INTRODUCTION

 

Diabetes-related eye disease is the most common cause of loss of vision in people of the working age (20-74 years) in developed countries¹. Functional abnormalities of the retina and vision may precede clinical signs of retinopathy in diabetes². Multifocal electroretinography (mfERG) is an objective and noninvasive method to measure retinal response to visual stimulation from small, essentially discrete patches of human retina³.

 

Aim of Work

The aim of this work is to evaluate the changes of the mfERG in the preclinical stage of diabetic retinopathy as a predictor of early onset of diabetic retinopathy and the correlation with diabetic control using glycosylated hemoglobin (HbA1c).

 

MATERIALS AND METHODS

 

Fifty-one eyes of 26 (5 males and 21 females) diabetic patients with more than five years of diagnosis of diabetes, and without diabetic retinopathy were included

in the study. The age range was 33 (48.1±8.137) years. The eyes were examined by ophthalmoscopic examination (indirect ophthalmoscopy and fundus biomicroscopy) and fluorescein angiography. Eyes with angiographic evidence of diabetic retinopathy or significant anterior or posterior segment diseases were excluded from the study.

Twelve eyes (23.5%) included in the study belonged to six patients with type I (insulin-dependent) diabetes and 39 eyes (76.5%) belonged to 20 patients with type II (non-insulin-dependent) diabetes. The diabetic control of the patients was tested by determination of HbA1c levels in blood. HbA1c levels were analyzed by chromatography techniques using commercially available kits in Cairo University Hospitals Laboratories, Faculty of Medicine, Cairo University, Cairo, Egypt. Normal values for HbA1c are less than 5.3%.

Thirty seven normal eyes of 21 subjects (13 males and 8 females) without ophthalmic or systemic abnormalities comprised the normal control group. The age range in the normal group was 43 (33.73±11.56) years.

 

Multifocal Electroretinography (mfERG) Recording Procedure

A written informed consent was obtained from all participants to perform the procedures used and mfERG test was done according to the International Society for Clinical Electrophysiology of Vision (ISCEV) Standards 2011 edition⁴. This was done by using HK Loop electrodes, on the RETIscan21mfERG version 07/01 (Roland Consult, Wiesbaden, Germany), after 10 minutes of light adaptation and pupil dilation with tropicamide. The stimulus consisted of 61 hexagons, covering 25-30° of visual field and presented on a 20-inch monitor at a viewing distance of 33 cm. Refractive errors were corrected for the viewing distance. A ground electrode was connected to the forehead. The fellow eye was occluded with light pressure to prevent blinking and the electrical artifacts it can introduce.

The mfERG recordings are performed using the ‘standard’ mfERG visual stimulus. The hexagon areas increased with eccentricity to compensate differences in cone density across the retina (leading to a fourfold size change). Each hexagon was temporally modulated between light and dark (frame rate: 60 Hz; maximum luminance: 120cd.m^-2). Subjects were instructed to fixate a small black cross in the center of the stimulus. Fixation was checked by means of online video monitoring during the 6 minutes lasting recording sessions and high-amplitude artifacts were automatically eliminated. To improve fixation stability, sessions were broken into 45-second segments, and eight trials were recorded in total. Signals were amplified with a gain of 100.000 and filtered with a band-pass filter (5–300 Hz). The surface electrode impedance was less than 10 kOhm. For each hexagon, the amplitude of P1 (defined as the difference between N1 trough and P1 peak) was calculated, and the implicit time of the P1 component determined from the onset of the stimulus till the peak of the P1 wave.

 

Data Output

Average responses were calculated for the fovea which corresponds to hexagon number 31 and for four retinal quadrants as in Figure (1). The orientation of the printout setting was set to "vertically mirrored" which sets the display and the printouts of the results according to the examiner (mirrored for the user who analyzes the data). The view in the printout corresponds to the retinal view. The responses and their measurements are plotted at corresponding retinal locations rather than in visual field orientation. The local waveforms of the diabetics and normal groups of the right eye responses were converted to left eye orientation to unify the locations in both eyes. Quadrant one (upper right) corresponds to the upper temporal retina, quadrant two (lower right) corresponds to the lower temporal retina, quadrant three (lower left) corresponds to the lower nasal retina and quadrant four (upper left) corresponds to the upper nasal retina.

The data was obtained from five elements from the output (four quadrant regions and the fovea) as well as the overall average. The amplitude and implicit time of the P1 wave of the local mfERG responses were analyzed.  The data is described as mean and standard deviation.

The data was entered into the Microsoft Excel 2003 (Microsoft Corporation, New York, USA) and MedCalc® software (version 10.2) (MedCalc Software, Mariakerke, Belgium) for Biomedical statistical analysis. The normality of the data acquired was tested by the chi-square test of normality. The distribution of data from quadrants and fovea did not appear to follow a normal distribution (Gaussian distribution) i.e. non-parametric data, so the Mann–Whitney U test was used to test for significance between the average traces of the diabetic and control groups in different quadrants and the fovea. A probability value (P value) less than 0.05 is considered statistically significant. The correlations between the non-parametric quantitative variables were done by Spearman’s (Rho or r) correlation test. Graphs and scatter diagrams were developed using Microsoft Excel and MedCalc® whenever appropriate.

 

RESULTS

 

Demographic Data:

Table (1) summarizes the clinical and demographic characteristics of both the diabetic and normal groups.

Spearman's coefficient of rank correlation (r) between the age, the amplitude and peak time (latency) of P1 is shown in Tables (2) and (3).

From the Tables (2) and (3) below there is a negative linear relationship between the amplitude of P1 and the patients age (i.e. as the age advances, the amplitude decreases); but no statistically significant relationship between age and average amplitude of P1 in both groups. A positive relationship was found between the peak time of P1 and age (i.e. as the age advances, the latency increases). The only statistically significant relationship (P<0.05) that was detected between the age and peak time in the diabetics group was in quadrant 4 (upper nasal) (P=0.0198) as well as the overall average (P=0.0372) (Figures 2 & 3). Quadrants 2 (lower temporal) showed a marginal significance (P=0.0572).

Effect of HbA1c level: The mean level of HbA1c for the diabetic patients was 8.265±1.265%. All the patients showed poor control of diabetes with the minimum value 6.5% and maximum value 11%. We did not find any significant correlation between HbA1c level and the P1 wave amplitude and peak time (Table 4).

Duration of diabetes: The mean duration of diabetes in the diabetic patients was 8.608±4.257 years (Table 5). The correlation between the duration of diabetes and the decrease in amplitude of P1 wave was less than the correlation with increase in peak time of P1 wave. The increase of peak time was noted more in quadrants 2, 3 and 4 (r=0.260, 0.201 and 0.292; respectively). However, the only statistically significant result (P<0.05) was in quadrant 4 (P=0.0391).

Quadrants data and statistics: Tables (6) and (7) show the results of the Mann-Whitney U test to compare the average amplitude and average peak times of both the normal and diabetic groups. Both the z scores from the data ranks of both study groups and the P value are reported. As for the amplitude, only the fovea showed a statistically significant difference between both study groups (P = 0.0357) (Figure 4). The peak time showed a difference to be statistically significant only in quadrant 2 (P=0.0308) (Figure 5).

Comparison between different regions in the diabetic eyes had been done and the results are presented in tables (8 and 9) below. The mean amplitude of the fovea was higher than other quadrants and this difference has been found to be statistically significant in 3 quadrants with the exception of quadrant 3 (Tables 6 and 8). On comparing the 4 quadrants with each other no statistically significant difference was detected except between quadrant 4 and 3 (P=0.0182) as the amplitude of quadrant 4 was much lower than quadrant 3 (Tables 6 and 8).

The mean peak time of the fovea was longer than other quadrants and this difference has been found to be statistically significant in all of the 4 quadrants (Tables 7 and 9). On comparing the 4 quadrants with each other no statistically significant difference was detected between any of the quadrants (Tables 7 and 9).

 

 

 

Figure 1. Quadrant orientation: Fovea corresponds to hexagon 31. All eyes were analyzed as left eye orientation.

 

 

 

Figure 2. Scatter diagram showing correlation between age and overall peak time in the normal control group.

 

 

Figure 3. Scatter diagram showing correlation between age and overall peak time in the  diabetics group. Notice the right skewing of the scatter plots, which indicate higher latency in diabetics in comparison to the normal group.

 

Figure 5. Box & Whiskers graph  showing comparison between both groups peak time of P1 at quadrant 2 which shows the mean peak time in the diabetics group being higher  than the normal controls.

 

Figure 4. Box & Whiskers Graph showing comparison between both groups amplitude of P1 at the fovea which shows the mean amplitude in the diabetics group being lower than the normal controls. The small dot represents a single case with large amplitude

 

Table 1. Clinical characteristics of the study groups.

 

	Diabetics	Normal
Number (M/F)	26 (5 / 21)	21 (13 / 8)
Age (year, mean [SD])	48.1 [± 8.137]	33.73 [± 11.56]
Eyes in study (OD/OS)	26 / 25	18 / 19
Eyes per Gender (M/F) (%)	10 / 41 (19.6 / 80.4 %)	22 / 15 (59.5 / 40.5 %)
Eyes per Type of Diabetes (I/II) (%)	12 / 39 (23.5 / 76.5 %)	NA
Duration of diabetes (years, mean [SD])	8.608 [± 4.257]	NA
HbA1c (%, mean [SD])	8.265 [± 1.265]	NA

F female, HbA1c glycosylated hemoglobin, M male, OD right, OS left; NA non-applicable, SD standard deviation

 

Table 2. Spearman's correlation coefficient (r) between age and the average amplitude of P1 wave in diabetics and normal patients.

 

	Diabetics	Normal
Quadrant  1	-0.217 (P=0.1252)	-0.163 (P=0.3278)
Quadrant  2	-0.230 (P=0.1037)	-0.252 (P=0.1300)
Quadrant  3	-0.141 (P=0.3185)	-0.285 (P=0.0875)
Quadrant  4	-0.120 (P=0.3953)	-0.231 (P=0.1652)
Fovea	-0.108 (P=0.4442)	-0.283 (P=0.0899)
Overall Average	-0.184 (P=0.1939)	-0.325 (P=0.0508)

 

 

Table 3. Spearman's correlation coefficient (r) between age and the average peak time of P1 wave in diabetics and normal patients.

 

	Diabetics	Normal
Quadrant  1	0.0709 (P=0.6163)	0.208 (P=0.2122)
Quadrant  2	0.269 (P=0.0572)	-0.0394 (P=0.8130)
Quadrant  3	0.261 (P=0.0645)	0.193 (P=0.2458)
Quadrant  4	0.330 (P=0.0198)	0.0128 (P=0.9387)
Fovea	0.168 (P=0.2344)	0.0473 (P=0.7764)
Overall Average	0.295 (P=0.0372)	0.119 (P=0.4758)

 

 

Table 4. Spearman's correlation coefficient (r) between HbA1c and the amplitude and peak time of P1 wave in diabetics.

 

	Amplitude	Peak time
Quadrant  1	-0.077 (P=0.5861)	-0.199 (P=0.1602)
Quadrant  2	-0.0519 (P=0.7139)	-0.229 (P=0.1054)
Quadrant  3	-0.0539 (P=0.7031)	-0.0356 (P=0.8011)
Quadrant  4	-0.0697 (P=0.6219)	-0.0377 (P=0.7896)
Fovea	0.00798 (P=0.9550)	-0.144 (P=0.3098)
Overall average	-0.0428 (P=0.7623)	-0.240 (P=0.0890)

 

Table 5. Spearman's correlation coefficient (r) between duration of diabetes and the amplitude and peak time of P1 wave in diabetics.

 

	Amplitude	Peak time
Quadrant  1	0.118 (P=0.4056)	0.174 (P=0.2191)
Quadrant  2	0.00511 (P=0.9712)	0.260 (P=0.0662)
Quadrant  3	-0.0271 (P=0.8483)	0.201 (P=0.1549)
Quadrant  4	-0.0297 (P=0.8335)	0.292 (P=0.0391)
Fovea	-0.145 (P=0.3052)	0.0313 (P=0.8249)
Overall average	-0.022 (P=0.8763)	0.159 (P=0.2622)

 

Table 6. The amplitude of P1 in both study groups (Mean±SD) and Mann-Whitney U test results (z score and P value).

 

	average amplitude of P1 wave in diabetic and normal subjects (µV)
	Diabetics (Mean±SD)	Normal (Mean±SD)	z Score	P value
Quadrant  1	0.5073 ± 0.2653	0.5438 ± 0.2318	0.524085	0.6002
Quadrant  2	0.5237 ± 0.2583	0.6105 ± 0.2314	1.715954	0.0862
Quadrant  3	0.5781 ± 0.2443	0.6269 ± 0.2628	0.612841	0.5400
Quadrant  4	0.4608 ± 0.2534	0.4694 ± 0.2190	0.409969	0.6818
Fovea	0.7144 ± 0.3758	0.8987 ± 0.3702	2.100565	0.0357
Overall average	0.5569 ± 0.2461	0.6299 ± 0.1978	1.398968	0.1618

SD standard deviation; µV microvolt

 

Table 7. The peak time of P1 in both study groups (Mean±SD) and Mann-Whitney U test results (z score and P value).

 

	Average peak time of P1 wave in diabetic and normal subjects (ms)
	Diabetics (Mean±SD)	Normal (Mean±SD)	z Score	P value
Quadrant  1	44.4667 ± 2.4055	44.1514 ± 2.0978	1.145378	0.2521
Quadrant  2	44.7020 ± 2.1190	43.6649 ± 1.5305	2.159736	0.0308
Quadrant  3	44.0765 ± 2.9090	43.7459 ± 1.4132	1.411647	0.0885
Quadrant  4	44.6039 ± 2.2893	43.8270 ± 2.7738	1.508856	0.1313
Fovea	45.0804 ± 6.8775	46.8270 ± 3.1224	1.018584	0.3084
Overall average	44.5349 ± 2.1039	44.4432 ± 1.2755	0.908695	0.3635

SD standard deviation, ms millisecond

 

Table 8. Mann-Whitney U test results of quadrants comparison in the amplitude of P1 in diabetics.

 

	Average amplitude of P1 wave in diabetic (quadrants comparison) ( z score and P value)
	Quadrant  1	Quadrant  2	Quadrant  3	Quadrant  4	Fovea
Quadrant  1	--------	0.334635 (P = 0.7379)	1.459010 (P = 0.1446)	0.829896 (P = 0.4066)	2.847747 (P = 0.0044)
Quadrant  2	0.334635 (P = 0.7379)	--------	1.174570 (P = 0.2402)	1.318463 (P = 0.1873)	2.690468 (P = 0.0071)
Quadrant  3	1.459010 (P = 0.1446)	1.174570 (P = 0.2402)	--------	2.362526 (P = 0.0182)	1.800338 (P = 0.0718)
Quadrant  4	0.829896 (P = 0.4066)	1.318463 (P = 0.1873)	2.362526 (P = 0.0182)	--------	3.490247 (P = 0.0005**)
Fovea	2.847747 (P = 0.0044*)	2.690468 (P = 0.0071)	1.800338 (P = 0.0718)	3.490247 (P = 0.0005**)	--------

* Significant at P<0.05  ** Significant at P<0.01

 

Table 9. Mann-Whitney U test results of quadrants comparison in the Peak time of P1 in diabetics.

 

	Average peak time of P1 wave in diabetic (quadrant comparison) ( z score and P value)
	Quadrant  1	Quadrant  2	Quadrant  3	Quadrant  4	Fovea
Quadrant  1	--------	0.0435026 (P = 0.9653)	0.351367 (P = 0.7253)	0.348021 (P = 0.7278)	3.430013 (P = 0.0006**)
Quadrant  2	0.0435026 (P = 0.9653)	--------	0.294479 (P = 0.7684)	0.301172 (P = 0.7633)	3.336315 (P = 0.0008**)
Quadrant  3	0.351367 (P = 0.7253)	0.294479 (P = 0.7684)	--------	0.632461 (P = 0.5271)	3.871731 (P = 0.0001**)
Quadrant  4	0.348021 (P = 0.7278)	0.301172 (P = 0.7633)	0.632461 (P = 0.5271)	--------	3.262695 (P = 0.0011*)
Fovea	3.430013 (P = 0.0006**)	3.336315 (P = 0.0008**)	3.871731 (P = 0.0001**)	3.262695 (P = 0.0011*)	--------

* Significant at P<0.05  ** Significant at P<0.01

 

 

DISCUSSION

 

In this study, we focused on detection of abnormalities in mfERG in 51 eyes of diabetic patients with no diabetic retinopathy and compare it to 37 normal control eyes. The quadrant group analysis and the fovea has been used to be more relevant clinically more than individual hexagons or ring analysis. The most important component of mfERG in diabetics is the P1 wave as P1 is most often abnormal in the eyes of individuals with diabetes⁵.

The amplitude and peak time of the P1 wave in the diabetic group was compared to those of the normal controls, and correlated with age, duration of diabetes and HbA1c levels. The mfERG in particular is a sensitive test for identifying neuronal deficits before the onset of vascular change. Functional abnormalities within the retina in diabetes can be caused by a variety of factors that lead to changes in one or more of the biochemical pathways altered during diabetes, or alternatively, hyperglycemia-induced changes in glial cell function⁶.

It has been established that mfERG peak time delays indicate areas of functional abnormalities in diabetes at different stages including the preclinical stage more than does the amplitude^{1,3,7-10. A} factor that is likely playing an important role is the type of pathology underlying the retinal dysfunction. Given that the early diabetic retinal disease first affects the microvasculature supplying inner retinal neurons including ganglion, amacrine and bipolar cells (and the Müller cells to some degree), then the time of signal generation and signal propagation through the retinal circuitry will be abnormally prolonged but the responses will not be extinguished¹¹. It is reasonable to expect that the local retinal responses would be abnormally delayed but not necessarily reduced in amplitude¹.

In a longitudinal study, Han and colleagues¹² studied the value of mfERG in predicting the development of diabetic retinopathy. They found that the main predictive parameter of the mfERG was the peak time of P1, and it was used with the combination of other risk factors; namely the duration of diabetes and state of diabetic retinopathy at presentation, to develop a predictive model for the development of diabetic retinopathy. In another study, it has been shown that abnormal mfERG peak times were associated with an almost eight times greater risk of development of recurring retinopathy over 3 years and combined with other risk factors for diabetes it can predict the development of diabetic retinopathy later on¹⁰.

In this study, we conducted; the amplitude of P1 wave showed a decrease in the diabetic group than in normal controls. This was statistically significant only in the fovea. However, the more important peak time was prolonged in the diabetics than normal group, and this was statistically significant in one quadrant only; quadrant 2 (lower temporal quadrant). On comparing the quadrants and fovea with each other, the mean peak time of the fovea was longer than other quadrants and this difference has been found to be statistically significant in all of the 4 quadrants. On comparing the 4 quadrants with each other no statistically significant difference was detected between any of the quadrants. From the above, the area that is more likely to develop diabetic retinopathy in the macular area later is the lower temporal quadrant. The fovea shows more peak time delay than all other regions in the diabetic retina, however the delay at the fovea was not found to be statistically different from the normal group.

In the review of Lai and colleagues¹³, the authors reported that the reductions in mfERG response amplitudes and delays in peak times with increasing age were found in most of the studies. Seiple and colleagues¹⁴ found that the P1 amplitude decreases with age about ten times more than the implicit times (P1 amplitude decreases by 10.5% per decade versus an increase in peak time of P1 by about 1% per decade). These results showed the importance to develop an age-matched normative data for mfERG analysis or corrections for age to be developed for the mfERG users to use by each laboratory.

In our study, the age of patients affected the output of mfERG. We have found that as the age advances, the amplitude of P1 decreases in both the diabetics and normal control groups. No statistically significant relationship was detected between age and average amplitude of P1 in the diabetics group. This runs in context with the results of Shimada and colleagues¹⁵, who found that age affected the amplitudes of P1 wave in both normal and diabetic groups but did not find any significant difference between both groups. In case of the peak time of P1, as the age advances, a delay in both diabetic and normal groups is noted. The effect of age on mfERG peak time and amplitude measurement was more pronounced in the diabetic group of subjects, which could be the result of either a combined effect of both aging and diabetes or a delay in diagnosis that prolonged the actual duration of the disease.

The duration of diabetes was correlated with the diabetic mfERG amplitude and peak time. We have found that there is a positive linear correlation between duration and peak time in all quadrants that was statistically significant in quadrant 4 (upper nasal). This correlation was not found between the duration of diabetes and the amplitude. It is logical that with the increase of the duration of diabetes, the severity of the changes of mfERG increases. This is supported by the study of Bronson-Castain and colleagues¹⁶ who have found that after a sufficient duration of the disease and the start of appearance of early NPDR, the functional integrity of the retina is compromised to the extent that there is a profound delay of the peak time first-order Kernel of the mfERG.

All of the diabetic patients had uncontrolled diabetes with high HbA1c levels. However, we did not find any significant correlation between the mfERG and HbA1c levels. This is consistent with the findings of Tyrberg⁸, Kim⁹, Holm¹⁷and their colleagues. These studies did not find any correlation between HbA1c and the increased amplitude or the peak time in the first-order component of the mfERG in diabetic patients.

Few clinical studies were conducted on the distribution of the diabetic retinopathy changes in the retina. They have reported that neovascularization and other changes in the form of microaneurysms and loss of pericytes were more on the temporal retina than the nasal retina^18,19. Feman and colleagues’ study reported a higher incidence of neovascularization in the supero-temporal quadrant of the retina than other quadrants¹⁸. Tang and coauthors studied the distribution of biochemical disturbances that occur in diabetic retinas, and found that the distribution of the changes in diabetic retinopathy follows the distribution of the biochemical disturbance; that occurred more on the temporal side of the retina¹⁹. These reports support and give strength to the findings of our study.

To our knowledge no previous study has reported the topographic incidence of diabetic retinopathy changes by retinal quadrants using the mfERG peak time delays. This study is the first to report that mfERG peak time delays are more in the lower temporal quadrant of the retina in patients with no retinopathy, which may have an impact later on the appearance of the diabetic changes afterwards.

 

[Disclosure: Authors report no conflict of interest]

 

REFERENCES

 

1.      Bearse MA, Adams AJ, Han Y, Schneck ME, Ng J, Bronson-Castain K, et al. A multifocal electroretinogram model predicting the development of diabetic retinopathy. ProgRetin Eye Res.2004; 25: 425–48.

2.      Bearse MA, Han Y, Schneck ME, Adams AJ. Retinal Function in Normal and Diabetic Eyes Mapped with the Slow Flash Multifocal Electroretinogram. Invest Ophthalmol Vis Sci. 2004; 45: 296–304.

3.      Fortune B, Schneck ME, Adams AJ. Multifocal Electroretinogram Delays Reveal Local Retinal Dysfunction in Early Diabetic Retinopathy. Invest Ophthalmol Vis Sci. 1999; 40: 2638–51.

4.      Donald CH,  Michael B,  Mitchell B, David K, Mineo K,  Jonathan SL, et al. ISCEV standard for clinical multifocal electroretinography 2011 edition. Doc Ophthalmol. 2012; 124: 1–13.

5.      Schneck ME, Bearse MA Jr, Han Y, Barez S, Jacobsen C, Adams AJ. Comparison of mfERG waveform components and implicit time measurement techniques for detecting functional change in early diabetic eye disease. Doc Ophthalmol. 2004; 108: 223–30.

6.      Fletcher EL, Phipps JA, Ward MM, Puthussery T, Wilkinson-Berka JL. Neuronal and Glial Cell Abnormality as Predictors of Progression of Diabetic Retinopathy. Curr Pharm Des. 2007; 13: 2699-712.

7.      Han Y, Bearse MA, Schneck ME, Barez S, Jacobsen C, Adams AJ. Towards optimal filtering of ‘‘standard’’ Multifocal electroretinogram (mfERG) recordings: findings in normal and diabetic subjects. Br J Ophthalmol.2004; 88: 543–50.

8.      Tyrberg M, Ponjavic V, Lövestam-Adrian M. Multifocal electroretinography (mfERG) in insulin dependent diabetics with and without clinically apparent retinopathy. Doc Ophthalmol. 2005; 110: 137–43.

9.      Kim SJ, Song S, Yu HG. Multifocal Electroretinogram Responses of the Clinically Normal Retinal Areas in Diabetes’. Ophthalmic Res.2007; 39: 282–8.

10.    Ng JS, Bearse MA Jr, Schneck ME, Barez S, Adams AJ. Local Diabetic Retinopathy Prediction by Multifocal ERG Delays over 3 Years. Invest Ophthalmol Vis Sci. 2008; 49: 1622–28.

11.    Hood DC, Frishman LJ, Saszik S,Viswanathan S. Retinal Origins of the Primate Multifocal ERG: Implications for the Human Response. Invest Ophthalmol Vis Sci. 2002; 43: 1673–85.

12.    Han Y, Schneck ME, Bearse MA Jr, Barez S, Jacobsen CH, Jewell NP, et al. Formulation and Evaluation of a Predictive Model to Identify the Sites of Future Diabetic Retinopathy. Invest Ophthalmol Vis Sci.2004; 45: 4106–12.

13.    Lai TYY, Chan W, Lai RYK, Ngai JWS, Li H, Lam DSC. The Clinical Applications of Multifocal Electroretinography: A Systematic Review. SurvOphthalmol.2007; 52: 61-96.

14.    Seiple W, Vajaranant TS, Szlyk JP, Clemens C, Holopigian K, PaligaJ, et al. Multifocal Electroretinography as a Function of Age: The Importance of Normative Values for Older Adults. Invest Ophthalmol Vis Sci. 2003; 44: 1783–92.

15.    Shimada Y, Li Y, Brease MA, Sutter EE, Fung W. Assessment of early retinal changes in diabetes using a new Multifocal ERG protocol. Br J Ophthalmol. 2001; 85: 414-9.

16.    Bronson-Castain KW, Bearse MA Jr, Han Y, Schneck ME, Barez S, Adams AJ. Association between Multifocal ERG Implicit Time Delays and Adaptation in Patients with Diabetes. Invest Ophthalmol Vis Sci. 2007; 48: 5250–6.

17.    Holm K, Larsson J, Lövestam-Adrian M. In diabetic retinopathy, foveal thickness of 300 µm seems to correlate with functionally significant loss of vision. Doc Ophthalmol.2007; 114: 117–24.

18.    Feman SS, Leonard-Martin TC, Semchyshyn TM. The Topographic Distribution of the First Sites of Diabetic Retinal Neovascularization. Am J Ophthalmol. 1998; 125: 704–706.

19.    Tang J, Mohr S, Du YP, Kern TS. Non-uniform distribution of lesions and biochemical abnormalities within the retina of diabetic humans. Curr Eye Res. 2003; 27: 7-13.