Multimodel images from patients 2, 9, and 13 of visits 0, 12, 24,and 36 showing for each visit the foveal avascular zone contour (FAZ), retinalleakage analyzer results, and retinal thickness analyzer results. The retinalleakage analyzer color-coded maps of the blood-retinal barrier permeabilityindexes are shown; retinal thickness analyzer views show white dot densitymaps of the percentage increases in retinal thickness. A, Pattern A, patient2. Note the little amount of retinal leakage over the 4 represented visitsand the normal FAZ contour. This patient showed a slow rate of microaneurysmformation. B, Pattern B, patient 9. Note the high retinal leakage showinga certain degree of reversibility and the normal FAZ contour. This patientshowed a high rate of microaneurysm accumulation over the 3-year follow-upperiod. C, Pattern C, patient 13. Note the reversible retinal leakage andthe development of an abnormal FAZ contour. This patient showed a high rateof micoraneurysm formation.
Lobo CL, Bernardes RC, Figueira JP, de Abreu JRF, Cunha-Vaz JG. Three-Year Follow-up Study of Blood-Retinal Barrier and Retinal ThicknessAlterations in Patients With Type 2 Diabetes Mellitus and Mild NonproliferativeDiabetic Retinopathy. Arch Ophthalmol. 2004;122(2):211-217. doi:10.1001/archopht.122.2.211
To examine the 3-year alterations of the blood-retinal barrier and changesin retinal thickness occurring in the macular region in 14 eyes of 14 patientswith type 2 diabetes mellitus (DM) and mild nonproliferative diabetic retinopathy.
We classified 14 eyes of 14 patients with type 2 DM and mild nonproliferativediabetic retinopathy, as having disease levels 20 (microaneurysms only) or35 (microaneurysm plus retinal hemorrhage[s] and/or hard exudates) of WisconsinCard-Sorting Test grading, by using 7-field stereoscopic fundus photographs.We examined them 7 times at 6-month intervals, using fundus photography, fluoresceinsodium angiography, the retinal leakage analyzer (RLA)–modified confocalscanning laser ophthalmoscope, and the retinal thickness analyzer. The retinalleakage and retinal thickness maps were aligned and integrated into 1 image.Data from the group of individuals with type 2 DM were compared with thoseof a healthy control population (n = 14; mean age, 48 years; age range, 42-55years) to establish reference maps for the RLA and retinal thickness analyzers.
Areas of abnormally increased fluorescein leakage were detected in alleyes examined at baseline. The sites of increased fluorescein leakage reachedvalues as high as 483% above normal levels, but in 20 of the total 95 examinationsperformed, fluorescein leakage returned to normal levels. Every eye that showedreversal to normal levels of fluorescein leakage showed stabilization or adecrease in glycosylated hemoglobin A1c values at the same visit.When comparing the RLA-leaking sites among the 7 examinations, they remained,in general, in the same locations, but there was a clear fluctuation in thepercentage of increases. No clear correlation was observed among the locationof areas of increased retinal thickness and RLA-leaking sites, the numberof microaneurysms, or the glycosylated hemoglobin A1c values. Microaneurysmson fundus photographs showed different cumulative incidences throughout thefollow-up period in the different eyes. Associations between these differentabnormalities suggest specific patterns of evolution of type 2 DM–relatedretinal disease.
The dominant alteration in the retina of patients with type 2 DM andmild nonproliferative retinopathy is the presence of RLA-leaking sites. Thisdamage seems to be reversible and directly associated with variations in glycemicmetabolic control. Together with the intensity and persistence of RLA-leakingsites, the rates of microaneurysm accumulation and alterations of the fovealavascular zone may characterize different genetically based phenotypes ofdiabetic retinopathy.
Retinopathy is a frequent microvascular complication of diabetes mellitus(DM) that remains a major therapeutic challenge.1 Toprevent and improve the treatment of diabetic retinopathy, it is fundamentalthat we know the evolution of the earliest changes that occur in the retinaaffected by DM and how these changes relate to the progression of retinopathy.
The accepted methods of assessing retinopathy are based on changes invisual acuity and on findings noted on fundus photographs. Examination ofthe initial alterations in retinal disease caused by DM must involve othermethods, such as measuring alterations of the blood-retinal barrier (BRB)and changes in retinal thickness.2 Methodsof multimodal imaging of the ocular fundus have recently been developed bycombining information obtained from different sources, such as fundus photographsand fluorescein sodium angiography, with measurements obtained with instrumentationsuch as the retinal leakage analyzer (RLA) and the retinal thickness analyzer(RTA).3,4 To our knowledge, weare providing the results of the first 3-year follow-up study of changes occurringin the macular region of eyes affected by type 2 DM and mild nonproliferativediabetic retinopathy using new methods of multimodal macula mapping.
Fourteen eyes from 14 patients (5 men and 9 women), aged 44 to 68 years(mean [SD], 55  years), with type 2 DM and mild nonproliferative diabeticretinopathy were studied. Characterized on 7-field stereoscopic fundus photographs,these 14 eyes met the inclusionary criteria of having either level 20 (microaneurysmsonly) or level 35 (microaneurysm plus retinal hemorrhage[s] and/or hard exudates)disease according to the Wisconsin Card-Sorting Test grading. We examinedthese 14 eyes 7 times using stereoscopic fundus photography, fluorescein sodiumangiography, RLA (Carl Zeiss International, Oberkochen, Germany), and RTA(Talia Technology Ltd, Neve Ilan Zion, Israel) at 6-month intervals duringa 3-year period. All patients had corrected visual acuity of 20/20. When botheyes met the inclusionary criteria, the right eye was chosen as the studyeye. An abridged version of the Early Treatment Diabetic Retinopathy Study(ETDRS) scale was used to characterize the fundus lesions..
The patients were followed-up by the same diabetologist and remainedwithin the limits of acceptable medical control. Their glycosylated hemoglobinA1c (HbA1c) values at baseline ranged from 5.4% to 9.4%(the percentage given for the total hemoglobin level), and the duration oftype 2 DM ranged from 2 to 15 years (Table1).
The tenets of the Declaration of Helsinki were followed and approvalof the institutional review board was obtained for this study. Informed consentwas obtained from the patients before they were enrolled in the study.
Stereoscopic fundus photography was performed according to the EarlyTreatment Diabetic Retinopathy Study protocol. Stereoscopic pairs of fieldswere obtained using a 30° fundus camera to be classified as level 20 orlevel 35 according to the Wisconsin Card-Sorting Test grading, although onlyfield 2 was used to examine and correlate with other data (retinal leakageand retinal thickness).
Fluorescein angiography was performed using a fundus camera with a 45°-fieldafter patients received an injection of fluorescein sodium of 14 mg/kg ofbody weight.
The retinal leakage analyzer (RLA) is a new method that quantifies localizedfluorescein leakage from the retina into the vitreous across the BRB in humans.The instrumentation and the entire image processing procedures were describedpreviously.2,4
The RLA obtains images of the fundus with real 3-dimensional information.Two types of information are obtained simultaneously, one for optical imagingand the other for fluorescence measurements being scanned. Axial graphicsof the fluorescein concentrations obtained from the vitreous, representinga volume of 75 × 75 × 2550 µm, were converted into RLA maps.Multiple measurements of retinal leakage can be graphically assembled in afalse-color RLA map that represents the distribution of alterations in theBRB in any chosen area of the total area of the posterior pole under examination.Intravisit and intervisit reproducibility values of the method are ±10.2%and ±13%, respectively.2
The retinal thickness analyzer (RTA) is a quantitative and reproduciblemethod for evaluating retinal thickness.5 Thesystem acquires optical images of the retina in sections by projecting thelaser into the retina at an angle, allowing the reflection or scattering ofthe laser light from the vitreoretinal and chorioretinal interfaces to beviewed. The separation between the reflections (and scatter) from these 2interfaces is a measure of retinal thickness. These data are analyzed andthe retinal thickness is given as a numerical value and as a color-coded map.
Data from the group of individuals with type 2 DM were compared withthose from a healthy control population (n = 14; mean age, 48 years; age range,42-55 years) to establish reference maps for the RLA and RTA.4 Becauseof the differences in resolution between the 2 instruments (RLA and RTA),a value of retinal leakage is obtained for each 75 × 75-µm area,while a single value of retinal thickness covers a 200 × 200-µmarea. To compare information from both the RLA and RTA, the maps are composedof 63 (9 × 7) values, each representing a 300 × 300-µm area,for a total of 2700 × 2100 µm.4 Bycomparing values from the patients with type 2 DM with those of a healthycontrol population, we can compute maps of increased retinal leakage or increasedretinal thickness as percentages.
All of the increases are computed considering the reference mean +2SDs of the healthy control population. An extended description of the dataanalysis method is available elsewhere.2
Correlations between the percentage of increases or decreases in RLAand RTA values occurring between examinations and HbA1c valueswere tested using the χ2 test to compare proportions and thenonparametric Spearman rank correlation coefficient (Base 11.0; SPSS Inc,Chicago, Ill).
Combined information is obtained only from field 2, using color stereoscopicfundus photography, fundus fluorescein angiography, RLA, and RTA. Lesionswere characterized using Early Treatment Diabetic Retinopathy Study definitions.The findings on fundus photographs and fluorescein angiograms were registeredindependently by 3 different observers (C.L.L., J.R.F.A., and J.G.C.-V.).Disagreements were resolved after joint examinations. The alterations recordedin field 2 with the different methods used were as follows.
Red dots considered to be microaneurysms were counted in field 2 ateach examination. Their location was identified and each time a new microaneurysmwas identified in a new location it was counted and added, giving the cumulativeincidence of microaneurysms throughout the study.
During the early phase the foveal avascular zone (FAZ) was outlined.A normal FAZ occurred when a smooth symmetrical outline of the FAZ was observed,with defects no larger than 10% of the entire outline.
Areas of increased fluorescein leakage (300 × 300 µm) areareas in which fluorescein leakage is above the reference made equal to themean + 2 SDs of a healthy control population. The increase is expressed asa percentage over the reference. Retinal leakage analyzer–leaking sitesare identified as the central sites (300 × 300 µm) of the maximumpercentage of increased fluorescein leakage occurring in each area.3,4
Localized measurements of retinal thickness are obtained from a 200× 200-µm area.
The results are summarized in Table1 and Table 2. Duringthe 3-year follow-up period, 12 patients completed the entire set of visits,with the study eye being examined 7 times at 6-month intervals. The other2 patients (patients 6 and 7) developed clinically significant macular edemaand received laser photocoagulation therapy, one during the period beforethe sixth visit and the other before the final visit. In total there were95 visits performed on the 14 patients included in the study.
The RLA–leaking sites, that is, sites of maximal fluorescein leakagefor each area of abnormally increased leakage, reached values as high as 483%above normal levels, but in 20 of the 95 examinations performed, the amountof fluorescein leakage recorded was within normal levels.
Reversibility of the abnormally increased fluorescein leakage was statisticallyassociated with documented improvement in HbA1c levels. There wasa good correlation between stabilization or a decrease in HbA1c levelsbetween successive visits and the occurrence of normal values of fluoresceinleakage. In all 13 visits in which a return to normal values was registered,there was an associated decrease or stabilization in HbA1c levels.There was also good correlation between increases or decreases in HbA1c levels between successive visits and increases or decreases in thepercentage of maximum increases in fluorescein leakage into the vitreous (P = .004).
When comparing the RLA-leaking sites among the 7 examinations in thedifferent eyes examined, they remained, in general, in the same location,but there was a clear fluctuation in the percentage of increase. Eyes showinga higher percentage of increase of abnormal fluorescein leakage showed higherleakage values (patients 3, 7, 8, and 9) throughout the 3-year follow-up period.
On successive visits special attention was given to counting microaneurysmsin a cumulative way. On the first visit, there were a total of 39 red dotsand at the end of the 3-year follow-up period, there was a cumulative countof 122 red dots. The rate of increase in microaneurysms varied between eyes,but increased significantly in the eyes of patients 3, 4, 7, 8, 9, and 13.
Special attention was given to changes in the contour of the FAZ. Changesin the contour of the FAZ were detected in the eyes of patient 6 on the fourthand subsequent visits, in patient 12 on the second and subsequent visits,and in patient 13 on the fifth and subsequent visits.
Increases in retinal thickness were detected in all eyes examined atsome time during the 3-year follow-up period. The sites of maximum percentageof increase in retinal thickness were generally lower than 35% above normalvalues. Values higher than 35% were observed on visit 4 of patient 1, visits3 and 4 of patient 14, and on the last visit of the 2 eyes that were laterdiagnosed as having clinical significant macular edema.
Changes were observed between the location of sites of maximum percentageof increase in retinal thickness when comparing the results of examinationsperformed on different visits in the same eye. No correlation could be establishedbetween changes in retinal thickness and changes in HbA1c values.Similarly, no correlation could be found between the percentage of increasein fluorescein leakage and the percentage of increase in retinal thickness.
When looking for associations between the main alterations found inthis group of eyes with mild nonproliferative diabetic retinopathy from patientswith type 2 DM, followed up for a 3-year period and examined at 6-month intervals,the following 3 major evolutionary patterns emerged (Figure 1).
Pattern A is defined as eyes showing reversible retinal leakage anda relatively low degree of abnormal fluorescein leakage in the RLA, a slowrate of microaneurysm accumulation, and a normal FAZ contour. The durationof the type 2 DM reported in these 7 eyes (patients 1, 2, 4, 5, 10, 11, and14) was 10, 2, 15, 9, 4, 5, and 2 years, respectively. The eye of patient4 has a relatively higher rate of microaneurysm accumulation, possibly associatedwith having a longer duration of type 2 DM.
Pattern B is defined as eyes showing persistent retinal leakage, withhigh fluorescein leakage values at some time during the 3-year follow-up period,relatively high rates of microaneurysm accumulation, and a normal FAZ contour.The duration of the type 2 DM registered in these 4 eyes (patients 3, 7, 8,and 9) was 10, 6, 8, and 2 years, respectively.
Pattern C is defined as eyes showing reversible retinal leakage anda low degree of abnormal fluorescein leakage during most of the 3-year follow-upperiod, variable degrees of progression of microaneurysm accumulation, andthe development of abnormalities in the capillary net surrounding the FAZ.The duration of type 2 DM registered in the 3 eyes fulfilling the requirements(patients 6, 12, and 13) was 10, 15, and 3 years, respectively.
Diabetic retinopathy is fundamentally a posterior pole disease, preferentiallyinvolving the macular area, since its initial stages. Multimodal imaging ofthe macular area was used in this study by combining maps of fluorescein leakageinto the vitreous, maps of changes in retinal thickness, and fundus images.The combination of these methods allows for an integrated view of the differentchanges occurring in the macular area. This prospective 3-year follow-up studyof the macular region, in patients with type 2 DM and mild nonproliferativediabetic retinopathy, using these combined imaging methods, opens new perspectivesfor our understanding of diabetic retinopathy by showing the progression ofthe initial alterations of the retinal microcirculation.
In a span of 3 years, eyes with minimal changes at the start of thestudy (level 20 or level 35 of Early Treatment Diabetic Retinopathy Study–WisconsinCard-Sorting Test grading for fundus lesions) were followed up at 6-monthintervals to monitor the progression of the retinal changes. The most frequentalterations observed were, by decreasing order of frequency, RLA-leaking sites,areas of increased retinal thickness, and the occurrence of microaneurysmsor hemorrhages.
The RLA-leaking sites were a most frequent finding and reached highBRB permeability values in some eyes (patients 3, 6, 7, 8, and 9). These sitesof alteration of the BRB, well identified in RLA-maps, maintained, in mostcases, the same location on successive examinations, but their BRB permeabilityvalues fluctuated greatly between examinations, indicating reversibility ofthis alteration.
There was, in general, a correlation between the BRB permeability valuesand the changes in HbA1c levels occurring in each patient. Thiscorrelation was particularly clear when looking at eyes that showed, at sometime during the 3-year follow-up period, BRB permeability values within thenormal range. A return to normal levels of BRB permeability was, in this studyand in each patient, always associated with a stabilization or a decreasein HbA1c values.
The frequent finding of RLA-leaking sites in these 14 eyes confirmsprevious reports using fluorescein angiography,6 vitreousfluorometry,7,8 and retinal leakageanalysis,4 which show that alteration of vascularpermeability is one of the most frequent alterations occurring in the initialstages of diabetic retinal disease.
Areas of increased retinal thickness were another frequent finding inthese study eyes. They were present in every eye at some time during the 3-yearfollow-up period and were absent, at baseline, in only 2 eyes (patients 2and 10). This confirms previous observations by our group4 andby others.9
However, the areas of increased retinal thickness varied in their locationover subsequent examinations, did not correlate directly with the locationof the RLA-leaking sites, and did not correlate in any clear way with changesin HbA1c levels. They may represent a delayed temporary responsein time to other changes occurring in the retina, such as increased leakage,as we have suggested previously.4
The areas of increased retinal thickness may, indeed, represent in mostcases zones of extracellular edema, an interpretation supported by the frequentshift observed in their location. However, there were eyes showing generalizedchanges in retinal thickness with little retinal leakage (patients 1, 4, 5,and 6). Association with capillary closure was well demonstrated in 1 eye,with clear alterations of the FAZ contour (patient 6), which may indicatethat in some instances intracellular edema associated with ischemia may bedominant.
Microaneurysms were counted in field 2 and their rate of accumulationover the 3-year follow-up period may indicate progression of retinopathy andremodeling of the retinal microvasculature. Increased rates of microaneurysmaccumulation were found in eyes that had more microaneurysms at baseline (patients3, 7, and 13) and higher values of BRB permeability during the 3-year study(patients 3, 7, 8, and 9). The study eye of patient 13 showed a clearly increasedrate of microaneurysm accumulation with little retinal leakage, but had ahigh number of microaneurysms at baseline and developed an abnormal FAZ contour,demonstrating capillary closure.
By combining different imaging techniques, multimodal imaging of themacula made apparent 3 major evolutionary patterns occurring during the 3-yearfollow-up period. One group included eyes with reversible and little abnormalfluorescein leakage, a slow rate of microaneurysm formation, and a normalFAZ contour. Patients included in this group had a longer duration of type2 DM. This group, designated as having pattern A, seems to consist of eyesthat will progress slowly and may never develop clinically significant macularedema or proliferative diabetic retinopathy.
A second group included eyes with persistently high leakage values,indicating an important alteration of the BRB, high rates of microaneurysmaccumulation, and a normal FAZ contour. The duration of type 2 DM in thesepatients was short. All these features suggest a rapid and progressive formof the retinal disease. This group, designated as having pattern B, couldalso identify a "wet" form of diabetic retinopathy.
The third group included eyes with reversible retinal leakage and anabnormal FAZ contour. This group is less well characterized considering thefew eyes that showed an abnormal FAZ contour. It may be that abnormalitiesof the FAZ may occur as a late development of those having patterns A or Bor progress rapidly as a specific "ischemic" form (pattern C).
It must be realized that levels of hyperglycemia and duration of type2 DM, that is, exposure to hyperglycemia, will certainly influence the evolutionand rate of progression tentatively classified in these 3 major patterns.If diabetic retinopathy is a multifactorial disease—in the sense thatdifferent factors or different pathways may predominate in different groupsof cases with diabetic retinopathy—then it is crucial that these differencesand the possible different phenotypes be identified.10
Diabetes mellitus is a familial metabolic disorder with equally stronggenetic and environmental causes. Familial aggregation is more common in type2 DM than in type 1 DM. Rema et al11 reportedthat familial clustering of diabetic retinopathy was 3 times higher in siblingsof patients with type 2 DM who have diabetic retinopathy. Polymorphisms involvedin differences in gene regulation of the receptor for advanced glycation end-productshave been linked to endothelial dysfunction in diabetes mellitus.12 Intercelullar adhesion molecule 1 polymorphism hasalso been proposed as a genetic risk factor for retinopathy in type 2 DM.13 Presence or absence of genetic factors may play afundamental role in determining specific pathways of vascular disease and,as a consequence, different progression patterns of diabetic retinal disease.It could be that certain polymorphisms would make the retinal circulationmore susceptible to an early breakdown of the BRB (pattern B) or microthrombosisand capillary closure (pattern C). The absence of these specific genetic polymorphismswould lead to an evolutionary pattern of pattern A.
It is clear from this study and from previous large studies such asthe Diabetes Control and Complications Trial–Epidemiology of DiabetesInterventions and Complications Research Group,14 andthe United Kingdom Prospective Diabetes Study Group15 thathyperglycemia plays a determinant role in the progression of retinopathy.It is interesting to note that HbA1c levels are also largely geneticallydetermined.16
An interesting perspective of our observations, analyzed in the lightof the available literature, depicts diabetic retinopathy as a microvascularcomplication of DM conditioned in its progression and prognosis by a varietyof different genetic polymorphisms, and modulated in its evolution by HbA1c levels, partly genetically determined and partly dependent on individualmanagement of DM. The interplay of these multiple factors and the durationof this interplay would finally characterize different clinical pictures orphenotypes of diabetic retinopathy.
The ultimate goal, therefore, should be the characterization of relationshipsbetween genetic factors (represented by distinct genotypes) and their medicallysignificant expression (distinct diabetic retinopathy phenotypes). Characterizationof different phenotypes of diabetic retinopathy has been suggested by previousdescriptions of extreme situations that may occur in diabetic retinal disease.10 We present herein, based on our observations of a3-year prospective study of eyes with mild nonproliferative diabetic retinopathyof patients with type 2 DM, an attempt to identify phenotypes of diabeticretinopathy: a wet or leaky type, an ischemic type, and finally, an apparentlymore common, slow progression type.
It must be realized, however, that our observations were made usingnew technologies, still in the research domain, and the sample size of thepopulation examined was small. A classification of diabetic retinopathy, basedon both relevant genotypes and disease phenotypes, is an ambitious goal. Webelieve that this route may help identify the particular form that threatensan individual patient and, consequently, offer an opportunity for specificand more effective therapies.
Corresponding author and reprints: Conceição L. Lobo,MD, PhD, Centre of Ophthalmology, University Hospital and Institute of BiomedicalResearch on Light and Image, Faculty of Medicine, University of Coimbra, 3000Coimbra, Portugal (e-mail: email@example.com).
Submitted for publication April 1, 2003; final revision received August17, 2003; accepted September 10, 2003.