Skip to main content

Multifocal pupillographic objective perimetry for assessment of early diabetic retinopathy and generalised diabetes-related tissue injury in persons with type 1 diabetes



To examine the potential utility of five multifocal pupillographic objective perimetry (mfPOP) protocols, in the assessment of early diabetic retinopathy (DR) and generalised diabetes-related tissue injury in subjects with type 1 diabetes (T1D).


Twenty-five T1D subjects (age 41.8 ± 12.1 (SD) years, 13 male) with either no DR (n = 13) or non-proliferative DR (n = 12), and 23 age and gender-matched control subjects (age 39.7 ± 12.9 years, 9 male) were examined by mfPOP using five different stimulus methods differing in visual field eccentricity (central 30° and 60°), and colour (blue, yellow or green test-stimuli presented on, respectively, a blue, yellow or red background), each assessing 44 test-locations per eye. In the T1D subjects, we assessed 16 metabolic status and diabetes complications variables. These were summarised as three principal component analysis (PCA) factors. DR severity was assessed using Early Treatment of Diabetic Retinopathy Study (ETDRS) scores. Area under the curve (AUC) from receiver operator characteristic analyses quantified the diagnostic power of mfPOP response sensitivity and delay deviations for differentiating: (i) T1D subjects from control subjects, (ii) T1D subjects according to three levels of the identified PCA-factors from control subjects, and (iii) TID subjects with from those without non-proliferative DR.


The two largest PCA-factors describing the T1D subjects were associated with metabolic variables (e.g. body mass index, HbA1c), and tissue-injury variables (e.g. serum creatinine, vibration perception). Linear models showed that mfPOP per-region response delays were more strongly associated than sensitivities with the metabolic PCA-factor and ETDRS scores. Combined mfPOP amplitude and delay measures produced AUCs of 90.4 ± 8.9% (mean ± SE) for discriminating T1D subjects with DR from control subjects, and T1D subjects with DR from those without of 85.9 ± 8.8%. The yellow and green stimuli performed better than blue on most measures.


In T1D subjects, mfPOP testing was able to identify localised visual field functional abnormalities (retinal/neural reflex) in the absence or presence of mild DR. mfPOP responses were also associated with T1D metabolic status, but less so with early stages of non-ophthalmic diabetes complications.

Peer Review reports


The chronic microvascular complications of type 1 diabetes (T1D), including diabetic retinopathy (DR), nephropathy and neuropathy, contribute profoundly to the burden of this disease [1,2,3,4]. In order to assess the subclinical stages of these complications, and the effects of new therapies for their prevention, more sensitive tests for detecting both organ-specific changes such as DR, as well as generalized diabetes-related tissue injury (e.g. from widespread hyperglycemia-induced tissue damage from accumulation of advanced glycation end products and oxidative stress), are needed.

Taking diabetic eye disease as an example, the traditional treatments of laser photocoagulation, anti-vascular endothelial growth factor injections and vitrectomy, reduce the risk of vision loss, but they target late disease stages such as proliferative DR and diabetic macular oedema. These late treatments are also associated with significant side effects [5,6,7,8]. Candesartan shows promise in the prevention of earlier stage retinopathy in T2D patients [9]. Fenofibrate is also gaining recognition as a therapy with potential to prevent progression and even reverse earlier stages of DR in T2D [10, 11]. Clearly, additional new treatments that target earlier stages of diabetic eye disease are needed and these will need monitoring tools for pre-clinical early-stage disease.

Multifocal pupillographic objective perimetry (mfPOP) assesses visual function by monitoring pupil responses to retinal stimuli presented to 44 retinal locations/eye. Both response sensitivity and delay are obtained at each location of both visual fields concurrently. The method is rapid, objective and requires minimal operator training. We have shown mfPOP to be clinically useful in early [12, 13] and later stage [14,15,16] age-related macular degeneration (AMD). We have also shown that mfPOP is able to identify localised visual field dysfunction in type 2 diabetes (T2D) subjects prior to the development of clinically detectable retinal vasculopathy [17, 18] and early-stage diabetic macular oedema [19]. Diabetic changes to the visual fields are similar in multifocal visual evoked potentials (mfVEPs) and mfPOP when tested in the same subjects [20]. The main advantage of mfPOP is the rapid, non-invasive nature of assessment.

Given those robust outcomes, we have hypothesised that mfPOP has potential to monitor early stage diabetes-related retinal and nerve dysfunction and, in addition might be useful as a clinical test to assess more generalised diabetes-related tissue injury, that may be superior to other measures such as skin advanced glycation end product (AGE) accumulation and the urine albumin excretion rate. In these ways, it could prove to be a useful clinical tool to assess effectiveness of interventions to prevent diabetes complications in their early stages of development.

In this pilot study, we have examined the potential utility of mfPOP in the assessment of early diabetic retinopathy (DR) and generalised diabetes-related tissue injury (GDTI) in subjects with type 1 diabetes. We compare diagnostic power of five novel mfPOP stimulus methods, which test either the peripheral or macular visual fields, with parameters of metabolic status, retinopathy using the Early Treatment of Diabetic Retinopathy Study (ETDRS) scales for macular and peripheral DR, and various assessments of non-ocular diabetic complications.



In this prospective study we recruited twenty-five T1D subjects with variable duration of diabetes with either no DR or non-proliferative DR, all non-smokers without other known diabetic complications from The Canberra Hospital Diabetes Clinic. Given past published results [17, 18], the required sample size for p = 0.05 and power of 0.8 was 10. Twenty-three age and gender matched non-diabetic control subjects were recruited from the general community. Controls were excluded if they had first-degree relatives with diabetes, previous gestational diabetes, or were smokers. T1D and control subjects were excluded if they were pregnant, had visual acuity worse than 6/12, past eye surgery, intraocular pressure > 21 mmHg, distance refraction ≥ ± 5 D or ≥ ±2 D cylinder, or had medications or comorbidities that would affect their pupillary responses.

Non-ocular clinical and biochemical assessments

On the first of 2 visits, T1D subject weight and height was measured to calculate body mass index (BMI). Standing systolic and diastolic blood pressure was measured (average of 2nd and 3rd of 3 readings). Venous blood (sodium fluoride and serum tubes) was taken from the T1D subjects before each mfPOP session to determine plasma glucose and serum potassium concentration levels performed within the clinical laboratories of ACT Pathology at The Canberra Hospital. The most recent measures of the T1D subjects’ HbA1c, serum creatinine, triglyceride, total cholesterol, high density lipoprotein (HDL) cholesterol and low density lipoprotein (LDL) cholesterol, urinary albumin creatinine ratio, as well as all available HbA1c measures over the previous 5 years, were recorded from the subjects’ medical records. Peripheral neuropathy was assessed using a biothesiometer (Bio-Medical Instrument Company, Newbury, Ohio, USA). Skin advanced glycated end-products (AGE) were assessed by an AGE reader (DiagnOptics, Gronungen, The Netherlands).

mfPOP stimuli and data acquisition

mfPOP testing was performed using a prototype of the Federal Drug Administration cleared objectiveField Analyzer (OFA) (Konan Medical USA Inc., Irvine CA) in all subjects. The OFA presented multifocal stimuli simultaneously to the two eyes via two liquid-crystal displays operating at 60 frames/s. Subjects fixated on a red cross centred within a dim starburst radial grating to aid binocular fusion. Trial lenses corrected the subject’s distance prescription and stimuli were presented at optical infinity. The mfPOP method had several quality control steps. First of all, the operator could see an image of the two fitted pupil diameters in real time, and its colour indicated quality. Poor quality would lead to a check of focus. Also a time-varying bar whose length indicated the quality of the diameter fit in real time was displayed. Blinks were monitored and data removed from the pupil record when blinks occurred. Each mfPOP test, or protocol, presented 9 test segments of 42.4 s duration (6.37 min total), with a short rest break in between segments. If more than 15% of the data from a segment was lost due to blinks it was repeated. The per-region responses were estimated by a regressive method, and that provided both t-statistics for each sensitivity and delay, and an overall goodness of fit statistic (r2) for each pupil.

Five stimulus protocols were tested in a randomised order, over two sessions, 2 weeks apart. The protocols differed in luminance, colour and visual extent (Fig. 1). Two of the protocols presented 44 stimuli/eye to the macular region (±15° eccentricity) that were either: yellow stimuli on a yellow background (Yellow Macula), or green stimuli on a red background (RG Macula). The other three protocols were wide-field methods that presented 44 stimuli/eye to a four-fold greater area (±30°). These presented yellow (Yellow Wide) or blue stimuli (Blue Wide), on the same colour background, or green stimuli on a red background (RG Wide).

Fig. 1
figure 1

The mfPOP stimulus protocols. All 5 protocols had 44 stimulus regions arranged in 5 rings, a and b. Pseudo-random sequences governed when any one stimulus region would be shown for 33 ms. Each region was presented 96 times. The presentation sequence meant that although the stimuli could overlap (c) in practice they never did. a to c show the layout of stimuli for the wide-field stimuli d,f,h. The layout of the macular stimuli, e,g, was the same, just isomorphically scaled by a factor of two. In d to h left and right halves of the 2- and 3-ring stimulus arrays are presented separately to aid visibility. d) The Blue Wide-field stimulus, e) Yellow Macular, f) Yellow Wide-field, g) red-green (RG) Macular, h) RG Wide-field. Different parts of the fields of normal persons differ in sensitivity, therefore individual regions varied in intensity to try to balance the size of the responses from each region

Pseudo-random trains of transient onset stimuli (33 ms duration) appeared at a mean interval of 4 s per stimulus-region. There were 44 stimulus-regions per eye (Fig. 1), providing 22 stimuli/s in total. Each region was tested 96 times. Pupil responses were measured at 60/s under infrared illumination using two video cameras. Fixation was monitored online and data obtained during blinks or fixation losses deleted. Segments with less than 85% of their data were repeated. The luminance of each stimulus for each protocol varied depending on its location in the visual field (Table 1, ESM). This luminance balancing method improves signal quality and has been described elsewhere [21].

Table 1 Characteristics of control and type 1 diabetes subjects, by retinopathy status

The average response waveforms for each test region were extracted from raw pupillary responses by multiple linear regression [22]. Pupil contractions were scaled by the mean pupil size over the protocol period and standardised to 3.5 mm and then transformed to decibel (dB) sensitivities [17, 22]. The times-to-peak contraction (delays) were expressed in milliseconds. With recording of both direct and consensual pupil responses to the visual stimuli to each of 44 regions of the retina of each eye, a total of 176 pairs of averaged sensitivities and delay measures was available for analysis [17, 22] for each subject.

Additional eye assessments

On one of the two visits, subjects had 24–2 SITA Fast perimetry (Humphrey Field Analyzer (HFA); Carl Zeiss Meditec, Inc., Dublin, CA) performed. Additionally, their pupils were dilated following mfPOP testing and five 45° fundus images/eye (CR-2 Retinal Camera, Canon Inc., Tokyo, Japan) were acquired equivalent to the seven 30° photos of the Early Treatment Diabetic Retinopathy study (ETDRS) [23]. ETDRS scores were determined by the Retinal Vascular Imaging Centre (East Melbourne, Vic). Classification of DR was according to the ETDRS guidelines, were recorded at baseline and ranged between L10 to L43 (Table 1) [24]. The reading centre also provided central scores (ETDRScen) using the central 45° fundus image, which corresponded more closely to the mfPOP stimulus area. We subsequently categorised the participants according to their grade of ETDRS score across both the ETDRS area and the central ETDRS area. This resulted in three severity categories (Table 1), severity 1 for ETDRS and ETDRScen corresponding to both eyes having a normal appearing fundus. Severity 2 representing T1D eyes with ETDRS and ETDRScen grading of no retinopathy and Severity 3 represents eyes with retinopathy. For the PCA factors the scores (how strong each factor was in each subject) were split into three groups according to the 33rd and 66th percentiles of the particular scores. A priori we did not know if positive or negative scores were associated with diabetic eye damage so were compared ROC analysis for positive and negative versions of the scores, and then selected the version that agreed with other disease severity markers for further analysis.

Statistical analysis

Was performed using MATLAB (2016b Mathworks Inc., Natick, MA). An objective of this study was to examine correlation between mfPOP performance and 16 T1D subject variables (Table 1), reduced to a small number of independent uncorrelated factors using principal components analysis that describe the cohort. Details of the method are given elsewhere [25]. We used multiple regression-linear models to examine the extent to which pupillary response amplitudes and delays, adjusted for age and gender, predicted the PCA factor scores, ETDRS, and ETDRScen scores. Receiver Operator Characteristic (ROC) analysis quantified the diagnostic power of the stimulus protocols as the percent area under the ROC curves (AUCs). The ROC analysis utilized deviations from control values for the 44 times to peak and sensitivities/eye, and also scores based upon combined delay and sensitivity using our published method from our earlier diabetes study [17]. The combined scores are linear combinations of the per-region sensitivities and delays differences from normal. Additional ROC analyses were also computed for three disease severity ratings based upon ETDRS, ETDRScen, and the PCA factor scores. Each of the severity rating levels was adjusted to include about equal numbers of subjects or eyes as appropriate.


Subject characteristics

Table 1 presents the control and T1D subject characteristics including demographics, ocular assessments other than mfPOP, and metabolic and diabetes complication assessments. The table segregates those with T1D patients into those without retinopathy (n = 13) and with retinopathy (n = 12) according to the ETDRS score (≤10 or > 10, respectively). The control and T1D subjects were well matched for age and gender. Overall, the T1D subjects had long-standing diabetes (24.0 ± 11.1 years; range, 7 to 46 years) with moderate diabetes control reflected in their mean HbA1c levels of the previous 5 years (61 ± 6 mmol/mol; 7.7 ± 0.7%). One T1D subject had an estimated glomerular filtration rate of < 60 ml/min (eGFR 57 ml/min) and one had microalbuminuria (urinary albumin creatinine ratio 8.0 mg/mmol), with all others having no clinically significant nephropathy. Five T1D subjects had evidence of peripheral neuropathy (biothesiometer score > 10), and 13 subjects had evidence of mild non-proliferative retinopathy (ETDRS score of > 10–43) in at least one eye. Correlation between patient ETDRS and ETDRScen scores was moderate at 0.64 (p < 0.01). Of the T1D subject variables, only the 5-year mean HbA1c levels and serum triglyceride concentrations were significantly higher in those subjects with compared to those without DR. Of note, BMI was also higher in subjects with DR, but this did not reach statistical significance (p = 0.056).

Prediction of retinopathy - regression modelling

We first explored which of the 16 patient variables might predict retinopathy. A stepwise regression model fitting DR vs no DR, based on the ETDRS scores, selected BMI as the only significant predictor (p = 0.014, F1,23 = 7.03). A similar model regressing upon the ETDRScen data selected serum creatinine and the 5-year HbA1c mean as significant predictors (model p = 0.005, F2,22 = 6.73). While these models indicated that some patient variables were important, stepwise regression can be unreliable [26]. To gain further insight we decided to use PCA based factor analysis to find a small number of factors within the patient variables that explain much of the variance in those measures.

Metabolic, tissue-injury and lipid PCA factors

The largest three PCA factors explained 59% of the variance in the 16 T1D subject variables. We correlated the scores from those 3 factors with the original T1D subject variables. The upper part of Table 2 shows the 9 T1D subject variables that had the highest correlations, and in bold those with the greatest factor loadings. PCA factor 1, which explained 23.8% of the variance, was most associated with HbA1c levels, BMI and triglyceride levels, such that we refer to it as the “Metabolic” variable based factor (Table 2, left 2 columns). PCA factor 2, explained 20.9% of the variance, was most associated with renal function, peripheral neuropathy, and AGE scores, as well as diabetes duration, and we refer to it as the “Tissue-injury” factor (Table 2, central 2 columns). PCA factor 3, explained 14.5% of the variance, and was associated with total cholesterol and low-density lipoprotein cholesterol, and so we refer to it as the “Lipid” factor (Table 2, right 2 columns). While eGFR is significantly correlated with each of the three PCA factors, the respective factor loadings for eGFR (possible range ± 100%) were 43.4% with PCA factor 1, 49.6% with factor 2 and 6.3% with factor 3; therefore eGFR was contributing mostly to PCA factor 2, the Tissue-injury factor. Only the nine patient variables most correlated with the PCA factors are presented in Table 2, as from 10 onward, both the correlation strengths and factor loadings dropped markedly.

Table 2 Candidate independent PCA factors of metabolic control/tissue-injury

Association of PCA factors with retinopathy

The lower part of Table 2 shows the correlations between the mean between-eye ETDRS and ETDRScen scores, and the Metabolic, Tissue-injury and Lipid factor scores. Of note, the correlations between both ETDRS scores and the Metabolic and Tissue-injury factor scores were at least moderate, with the ETDRScen score also being significantly associated with the Lipid factor.

Associations of mfPOP response delays and sensitivities with retinopathy and PCA factors

The results of multiple linear regression models are presented for response delays (Table 3) and response sensitivity (Table 4). For these models we split the values of the ETDRS and ETDRScen ratings, and the three types of PCA factor scores, into three severity categories (Methods). This was done to characterise the effects of disease severity as measured by those methods. Only models containing ETDRS, ETDRScen, and the Metabolic PCA factor showed strong associations with mfPOP delays (Table 3). The models were referenced to normal males (top row for each model). Age effects were non-significant and so are not shown. Females had shorter times to peak. The occasional non-significant results of the disease severity ratings are in bold, all the others were significant (p < 0.05). The goodness of fit (R2) increased somewhat from the ETDRS model through to the Metabolic factor, rising from 0.359 to 0.420. Delays increased more uniformly with increasing ETDRScen and Metabolic severity ratings. The only significant results for the Tissue-Damage and Lipid factors were for their middle severity rating and paradoxically both indicated decreased delays. Since the primary objective of this study was to compare the diagnostic power of the methods, we compare that for ETDRS, ETDRScen, Tissue-Damage and Metabolic factors below.

Table 3 Summary of multiple linear regression models of mfPOP response delays for each stimulus protocol against disease severity ratings based on ETDRS and PCA factor 1 (Metabolic) scores, corrected for age and sex
Table 4 Summary of multiple linear regression models of mfPOP sensitivities against disease severity ratings based on ETDRS and PCA factor 1 (Metabolic) scores, corrected for age and sex

Table 4 provides results of similar models for decibel sensitivity for each eye. Here the (infrequent) significant values are in bold (p < 0.05), and non-significant figures are in grey. Marginally significant values (p > 0.05 and < 0.08) are in plain text and are marked with an*. Negative values indicate suppressed sensitivity. As for delays, the mean R2 increased from the ETDRS to the Metabolic factor models, rising from 0.220 to 0.282. In general response suppression increased with increasing severity ratings. The model for the Metabolic factor severity ratings had the largest number of significant coefficients, six. Yellow-Wide was perhaps the best performing protocol, with 2 severities being significant for each severity type. The blue protocol performed worst, having R2-values < 0.180 for all protocols. No significant results were obtained for the Tissue-Damage or Lipid factors.

To reduce the effects of multiple comparisons the models of Tables 4 and 5 were based on means of mfPOP functional performance of each eye. Previous mfPOP studies of T2D indicated that a few severely affected regions can be highly diagnostic in diabetes [17, 18]. We explored this using ROC analysis, the inputs for which were deviations from the normative data at each point in the visual field of each eye.

Table 5 AUC for discriminating No DR vs. mild to moderate DR in eyes classed as Metabolic severity rating 3

We generated 20 ROC plots of sensitivity (true-positive rate) on the false-positive rate for sensitivities and delays for each protocol for each of the three-level disease severity ratings. The first to 20th ROC plots were based on the number (N) of the worst deviations from normal used, where N = 1 is the single worst deviation from normal in each field, N = 2 is the mean of the worst two deviations, and so on up to the mean of the worst 20 deviations. This simple method provides unbiased insights into the reliability of small numbers of deviating points.

The panels of Fig. 2 illustrate the results of such calculations for discriminating normal control eyes from T1D eyes, comparing control subjects and eyes in the most severe disease rating as determined by: ETDRScen scores (left column, Fig. 2a), the Metabolic scores (middle column, Fig. 2a), and the Tissue-injury scores (right column, Fig. 2c). Across protocols the Metabolic ratings gave the largest number of high AUC values for both sensitivities and delays. N worst deviations in the range 1 to 3 generally gave the best performance. All protocols appeared to be discriminatory for the Metabolic factor, although green on red appeared best (Fig. 2b). Blue on blue was the worst with respect to detecting retinopathy (Fig. 2a).

Fig. 2
figure 2

Percent Areas under ROC plots (AUCs ± SE) for the mean of N worst amplitude and time to peak deviations for the five protocols (rows). Discrimination was between the eyes of T1D subjects in the highest of three severity categories of Column a) ETDRScen; Column b) the Metabolic factor; and Column c) the Tissue-injury factor and normal control eyes. For most protocols the best diagnostic power was achieved for the first few worst regions. For ETDRScen the eyes were classed as moderate non-proliferative diabetic retinopathy, Levels 35 and 43. For the Metabolic and Tissue-injury severities the included eyes could range from ETDRS 10 to 43

Since there are a large number of such plots, we sought to summarise the outcomes by reducing each to one summary AUC value by taking the value for the single worst region (N worst =1 in Fig. 2). Figure 3 shows all results for the Metabolic and Tissue-injury PCA factors and the ETDRScen and ETDRS scores. Data are presented for delays, and the somewhat better performing combined sensitivity and delay scores (Methods). In order to compare AUC for the patient-wise PCA factors and the eye-wise ETDRS scores, we combined the latter by taking the maximum score between eyes for an individual. Thus, Severity 1 for ETDRS and ETDRScen corresponds to both eyes having a normal appearing fundus. Severity 2 represents T1D eyes with ETDRS and ETDRScen grading of no retinopathy and Severity 3 represents eyes with retinopathy.

Fig. 3
figure 3

Each bar is the ROC values for the mean of the single worst region for each of the T1D eyes and control eyes, i.e. the first point in plots like those in Fig. 2. The error bars are the SE. The performance of the 5 protocols (legend) to discriminate between eyes of T1D subjects in each of the three severity ratings (Sev1 to Sev3) of the Metabolic scores, Tissue-injury scores, ETDRScen, and ETDRS (rows) and normal control eyes. Panel (a) gives the outcomes for the combined sensitivity and delay scores (Methods), (b) is delay data as in Fig. 2. Generally, the Metabolic scores were best at segregating eyes with minor damage from those with more severe functional change from control eyes, i.e. the AUC values

Outcomes for the Metabolic ratings were possibly more concordant with those for ETDRScen than with ETDRS. As compared to the results for prediction of the different severities of the Tissue-injury and ETDRS scores, functional assessment as assessed by mfPOP using combined sensitivity and delay scores most strongly predicted subjects in the most severe Metabolic factor category, achieving AUCS over 90%, over the least severe category (Fig. 3a). Of note, greater change in AUCs across severity categories for predicting the Metabolic Factor, ETDRScen and ETDRS scores was evident with the blue on blue stimuli (Fig. 3a). Similar findings were evident using the mfPOP delays only, but with less evident change in AUCs across severity categories using the blue stimuli (Fig. 3b). None of the mfPOP protocols were able to better separate T1D subjects with the most severe Tissue-factor score from those with the least severity category from control subjects.

An important issue for prognostication is how well mfPOP distinguishes patients with normal fundus appearance from those with mild to moderate retinopathy. We did ROC analysis discriminating these patients, repeating it for the ETDRS, ETDRScen, Metabolic and Tissue-injury severity ratings. The Metabolic biomarker produced the best discrimination and the data are given in Table 5. The tabulated values are for the mean of the AUCs for the 6 to 10 worst regions in the field, for Metabolic severity rating 3 (Sev3). The Wide-field protocols seemed to perform better than macular ones, Yellow Wide the best overall producing AUC values across all three methods of ≥79%.


Some of the stimulus protocols were motivated by previous results, e.g. red-green stimuli for retinal ganglion cell death [21], which has been reported in early DR [27]; and yellow wide-field [17, 18] and macular stimuli in early DR [18] and AMD [12, 13]. Reports that diabetes damages short-wavelength cones motivated the transient blue stimuli [28]. The transient blue stimuli do not drive the slowly responding melanopsin-containing retinal ganglion cells [29, 30].

The purpose of this study was to measure the effects of T1D upon retinal function in terms of relative diagnostic power of the five methods. Unique to this study is the influence of 16 patient variables on localised retinal function. The Metabolic Biomarker was best able to segregate controls from patients (Tables 3 and 4, and Fig. 3), and also patients with normal fundi from those with mild to moderate DR (Table 5 and Fig. 3 cf. Sev1 and Sev2/3). The other two potential biomarkers performed less well in ROC analysis (Figs. 2 and 3). Combining per-region sensitivity and delay data provided the best diagnostic power (Fig. 3a, Table 5), as we have reported for T2D [17]. We have also reported that wide-field stimuli outperformed macular stimuli in early-stage T2D [18], in agreement with the findings here for the Yellow and RG protocols. That being said the RG Macular stimulus performed well (Figs. 2 and 3). The Metabolic Biomarker was correlated with HbA1c (Table 2) and in a previous study ranking the severity of the T2D patients by severity of their HbA1c levels produced good correlation between AUC achieved and HbA1c levels [18].

We have previously reported AUCs for discriminating T2D patients and controls 87.1 ± 6.3%. Here the result was more modest at 76.7 ± 8.8 for classification of severity by ETDRScen (Fig. 3 Sev1). In both cases this was for persons with two normal appearing eyes. Classification by eye, where a fellow eye could be worse, produced AUCs about 2% higher. Aside from our mfPOP studies of diabetic eye disease [17,18,19], other functional measures indicate that the retinal neuropathy can precede classical diabetic retinopathy [28, 31,32,33,34,35]. Here we found mfPOP discriminated patients with and without DR, the Yellow Wide protocol achieving AUCs between 79.0 and 85.9% (Table 5). In a clinical setting with a high prior probability of encountering a patient with retinal damage this level of diagnostic power might be useful to indicate which patients should be treated with Fenofibrate or other interventions.

AUC values in our previous studies [17, 18] were not significantly different to those reported here. In those studies, asymmetry between eyes produced significantly higher AUC values. We did not see that here and have not reported on asymmetries due to lack of space. This might suggest a difference between type 1 and type 2 diabetes eyes. Alternatively, the results might be a property of the particular subject groups.

The effect pupillary autonomic neuropathy is relatively minor because frequently the diagnostic power is carried by the few most deviating points in a field, as we have shown before [17,18,19], and in Fig. 2. Obviously neuropathy of the pupil cannot change the sensitivity or delay of just a few parts of the visual field. Pupil neuropathy would mimic global effects. At the request of a reviewer we examined the correlation between eyes of the regional data. In fact larger and smaller than normal sensitivities or delays can be seen frequently in the same retina [19]. We calculated the correlation between the 1672 to 2200 field points for each of the two eyes. Anatomically equivalent naso-temporal locations in the two eyes were compared. Across the 5 protocols (order as in Table 5) the values for sensitivities were: 0.421, 0.519, 0.476, 0.481, 0.487, and delays: 0.658, 0.487, 0.502, 0.489, 0.498. The low correlations go to illustrate the relative independence of regional values in the two eyes.

The blue stimulus was among the worst performing. Much work on pupillography has focused on blue stimuli with flash durations around a second (reviewed, [29, 30]). This is because things like the steady-state diameter of the pupils are driven by the very slow melanopsin driven responses of special retinal ganglion cells. We have created slow mfPOP designed to drive these cells [29, 30]. Like the faster blue stimuli used here they had lower diagnostic power than yellow stimuli. We have shown that transient blue-containing mfPOP stimuli like those used here, are more affected by changes in visual attention than yellow stimuli [36]. Blue stimuli are also more likely to be affected by lens brunescence and light scattering.

Although the results here were promising we have recently introduced a new mfPOP stimulus variant called clustered-volleys that outperforms stimuli like those used here. The first demonstration was in early macular degeneration [37]. Another report on 6 studies of 96 normal subjects showed that the signal to noise ratios for the clustered-volleys method was between 35 and 57% larger than methods like those here (p < 0.001 in 5/6 studies) (Carle et al. Clustered volleys stimulus presentation for multifocal pupil perimetry, In revision). Using those stimuli we have compared mfPOP with structural data from optical coherence tomography and macular perimetry in T2D patients with and without mild macular oedema [19], and have shown that mfPOP functional change over time correlates well with structural change (Sabeti et al. Objective Perimetry identifies functional progression and recovery in mild diabetic macular oedema, In revision).

Published investigations of retinal function in diabetes have utilized only ETDRS-like measures and blood glucose control [38,39,40,41,42]. To our knowledge, this is the first report on retinal function in T1D and its association with tissue-damage and metabolic patient variables and moderate ETDRS scores. Measuring visual function in combination with Biomarkers derived from such patient data may be useful in assessing the efficacy of new treatments aimed at early stages of DR, and identifying eyes at risk of progression of retinopathy. Longitudinal studies will be required to confirm this. The Metabolic and Tissue-Damage factors are mutually uncorrelated summary measures from the patients that may reflect independent tissue-damage processes. The validity of these Biomarkers will require further study to determine if in larger cohorts the form of these markers are stable. In addition to the PCA based factors we also checked if non-orthogonal factors might explain the patient variables better, they did not. The PCA method has the advantage that the major independent sources of variation in the data are identified in an assumption free manner. Future studies with more subjects should re-examine the PCA factors and scores, and that information could underpin the creation of models for more formal severity markers. Table 2 is a start on examining what variables should be included in such models, which would likely include the effects of age and sex, and possibly the type of diabetes.

Limitations of our study were the relatively long duration of T1D (23.3–24.7 years) and the sample size for each retinopathy severity. Also, we did not take blood samples from control subjects. Finally repeatability was not investigated here, however we have shown good repeatability previously in T2D [17]. Using more modern mfPOP methods on 40 glaucoma patients and 95 match controls test-retest variability was half that of standard automated perimetry [43]. Further investigations of whether improvements in retinal function measured by mfPOP can be achieved by improving long-term blood glucose levels are warranted.

Availability of data and materials

The datasets used and analysed in the current study are available from the corresponding author on reasonable request due to confidentially requirements on patent applications pending.



Age-related macular degeneration


Advanced glycation end-products


Anti-vascular endothelial growth factor


Blood glucose level


Body mass index


Diabetic retinopathy


Early treatment of diabetic retinopathy study


Generalised diabetes-related tissue injury


Multifocal pupillographic objective perimetry


Multifocal visual evoked potentials


Pattern deviations


Principle component analysis


Type 1 diabetes


Type 2 diabetes


  1. The Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med. 1993;329(14):977–86.

    Article  Google Scholar 

  2. Nathan DM, Group DER. The diabetes control and complications trial/epidemiology of diabetes interventions and complications study at 30 years: overview. Diabetes Care. 2014;37(1):9–16.

    Article  CAS  PubMed  Google Scholar 

  3. Klein R, Knudtson MD, Lee KE, Gangnon R, Klein BE. The Wisconsin epidemiologic study of diabetic retinopathy: XXII the twenty-five-year progression of retinopathy in persons with type 1 diabetes. Ophthalmology. 2008;115(11):1859–68.

    Article  PubMed  Google Scholar 

  4. Tesfaye S, Boulton AJ, Dyck PJ, et al. Diabetic neuropathies: update on definitions, diagnostic criteria, estimation of severity, and treatments. Diabetes Care. 2010;33(10):2285–93.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Sivaprasad S, Elagouz M, McHugh D, Shona O, Dorin G. Micropulsed diode laser therapy: evolution and clinical applications. Surv Ophthalmol. 2010;55(6):516–30.

    Article  PubMed  Google Scholar 

  6. van Wijngaarden P, Coster DJ, Williams KA. Inhibitors of ocular neovascularization: promises and potential problems. JAMA. 2005;293(12):1509–13.

    Article  PubMed  Google Scholar 

  7. Cheung N, Mitchell P, Wong TY. Diabetic retinopathy. Lancet. 2010;376(9735):124–36.

    Article  PubMed  Google Scholar 

  8. Tan GS, Cheung N, Simo R, Cheung GC, Wong TY. Diabetic macular oedema. Lancet Diabetes Endocrinol. 2017;5:143–55.

  9. Chaturvedi N, Porta M, Klein R, et al. Effect of candesartan on prevention (DIRECT-prevent 1) and progression (DIRECT-protect 1) of retinopathy in type 1 diabetes: randomised, placebo-controlled trials. Lancet. 2008;372(9647):1394–402.

    Article  CAS  PubMed  Google Scholar 

  10. Accord Eye Study Group, Chew EY, Ambrosius WT, et al. Effects of medical therapies on retinopathy progression in type 2 diabetes. N Engl J Med. 2010;363(3):233–44.

    Article  CAS  PubMed Central  Google Scholar 

  11. Keech AC, Mitchell P, Summanen PA, et al. Effect of fenofibrate on the need for laser treatment for diabetic retinopathy (FIELD study): a randomised controlled trial. Lancet. 2007;370(9600):1687–97.

    Article  CAS  PubMed  Google Scholar 

  12. Sabeti F, James AC, Essex RW, Maddess T. Multifocal pupillography identifies retinal dysfunction in early age-related macular degeneration. Graefes Arch Clin Exp Ophthalmol. 2013;251(7):1707–16.

    Article  PubMed  Google Scholar 

  13. Sabeti F, Maddess T, Essex RW, et al. Multifocal pupillography in early age-related macular degeneration. Optom Vis Sci. 2014;91(8):904–15.

    Article  PubMed  Google Scholar 

  14. Sabeti F, Maddess T, Essex RW, James AC. Multifocal pupillographic assessment of age-related macular degeneration. Optom Vis Sci. 2011;88(12):1477–85.

    Article  PubMed  Google Scholar 

  15. Sabeti F, Maddess T, Essex RW, James AC. Multifocal pupillography identifies ranibizumab-induced changes in retinal function for exudative age-related macular degeneration. Invest Ophthalmol Vis Sci. 2012;53(1):253–60.

    Article  CAS  PubMed  Google Scholar 

  16. Rosli Y, Bedford SM, James AC, Maddess T. Photopic and scotopic multifocal pupillographic responses in age-related macular degeneration. Vis Res. 2012;69:42–8.

    Article  CAS  PubMed  Google Scholar 

  17. Bell A, James AC, Kolic M, Essex RW, Maddess T. Dichoptic multifocal pupillography reveals afferent visual field defects in early type 2 diabetes. Invest Ophthalmol Vis Sci. 2010;51(1):602–8.

    Article  PubMed  Google Scholar 

  18. Sabeti F, Nolan CJ, James AC, Jenkins A, Maddess T. Multifocal pupillography identifies changes in visual sensitivity according to severity of diabetic retinopathy in type 2 diabetes. Invest Ophthalmol Vis Sci. 2015;56(8):4504–13.

    Article  PubMed  Google Scholar 

  19. Rai BB, Maddess T, Carle CF, et al. Comparing objective perimetry, matrix perimetry, and regional retinal thickness in early diabetic macular oedema. Trans Vis Sci Tech. 2021;10(13(32)):1–12.

    Google Scholar 

  20. Sabeti F, James AC, Carle CF, et al. Comparing multifocal pupillographic objective perimetry (mfPOP) and multifocal visual evoked potentials (mfVEP) in retinal diseases. Sci Rep. 2017;7:45847.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  21. Carle CF, James AC, Kolic M, Essex RW, Maddess T. Luminance and colour variant pupil perimetry in glaucoma. Clin Exp Ophthalmol. 2014;42(9):815–24.

    Article  PubMed  Google Scholar 

  22. Carle CF, James AC, Maddess T. The pupillary response to color and luminance variant multifocal stimuli. Invest Ophthalmol Vis Sci. 2013;54(1):467–75.

    Article  PubMed  Google Scholar 

  23. Early treatment diabetic retinopathy study research group. Fundus photographic risk factors for progression of diabetic retinopathy. ETDRS report number 12. Ophthalmology. 1991;98(5 Suppl):823–33.

    Google Scholar 

  24. Early photocoagulation for diabetic retinopathy. ETDRS report number 9. Early treatment diabetic retinopathy study research group. Ophthalmology. 1991;98(5 Suppl):766–85.

    Google Scholar 

  25. Seamons JW, Barbosa MS, Bubna-Litic A, Maddess T. A lower bound on the number of mechanisms for discriminating fourth and higher order spatial correlations. Vis Res. 2015;108:41–8.

    Article  PubMed  Google Scholar 

  26. Sabeti F, Lane J, Rohan EMF, Essex RW, McKone E, Maddess T. Relationships between retinal structure and function and vision-related quality of life measures in advanced age-related macular degeneration. Graefes Arch Clin Exp Ophthalmol. 2021;259:3687–96.

  27. van Dijk HW, Kok PH, Garvin M, et al. Selective loss of inner retinal layer thickness in type 1 diabetic patients with minimal diabetic retinopathy. Invest Ophthalmol Vis Sci. 2009;50(7):3404–9.

    Article  PubMed  Google Scholar 

  28. Nitta K, Saito Y, Kobayashi A, Sugiyama K. Influence of clinical factors on blue-on-yellow perimetry for diabetic patients without retinopathy: comparison with white-on-white perimetry. Retina. 2006;26(7):797–802.

    Article  PubMed  Google Scholar 

  29. Carle CF, James AC, Kolic M, Essex RW, Maddess T. Blue multifocal pupillographic objective perimetry in glaucoma. Invest Ophth Vis Sci. 2015;56:6934–403.

    Google Scholar 

  30. Ali EN, Carle CF, Lueck CJ, Kolic M, Maddess T. Assessing migraine patients with multifocal pupillographic objective perimetry. BMC Neurol. 2021;21(211):1–12.

    Google Scholar 

  31. Henricsson M, Heijl A. Visual fields at different stages of diabetic retinopathy. Acta Ophthalmol. 1994;72(5):560–9.

    Article  CAS  Google Scholar 

  32. Verrotti A, Lobefalo L, Petitti MT, et al. Relationship between contrast sensitivity and metabolic control in diabetics with and without retinopathy. Ann Med. 1998;30(4):369–74.

    Article  CAS  PubMed  Google Scholar 

  33. Verma A, Rani PK, Raman R, et al. Is neuronal dysfunction an early sign of diabetic retinopathy? Microperimetry and spectral domain optical coherence tomography (SD-OCT) study in individuals with diabetes, but no diabetic retinopathy. Eye (Lond). 2009;23(9):1824–30.

    Article  CAS  Google Scholar 

  34. Harrison WW, Bearse MA Jr, Ng JS, et al. Multifocal electroretinograms predict onset of diabetic retinopathy in adult patients with diabetes. Invest Ophthalmol Vis Sci. 2011;52(2):772–7.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Montesano G, Gervasoni A, Ferri P, et al. Structure-function relationship in early diabetic retinopathy: a spatial correlation analysis with OCT and microperimetry. Eye (Lond). 2017;31(6):931–9.

    Article  CAS  Google Scholar 

  36. Rosli Y, Carle CF, Ho Y, et al. Retinotopic effects of visual attention revealed by dichoptic multifocal pupillography. Sci Rep. 2018;8:2991.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  37. Sabeti F, Maddess T, Saikal A, et al. Multifocal pupillography in early age-related macular degeneration. Opt Vis Sci. 2014;91:904–15.

    Article  Google Scholar 

  38. Jackson GR, Scott IU, Quillen DA, Walter LE, Gardner TW. Inner retinal visual dysfunction is a sensitive marker of non-proliferative diabetic retinopathy. Br J Ophthalmol. 2012;96(5):699–703.

    Article  PubMed  Google Scholar 

  39. Tyrberg M, Lindblad U, Melander A, et al. Electrophysiological studies in newly onset type 2 diabetes without visible vascular retinopathy. Doc Ophthalmol. 2011;123(3):193–8.

    Article  CAS  PubMed  Google Scholar 

  40. Juen S, Kieselbach GF. Electrophysiological changes in juvenile diabetics without retinopathy. Arch Ophthalmol. 1990;108(3):372–5.

    Article  CAS  PubMed  Google Scholar 

  41. Schneck ME, Bearse MA Jr, Han Y, et al. Comparison of mfERG waveform components and implicit time measurement techniques for detecting functional change in early diabetic eye disease. Doc Ophthalmol. 2004;108(3):223–30.

    Article  PubMed  Google Scholar 

  42. Wolff BE, Bearse MA Jr, Schneck ME, Barez S, Adams AJ. Multifocal VEP (mfVEP) reveals abnormal neuronal delays in diabetes. Doc Ophthalmol. 2010;121(3):189–96.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Maddess T, van Kleef JP, Kolic M, et al. Comparing macular and wide-field objective perimetry. In: World Glaucoma Congress. vol. ABSUB-824; 2021.

    Google Scholar 

Download references


Not applicable.


This research was supported by the Australian Research Council through the ARC Centre of Excellence in Vision Science (CE0561903) and the Our Health in Our Hands (OHIOH) ANU intramural grant.

Author information

Authors and Affiliations



Project proposal and planning: F.S, C.J.N, A.J.J, T.M. Data collection: F.S, L.B, C.E.C, V.C, M.C. Analysis: C.F.C, T.M. Manuscript preparation and editing: F.S, C.E.C, V.C, M.C, and C.J.N, T.M. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Faran Sabeti.

Ethics declarations

Ethics approval and consent to participate

The study was approved by the ACT Health and the Australian National University Human Research Ethics Committees and informed written consent was obtained from all subjects. All methods were carried out in accordance with relevant guidelines and regulations. The study was approved by the ACT Health and the Australian National University Human Research Ethics Committees, adhered to the tenets of the Declaration of Helsinki, and informed written consent was obtained from all subjects.

Consent for publication

Not applicable.

Competing interests

T.M. Stock: EyeCo Pty Ltd., Royalties: Konan Medical USA, Patent: Konan Medical USA, Equipment: ColorDx and EyeKinetix.

C.F.C Royalties: Konan Medical USA, Patent: Konan Medical USA,

A.C.J Royalties: Konan Medical USA.

Other Authors do not have any competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Sabeti, F., Carle, C.F., Nolan, C.J. et al. Multifocal pupillographic objective perimetry for assessment of early diabetic retinopathy and generalised diabetes-related tissue injury in persons with type 1 diabetes. BMC Ophthalmol 22, 166 (2022).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: