Diabetes mellitus without retinopathy

Retinal microvascular alterations visible on ophthalmoscopy typically develop years following the onset of DM. Chapter 46 (Diabetic retinopathy: Genetics and etiologic mechanisms) reviews what we know about the early biochemical and cellular alterations leading to diabetic retinopathy, while Chapter 45 (The epidemiology of diabetic retinopathy) discusses the incidence and prevalence of retinopathy among diabetics. Experimental models of diabetic retinopathy in dogs and rats, and studies of human autopsy eyes, indicate that early alterations in retinal blood vessels include the loss of capillary pericytes (Fig. 47.1) and thickening of the capillary basement membrane.^1–21

Fig. 47.1 Photomicrographs of mounted retinal capillaries following trypsin digestion and staining with periodic acid-Schiff-hematoxylin, demonstrating retinal capillary alterations in a canine model for diabetic retinopathy. (A) Normal capillaries showing a typical distribution of endothelial cells and pericytes from a control dog fed normal chow (original magnification, ×825). (B,C) Pericyte ghosts (G), focal proliferation of endothelial cells (arrows), and acellular capillaries (A) are visible in a dog fed a galactose diet for 24 months (original magnification, ×825 and ×925, respectively). (D) A microaneurysm is present in a dog fed a galactose diet for 27 months (original magnification, ×570).

(Courtesy of Dr Peter Kador.)

There is preliminary evidence that parenchymal cells of the retina exhibit changes in the early stages of disease, including glial cell reactivity, alterations in glutamate metabolism, and neuron cell death.^22–28 Several studies have documented subtle changes in contrast sensitivity and color perception in diabetics in the absence of visible retinopathy,^29–34 but it remains unclear whether these effects result from deficits in retinal function, or from other diabetic alterations such as cataractogenesis. Central vision (measured by visual acuity) and peripheral vision (measured by common perimetric tests) typically remain normal in DM prior to onset of clinically evident retinopathy, in the absence of other factors such as cataract.

Microaneurysms

Microaneurysms, identified clinically by ophthalmoscopy as deep-red dots varying from 25 to 100 µm in diameter, are usually the first visible sign of diabetic retinopathy. Although microaneurysms are occasionally seen in normal aging adults and also occur in other retinal vascular diseases, such as retinal vein occlusion and radiation retinopathy, they are a hallmark of NPDR.

Microaneurysms arise as hypercellular saccular outpouchings of the capillary wall that can be well visualized in trypsin-digest retinal mounts (see Fig. 47.1D).¹ Their lumina are sometimes occluded by agglutinated erythrocytes or thrombus. Over time they sometimes become acellular, just as damaged retinal capillaries can evolve into “ghost” vessels devoid of endothelial cells and pericytes. The mechanism for microaneurysm formation is unknown. Possible contributing factors may include alterations in the retinal microenvironment from metabolic effects on neurons, glial cells, and endothelial cells; endothelial cell injury secondary to leukostasis from altered interaction between endothelial cells and leukocytes; response of endothelial cells to altered balance between proliferative and antiproliferative factors; structural changes in the capillary wall (such as from loss of pericytes); or increase in intraluminal pressure.

Microaneurysms visualized by ophthalmoscopy or angiography commonly appear and disappear over time, though some retain a stable appearance for years. The presence of microaneurysms alone, in the absence of other features of diabetic retinopathy, remains compatible with normal vision. However, as the number of microaneurysms increases, there is a greater risk of retinopathy progression.^35–37

Retinal vascular hyperpermeability

Subtle compromise of the blood–retinal barrier may begin at an early stage of disease, even preceding the appearance of retinopathy, but clinically appreciable retinal vascular hyperpermeability typically follows the appearance of microaneurysms. Visualized clinically by angiography, leakage may arise from microaneurysms, retinal capillaries, or other microvascular abnormalities, and can be highly variable in magnitude and extent. Retinal vascular incompetence may or may not result in localized areas of thickening of the retina. Hard exudates, extravascular deposits of lipid-rich material that result from spillage and incomplete resorption of plasma lipoproteins, may accumulate. Intraretinal hemorrhages appear in the posterior pole and in the retinal periphery. The vascular alterations responsible for hyperpermeability in NPDR remain incompletely understood, but may involve dysfunction of the tight junctions between retinal capillary endothelial cells. Possible mechanisms for breakdown of the blood–retinal barrier are discussed in Chapter 27 (Blood–retinal barrier, immune privilege, and autoimmunity).

Diabetic macular edema

Diabetic macular edema (DME), defined as retinal thickening in the posterior pole, results from retinal vascular hyperpermeability and other alterations in the retinal microenvironment, and represents a common cause of vision loss among diabetics. DME can occur in eyes with a wide spectrum of underlying retinopathy, from mild NPDR to PDR. It can occur in areas of demonstrable retinal vascular incompetence, as visualized by angiography, and also in regions of retinal ischemia. In areas of vascular incompetence, DME may result from leakage of microaneurysms, or it may evolve from diffuse leakage of hyperpermeable capillaries. In areas of capillary nonperfusion on angiography, retinal thickening may result from ischemia in the absence of prominent vascular leakage, though hyperpermeable microvascular abnormalities at the borders of such regions may contribute to swelling. Macular edema may or may not be characterized by intraretinal cyst formation. In some cases, usually in the setting of severe cystic thickening involving the fovea, subretinal fluid may also be present.

The pathogenesis of DME remains poorly understood, partly because of the absence of a good animal model. Chapter 46 (Diabetic retinopathy: Genetics and etiologic mechanisms) reviews what we know about the early biochemical and cellular alterations leading to DME, while Chapter 28 (Mechanism of macular edema) discusses factors involved in the pathogenesis of macular edema of different causes.

Extrafoveal foci of retinal thickening and hard exudates may not cause symptoms or affect visual acuity, but DME that involves or threatens the center of the macula carries a significant risk of vision loss. In the Early Treatment of Diabetic Retinopathy Study (ETDRS), the 3-year risk of moderate visual loss (a decrease of three lines or more on a logarithmic visual acuity chart, corresponding to a doubling of the initial visual angle) among untreated eyes with DME involving or threatening the center of the macula was 32%.³⁸ The natural history of DME is variable. In some eyes, it can persist for years, while in others it may spontaneously resolve. Chapter 45 (The epidemiology of diabetic retinopathy) discusses the incidence and prevalence of DME and consequent vision loss.

Capillary closure, microvascular remodeling, and retinal ischemia

One of the most serious consequences of diabetic retinopathy is progressive loss of functional retinal capillaries. Trypsin-digest preparations of the retina show areas of acellular capillaries, or “ghost” vessels, which have lost the endothelial cells and pericytes that once lined them (see Fig. 47.1C). When patches of such acellular capillaries, first seen early in the course of NPDR, increase and become confluent, the terminal arterioles that supply these capillaries often become occluded. Regions of acellular capillaries in histologic sections have been shown to correspond to areas of capillary nonperfusion visualized by fluorescein angiography.³⁹ Adjacent to these areas of retinal ischemia, clusters of microaneurysms and hypercellular vessels often develop. It has been difficult to determine whether such vessels represent altered pre-existing capillaries or neovascularization within the retina. Such vessels are described clinically as intraretinal microvascular abnormalities (IRMA), a term intended to accommodate both possibilities.

Progressive capillary closure and resulting retinal ischemia are commonly associated with increasing IRMA, intraretinal hemorrhages, and venous abnormalities such as segmental dilation (venous beading). Occasionally, in cases of extensive capillary nonperfusion, the retina acquires a featureless appearance with a relative dearth of visible vessels, hemorrhages, or microvascular abnormalities. Retinal ischemia represents another cause for vision loss in NPDR, and also plays a central role in the pathogenesis of PDR by stimulating elaboration of vascular endothelial growth factor A (VEGF-A) and other factors promoting the growth of extraretinal neovascularization.^40–47 Chapter 48 (Proliferative diabetic retinopathy) discusses the natural course of PDR.

Alterations of the vitreous gel and vitreoretinal interface

The vitreous gel plays a key role in the fibrovascular proliferation of PDR, but it may also exert important effects at earlier stages of retinopathy. Epiretinal membrane formation, arising from liquefaction of the vitreous gel and consequent effects at the vitreoretinal interface, can occur with advancing age in otherwise healthy eyes but is more common in diabetic eyes.^48–52 It has been hypothesized that liquefaction of the vitreous gel out of proportion to diminution of posterior vitreoretinal adhesion may underlie the pathophysiology of many vitreoretinal diseases,⁵³ and observations at vitrectomy suggest that the posterior cortical vitreous is frequently more adherent to the retina in the setting of diabetic retinopathy. Studies have documented a number of biochemical changes to the vitreous gel in the setting of DM, including increased collagen fibril cross-linking, accumulation of advanced glycation end products that may augment vitreoretinal adhesion and incite retinal glial cell reactivity, and alterations in the concentration of various soluble proteins.^54–58

The role of the vitreous gel in the pathophysiology of diabetic retinopathy may extend beyond its capacity to exert mechanical effects on the retina. For example, there is recent evidence that the vitreous may function as an important regulator of intraocular oxygen tension, a finding that could have important implications for diabetic retinopathy and other diseases involving retinal hypoxia.^59,60

Duration of diabetes mellitus

The duration of DM is strongly associated with risk of retinopathy. Cross-sectional and longitudinal analyses from population-based epidemiologic studies have established the association between prevalence of retinopathy and duration of disease.^61–70 The Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR) examined prevalence and severity of diabetic retinopathy among diabetics categorized into a younger-onset group consisting of those diagnosed with DM prior to 30 years of age who were taking insulin at the time of the evaluation (predominantly type 1 DM) and an older-onset group consisting of those diagnosed with DM at 30 years of age or older (predominantly type 2 DM), the latter subdivided according to whether insulin was being used at the time of evaluation. The prevalence of retinopathy may have changed somewhat since the 1980s when WESDR data was collected, given interim advances and evolving standards of care in management of DM, but the WESDR remains one of our most definitive sources of information about the epidemiology of diabetic retinopathy in the United States. In the younger-onset group, retinopathy (consisting of NPDR or PDR) was seen in 13% of those with less than a 5-year duration of DM and in 90% of those with a duration of 10–15 years.⁶⁵ In the older-onset group using insulin, retinopathy was seen in 40% of those with less than a 5-year duration of disease and in 84% of those with a duration of 15–19 years, while the corresponding rates in the older-onset group not taking insulin were 24% and 53%, respectively.⁶⁴

Among those with type 2 DM, in which disease onset is often more insidious than in type 1 DM and in which hyperglycemia can remain asymptomatic for years, age at diagnosis may not always accurately reflect disease duration. The Centers for Disease Control and Prevention estimate that 25.8 million people in the United States have DM, of whom 7.0 million do not know they have it.⁷¹ The prevalence of undiagnosed type 2 DM is felt to be the key factor explaining the higher rates of retinopathy noted soon after diagnosis among type 2 diabetics (24% in the older-onset group not taking insulin with less than a 5-year duration of disease in the WESDR) compared with type 1 diabetics (13% in the younger-onset group with less than a 5-year duration of disease).

Hyperglycemia

The relationship between the degree of hyperglycemia and the presence and severity of diabetic retinopathy has been extensively studied in observational studies and clinical trials. The observational studies have demonstrated that greater hyperglycemia is associated with increased prevalence and severity of diabetic retinopathy.^{62,64–67,69,70} Several key randomized controlled clinical trials have demonstrated that better glycemic control is associated with a decreased risk of secondary complications of DM, including diabetic retinopathy.^72–79

In the Diabetes Control and Complications Trial (DCCT), 1441 participants with type 1 DM were randomly assigned to either conventional or more intensive insulin treatment and followed for a period of 4–9 years.^72,73,76,79 The average difference in glycosylated hemoglobin (hemoglobin A1C [HbA1C]) between the two groups was almost 2%. Intensive insulin treatment as defined by the DCCT was associated with a decreased risk of both the development and progression of diabetic retinopathy. In patients without retinopathy at enrollment, the 3-year risk of developing retinopathy was reduced by 75% in the intensive insulin treatment group compared with the conventional treatment group. The benefit of better glycemic control was also evident in patients with existing retinopathy at baseline, as shown by a 50% reduction in the rate of progression of retinopathy compared with controls. At the 6- and 12-month visits, more intensive insulin treatment exerted a small adverse effect on retinopathy progression, similar to that described in other trials of glycemic control. However, among eyes with little or no retinopathy at the time of initiating better control of hyperglycemia, it was found that such “early worsening” of retinopathy was unlikely to threaten vision. When the DCCT results were stratified by levels of glycosylated hemoglobin, there was a 35–40% reduction in the risk of retinopathy progression for every 10% decrease in HbA1C (e.g., from 8% to 7.2%). This represented a fivefold increase in risk of progression for patients with HbA1C around 10% compared with those with HbA1C around 7%. Notably, there was also a statistically significant reduction in other microvascular complications of DM, including nephropathy and peripheral neuropathy, with more intensive glycemic control in the DCCT.

When the randomized controlled clinical trial was completed, DCCT participants were informed of the results and enrolled in a follow-up phase of the study, known as the Epidemiology of Diabetes Intervention and Complications (EDIC) study.⁷⁶ After an additional 7 years of follow-up, during which the HbA1C values in both treatment groups did not differ significantly (8.1% vs 8.2%, P = 0.09), the rate of retinopathy progression remained significantly lower in those who had received more intensive treatment in the DCCT than in those who had received conventional therapy. Thus, more intensive glycemic control over a period of 6.5 years conferred benefits well beyond the period of treatment.

In the United Kingdom Prospective Diabetes Study (UKPDS), 3867 patients with newly diagnosed type 2 DM were randomly assigned to conventional therapy or to more intensive glycemic control with either insulin or a sulfonylurea.^74,75 Among participants treated with conventional therapy, those who were overweight were given metformin and those who were not overweight did not receive this medication. After 12 years, the rate of retinopathy progression was reduced by 21% and the use of laser photocoagulation was reduced by 29% in those getting intensive glycemic control compared with those getting conventional treatment. For every percentage point decrease in HbA1C (e.g., 9% to 8%), there was a 35% reduction in the risk of microvascular complications of disease.

The results of the UKPDS have recently been corroborated by those of another large randomized controlled clinical trial, the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study. The ACCORD study randomly assigned 10 251 patients with type 2 DM to very intensive glycemic control (targeting HbA1C less than 6%), or to standard treatment (targeting HbA1C between 7 and 7.9%). A subset of 2856 participants was evaluated for progression of retinopathy by comparison of fundus photographs taken at baseline and at 4 years.⁷⁸ In this subset at one year, median HbA1C in the intensive treatment cohort was 6.4% compared with 7.5% in the cohort getting conventional therapy. In the intensive treatment group, the rate of retinopathy progression was 7.3%, compared with 10.4% in the standard therapy group (adjusted odds ratio, 0.67; 95% confidence interval [CI] 0.51–0.87; P = 0.003). The glycemia trial, performed alongside other studies evaluating control of blood pressure and plasma lipids, was halted early at a median of 3.7 years because of an increased rate of death from all causes in participants treated with intensive control compared with those managed with standard control (5% vs 4%, respectively), and the very intensive control approach was abandoned.

Hypertension

In addition to assessing the effects of differential glycemic control, the UKPDS provided a randomized comparison between more intensive blood pressure control (targeting a systolic blood pressure less than 150 mmHg) and less intensive blood pressure control (targeting a systolic blood pressure less than 180 mmHg) in a large cohort of participants with newly diagnosed type 2 DM.⁸⁰ Study results demonstrated that intensive blood pressure control was associated with a decreased risk of retinopathy progression. Of 1148 hypertensive participants in the UKPDS, 758 were allocated to more intensive control of blood pressure and 390 to less intensive control with a median follow-up of 8.4 years. More intensive blood pressure control resulted in a 37% reduction in microvascular complications of DM, predominantly a reduced risk of retinal photocoagulation, compared with less intensive control. An earlier report assessing the effects of various blood pressure medications on diabetic retinopathy had suggested that there might be a specific benefit of angiotensin-converting enzyme (ACE) inhibition, even in normotensive persons. The UKPDS included a randomized comparison of beta-blockers and ACE inhibitors within the more intensive blood pressure control arm. Benefit from more intensive blood pressure control was noted in both the beta-blocker and ACE inhibitor treatment groups, with no statistically significant difference between them, suggesting that the treatment effect was probably attributable to blood pressure reduction and not to a specific effect of ACE inhibition.

The ACCORD study attempted to extend the results of the UKPDS by assessing the effect of very tight blood pressure control on retinopathy. In this study, 4733 of the 10 251 participants with type 2 DM were randomly assigned to very intensive control of blood pressure (targeting a systolic blood pressure less than 120 mmHg) or standard blood pressure control (targeting a systolic blood pressure less than 140 mmHg). A subset of 1263 participants in the blood pressure trial was evaluated for progression of retinopathy just as in the glycemia arm of the study, with comparison of fundus photographs at baseline and 4 years.⁷⁸ Rate of progression of retinopathy was not significantly different in the two groups (10.4% of those treated intensively compared with 8.8% of those treated with standard care, with an adjusted odds ratio of 1.23; 95% CI 0.84–1.79; P = 0.29). The results of the UKPDS and ACCORD study are not necessarily contradictory in this regard, given the very different targets used for more and less intensive control of blood pressure between the two trials.

Dyslipidemia

Elevated levels of plasma cholesterol were associated with greater severity of retinal hard exudates in the WESDR and in the ETRDS.^81,82 Independent of coincident retinal thickening, the severity of retinal hard exudates at baseline was associated with decreased visual acuity in the ETDRS. The severity of retinal hard exudates was also a significant risk factor for moderate visual loss during the course of the study. Elevated levels of plasma triglycerides were associated with a greater risk of developing high-risk PDR in the ETDRS patients.⁸³ Two recent randomized controlled clinical trials evaluating the plasma lipid-modulatory agent fenofibrate in combination with statins in individuals with elevated plasma lipids have demonstrated that fenofibrate reduces the risk of diabetic retinopathy progression while only providing modest alterations in the plasma lipid profile.^78,84 It is presently uncertain whether the effects of fenofibrate on diabetic retinopathy are secondary to its plasma lipid-modulatory activity or to some other mechanism.

Other extraocular factors

Diabetic retinopathy can worsen precipitously in the setting of pregnancy. Chapter 92 (Pregnancy-related diseases) reviews the natural history of retinopathy in pregnancy. Diabetic nephropathy, as measured by albuminuria, proteinuria, or manifestations of renal failure, has been inconsistently associated with progression of retinopathy.^83,85,86 Anemia has been associated with progression of diabetic retinopathy in two small case series and two epidemiologic studies.^83,87–89 The ETDRS found association between decrease in hematocrit and increase in the incidence of high-risk PDR in an adjusted multivariate model. A few reports have suggested an association between diabetic neuropathy or cardiovascular autonomic neuropathy and progression of retinopathy.^83,90,91 Significant progress has been made in understanding the heritable contributions to the development of DM, particularly the type 1 form and certain familial and syndromic variants of the disease, but the identification of genetic risk factors for development and progression of diabetic retinopathy has been challenging. Chapter 46 (Diabetic retinopathy: Genetics and etiologic mechanisms) reviews our present knowledge about the genetics of diabetic retinopathy.

Ophthalmic evaluation

Ophthalmic evaluation of a patient with DM begins with identification of pertinent features of ocular history, including previous diagnosis of diabetic retinopathy or DME, comorbid ophthalmic conditions, and previous medical or surgical treatment. The severity of existing retinopathy is a powerful predictor of risk of retinopathy progression and vision loss, as discussed below in the section “Classification of diabetic retinopathy.” Knowledge of any previous treatment has recently become more relevant for management of NPDR and DME. A proliferating array of therapeutic options and the variability of treatment efficacy among individuals have made management algorithms more complex and clinical decision-making more contingent on assessment of the success of previous treatment strategies. Most pharmacologic therapies leave no lasting marks like the telltale scars of laser photocoagulation, making clinicians more reliant on history-taking and review of records for evidence of prior treatment.

A comprehensive eye examination in a person with DM includes: measurement of visual acuity and intraocular pressure; evaluation of the anterior segment by slit-lamp biomicroscopy; gonioscopy when warranted (such as in the setting of elevated intraocular pressure, neovascularization of the iris, or glaucoma); and dilated funduscopic examination.⁹² Evaluation of the anterior segment prior to dilation of the pupil may allow visualization of neovascularization of the iris or anterior chamber angle not visible following mydriasis. Dilation of the pupil is important for adequate assessment of the posterior segment. In the absence of pupil dilation, only 50% of eyes are correctly diagnosed for the presence and severity of retinopathy.⁹³

Ophthalmoscopy, including stereoscopic evaluation of the posterior pole and visualization of the vitreous gel and peripheral retina, remains the standard for clinical diagnosis and characterization of NPDR and DME. Examination of the posterior pole is best achieved using slit-lamp biomicroscopy with accessory lenses. A hand-held accessory lens may provide sufficient visualization of the posterior pole and midperipheral retina, but in cases where superior stereopsis and discrimination are desired, an examination contact lens coupled to the eye surface after application of topical anesthetic drops can be used. The peripheral retina is typically surveyed using indirect ophthalmoscopy, but slit-lamp biomicroscopy with an accessory lens may serve as a supplement or substitute when visualization with high magnification is warranted. A three- or four-mirror contact lens coupled to the eye surface can also be used to examine peripheral lesions under high magnification. Red-free illumination can be useful at highlighting retinal vessels and associated lesions of NPDR such as microaneurysms, hemorrhages, and IRMA.

Ancillary ocular imaging

Imaging modalities in common clinical use for management of NPDR and DME include fundus photography, fluorescein angiography (FA), and optical coherence tomography (OCT).

Fluorescein angiography

Fluorescein angiography (FA), consisting of photography or videography of the fundus using cameras equipped with appropriate filters following intravenous injection of the hydrocarbon dye fluorescein sodium, has been used clinically in ophthalmology for over 40 years and is described in detail in Chapter 1 (Fluorescein angiography). FA has multiple uses in the clinical evaluation of diabetic retinopathy. Because intravenous injection of fluorescein sodium can be associated with life-threatening adverse reactions (with risk of death estimated at approximately 1/200 000),⁹⁴ FA for evaluation of diabetic retinopathy is largely confined to settings in which important aspects of management hinge on the results of the test. FA has been used extensively for evaluation of diabetic retinopathy, and is not known to pose any special hazards in DM. Nephropathy or renal failure is not a contraindication to testing.⁹⁵ In NPDR, FA is most commonly indicated for further characterization of DME diagnosed on ophthalmoscopy (Fig. 47.2), and its utility in this setting is discussed below in the section “Clinical evaluation of diabetic macular edema.”

Fig. 47.2 Moderate NPDR and DME. (A) Photograph illustrating microaneurysms, intraretinal hemorrhages, and hard exudates. Retinal thickening visible with stereoscopic viewing on biomicroscopy is present temporal to the foveal center. Microaneurysms appear as variably-sized small red dots most prominent temporal to the fovea. Hemorrhages (arrow) can sometimes be distinguished from microaneurysms because of their size and less sharply demarcated borders. Mild hard exudates appear as small yellow-white intraretinal deposits with sharply demarcated borders. (B) Early frame of the FA. Microaneurysms are visible as variably sized hyperfluorescent dots. Note that angiography allows visualization of microaneurysms not readily seen in the color photograph. Hemorrhages (arrowhead) and hard exudates are hypofluorescent. Microvascular abnormalities not visible in the color photograph are present temporal to the fovea. Degree of capillary nonperfusion is minimal. (C) Late frame of the FA. Mild late leakage is visible as poorly demarcated hyperfluorescence surrounding clusters of microaneurysms temporal to the fovea.

(National Eye Institute, Bethesda, MD.)

FA can also be used to investigate vision loss incompletely explained by ophthalmoscopy and other less-invasive testing. In some cases, significant capillary nonperfusion involving the fovea and parafovea, best visualized during the arteriovenous transit phase of the angiogram, implicates macular ischemia as a cause for vision loss. In other cases, FA may suggest other causes for vision loss unrelated to diabetic retinopathy.

FA is occasionally used in eyes with presumed NPDR to search for early lesions of extraretinal neovascularization not detected by ophthalmoscopy or fundus photography, such as in cases where preretinal or vitreous hemorrhage suggests PDR but no cause for bleeding is seen by ophthalmoscopy. The technique in this setting requires a view of the fundus not overly obscured by hemorrhage, and use of either wide-angle lenses or “sweep” fields to image as much of the retina as possible. Its success depends on the prominent leakage of fluorescein dye characteristic of extraretinal neovascularization, typically readily visible against the background of other vascular alterations, but it is important to note that intraretinal alterations such as IRMA can exhibit similar leakage. Sometimes stereoscopic images or further ophthalmoscopy may clarify the extraretinal nature of neovascularization and associated leakage.

Commonly used grading systems for severity of diabetic retinopathy do not consider findings on FA, and FA is not indicated for classification of disease. While FA is a sensitive means for detection of early features of NPDR such as microaneurysms and retinal capillary hyperpermeability, it is not clinically indicated to screen for mild retinopathy.⁹²

Optical coherence tomography

Optical coherence tomography (OCT), which utilizes low-coherence interferometry involving near-infrared light for cross-sectional imaging of intraocular structures, has emerged over the last decade as a fast, noninvasive means of imaging the retina, vitreoretinal interface, and the retinal pigment epithelium (RPE) in NPDR and other macular diseases. The basic principles of OCT and the applications of this technology to evaluation of the retina are discussed in Chapter 3 (Optical coherence tomography). In NPDR, OCT is commonly used to characterize DME (Fig. 47.3) and abnormalities of the vitreoretinal interface.

Fig. 47.3 OCT characterization of macular thickening in NPDR and DME. (A) A 6 mm horizontal line scan obtained with a spectral domain OCT system (Cirrus, Carl Zeiss Meditec, Dublin, CA) imaging the macula of the eye illustrated in Fig. 47.2 shows mild retinal thickening temporally, corresponding to the area of microaneurysms leaking on FA. (B) The retinal thickness map generated by automated segmentation of a 6×6 mm 512×128 Macular Cube scan shows mean thickness values for nine standard subfields. Mean thickness is greater than in 99% of normals in the central and inner temporal subfields (shaded purple) and greater than in 95% of normals in the outer temporal subfield (shaded red), based on a comparison to data in the Cirrus normative database.

(National Eye Institute, Bethesda, MD.)

Investigators of the Diabetic Retinopathy Clinical Research Network (DRCR.net),⁹⁶ a collaborative network of over 100 sites in the United States dedicated to multicenter clinical research of diabetic retinopathy, have rigorously evaluated OCT as a test yielding measures of macular thickness in diabetic eyes. The standardization required for DRCR.net clinical trials has necessitated use of a common OCT platform and scanning technique. OCT scanning in DRCR.net studies to date involves acquisition of images of the macula by certified operators after pupil dilation, with use of a time-domain system called the Stratus OCT (Carl Zeiss Meditec, Inc., Dublin, CA) and a scanning algorithm known as the fast macular thickness map, which obtains 128 axial scans (A scans) along each of six radial lines (each 6 mm in length) intersecting at a common center, with a total scan time of 1.9 seconds. Stratus processing software includes an automated segmentation algorithm that identifies an inner and outer retinal border on line scans and calculates retinal thickness. A retinal thickness map is generated using data from all line scans obtained by the scanning algorithm. Output includes center-point thickness, total macular volume, and mean values for retinal thickness in a grid comprised of a central subfield, four inner subfields, and four outer subfields. Assessment of the quality of output depends on assessment of other parameters such as signal strength and standard deviation of center-point thickness. Software offers a comparison of thickness values to those from a normative database of measurements obtained from normal individuals.

DRCR.net investigators compared central subfield mean thickness (CSMT) values in 97 diabetics with minimal or no retinopathy and no central macular thickening on examination to normal values for nondiabetics in the Stratus OCT database.⁹⁷ Results showed no significant difference between CSMT of diabetic eyes with minimal or no retinopathy and normal values imputed from nondiabetic individuals in the database, suggesting that the presence of DM is not associated with clinically important changes in CSMT in the absence of retinopathy. The study noted a statistically significant difference between CSMT in men and women, similar to findings in other reports.

DRCR.net investigators evaluated the reproducibility of OCT in eyes with DME, and established thresholds indicative of meaningful change in retinal thickness. In a prospective multicenter one-day study evaluating diurnal variation of DME, 212 eyes of 107 participants with DME involving the foveal center on biomicroscopy and CSMT of 225 microns or greater were imaged multiple times on Stratus OCT and the scans were evaluated by a reading center.⁹⁸ Both eyes were imaged six times during the day, and at each of these six sittings a pair of scans was generated, with repeat imaging done for scans of suboptimal quality. Of a possible 1284 pairs, 1223 were analyzed. Reproducibility was better for CSMT than for center-point thickness, not surprising considering that CSMT incorporates more data points. The median absolute difference between replicate measurements of CSMT was 7 µm. Expressed as a percentage difference between the two measurements, the half-width of the 95% CI for a change in retinal thickness for CSMT was 10% for eyes with CSMT less than 400 µm, and 13% for eyes with CSMT equal to or greater than 400µm. The authors concluded that a change in CSMT greater than 11% using the Stratus OCT for DME is likely to be real.

The rapid pace of technological progress, resulting in recent improvements such as spectral domain imaging and image-registration capability in some systems, makes for substantial variability in image acquisition, processing, and output metrics among presently available OCT platforms. For example, the spectral domain Cirrus OCT (Carl Zeiss Meditec, Dublin, CA), though its output is similar in many ways to that of the time domain Stratus OCT, uses a distinct automated segmentation algorithm that results in retinal thickness values that are approximately 30–55µm greater than those reported by the Stratus OCT in eyes with DME.⁹⁹ Meaningful interpretation of an OCT scan requires familiarity with the system utilized, including understanding of scan acquisition, indicators of scan quality, processing of raw data, algorithms for automated segmentation, and format of output metrics, as well as knowledge of any normative database used by system software to compare a given set of measurements to a normal range. Caution is warranted in comparing OCT scans obtained using different systems, and clinical researchers using OCT must grapple with such issues in planning studies that require standardization of procedures among clinical sites. The DRCR.net is presently evaluating issues related to integrating new OCT systems into future research studies.

Funduscopic lesions of nonproliferative diabetic retinopathy

Small or nonperfused microaneurysms may not be discernible on ophthalmoscopy, but those that are visible appear as small deep-red dots between 25 and 100µm in diameter within the retina (see Figs 47.2A, 47.4). Microaneurysms in NPDR typically arise in the posterior pole, but they can also be present in the midperipheral and peripheral retina, particularly in more severe retinopathy. They may be solitary or may appear in clusters. They may remain stable across months or even years, but many appear and eventually disappear. On FA, microaneurysms are visible during arteriovenous transit as hyperfluorescent dots within the retina (see Fig. 47.2B). This hyperfluorescence typically persists in later phases of the angiogram, and may or may not be associated with leakage in mid- and late frames (see Fig. 47.2C). FA is a sensitive means of detecting microaneurysms, and may reveal lesions that were poorly visible or not visible on ophthalmoscopy.

Fig. 47.4 Standard photograph 2A, intermediate standard for hemorrhages/microaneurysms. ETDRS extension of the Modified Airlie House classification of diabetic retinopathy.

(Courtesy of the Diabetic Retinopathy Study Research Group.)

Intraretinal hemorrhages in NPDR are variable in appearance, just as in other retinal diseases (Figs 47.2A, 47.4). Dot-blot hemorrhages are typically small with sharply demarcated borders, and are sometimes indistinguishable from microaneurysms on ophthalmoscopy. Flame hemorrhages can be larger, and manifest wispy margins as a consequence of their location in the nerve fiber layer. Intraretinal hemorrhages appear hypofluorescent on FA, blocking normal fluorescence from the underlying choroid, and consequently FA offers ready distinction between microaneurysms and hemorrhages (Fig. 47.2B). Intraretinal hemorrhages can be present in the posterior pole and in more peripheral retina, and frequently appear and disappear over weeks or months. Hemorrhages on the optic disc are not typical for diabetic retinopathy, and should raise suspicion for neovascularization or a comorbid condition affecting the optic nerve head. Variability in the density of hemorrhages between one sector of the retina and another is common, but striking asymmetry sometimes suggests a superimposed process such as a branch retinal vein occlusion.

Hard exudates are visible on ophthalmoscopy as sharply demarcated yellow-white deposits within the retina (Figs 47.2A, 47.5). They are visible on OCT line scans as hyperreflective foci within the retina. With stereoscopic viewing they are readily distinguishable from drusen, which reside external to the retina. Hard exudates are often distributed at the border between edematous and nonedematous retina. They may form a circinate ring around areas of prominent vascular hyperpermeability such as a cluster of microaneurysms. They tend to form in the posterior pole in association with macular thickening, but small collections are sometimes present in more peripheral retina. They are hypofluorescent on FA, blocking underlying choroidal fluorescence (see Fig. 47.2B). Hard exudates may appear and disappear over months or years. When severe, they may undergo organization and may cause subretinal fibrosis.

Fig. 47.5 Standard photograph 4, severe standard for hard exudates. ETDRS extension of the Modified Airlie House classification of diabetic retinopathy. Hard exudates in the temporal macula form a circinate ring.

(Courtesy of the Diabetic Retinopathy Study Research Group.)

Cotton-wool spots, patches of relative ischemia affecting the nerve fiber layer of the retina, are visible on ophthalmoscopy as small white patches with wispy borders situated in the inner retina (Fig. 47.6). They are hypofluorescent on FA, blocking underlying choroidal fluorescence. They commonly appear and disappear over weeks or months.

Fig. 47.6 Standard photograph 8A, less severe of two standards for cotton-wool spots and IRMA. ETDRS extension of the Modified Airlie House classification of diabetic retinopathy. Cotton-wool spots are visible in two areas as subtle white patches with wispy borders (arrows). Several areas of IRMA are visible as abnormal, tortuous, dilated retinal vessels (arrowheads).

(Courtesy of the Diabetic Retinopathy Study Research Group.)

The major retinal vessels can exhibit changes in appearance in the setting of advanced retinopathy. Arterioles may appear thin or white. Venules may appear dilated and tortuous. Venous loops are occasionally present. Venous beading consists of localized areas of change in vessel caliber, visible as alternating regions of relative dilation and constriction (Fig. 47.7).

Fig. 47.7 Standard photographs 6A and 6B, less and more severe standards for venous beading. ETDRS extension of the Modified Airlie House classification of diabetic retinopathy. (A) Less severe standard (6A). Two branches of the superior temporal venule show beading that is definite but not severe. (B) More severe standard (6B). Most large and small venule branches show severe beading.

(Courtesy of the Diabetic Retinopathy Study Research Group.)

Intraretinal microvascular abnormalities (IRMA) appear as segments of dilated and tortuous retinal vasculature amidst retinal vessels that are normally too small to be visible on ophthalmoscopy (see Fig. 47.6). They are usually readily distinguishable from extraretinal neovascularization on careful biomicroscopy. On FA they appear hyperfluorescent during the arteriovenous transit phase, may leak in later phases, and are often situated at the borders of areas of capillary nonperfusion. They may persist for months or years.

In occasional cases, typical lesions of severe NPDR such as intraretinal hemorrhages, venous abnormalities, and IRMA, may be sparse or absent in regions of the retina despite profound underlying microvascular alterations poorly visible on ophthalmoscopy. When such areas of “featureless” retina are widespread, the bland appearance may belie the severity of disease. Careful ophthalmoscopy usually reveals manifestations of severe retinal ischemia such as arteriolar narrowing and sheathing, absence of normal vessel markings, and retinal thinning in the setting of an eye at high risk for advanced retinopathy. FA can be used to confirm clinical suspicion when necessary, characteristically revealing widespread capillary nonperfusion in areas of featureless retina.

Classification of diabetic retinopathy

The Diabetic Retinopathy Study, a landmark clinical trial that established the efficacy of scatter laser photocoagulation for reduction of severe vision loss secondary to PDR, used an adaptation of the Airlie House classification of diabetic retinopathy originally developed in 1968.¹⁰⁰ This modified Airlie House system was extended for grading severity of retinopathy in the ETDRS.¹⁰¹ Grading of retinopathy in the ETDRS involved evaluation of seven-field 30-degree nonsimultaneous stereo color fundus photographs by trained readers. Over 30 characteristics were separately graded, using standard photographs to define thresholds for scoring (see Figs 47.4, 47.5, 47.6, 47.7 for some of the standard photographs utilized in the system), and a summary grade was assigned. The ETDRS diabetic retinopathy severity scale was evaluated for its reproducibility as part of the study and was validated as a grading system with prognostic power.^101–103 ETDRS investigators assigned thresholds to define mild NPDR, moderate NPDR, severe NPDR, early PDR, and high-risk PDR (Box 47.1). The 5-year rates of development of high-risk PDR in eyes randomly assigned to deferral of photocoagulation (no laser treatment unless features of high-risk PDR developed) for eyes with mild, moderate, and severe NPDR at baseline were 15.5%, 26.5%, and 56%, respectively.

The prognostic utility of a 1-step or a 2-or-greater-step progression in the ETDRS diabetic retinopathy severity scale has been evaluated using data from the WESDR.¹⁰⁴ Findings indicated that 1-step or 2-or-greater-step worsening in level of retinopathy over 4 years strongly predicted the development of PDR during the subsequent 6 years. The ETDRS diabetic retinopathy severity scale remains the standard for photographic grading of retinopathy in clinical trials. The schema is impractical for routine clinical use because of its complexity and its basis in photographic – not ophthalmoscopic – assessment of the fundus, but the ETDRS definitions for mild, moderate, and severe NPDR and early and high-risk PDR were incorporated into clinical practice. A 4–2–1 rule was popularized to simplify the definition of severe NPDR.¹⁰⁵ Upon examination of the four midperipheral quadrants of the retina, the presence of any one of the following features was considered sufficient for diagnosis of severe NPDR (in the absence of evidence for PDR): (1) severe intraretinal hemorrhages and microaneurysms in all four quadrants (≥ standard photograph 2A [see Fig. 47.4]); (2) venous beading in two or more quadrants; or (3) moderate IRMA in at least one quadrant (≥ standard photograph 8A [see Fig. 47.6]). If any two of these features were present, the retinopathy was considered very severe. It is worth pointing out that the 4–2–1 rule uses a threshold for definition of severe NPDR inclusive of milder retinopathy than the definition of severe NPDR used in the ETDRS.

In answer to the need for a simplified classification of diabetic retinopathy to facilitate communication among clinicians worldwide, the Global Diabetic Retinopathy Project Group published proposed International Clinical Diabetic Retinopathy and Diabetic Macular Edema Severity Scales in 2003 (Table 47.1).¹⁰⁶ The classification was developed using evidence from studies such as the ETDRS and WESDR to draw distinctions in retinopathy severity most important for prognosis and management, and approximates the definitions used in the ETDRS.

Table 47.1 Disease severity scales for diabetic retinopathy and diabetic macular edema

Distribution of retinal thickening and hard exudates

The ETDRS defined clinically significant macular edema (CSME) as any of the following noted on biomicroscopy: (1) thickening of the retina at or within 500 µm of the center of the macula; (2) hard exudates at or within 500 µm of the center of the macula, if associated with thickening of the adjacent retina (not residual hard exudates remaining after the disappearance of retinal thickening); or (3) a zone or zones of retinal thickening one disc area or larger, any part of which is within one disc diameter of the center of the macula.³⁸

The definition for CSME was based on observation that retinal thickening or hard exudation involving or threatening the fovea frequently leads to vision loss. Careful assessment of the distribution of retinal thickening and hard exudates and their relation to the center of the macula remains paramount to management of DME. Severity of DME in the International Clinical Diabetic Macular Edema Disease Severity Scale is based solely on whether retinal thickening and hard exudates involve or threaten the center of the macula, reflecting the importance of foveal involvement for prognosis and management.¹⁰⁶

While diagnosis on biomicroscopic examination remains the clinical standard for detection of DME, OCT is increasingly used as a fast and noninvasive tool to quantitatively map areas of macular thickening (see Fig. 47.3). Subtle changes in distribution of thickening over time and relationship to the fovea can be documented with good-quality scans that serially image the same region of the macula (Fig. 47.8). The standardization afforded by OCT has proven valuable in clinical practice and in research. Hard exudates can be visualized on OCT as hyperreflective foci within the retina, but biomicroscopic examination and fundus photography remain important to document their extent and proximity to the fovea.

Fig. 47.8 OCT quantification of change in retinal thickness in DME. (A) A retinal thickness map generated using a spectral domain OCT system (Cirrus, Carl Zeiss Meditec, Dublin, CA) shows areas of severe macular edema juxtafoveally in an eye responsive to previous focal/grid laser photocoagulation and serial intravitreous bevacizumab injections. Another intravitreous bevacizumab (1.25 mg) injection was given at this time. (B) The retinal thickness map 5 weeks following injection shows near-resolution of macular edema, with only subtle areas of residual retinal thickening (areas shaded red and yellow) still present.

(National Eye Institute, Bethesda, MD.)

Magnitude of retinal thickening

The degree of thickening at any given point in the retina has traditionally been estimated on biomicroscopy and stereoscopic photographs, but OCT has recently emerged as a superior means of quantifying retinal thickness. The sensitivity of OCT for retinal thickening exceeds that of contact lens-assisted biomicroscopy performed by experienced examiners.¹⁰⁷ Correlation between the estimated degree of retinal thickening at the center of the macula on stereoscopic photographs and the center-point thickness measured on time domain OCT is modest, reflecting the limitations of grading retinal thickness on photographs.¹⁰⁸

OCT processing yields several metrics useful in characterizing the thickness of the retina in DME. For example, the Stratus OCT used by DRCR.net sites calculates a center-point thickness, CSMT, inner and outer subfield mean thicknesses, and macular volume for commonly used scanning algorithms. A DRCR.net study concluded that CSMT is well suited as a metric for central macular thickness in clinical research, with high reproducibility and good correlation to other metrics in the setting of DME.¹⁰⁹ Mean thickness in other subfields and total macular volume are useful in assessing extrafoveal macular edema.

OCT is capable of detecting areas of subtle macular edema that are sometimes not suspected on ophthalmoscopy (Fig. 47.8B). Easy detection of such “subclinical” thickening has added a new and provocative facet to management of DME, challenging clinicians to incorporate knowledge of more mild pathology into clinical decision-making. Conversely, OCT can illustrate areas of retinal thinning that sometimes result from advanced retinopathy. Occasionally, the presence of mild macular edema in retina that would be thinner than normal in the absence of such edema can result in a retinal thickness that appears “normal” relative to typical thickness in diabetics without retinopathy. Such abnormal thickening of retina that would otherwise be abnormally thin can make it challenging to define what constitutes complete resolution of macular edema in some eyes.

Retinal microvascular alterations and vascular hyperpermeability

Findings on ophthalmoscopy may suggest areas of vascular incompetence or retinal ischemia responsible for macular edema. For example, a circinate lipid ring may imply leakage from a particular cluster of microaneurysms, or intraretinal microvascular abnormalities may highlight the borders of a region of capillary closure. But more definitive information about the extent of microvascular alterations and any associated vascular hyperpermeability can be gained using FA. Stereoscopic images are sometimes helpful, allowing visualization of areas of retinal thickening and localization of angiographic features to various depths within or external to the retina.

The retinal microvasculature and any areas of capillary nonperfusion are best imaged in the arteriovenous transit phase of the angiogram. Microaneurysms and microvascular abnormalities, sometimes unsuspected on ophthalmoscopy, are readily visualized (Fig. 47.2). The absence of normal hyperfluorescence from capillaries in a region where they should usually exist indicates capillary nonperfusion (Fig. 47.9). Areas of foveal capillary nonperfusion manifest as an abnormally large foveal avascular zone (FAZ) or as irregularity in the borders of the FAZ. Unfortunately, visualization of capillary nonperfusion requires resolution near the limit of present camera systems, and even slight decreases in image quality (such as from poor focus or cataract) can impair assessment. The retina in areas of capillary nonperfusion may be edematous, normal thickness, or thin.

Fig. 47.9 Visualization of capillary nonperfusion and microvascular hyperpermeability by FA. (A) An early frame during arteriovenous transit highlights areas of capillary nonperfusion (visualized as a patchy absence of normal faint diffuse hyperfluorescence from capillaries, sometimes bordered by hyperfluorescent microvascular abnormalities [arrow]), microaneurysms, and microvascular abnormalities. (B) A late frame (5 minutes) illustrates focal fluorescein leakage from microaneurysms and patchy diffuse leakage from retinal capillaries. (C) A montage image created with frames from the mid-phase of the angiogram demonstrates widespread peripheral capillary nonperfusion and leakage of fluorescein from venules. Note that despite significant microvascular alterations and fluorescein leakage, visual acuity measured 20/16 and no macular thickening was present. Also note hyperfluorescence on the temporal aspect of the disc from early neovascularization.

(National Eye Institute, Bethesda, MD.)

FA is useful in evaluating retinal vascular competence, illustrating areas of hyperpermeability where dye leaks into the extravascular space (Fig. 47.9). Fluorescein leakage may emanate from discrete microaneurysms or microvascular abonormalities visible on the angiogram, or it may accumulate in areas of diffuse retinal capillary incompetence. Microaneurysms, microvascular abnormalities, and capillary telangiectasis visualized in the early phase of the angiogram can exhibit progressive leakage best appreciated in later phases, between 5 and 10 minutes after injection of dye. Fluorescein leakage may be present in a region of the retina that is edematous, normal thickness, or thin, and is therefore not synonymous with macular edema. However, ETDRS investigators and others have demonstrated correlation between area of fluorescein leakage and extent of retinal thickening on stereoscopic photographs and OCT.^110–112

Traction by vitreous gel and epiretinal proliferation

Macular edema can occur secondary to traction exerted by cortical vitreous gel and epiretinal membranes at points of attachment to the retina. In some cases retinal thickening may result solely from the mechanical forces transmitted to the retina, without significant secondary alterations in vascular permeability. Macular edema in such cases can result from traction even in the absence of retinal vascular leakage demonstrable by FA. In other cases, mechanical traction may exert an effect on the competence of retinal capillaries. Macular edema in these cases results from the combined effects of mechanical distortion and retinal microvascular leakage. Complicating clinical assessment of the role of mechanical traction in diabetic eyes is the backdrop of microvascular alterations present as a consequence of the metabolic disease. The clinician is faced with presence of macular edema, a visible hyaloid or epiretinal membrane with attachment to the retina, and a pattern of retinal vascular leakage on FA, and must judge how significant a role the membrane plays in the retinal thickening and vascular leakage.

In some cases, findings on biomicroscopy, OCT, and FA strongly suggest a tractional component to DME (Fig. 47.10). A thickened posterior hyaloid membrane or epiretinal membrane with attachment in the area of macular edema, with or without retinal striae, may be visible on biomicroscopic examination. OCT may show a hyaloid or epiretinal membrane stretched taut over an area of retinal thickening. This membrane may appear uniformly attached to the retina, or may show multiple points of focal attachment, the latter frequently associated with focal tentings of the inner aspect of the retina. The region of macular edema may correspond closely with the area in which the membrane has attachment to the retina, and retina outside this area of attachment may be uninvolved, with an appreciable step-off in thickness at the borders of membrane attachment. Fluorescein leakage, if present, may arise diffusely from telangiectatic retinal capillaries in the region of macular edema. These features are similar to those seen in vitreomacular traction in the setting of incomplete posterior vitreous separation or epiretinal membrane formation with macular edema in nondiabetic eyes.

Fig. 47.10 Macular edema predominantly secondary to vitreomacular traction and epiretinal proliferation. (A) A photograph of the left macula shows a thick epiretinal membrane in a 61-year-old woman with history of PDR treated with scatter laser photocoagulation. Retinal striae are visible in the nasal macula. There was no visible separation of the posterior hyaloid on biomicroscopy. Visual acuity measured 20/80. (B) An early frame of the FA shows relatively mild microvascular alterations with only occasional microaneurysms. Later frames (not pictured) showed only very mild diffuse leakage from capillaries. (C) An OCT 6 mm horizontal line scan through the foveal center shows presence of a thick epiretinal membrane with multiple focal attachments to the retina, resulting in a serrated appearance to the inner retinal surface. Severe central thickening is present. (D) A corresponding OCT line scan following vitrectomy with membrane peeling shows relief of traction and near-resolution of retinal thickening. Intraoperative medications included subconjunctival dexamethasone and cefazolin and postoperative medications included topical prednisolone acetate 1% and moxifloxacin. No laser photocoagulation was performed. Visual acuity improved to 20/40.

(National Eye Institute, Bethesda, MD.)

In other cases, a posterior hyaloid or epiretinal membrane visible on ophthalmoscopy or OCT may not appear to exert a predominant mechanical effect on the retina, and findings may suggest that retinal microvascular alterations from the metabolic disease are responsible for DME (Fig. 47.11). A subtle glistening of the vitreoretinal interface on biomicroscopy may belie the presence of the membrane, or the membrane may only be visible on OCT. A membrane visible on OCT may parallel the contour of the inner aspect of the retina with areas of attachment, but without any serrated distortion of the inner retina or retinal striae visible on biomicroscopy. The area of macular edema may exhibit a gradually tapering convexity, with or without presence of cysts, without any abrupt step-off to noninvolved areas at its borders. Biomicroscopy may show discrete microaneurysms in the area of macular edema, and FA may show areas of focal leakage from such microaneurysms or other microvascular abnormalities. Incomplete posterior vitreous separation or epiretinal membrane formation in the absence of retinal distortion or thickening in nondiabetic eyes is often compatible with normal vision. Such membranes, when present in eyes with DME, may be incidental to the macular edema associated with the retinopathy.

Fig. 47.11 Macular edema predominantly secondary to retinal vascular alterations from DM, with presence of an epiretinal or posterior hyaloid membrane exerting only mild mechanical effects. (A) A photograph of the right macula shows microaneurysms, intraretinal hemorrhages, cotton-wool spots, and hard exudates in a 74-year-old woman with severe NPDR. There is no visible epiretinal membrane or retinal striae formation. The pseudo-hole appearance at the foveal center is likely secondary to presence of shallow subfoveal fluid (see panel C). Visual acuity measured 20/50. (B) An early frame of the FA shows extensive vascular alterations, including presence of microaneurysms, patches of capillary nonperfusion, and microvascular abnormalities. Later frames (not pictured) showed severe leakage, most prominent in the areas of greatest retinal thickening inferonasal to the fovea. (C) An OCT 6 mm horizontal line scan through the foveal center shows presence of severe retinal thickening nasal to the fovea and mild retinal thickening temporal to the fovea, with presence of intraretinal cysts and scant subfoveal fluid. A thin epiretinal or posterior hyaloid membrane is present nasal and temporal to the foveal center. There is minimal distortion of the inner retinal surface, except at the temporal margin of the foveal depression, where an abrupt step-off in thickness suggests a local mechanical effect. (D) A corresponding OCT line scan 2 years later following treatment with multiple sessions of focal/grid laser photocoagulation, one bevacizumab injection, and one subtenon triamcinolone acetonide injection shows significant improvement in retinal thickening, with resolution of intraretinal cysts and subretinal fluid. Persistence of an abrupt step-off of the inner retinal surface at the temporal margin of the foveal depression suggests a mild local mechanical effect. No vitrectomy was performed.

(National Eye Institute, Bethesda, MD.)

In many eyes with a visible membrane and at least some evidence for mechanical distortion of the retina, DME may result from a combination of tractional effects and microvascular alterations associated with the diabetic retinopathy. It may be unclear which factor predominates, and in such cases, evaluation across time can be helpful. In clinical practice, this often involves treatment of DME, starting with less-invasive nonsurgical approaches, with careful assessment of the response to treatment.

Alterations in the retinal pigment epithelium

In most eyes with untreated DME, the RPE appears normal on ophthalmoscopy, OCT, and FA. Longstanding macular edema is occasionally accompanied by pigmentary alterations or atrophy of RPE. More commonly, RPE changes reflect stigmata of previous laser photocoagulation in eyes previously treated for DME (Fig. 47.12A). Visualization of discrete laser scars on FA, not always visible on ophthalmoscopy, allows assessment of the adequacy and extent of past treatment and can be helpful in planning further therapy (Fig. 47.12B). Autofluorescence photography has recently emerged as a less invasive means of evaluating such RPE alterations, and may largely supplant angiography for this application in DME (Fig. 47.12C).^113,114

Fig. 47.12 Pigmentary alterations of the RPE secondary to laser photocoagulation. (A) Color photograph showing scattered microaneurysms and intraretinal hemorrhages in the setting of NPDR and DME previously treated with focal/grid laser photocoagulation, with focal hypo- and hyperpigmented alterations of the RPE most prominent temporal to the foveal center. (B) An early frame of the FA shows multiple circular spots of hyperfluorescence, some with hypofluorescent centers, corresponding to laser scars. Note that more laser scars are visible on the FA than on the color photograph. (C) A fundus autofluorescence photograph (modified fundus camera, single-flash, 580/700 nanometers) demonstrates circular spots of hypofluorescence, some with hyperfluorescent centers, corresponding to the same laser scars visualized by FA. Intraretinal hemorrhages also appear hypofluorescent.

(National Eye Institute, Bethesda, MD.)

Subretinal fibrosis

Subretinal fibrosis is an uncommon feature of DME associated with severe hard exudates or disruption of the RPE by overaggressive laser treatment. In the ETDRS, investigators identified 109 eyes with subretinal fibrosis, defined as a mound or sheet of gray-to-white tissue beneath the retina at or near the center of the macula.¹¹⁵ Very severe hard exudates were present in the macula of 74% of these eyes prior to appearance of subretinal fibrosis; by comparison, exudates this severe were seen only in 2.5% of eyes with CSME that did not develop subretinal fibrosis (P <0.001). Among 264 eyes with very severe hard exudates in the macula at any time during the ETDRS, subretinal fibrosis developed in 31%, while among 5498 eyes with CSME and lesser levels of hard exudation, such fibrosis developed in only 0.05%. In the 4823 eyes that underwent laser photocoagulation for treatment of DME, only 9 eyes developed subretinal fibrosis adjacent to a laser photocoagulation scar, making this at worst a rare complication of such treatment.

Visual acuity and its correlation to retinal thickening and fluorescein leakage

Retinal thickening that involves or threatens the fovea frequently leads to vision loss. In the ETDRS, the 3-year risk of moderate vision loss was 32% among eyes observed with CSME.³⁸ However, studies have shown variable but generally modest correlation between central retinal thickness and concurrently measured visual acuity.^116–124 In one study of 251 eyes of 210 participants in a randomized clinical trial evaluating laser techniques, in which vision was tested with an electronic ETDRS protocol following standardized refraction and central retinal thickening was measured using time domain OCT and evaluated by a reading center, the correlation between visual acuity and center-point thickness was 0.52 at the baseline time point (Fig. 47.13).¹²⁵ The slope of the corresponding best-fit line was 4.4 letters (95% CI 3.5–5.3) of better visual acuity for every 100 µm decrease in center-point thickness.

Fig. 47.13 Scatterplot comparing concurrently measured OCT center-point thickness and visual acuity at a single timepoint (baseline) in 251 eyes (210 participants) with DME in a randomized clinical trial evaluating laser techniques. The solid line represents the regression line and the dotted lines represent the 95% CI for the mean.

(Adapted with permission from Diabetic Retinopathy Clinical Research Network. Relationship between optical coherence tomography-measured central retinal thickness and visual acuity in diabetic macular edema. Ophthalmology 2007;114:525–36.)

Fluorescein leakage on FA has low correlation with concurrently measured visual acuity. Among 422 eyes (a mix of study and nonstudy eyes) of participants in the randomized clinical trial evaluating laser techniques mentioned previously, the area of fluorescein leakage within an ETDRS macular grid was graded at baseline and correlated to several other baseline measures.¹¹¹ Correlation of the area of fluorescein leakage with visual acuity was 0.33, compared with a value of 0.38 for OCT CSMT and 0.58 for OCT total macular volume.

Diurnal variation of DME

On average, retinal thickness in DME decreases slightly during the day, but the proportion of eyes with DME exhibiting clinically meaningful changes is small. In the largest study to date, 156 eyes of 96 participants with center-involved DME on ophthalmoscopy and CSMT of 225 microns or greater were evaluated at six time points between 8am and 4pm using the Stratus OCT.¹²⁶ Two scans of adequate quality were obtained at each time point and sent to a reading center. The mean change in relative central subfield thickening, defined to represent the change in excess retinal thickness, was a decrease of 6% (95% CI −9% to −3%) between the 8am and 4pm time points. The mean absolute change was a decrease in 13 µm (95% CI −17 to −8 µm). Three percent (5 of 156) of eyes met a composite endpoint of 25% or greater decrease in relative central subfield thickening and 50 µm or greater decrease in CSMT at two consecutive time points, and 1% (2 of 156) of eyes exhibited increases in both measures by at least these amounts.

Modification of systemic risk factors

Control of hyperglycemia is critical to minimizing risk of onset and progression of diabetic retinopathy. The benefits of better glycemic control in reducing risk for retinopathy progression have been demonstrated in multiple randomized controlled clinical trials, including the DCCT, EDIC study, UKPDS, and ACCORD study mentioned previously and further discussed in Chapter 45 (The epidemiology of diabetic retinopathy).^{72,73,75,76,78,79} The American Diabetes Association recommends glycemic control targeting a hemoglobin A1C of 7.0% or lower for most diabetics.¹²⁷ Not only does such a target lower risk of retinopathy onset and progression, it also lowers risk of other microvascular complications of disease, such as neuropathy and nephropathy. The ACCORD study has provoked recent controversy about whether more aggressive targeting of near-normal hemoglobin A1C values is desirable.^128,129 As discussed earlier, the ACCORD study included a trial that randomly assigned 10 251 patients with type 2 DM to intensive glycemic control targeting HbA1C of less than 6% or to standard treatment targeting HbA1C between 7 and 7.9%. The glycemia trial was stopped after a median of 3.7 years because of an increase in all-cause mortality in the intensive treatment group (5% in the intensive treatment group versus 4% in the standard treatment group; hazard ratio 1.21; 95% CI 1.02–1.44).¹³⁰ Rates of hypoglycemia requiring assistance were significantly higher in the intensive treatment group than in the standard treatment group (10.5% vs 3.5%; P = 0.001), corroborating similar findings among intensively treated participants in other clinical trials of glycemic control, but the higher mortality rate noted in participants in the ACCORD study could not be readily attributed to complications of hypoglycemia. Despite evidence that aggressive targeting of near-normal glycosylated hemoglobin levels may offer benefit in reducing microvascular complications of DM, there is consensus that such an approach should be pursued cautiously based on the mortality findings of the ACCORD study.¹²⁷

Control of hypertension is also beneficial in lowering risk of progression of diabetic retinopathy, as demonstrated in the UKPDS.^74,80 Hypertensive participants with newly diagnosed type 2 DM in the UKPDS were randomly assigned to more intensive blood pressure control (targeting a systolic blood pressure less than 150 mmHg) and less intensive blood pressure control (targeting a systolic blood pressure less than 180 mmHg) and also randomly assigned to treatment with beta-blockers or ACE inhibitors. The UKPDS showed benefit of better control of blood pressure, as discussed previously. The ACCORD study, which included a blood pressure trial that randomly assigned participants to more or less intensive control of blood pressure (targeting a systolic blood pressure less than 120 mmHg or less than 140 mmHg, respectively), did not show benefit in reducing risk of retinopathy progression in a subset of participants evaluated for eye disease.⁷⁸ While provocative, the results of the ACCORD study do not necessarily contradict the findings of the UKPDS, given the very different targets used for more and less intensive control of blood pressure between the two trials. It is possible that modest control of severe hypertension (as done in the UKPDS) lowers retinopathy progression, while very aggressive lowering of blood pressure below present standard of care (as done in the ACCORD study) lends no additional benefit. Alternatively, the disparity in findings could be attributable to differences in the study populations.

Treatment of dyslipidemia may be beneficial to retinopathy, based on consideration of observational data from studies like the WESDR and the ETDRS.^81,82 However, there is no evidence from randomized controlled clinical trials that plasma lipid-lowering reduces risk of retinopathy onset or progression, largely because persons with dyslipidemia need to lower their lipid levels to reduce the risk of cardiovascular disease, making such trials difficult or impossible. The ACCORD study, together with another randomized controlled clinical trial called the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study, suggests a possible role for fenofibrate in reducing risk of retinopathy progression, but it is unclear whether the mechanism of action involves alteration of the plasma lipid profile.^78,84 Fenofibrate is discussed further below as a potential systemic treatment for diabetic retinopathy.

Retinopathy screening and surveillance

Sight-threatening retinopathy may not cause symptoms prompting evaluation until disease is advanced. Treatment to reduce risk of vision loss in eyes with sight-threatening complications of diabetic retinopathy is most effective when initiated before severe vision loss has occurred. These facts underpin the importance of screening and surveillance for retinopathy, but reports suggest that many diabetics do not receive eye examination on the schedule recommended by organizations such as the American Diabetes Association and the American Academy of Ophthalmology.^92,131–135 For example, an analysis of Medicare claims data for beneficiaries 65 years of age and older revealed that only between 50 and 60% of diabetics received annual eye examinations over a 15-month period.¹³³ In many less-developed countries the situation is much worse, with only a small fraction of diabetics receiving any eye evaluation at all.

The recommended schedule for screening and surveillance for NPDR reflects knowledge about the epidemiology and natural history of disease. Initial eye examination is recommended 3–5 years following diagnosis of type 1 DM, and at time of diagnosis for those with type 2 DM.⁹² Recommended follow-up examination for type 1 and type 2 diabetics with no retinopathy is yearly. In the absence of DME, those with mild to moderate NPDR should be evaluated every 6–12 months, and those with severe NPDR should be seen every 2–4 months. Diabetics with DME merit frequent follow-up, generally at least every 2–4 months, and sometimes monthly depending on treatment. Any new ocular symptoms should prompt timely evaluation tailored to the circumstances.

In the setting of pregnancy, eye examinations are recommended prior to conception and early during the first trimester.⁹² Follow-up for pregnant patients with no retinopathy, mild NPDR, or moderate NPDR should be individualized based on the severity and recent changes in retinopathy. Pregnant patients with severe NPDR should be evaluated every 1–3 months. Specific circumstances, such as presence of DME, may dictate need for more frequent follow-up.

A comprehensive eye evaluation including dilated funduscopic examination by an ophthalmologist experienced in management of diabetic eye disease remains the standard of care for retinopathy screening in areas with adequate access to ophthalmic care. However, as imaging and information-sharing technologies continue to improve, remote screening for diabetic retinopathy, such as review of digital fundus photographs acquired outside the ophthalmologist’s office, may offer a cost-effective alternative. In such a model, patients manifesting findings indicative of a certain level of retinopathy and patients for whom adequate images cannot be obtained are referred to an ophthalmologist for full evaluation.¹³⁶ Such an approach has particular promise in settings where access to ophthalmic care is limited. A number of studies have demonstrated reasonable sensitivity and specificity of various remote screening platforms compared with the standard of dilated funduscopic examination by an ophthalmologist or expert evaluation of seven-field fundus photographs obtained by an ophthalmic photographer.^137–143

Ocular treatment for diabetic macular edema

For years following the ETDRS, the standard of care for treatment of clinically significant diabetic macular edema has been focal/grid laser photocoagulation, with consideration of other measures such as vitrectomy in select circumstances. Recent studies and clinical experience with intravitreous injection of agents inhibiting the activity of vascular endothelial growth factor A (VEGF-A) isoforms, often done in combination with focal/grid laser photocoagulation, have demonstrated short-term efficacy superior to laser photocoagulation alone in certain circumstances. An expanding array of corticosteroid formulations and delivery systems has also been evaluated in recent clinical trials, adding to our knowledge about the efficacy and hazards of these medications.

Recent success of certain pharmacologic agents for treatment of DME has been exciting, but even the most informative studies raise important unsettled questions. Management algorithms for DME have been made more complex by the availability of alternatives to laser photocoagulation, resulting in decision-making individualized to a variety of factors. The value of information gained from high-quality randomized controlled clinical trials and the importance of long-term outcomes in evaluating new therapies cannot be overstated. Recent experience with use of intravitreous injection of triamcinolone acetonide for treatment of DME illustrates both points. Early clinical experience, case series, and small trials with short-term follow-up suggested significant and occasionally dramatic improvement in macular edema following intravitreous administration of triamcinolone, and the treatment gained popularity, particularly in diabetics refractory to laser photocoagulation therapy.^{121,144–150} A randomized controlled clinical trial was carried out comparing focal/grid laser photocoagulation and intravitreous injections of triamcinolone acetonide (1 mg or 4 mg) for treatment of DME involving the center of the macula.^151,152 At 4 months, mean visual acuity was significantly better in eyes treated with triamcinolone (4 mg) than in eyes treated with laser photocoagulation. However, at evaluation of the primary outcome at 2 years, mean visual acuity was significantly better in eyes treated with focal laser than in eyes from either of the triamcinolone groups, a difference that could not be attributed solely to cataract progression in eyes receiving triamcinolone. The results of this trial offered critical clarification of the benefits and limitations of intravitreous triamcinolone not discernible from short-term studies and anecdotal experience.

Despite important progress in treatment, DME remains a significant cause of vision loss. This reflects shortcomings in diagnosis and treatment of DM, inadequacies in screening and surveillance of diabetic retinopathy, and limitations of present treatment for DME. Ocular therapy for DME reduces risk of future vision loss, and may improve vision, but often does not restore normal macular function or structure. In a recent randomized controlled clinical trial demonstrating some of the most significant benefits seen to date for any treatment of DME, groups gaining most vision (those treated with intravitreous injections of the VEGF antagonist ranibizumab plus prompt or deferred focal/grid laser photocoagulation) demonstrated an 8- or 9-letter mean gain in visual acuity at the 2-year visit compared with a baseline mean visual acuity of 63 letters.¹⁵³ Such results indicate that on average vision in these eyes improved from approximately 20/50− to 20/40+. At 2 years, the percentage of eyes with CSMT equal to or greater than 250 µm (using a time domain system for which normal CSMT measures approximately 200µm) was 43% in the ranibizumab plus prompt laser group and 42% in the ranibizumab plus deferred laser group, reflecting a significant number of eyes with residual macular edema despite treatment.

The ETDRS concept of clinically significant macular edema (CSME), discussed earlier in this chapter, remains relevant to management of DME. The ETDRS demonstrated that eyes with DME involving or threatening the center of vision are at high risk of vision loss,³⁸ and there is consensus that such eyes merit treatment to reduce the chance of further vision loss. However, eyes with clinically significant macular edema that threatens but does not involve the macular center have often been excluded from recent clinical trials, making for greater uncertainty about the best treatment options in this subset of eyes. Such eyes received focal/grid laser photocoagulation in the ETDRS, but it is not known how laser treatment compares to other therapeutic options. Likewise, recent trials have frequently excluded eyes with very good and very poor visual acuity, making evidence-based decision-making more difficult in these cases as well. Finally, as mentioned previously, the sensitivity of OCT for retinal thickening allows identification of very mild disease, often termed “subclinical” macular edema because it is not easily visible on biomicroscopy. OCT thickness parameters designated in eligibility criteria and retreatment algorithms from clinical trials indicate what threshold of thickening was considered significant enough to treat and retreat in a given study; however, such information provides no guidance on whether outcomes would have differed if other thickness thresholds had been used instead. Areas of macular thickening and hard exudate not meeting criteria for CSME can usually be observed vigilantly, with institution of treatment if CSME evolves.³⁸

Attention to the context of DME is important for management. The principles of treatment of DME remain the same regardless of the underlying severity of diabetic retinopathy, but management may differ depending on the circumstances. There are special considerations for treatment of DME in the context of PDR, given the imperative to administer any necessary treatment to lower risk of vision loss from the complications of proliferative disease. Chapter 48 (Proliferative diabetic retinopathy) discusses considerations for management of DME in the setting of PDR. Other ocular conditions may impact DME or the choice of treatment. For example, postoperative cystoid macular edema may complicate DME following intraocular surgery and may warrant different treatment. Presence of glaucoma may contraindicate use of local corticosteroids. Some such considerations are included in the following discussion, but an exhaustive review is beyond the scope of this chapter.

Pharmacotherapy with vascular endothelial growth factor (VEGF) antagonists

VEGF-A has been demonstrated to exert potent effects on retinal vascular permeability, and concentrations of certain VEGF-A isoforms in the retina and vitreous are elevated in diabetic retinopathy.^40,41,170 Various VEGF antagonists have been developed for ophthalmic and nonophthalmic uses, including bevacizumab, a humanized murine monoclonal antibody binding VEGF-A; ranibizumab, a humanized murine monoclonal antibody fragment, also binding VEGF-A; pegaptanib sodium, an aptamer specifically inhibiting the VEGF-A 165 isoform; and aflibercept, a human fusion protein incorporating ligand-binding elements from VEGF receptors and the Fc region of an IgG1 molecule. Intravenous administration for ophthalmic disease has been studied for some agents,^171–173 but intravitreous injection has rapidly become the most common mode of delivery to the eye following seminal clinical trials in neovascular age-related macular degeneration and recent clinical experience demonstrating a reasonable safety profile. Following demonstration of efficacy in neovascular age-related macular degeneration, choroidal neovascularization from other causes, and macular edema following retinal vein occlusion, various VEGF antagonists have been evaluated for treatment of DME in a number of small studies with short-term follow-up.^174–180 Encouraging results have led to several larger clinical trials.

The first major clinical trial to report results was designed by DRCR.net investigators to compare four strategies for treatment of DME including focal/grid laser photocoagulation alone, intravitreous injection of ranibizumab (0.5 mg) with deferral of early laser, intravitreous injection of ranibizumab (0.5 mg) with early laser, and intravitreous injection of triamcinolone acetonide (4 mg) with early laser.^153,169 A total of 854 eyes of 691 participants with DME involving the center of the fovea and visual acuity of 20/32 to 20/320 were randomly assigned to one of four treatments. The main outcome measure, best-corrected visual acuity, was evaluated at one year, with follow-up planned for 3 years. Eyes in a prompt laser plus sham injection group received modified-ETDRS focal/grid laser photocoagulation within 3–10 days following sham injection. Those in a ranibizumab plus prompt laser group were given an intravitreous injection of ranibizumab followed by focal/grid laser photocoagulation within 3–10 days. Those in a ranibizumab and deferred laser group received intravitreous injection of ranibizumab as required by protocol without any laser treatment for at least 24 weeks. Finally, those in a triamcinolone plus prompt laser were given an intravitreous injection of a preservative-free formulation of triamcinolone acetonide followed by focal/grid laser photocoagulation within 3–10 days. Retreatment algorithms were complex, but were generally intended to require further therapy in eyes with residual disease in order to maximize treatment effects in the first year. Focal/grid laser photocoagulation was repeated as often as every 13 weeks, ranibizumab was readministered as frequently as every 4 weeks, and triamcinolone acetonide was reinjected as often as every 16 weeks. At one year, mean change in visual acuity was significantly better in the ranibizumab plus prompt laser (+9 letters) and ranibizumab and deferred laser (+9 letters) groups compared with the prompt laser plus sham injection group (+3 letters; both P <0.001). Mean change in visual acuity was not significantly different from the prompt laser plus sham injection group in the triamcinolone plus prompt laser group (+4 letters; P = 0.31). Results at 2 years were similar (Fig. 47.14). Median number of injections in the first year was 8 in the ranibizumab plus prompt laser group, 9 in the ranibizumab and deferred laser group, and 3 in the triamcinolone plus prompt laser group. Median number of laser treatments in the first year in these groups was 2, 0, and 2, respectively, compared with 3 in the prompt laser plus sham injection group. Median number of injections in the second year in these groups decreased to 2, 3, and 1, respectively. There were 3 cases of postinjection endophthalmitis among 3973 ranibizumab injections (0.08%; 95% CI 0.02–0.22%), and 1 case of progression of pre-existing tractional retinal detachment in the ranibizumab and deferred laser group. There were no significant differences in rates of systemic adverse events between groups, and specifically no increase in cardiovascular or cerebrovascular events in the ranibizumab groups.

Fig. 47.14 Two-year results of a randomized clinical trial of 854 eyes (691 participants) evaluating intravitreous injection of ranibizumab (0.5 mg) or triamcinolone acetonide (4 mg) combined with focal/grid laser photocoagulation compared to focal/grid laser photocoagulation alone for treatment of DME. The primary outcome was measured at 52 weeks (corresponding to vertical line on graphs). (A) Mean change in visual acuity at follow-up visits for cohort that completed 2-year visit. Values that were ±30 or more letters were assigned a value of 30. P values for difference in mean change in visual acuity from sham+prompt laser at the 104-week study visit: ranibizumab+prompt laser = 0.03, ranibizumab+deferred laser <0.001, and triamcinolone+prompt laser = 0.35. (B) Mean change in OCT mean central subfield thickness at follow-up visits for cohort that completed 2-year visit. P values for difference in mean change in OCT mean central subfield thickness from sham+prompt laser at the 104-week visit: ranibizumab+prompt laser = 0.003, ranibizumab+deferred laser = 0.01, and triamcinolone+prompt laser = 0.37.

(Adapted with permission from Elman MJ, Bressler NM, Qin H, et al. Expanded 2-year follow-up of ranibizumab plus prompt or deferred laser or triamcinolone plus prompt laser for diabetic macular edema. Ophthalmology 2011;118:609–14.)

The results of this study have established a role for ranibizumab in treatment of DME, and have already been supplemented by findings from other trials, including one-year results of a small randomized controlled trial evaluating use of intravitreous bevacizumab.^181–183 The RESTORE study randomly assigned 345 participants with DME and visual acuity of 20/32 to 20/160 to intravitreous injection of ranibizumab (0.5 mg) and sham laser, injection of ranibizumab (0.5 mg) and focal/grid laser photocoagulation, or sham injection and focal/grid laser photocoagulation.¹⁸² Participants received ranibizumab or sham injections monthly for 3 months, then as needed according to retreatment criteria; participants received laser photocoagulation at baseline and then as needed as often as every 3 months at the discretion of investigators. At 12 months, the mean change in visual acuity in the group getting ranibizumab alone (+6.1 letters) and in the group getting ranibizumab plus laser (+5.9 letters) was significantly better compared with the group getting laser alone (+0.8 letters; both P <0.0001). The BOLT study randomly assigned 80 eyes of 80 participants with DME involving the foveal center and visual acuity of 20/40 to 20/200 to intravitreous injection of bevacizumab (1.25 mg) or modified-ETDRS focal/grid laser photocoagulation.¹⁸³ Participants getting bevacizumab received three injections spaced apart by 6 weeks each, followed by repeat injections every 6 weeks as needed according to retreatment criteria. Participants getting laser treatment received focal/grid laser photocoagulation at baseline and as often as every 4 months thereafter as needed based on ETDRS guidelines. At 12 months, the mean change in visual acuity was significantly better in bevacizumab group (+5.6 letters) than in the laser group (−4.6 letters; P = 0.0006). The results from these studies should soon be supplemented by longer-term follow-up data and findings from other large clinical trials presently underway.

A host of new questions heralds the beginning of what will probably be a period of rapid evolution in treatment paradigms involving VEGF antagonists for DME. What will long-term follow-up of eyes receiving ranibizumab injections reveal? Can we identify factors that reliably predict which eyes stand to benefit most or least from treatment? What retreatment criteria are sensible in different circumstances? What should be the role of focal/grid laser photocoagulation relative to pharmacotherapy in a variety of circumstances? Does treatment with bevacizumab or other VEGF antagonists yield results appreciably different from those of ranibizumab? Can we devise methods superior to frequent intravitreous injection for drug delivery to the retina? Can we identify other pharmacotherapies that complement anti-VEGF treatment and allow better treatment of residual macular edema?

Pharmacotherapy with corticosteroids

Corticosteroids have immune modulatory and antiangiogenic properties and have been utilized for treatment of ophthalmic disease since the 1950s.^184–189 Common adverse effects, including hyperglycemia and other metabolic alterations difficult to manage in the setting of DM, limit their long-term systemic use, but a number of formulations allowing local delivery to the eye have been developed. The poor intraocular penetration of existing topical preparations makes them unsuitable for treatment of most retinal diseases, but injectable corticosteroids and sustained-release formulations for intraocular use have been evaluated for efficacy in DME.

Our most definitive knowledge about intravitreous injection of triamcinolone acetonide for treatment of DME comes from two randomized controlled clinical trials conducted by the DRCR.net.^{151,152,153,169} The first study enrolled 840 eyes of 693 participants with DME involving the foveal center and visual acuity of 20/32 to 20/320, with a primary outcome measure of mean change in best-corrected visual acuity at two years.^151,152 Eyes were randomly assigned to receive modified-ETDRS focal/grid laser photocoagulation, intravitreous injection of triamcinolone acetonide (1 mg), or intravitreous injection of triamcinolone acetonide (4 mg), both corticosteroid arms utilizing a preservative-free formulation of the drug. Persistent or new macular edema was retreated every 4 months. Mean change in visual acuity at two years was significantly better in laser-treated eyes (+1 letter) than in eyes receiving 1 mg triamcinolone (−2 letters; P = 0.02) and likewise significantly better than in eyes receiving 4 mg triamcinolone (−3 letters; P = 0.002). The mean number of treatments over two years was 2.9 in the laser group, 3.5 in the 1 mg triamcinolone group, and 3.1 in the 4 mg triamcinolone group. Corticosteroids are well-known to promote cataract and to elevate intraocular pressure in some eyes, and such effects were carefully documented in this study. Cataract surgery was performed in 13% of eyes in the laser group, 23% of eyes in the 1 mg triamcinolone group, and 51% of eyes in the 4 mg triamcinolone group by two years. Intraocular pressure elevation of 10 mmHg or more from baseline at any study visit was noted in 4, 16, and 33% of eyes in the three groups, respectively, by two years. There were no cases of endophthalmitis or inflammatory pseudoendophthalmitis among 1649 intravitreous injections. Addressing whether cataract formation obscured benefit of treatment of DME in eyes treated with corticosteroids, subgroup analysis of 145 eyes pseudophakic at baseline was performed and at two years showed mean change in visual acuity of +2 letters in the laser group, +2 letters in the 1 mg triamcinolone group, and −1 in the 4 mg triamcinolone group. Investigators judged that the superiority of focal/grid laser photocoagulation could not be explained solely by cataract progression in triamcinolone-injected groups. This conclusion was supported by data on change in retinal thickness. At 2 years, mean improvement in CSMT was significantly greater in the laser group (139 µm) compared with the 1 mg triamcinolone group (86 µm; P <0.001) and the 4 mg triamcinolone group (77 µm; P <0.001).

The other study, already discussed above in the section “Pharmacotherapy with VEGF antagonists,” compared a combination of intravitreous injection of preservative-free triamcinolone (4 mg) and focal/grid laser photocoagulation to three other strategies, including focal/grid laser photocoagulation alone, intravitreous injection of ranibizumab plus early focal/grid laser photocoagulation, and intravitreous injection of ranibizumab and deferral of early laser.^153,169 At the primary endpoint of one year, mean change in visual acuity was not significantly different in the triamcinolone plus laser group (+4 letters) compared with the prompt laser plus sham injection group (+3 letters; P = 0.31). However, subgroup analysis of eyes pseudophakic at baseline showed a trend toward benefit in the triamcinolone plus laser group comparable to that seen in ranibizumab groups and superior to that seen in the laser group (Fig. 47.15). Investigators judged that cataract formation, cataract surgery, or both may have adversely impacted benefit of combination therapy with intravitreous injection of triamcinolone and focal/grid laser photocoagulation in phakic eyes. Taken together, the results of these two trials suggest that while monotherapy with intravitreous injection of triamcinolone acetonide is inferior to monotherapy with focal/grid laser photocoagulation, the combination of the two treatments in pseudophakic eyes might possibly be superior to laser alone.

Fig. 47.15 Two-year mean change in visual acuity for eyes pseudophakic at baseline in a randomized clinical trial evaluating intravitreous injection of ranibizumab (0.5 mg) or triamcinolone acetonide (4 mg) combined with focal/grid laser photocoagulation compared to focal/grid laser photocoagulation alone for treatment of DME. The primary outcome was measured at 52 weeks (corresponding to vertical line on graph). Values that were ±30 or more letters were assigned a value of 30. P values for difference in mean change in visual acuity from sham+prompt laser at the 104-week study visit: ranibizumab+prompt laser = 0.83, ranibizumab+deferred laser = 0.15, and triamcinolone+prompt laser groups = 0.53.

(Adapted with permission from Elman MJ, Bressler NM, Qin H, et al. Expanded 2-year follow-up of ranibizumab plus prompt or deferred laser or triamcinolone plus prompt laser for diabetic macular edema. Ophthalmology 2011;118:609–14.)

Sustained-release devices designed for intraocular delivery of fluocinolone acetonide and dexamethasone have been developed and tested for various ophthalmic applications, including treatment of DME. The first to achieve wide clinical use was the fluocinolone acetonide intravitreal implant (Retisert, Bausch & Lomb, Rochester, NY), consisting of a 0.59 mg pellet embedded in a nonbiodegradable scaffold designed to be implanted in the vitreous cavity via a sclerotomy and anchored by a suture to the eye wall. The Retisert, which releases drug at steady state between 0.3 and 0.4µg/day for approximately 30 months, has been used most commonly for treatment of chronic non-infectious posterior uveitis.^190–193 Though small studies were initiated for treatment of DME, the requirement for incisional surgery and the high rates of cataract progression and intraocular pressure elevation associated with the device have limited its use for this indication.¹⁹⁴

An intravitreous insert (Iluvien, Alimera Sciences, Alpharetta, GA) releasing stable quantities of fluocinolone acetonide for at least a year, consisting of a non-biodegradable cylinder (3.5×0.37 mm) containing the same polymer matrix as the Retisert implant and introduced into the vitreous cavity via injection by 25-gauge needle, was recently evaluated for treatment of DME in a large randomized controlled clinical trial.¹⁹⁵ Statistically significant benefit was demonstrated at 2 years for inserts releasing 0.2 and 0.5µg/day compared with sham injection in a study design that involved use of “rescue” focal/grid laser photocoagulation for persistent macular edema in all three groups. Progression of cataract and elevation of intraocular pressure were common in eyes receiving steroid inserts. Visual gain in eyes receiving inserts was greater in those that were pseudophakic at baseline and similar in magnitude to benefit seen in eyes treated with intravitreous triamcinolone and prompt focal/grid laser photocoagulation in the DRCR.net trial.

A dexamethasone intravitreous implant (Ozurdex, Allergan, Irvine, CA) consisting of a biodegradable pellet releasing high levels of drug for approximately 60 days and delivered via surgical sclerotomy or 22-gauge needle-injector system has been evaluated in a randomized clinical trial for macular edema.¹⁹⁶ Two versions, a 350 µg and 700 µg version, were tested for efficacy in 171 eyes with DME comprising a subgroup of a short-term larger study also enrolling eyes with macular edema secondary to branch and central retinal vein occlusion, uveitis, and cataract surgery. At the primary endpoint at 90 days, the percentage of eyes gaining 10 or more letters was significantly higher in the dexamethasone 700 µg microgram implant group than in the observation group (33% vs 12%; P = 0.007).

Although there is ample evidence for benefit of various corticosteroids in treatment of DME, none has shown superiority to focal/grid laser photocoagulation or intravitreous injection of VEGF antagonists when used as monotherapy. However, these medications show promise as components of combination therapy, particularly in the setting of DME refractory to other therapies and in the setting of pseudophakia. The need for less frequent administration is a present advantage over monthly intravitreous injection of ranibizumab or bevacizumab, but frequent follow-up is still necessary to monitor for intraocular pressure elevation and glaucoma.

Ocular treatment for nonproliferative diabetic retinopathy

Various local therapies have been evaluated for their effects on altering the course of NPDR. The efficacy of scatter laser photocoagulation for reduction of severe vision loss among eyes with advanced retinopathy (severe NPDR and PDR) was established in the Diabetic Retinopathy Study.²¹³ The ETDRS was designed partly to clarify the optimal point during the course of retinopathy for administration of scatter laser photocoagulation. In the ETDRS, eyes were randomly assigned to early laser photocoagulation or deferral of laser treatment. Eyes were categorized according to presence or absence of macular edema and according to severity of retinopathy, and further randomly assigned to various laser strategies in a complex study design. In addition to testing the effects of focal/grid laser photocoagulation for macular edema, the ETDRS also evaluated the effects of immediate “mild” and “full” scatter photocoagulation in eyes with more severe retinopathy (severe NPDR and non-high-risk PDR) with and without macular edema and delayed mild and full scatter photocoagulation in eyes with less severe retinopathy and macular edema. At 5 years, the rate of severe vision loss (defined as visual acuity less than 5/200 at two consecutive visits) was low in eyes receiving early laser treatment (2.6%) and in those with deferral of laser (3.7%) (Fig. 47.16).¹⁰³ Rates of severe vision loss in the subset of eyes with mild or moderate NPDR were even lower. On the basis of these findings, ETDRS investigators recommended against scatter laser photocoagulation for mild and moderate NPDR, provided that adequate follow-up could be anticipated. They suggested consideration of scatter laser photocoagulation for severe NPDR and non-high risk PDR, with potential benefit balanced against adverse effects of laser on visual acuity and visual fields. The use of scatter laser photocoagulation in diabetic retinopathy is further discussed in Chapter 48 (Proliferative diabetic retinopathy).

Fig. 47.16 Life table cumulative event rates of severe vision loss in all eyes assigned to immediate laser photocoagulation (blue) or deferral of laser photocoagulation (black) in the ETDRS.

(Adapted with permission from Early Treatment Diabetic Retinopathy Study Research Group. ETDRS report number 9. Early photocoagulation for diabetic retinopathy. Ophthalmology 1991;98(5 Suppl):766–85.)

There is preliminary evidence suggesting that local administration of VEGF antagonists or corticosteroids for treatment of DME may also benefit the course of NPDR. In the aforementioned DRCR.net study comparing various strategies of focal/grid laser photocoagulation and intravitreous administration of ranibizumab or triamcinolone acetonide combined with laser for treatment of DME, ranibizumab- and triamcinolone-treated eyes appeared less likely to exhibit progression of retinopathy and more likely to exhibit improvement of retinopathy than eyes treated with focal/grid laser photocoagulation alone, as assessed by reading-center evaluation of fundus photographs taken at baseline and one year.^153,169 The trend in most subgroups was not statistically significant. Ranibizumab- and triamcinolone-treated eyes were significantly less likely to manifest vitreous hemorrhage or receive scatter laser photocoagulation during the first year than eyes treated with focal/grid laser photocoagulation alone. A similar exploratory analysis was performed on data from the DRCR.net study comparing focal/grid laser photocoagulation to intravitreous injection of triamcinolone acetonide (1 mg or 4 mg), with worsening of retinopathy defined as a composite measure including evidence of progression on fundus photographs evaluated by a reading center, any documentation of vitreous hemorrhage, or receipt of scatter laser photocoagulation.²¹⁴ At 2 years, the cumulative probability of retinopathy progression was significantly lower in the 4 mg triamcinolone group (21%) but not in the 1 mg triamcinolone group (29%) compared with the laser group (31%; P = 0.005 and P = 0.64, respectively). Such preliminary findings are intriguing, but merit more definitive investigation to characterize any benefit and its clinical significance.

Other systemic treatment for nonproliferative diabetic retinopathy

Antiplatelet agents have been evaluated for treatment of diabetic retinopathy in a few randomized controlled clinical trials. In a 3-year study, 475 participants with early retinopathy were randomly assigned to daily aspirin, aspirin plus dipyridamole, or placebo, and progression of retinopathy was measured as the change in the number of microaneurysms visualized on FA.²¹⁵ The annual increase in macroaneurysms was significantly greater in the placebo group than in the treatment groups, but the growth rate was low (less than two microaneurysm per year) in all groups. In an analogous 3-year study involving 435 participants randomly assigned to daily ticlopidine or placebo, the annual increase in microaneurysms was greater in the placebo group than in the treatment group in an analysis that showed low appearance rates and borderline statistical significance.²¹⁶ Neither study showed a change in severity level of retinopathy, and the clinical significance of differences in microaneurysm counts remains uncertain.

The most definitive evidence for effects of antiplatelet therapy comes from the ETDRS, in which 3711 participants were randomly assigned to aspirin (650 mg per day) or placebo. Eyes assigned to deferral of laser photocoagulation were assessed for the effects of aspirin on progression of diabetic retinopathy. Aspirin use did not affect the severity of retinopathy or the risk of visual loss over 7 years.²¹⁷ Aspirin use did not increase the incidence, severity, or duration of preretinal or vitreous hemorrhage in eyes with deferral of photocoagulation or in eyes randomly assigned to early laser treatment.^217,218 Participants randomly assigned to aspirin exhibited a 17% reduction in morbidity and mortality from cardiovascular disease compared with participants receiving placebo, corroborating benefits seen in other studies. Aspirin remains an important therapy for control of cardiovascular risk in many diabetics, and no level of retinopathy severity, including PDR, should contraindicate its use. The ETDRS results apply to persons with significant diabetic retinopathy, and some uncertainty remains about any effects of antiplatelet agents in very mild NPDR.

Sorbinil, an inhibitor of the enzyme aldose reductase, which catalyzes the conversion of glucose to sorbitol, was evaluated in a randomized controlled clinical trial following animal studies, suggesting that these drugs might slow the development of diabetic retinopathy.^7,11,12 The Sorbinil Retinopathy Trial randomized 497 participants with type 1 DM and mild or no retinopathy to sorbinil or placebo.²¹⁹ The percentage of participants showing significant progression of diabetic retinopathy, defined as a two-step or greater change in the ETDRS diabetic retinopathy severity scale on fundus photographs, was not significantly different between the two groups after 3–4 years of follow-up.

Inhibitors targeting the protein kinase C family, which is implicated in VEGF-mediated vasopermeability and endothelin-mediated vasoconstriction in animal models of diabetic retinopathy,^220–223 have been developed and evaluated for effects on diabetic retinopathy. Ruboxistaurin, an inhibitor of beta-isoforms of protein kinase C, has been evaluated in two randomized controlled clinical trials. In the first, which randomly assigned 252 participants with moderate to very severe NPDR to receive ruboxistaurin or placebo for 36–46 months, rates of retinopathy progression were not significantly different between groups, but participants taking ruboxistaurin (32 mg/day) showed a significant delay in time to moderate vision loss (doubling of the visual angle) compared with those taking placebo (P = 0.038).²²⁴ In a second larger study, 685 participants with moderate to very severe NPDR were randomly assigned to ruboxistaurin or placebo and evaluated at 3 years for rates of sustained moderate vision loss, defined as decrease in ETDRS visual acuity score of 15 letters or more (doubling of the visual angle) for 6 months or longer.²²⁵ The rate of sustained moderate vision loss was significantly lower in the ruboxistaurin group (5.5%) than in the placebo group (9.1%), representing a 40% risk reduction (P = 0.034). Focal/grid laser photocoagulation was initiated 26% less frequently among those treated with ruboxistaurin than among those receiving placebo (P = 0.008). There was no significant difference between rates of retinopathy progression between the two groups. Despite intriguing results, further clinical testing necessary for regulatory approval in the United States and in Europe has not been pursued and the drug remains unavailable.

Fenofibrate, a peroxisome proliferator-activated receptor (PPAR) alpha agonist, has been evaluated for efficacy in treatment of diabetic retinopathy in combination with statin therapy as a plasma lipid-modulating agent capable of lowering triglyceride levels and raising high-density lipoprotein (HDL) cholesterol levels. Two randomized controlled clinical trials have suggested benefit of fenofibrate for diabetic retinopathy. The Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study was a large trial randomizing 9795 participants with type 2 DM to fenofibrate (200 mg/day) or placebo.⁸⁴ The cumulative percentage of participants receiving initiation of laser treatment for retinopathy (including focal/grid and scatter laser photocoagulation), a prespecified tertiary endpoint in the main study, was significantly lower in the fenofibrate group (3.4%) than in the placebo group (4.9%; hazard ratio, 0.69; 95% CI 0.56–0.84; P = 0.0002) over six years. Progression by two steps or greater on an adapted ETDRS diabetic retinopathy severity scale, the primary endpoint of a substudy involving analysis of two-field 45-degree fundus photographs for 1012 participants, did not differ significantly between fenofibrate and placebo groups. Investigators commented on a significantly higher rate of retinopathy progression in the placebo group compared with the fenofibrate group among participants with retinopathy at baseline, but the number of incident cases was small in both groups. The ACCORD study, discussed above, was a complex trial that included a lipid study randomizing 5518 participants with type 2 DM and dyslipidemia to receive simvastatin plus fenofibrate (160 mg/day) or simvastatin plus placebo. In the lipid study 1593 participants were evaluated for effects of fenofibrate on retinopathy.⁷⁸ At 4 years, the rate of progression of retinopathy, defined as a composite measure of three-step or greater progression on the ETDRS diabetic retinopathy severity scale or worsening requiring laser photocoagulation or vitrectomy, was significantly lower in the fenofibrate group (6.5%) compared with the placebo group (10.2%; adjusted odds ratio 0.60; 95% CI 0.42–0.87; P = 0.006).

The results of the FIELD and ACCORD studies are intriguing, particularly in the setting of some uncertainty about whether the benefit of fenofibrate on retinopathy is secondary to its effects on plasma lipid alterations or not given relatively modest effects on plasma HDL cholesterol and triglyceride levels among fenofibrate-treated participants in both trials.

A number of other agents have been evaluated for benefit in treatment of diabetic retinopathy, including ACE inhibitors, inhibitors of advanced glycation endproduct formation, growth hormone antagonists, antioxidants, and others. At the present time, the systemic strategies with a definite role in the management of diabetic retinopathy involve those interventions effective in controlling underlying hyperglycemia, hypertension, and dyslipidemia.