 95% CI 54.77 to 58.05) μm, p = 0.0092) and superior average (62.57 (SD 5.41; 95% CI 60.23 to 64.91 vs 67.01 (SD 7.17; 95% CI 64.34 to 69.69) μm, p = 0.017) were found to be significantly reduced in DM1 subjects with HbA_1c ⩾7% compared with controls. No significant differences were found between DM1 subjects with HbA_1c <7% and controls for any parameter.

The proportion of subjects with functional or structural parameters outside 95% CI based on control data were calculated and are shown in table 4. The threshold of normality was calculated as +0.58 dB for Matrix MD, +2.78 dB for Matrix PSD, −1.31 dB for HFA MD and +1.72 dB for HFA PSD.

The highest proportion of DM1 subjects with test outside 95% CI was found for Matrix MD (73.3%) followed by GDx TSNIT average (70%). Full details are given in table 4.

No statistically or clinically significant correlations were found between functional and structural study variables in either group.

DISCUSSION

This study supports the evidence of an early retinal functional impairment associated with a reduction of RNFL in DM1 subjects without any detectable sign of retinal vasculopathy. DM has been proposed not just as a disease of the vasculature, but as a multifactorial disease involving retinal neurons and glia. The functional and structural impairment of neural tissue may precede and may ultimately induce the earliest morphological alterations of vascular tissue.5 25

Several experimental animal studies reported retinal neurodegeneration in early DM, with apoptosis of neurons and retinal thinning, reduction of RGC and photoreceptor layer.6 Extensive analysis by electroretinogram (ERG) showed a general and early dysfunction of retinal neurons in DM, before retinal vasculopathy, with a wide variety of functionally abnormal neuronal and glial classes. A sensitivity loss of phototransduction for both rod and cones, causing reduction in dark adaptation and loss of colour determination and contrast sensitivity, and an altered waveform response for both amacrine and bipolar cells, reflecting a diffused alteration in neuronal synaptic transmission, has been described.5 12 14^–18

In early DM without signs of retinopathy, functional and structural alterations of Müller cells with dysfunction and gliosis were noted. This could explain the early neuronal functional loss and death, and the late development of vascular abnormalities.6

In our study, DM1 subjects showed a decreased threshold of retinal sensitivity as measured by HFA and Humphrey Matrix.

While the functional differences found in this study between DM1 and control subjects might improve the understanding of the pathophysiology of early retinal impairment in diabetes, the small magnitude of differences of functional and structural parameters between DM1 and control subjects makes it difficult to translate these data in the clinical practice. In order to give a more clinical perspective to these results, we specified 95% CI for each parameter based on data from control subjects and examined the proportion of DM1 subjects showing values outside normal limits (sensitivity). The best functional parameter to correctly classify DM1 subjects was Matrix MD (sensitivity 73.3%), suggesting that this parameter deserves future diagnostic studies as it probably represents a good candidate for the early classification of functional retinal impairment in DM without retinopathy.

By testing 69 retinal locations with spatial frequency doubled stimuli, Humphrey Matrix visual field allows a selective high-resolution measurement of non-redundant MGC activity, which represents 3–5% of all RGC.

The finding of a generalised reduction of retinal sensitivity to frequency doubled stimuli in DM1 subjects might be related to a higher susceptibility of the magnocellular RGCs to hyperglycaemia; this might be responsible for the early selective loss of this cell component. Nonetheless, another possible explanation might be that magnocellular RGCs have similar susceptibility to hyperglycaemia compared with other RGCs, but are simply less redundant than the whole RGC population, with no overlap between receptive fields, allowing the earlier detection of losses if selectively stimulated with frequency doubled stimuli.

This result is in agreement with a previous finding of abnormalities in frequency doubling perimetry using low-resolution screening test strategies (19 retinal locations),19 and with several reports of a significant reduction in contrast sensitivity as measured by stationary gratings at different spatial frequencies in subjects with DM without any clinical sign of diabetic retinopathy.14

Besides retinal function, retinal morphology of DM1 subjects was explored for early changes of peripapillary RNFL thickness by GDx VCC. The average peripapillary RNFL thickness was found to be significantly thinner in DM1 subjects, although the difference with regard to control subjects was of small magnitude. A previous report in the literature showed no differences in RNFL thickness between controls and diabetic subjects. This discrepancy might be explained by considering the different characteristics between the two sample populations: although both studies enrolled subjects with similar disease duration, only younger type 1 (36.77 (SD 9.78) years) diabetic subjects were studied in the present study, while the previous study in the literature enrolled older (65.1 (SD 6.9) years) and type 2 subjects. As RNFL thickness is known to decrease with increasing age, the ageing process might smooth out small differences between controls and DM subjects. Moreover, a different impact on RNFL thickness exerted by different types of diabetes (type 1 vs type 2) cannot be excluded.10 TSNIT average was able to correctly classify DM1 eyes with good sensitivity (sensitivity 70%), and might thus represent a good candidate among GDx parameters for the detection of early retinal changes.

Stratifying DM1 by HbA_1c levels enabled the observation that short-term metabolic control, expressed as a single value of HbA_1c, is able to influence functional and structural retinal testing. Functional data, in fact, were found to be significantly different from control subjects only in DM1 subjects with poorer metabolic control (HbA_1c ⩾7%) suggesting that sustained hyperglycaemia may be able to influence functional data and that functional testing in subjects with “good” metabolic control, according to our results, is likely to be irrelevant.

It would be of interest to investigate in future perspective studies whether fluctuations of short-term metabolic control over time may be associated with fluctuations of retinal functional parameters.

In addition, structural parameters as measured in our study by GDx VCC were found to be worse only in DM1 with HbA_1c ⩾7% compared with healthy subjects. This finding is in agreement with a previous report, where a significant relationship was found between RNFL thinning, measured by the first generation GDx-NFA, and impairment of metabolic regulation.8

Our results contribute to the hypothesis that DM1 is likely to produce an early functional and structural retinal impairment, before the onset of any clinically relevant retinal vasculopathy, probably due to the neuronal alteration, which may be considered as part of a systemic diabetic neuropathy. Nevertheless the relationship between the observed early functional retinal impairment and potential dynamic abnormalities of retinal perfusion that take place in early diabetes, and not detectable by fluorescein angiography, may not be excluded and should be considered when interpreting our results.

In conclusion functional and structural retinal testing by frequency doubling perimetry and scanning laser polarimetry could be useful for the identification of early retinal impairment in subjects with DM1 without any sign of clinically detectable retinal vasculopathy, although broader diagnostic studies are required to assess their sensitivity and specificity accurately in clinical settings and for screening purposes.