Defining Diabetic Retinopathy Severity

The fundus abnormalities seen in diabetic retinopathy can conceptually be split into three categories

– those findings resulting from leaking microvasculature (hemorrhages, lipid exudates, retinal edema); those findings resulting from structural damage to the microvasculature wall (microaneurysms); and those findings resulting from ischemia with a subsequent overproduction of vascular growth factors (cottonwool patches, intraretinal microvascular abnormalities [IRMA], preretinal neovascularization, fibrous proliferation, and vitreous hemorrhage). The severity of each of these findings can be classified and quantified based on the degree of retina involvement, e.g., the number of microaneurysms and hemorrhages in each quadrant or photographic field, the area of retina affected by neovascular tissue or IRMA, or the area of macula involved with retinal thickening. Classification schemes are a means to categorize the varying degrees of such findings to facilitate communication. As with all specialized languages, the value of these languages depends upon who is speaking to whom. The ultimate goal of classification schemes is to provide a system to improve patient care. This may be achieved initially by precise communication among researchers to define categories by which the natural history and subsequent response to intervention(s) can be identified. Once these research goals have been achieved these classification schemes should be understandable and reproducible by practicing clinicians. If not, then modification of these classification schemes is desirable to allow clinicians to communicate among

K. Wong (*)

University of South Florida, Sarasota, FL 34242, USA e-mail: iskeye@yahoo.com

themselves and with their patients the expected results of therapies demonstrated to be beneficial in clinical trials. One should realize that the goals of communication between researchers may be different from the goals of communication between clinicians and may likewise be different from the goals of communication between clinician and patient. Evolution of classification schemes is desirable dependent upon who is communicating with whom.

Historically, the initial classification schemes for diabetic retinopathy were developed to facilitate communication among researchers. In 1968 the US Public Health Service held a symposium at the Airlie House in Warrington, PA, to develop a classification system for diabetic retinopathy.1 The ‘‘Airlie House’’ classification system used five standard photographic fields (field 2 centered on the foveola; fields 4 and 5 just above and below the foveola respectively; fields 1 and 3 just nasal and temporal to the foveola respectively). Fundus lesions (hemorrhages, microaneurysms, lipid, neovascularization, etc.) were classified into one of three categories: absent, mild to moderate, or severe. However, with only three levels of severity graders using the Airlie House classification often found it difficult to reach consensus. This system therefore fell short in its ability to communicate with consistency between researchers. The Airlie House classification scheme did not define enough distinction between ‘‘mild to moderate’’ lesions and ‘‘severe’’ lesions so that most lesions were lumped into the ‘‘mild to moderate’’ category. In the ‘‘lumpers’’ vs. ‘‘splitters’’ distinction, there was not enough ‘‘splitting.’’

The Diabetic Retinopathy Study (DRS) subsequently created the modified Airlie House

Fig. 5.1 Seven standard fields of the modified Airlie House classification (shown for the right eye). Field 1 is centered on the optic disc, field 2 on the macula. Field 3 is temporal to the macula. Fields 4–7 are tangential to the horizontal lines passing through the upper and lower poles of the disc and to a vertical line passing through its center. Reproduced from ETDRS Research Group3

classification system with a greater degree of ‘‘splitting.’’ Two additional photographic fields were analyzed (field 6 superonasal and field 7 inferonasal) to the optic nerve head (see Fig. 5.1).

In addition, more gradations of severity for individual components were added.2 The primary question asked by the DRS was whether photocoagulation helped to prevent severe visual loss from proliferative diabetic retinopathy. Eligible patients had either proliferative diabetic retinopathy or severe nonproliferative diabeticretinopathy (defined asfollows)(see Table5.1).

One eye received photocoagulation and the other eye received observation only during the study period. The principal outcome measure was visual acuity of <5/200 at two consecutive 4-month intervals (defined as ‘‘severe visual loss’’). For the targeted study population, the modified Airlie House classification scheme allowed researchers to classify lesions consistently and to demonstrate the benefit of immediate scatter laser photocoagulation in patients with ‘‘high-risk’’ proliferative diabetic retinopathy. These ‘‘high-risk’’ characteristics were (1) neovascularization of the disc, (2) neovascularization of the disc > standard photograph 10A (see Fig. 5.2), (3) neovascularization elsewhere, and (4) vitreous hemorrhage (defined as hemorrhage more forward in the vitreous cavity than preretinal hemorrhage – standard photograph 13)3 (see Fig. 5.3).

However, for patients with lesion characteristics which approached but did not yet reach the definitions of ‘‘high-risk’’ proliferative diabetic retinopathy,

Table 5.1 DRS definition of severe nonproliferative diabetic retinopathy

Any three of the following characteristics: Soft exudates definitely present in 2 of

photographic fields 4–7

IRMA definitely present in 2 of photographic fields 4–7

Venous beading definitely present in 2 of photographic fields 4–7

Hemorrhages and/or microaneurysms severe

( std photo 2A) in 1 of photographic fields 4–7

Or

IRMA standard photo 8A in 2 of photographic fields 4–7 and definitely present in the remaining two fields

IRMA ¼ intraretinal microvascular abnormalities (dilated preexisting vessels or intraretinal new vessels).

Adapted with permission from ETDRS Report No. 12.4

Fig. 5.2 Standard photograph 10A, defining the lower boundary of Grade 3 (moderate) NVD. NVD covers approximately one-third the area of the standard disc (4.7 mm diameter on the original transparency) and about one-fifth the area of this unusually large disc (6.0-mm diameter on the original transparency). The unusual size of this disc was not taken into consideration when it was chosen as a standard, and is disregarded in grading (i.e., the area of the NVD in the photograph being graded is compared with that in the standard, without regard to the size of the discs). Reproduced with permission from the Early Treatment Diabetic Retinopathy Study Research Group

Fig. 5.3 Standard photograph 13, one of two standards used (interchangeably) to define the lower boundary of Grade 3 (moderate) preretinal hemorrhage. Reproduced with permission from the Early Treatment Diabetic Retinopathy Study Research Group

the DRS results were not helpful in determining whether immediate photocoagulation was a better treatment strategy as compared to deferral of scatter photocoagulation until ‘‘high-risk’’ characteristics developed. The Early Treatment Diabetic Retinopathy Study (ETDRS) therefore was designed to answer whether a strategy of earlier treatment with scatter photocoagulation was beneficial. As in the DRS, the primary end point was the rate of development of visual acuity <5/200 at two consecutive 4-month visits. However, since DRS results had already indicated that eyes with proliferative diabetic retinopathy and high-risk characteristics were at increased risk of ‘‘severe visual loss,’’ ETDRS data were also analyzed to determine rates of progression to proliferative diabetic retinopathy.

The ETDRS additionally asked whether laser photocoagulation of diabetic macular edema was helpful. The primary end point for this group of patients with macular edema was the rate of development of loss of 15 or more letters from baseline (defined as ‘‘moderate visual loss’’).

In attempting to answer these two study questions, the classification system of the modified Airlie House system was felt to have not enough detail in

the categories of moderate to severe NPDR and the ‘‘interim’’ ETDRS scale was designed. Besides the modified Airlie House lesions of microaneurysms, IRMA, NVE, and fibrous proliferations, the ETDRS asked that photographs be graded additionally for hard exudates, cottonwool patches, arteriovenous nicking, retinal elevation, and vitreous hemorrhage.3 Venous abnormalities were further subdivided into venous beading, venous narrowing, venous loops, venous sheathing, and perivascular exudate. The grading of each of these 17 factors in multiple stereoscopic fields with several steps of severity presents a rather complicated grading scheme. The ETDRS ‘‘interim’’ scale classification is shown in Table 5.2.

Table 5.2 ETDRS interim scale

Level Definition

30 A Venous loops definitely present

30 B Soft exudates, IRMA, or venous beading questionable

30 C Hemorrhages present but less than severe in 1–3 fields

30 D Definite hard exudates

30 E Hard exudates > moderate in 1 field

41 A Definite IRMA in 1–3 fields

41 B Definite soft exudates in 1–3 fields

45 A Definite soft exudates in 4–5 fields

45 B Definite IRMA in 4–5 fields or moderately severe IRMA in 1 field

45 C Definite venous beading in 1 field

45 D Severe hemorrhages in 1–3 fields

51 A Severe hemorrhages in 4–5 fields

Reproduced from ETDRS Report No. 12.4

Subsequent analysis of ETDRS data validated the prognostic power of this ‘‘interim’’ scale and this interim scale was thereby modified to a ‘‘final’’ scale4 (see Table 5.3).

One may notice that the numbers used are different between the interim and final scales (see Table 5.4).

In contrast to the numerical severity levels (of use principally to professional graders), the verbal definitions5 of the ETDRS classification scheme in general are more easily understood (e.g., macular edema, mild NPDR, high-risk PDR) (see Table 5.5).

However, some of the definitions (e.g., clinically significant macular edema, severe NPDR) remain so complex as to make clinical application problematic (as evidenced by the impetus behind the

Table 5.3 ETDRS final scale

Level Definition

10 Microaneurysms and other characteristics absent

20 Microaneurysms definite, other characteristics absent

35 A Definite venous loops in 1 field

35 B Questionable soft exudates, IRMA, or venous beading

35 C Hemorrhage present

35 D Definite hemorrhage in 1–5 fields

35 E Moderately severe hemorrhages in 1 field

35 F Definite soft exudates in 1 field

43 A Moderately severe hemorrhages in 4–5 fields or severe hemorrhages in 1 field

43 B Definite IRMA in 1–3 fields

47 A Both 43 A and 43 B definitions

47 B Definite IRMA in 4–5 fields

47 C Severe hemorrhages in 2–3 fields

47 D Definite venous beading in 1 field

53 A 2 level 47 definitions

53 B Severe hemorrhages in 4–5 fields

53 C Moderately severe IRMA in 1 field

53 D Definite venous beading in 2–3 fields

53 E 2 or more level 53 definitions

Reproduced from ETDRS Report No. 12.4

proposed International Clinical Diabetic Retinopathy and Diabetic Macular Edema Severity Scales).

Although complicated the ETDRS severity scale has become the de facto gold standard for grading diabetic retinopathy severity in clinical trials. In the recently described International Classification Scheme (see below) the Global Diabetic Retinopathy Project Group believed that ‘‘retina specialists were considered to be familiar with the ETDRS classification system and expected to continue using either it or their personal customized modifications.’’6 They are likely making this statement in reference to the verbal definitions as the numerical levels are of value only for communication among researchers or the extremely obsessive–compulsive clinician.

The additional value which the ETDRS classification provides in ‘‘splitting’’ categories can be found in natural history data. Based on a patient’s current ETDRS retinopathy level, one can predict the chance of developing high-risk proliferative retinopathy7 (see Table 5.6).

In retrospect, a careful inspection of this data demonstrates a deficiency of the ETDRS classification scheme, whereas moderate NPDR (level 47) carries an 8.6% 1-year risk of developing high-risk

Table 5.4 Comparison of ETDRS interim vs. final scales Interim scale Final scale Severity

30 A

30 B

30 C

30 D

30 E

Adapted from ETDRS Report No. 12.4

PDR, severe NPDR (level 53e) carries a 45% risk of developing high-risk PDR. Therefore, a jump in a single numerical level of severity (47–53) results in a big jump in risk (8.6–45%). Ideally, the risk of developing high-risk PDR will show a linear change in risk as one progresses along the severity scale.

The demonstration by the ETDRS that a patient’s 1- and 3-year risk of progressing to high-risk proliferative retinopathy increases dramatically when that patient’s severity level increases from moderate to severe NPDR has led to a concerted effort by the

American Academy of Ophthalmology to educate ophthalmologists to attempt differentiate severe NPDR from lesser levels. A simplified technique to identify severe NPDR is the 4:2:1 rule.8 The 4:2:1 rule classifies severe NPDR as eyes with hemorrhages in four quadrants > standard photograph 2A (see Fig. 5.4); or venous beading in two quadrants > standard photograph 6A (see Fig. 5.5); or IRMA > standard photograph 8A in one quadrant (see Fig. 5.6).

Fig. 5.4 Standard photograph 2A, intermediate standard for hemorrhages/microaneurysms. This is the minimum grade of hemorrhages and microaneurysms necessary for diagnosing severe nonproliferative diabetic retinopathy. Reproduced with permission from the Early Treatment Diabetic Retinopathy Study Research Group

Fig. 5.5 Standard photograph 6A, less severe of two standards for venous beading. Two main branches of the superior temporal vein show beading that is definite, but not severe. Reproduced with permission from the Early Treatment Diabetic Retinopathy Study Research Group

Fig. 5.6 Standard photograph 8A, less severe of two standards for grading soft exudates and IRMA. This is the minimum standard for grading IRMA sufficient to diagnose severe nonproliferative diabetic retinopathy. Reproduced with permission from the Early Treatment Diabetic Retinopathy Study Research Group

One should recognize, however, that the 4:2:1 description of severe NPDR describes a level of disease severity that is less than the DRS definition

of severe NPDR (see Table 5.1) and also less than the ETDRS definition of severe NPDR. The ETDRS definition of severe NPDR is characterized by any one of three clinical descriptions.9 In the first description, soft exudates, venous beading, and IRMA all have to be present in at least two photographic fields (quadrants). The 4:2:1 rule makes no mention of soft exudates and requires IRMA in only one quadrant. The second ETDRS description most closely approximates the 4:2:1 rule with hemorrhages in all four quadrants (but only greater than standard photograph 2A in one of them) and also with two of the following three lesions (soft exudates, venous beading, or IRMA) present in two fields. The 4:2:1 rule requires only hemorrhages in four quadrants or venous beading in two quadrants or IRMA in one quadrant. In contrast, the second ETDRS description requires hemorrhages in four quadrants and two of the three lesions in two quadrants. The third ETDRS description of severe NPDR stipulates IRMA in four quadrants with at least two of the quadrants having severity standard photograph 8A. The 4:2:1 rule therefore describes a level of diabetic retinopathy severity which is probably intermediate between the ETDRS definition of moderate and severe NPDR. Although probably reasonable, one must realize that the 4:2:1 rule is an extrapolation but not identical to ‘‘severe NPDR’’ as defined in the ETDRS. The scant evidence available suggests that clinicians are poor at correctly discerning severe NPDR despite the intensive educational efforts toward the goal of accurate diagnosis.10

The second question asked by the ETDRS was whether laser photocoagulation of diabetic macular edema was helpful in reducing the rate of moderate visual loss. Enrolled patients could have 20/20 visual acuity in which vision gain was not likely. The ETDRS was conducted in an era prior to optical coherence tomography and defined macular edema and clinically significant macular edema based on examination of stereo fundus photographs. When macular edema was defined as retinal thickening within 1 disc diameter of the center of the macula, protocol laser treatment was not shown to reduce the risk of moderate visual loss within the initial 2 years following treatment (see Fig. 5.7).

When macular edema was more strictly defined to be closer to the fovea (clinically significant

Fig. 5.7 Comparison of percentages of eyes that experienced visual loss of 15 or more letters (equivalent to at least doubling of the initial visual angle) in eyes classified by severity of macular edema in baseline fundus photographs and assigned to either immediate photocoagulation for macular edema (broken line) or deferral of photocoagulation unless high-risk characteristics develop (solid line). Reproduced with permission from ETDRS Research Group11

Macular Edema–Not Clinically Significant

Macular Edema–Clinically Significant without Center Involvement

Macular Edema–Clinically Significant with Center Involvement

Months of Follow-up

macular edema), the rate of moderate visual loss was decreased by 50% (24% of untreated eyes compared to about 12% of treated eyes at 3 years).11 Based on this data, the ETDRS recommended that ‘‘eyes with CSME with center involvement should be considered for immediate laser treatment.’’ Since the ETDRS demonstrated that subclassifying diabetic macular edema severity as ‘‘clinically significant’’ or ‘‘not clinically significant’’ resulted in a

differential response to laser therapy, it is important for the clinician to recognize this distinction. However, like ETDRS severity levels, the definition of CSME is a research definition initially designed for use by professional graders reviewing stereoscopic fundus photographs. In clinical practice the ability to distinguish ‘‘clinically significant macular edema’’ from ‘‘macular edema’’ may be difficult to reproduce. Furthermore, following laser therapy it

may also be difficult to clinically determine improvement or resolution of clinically significant macular edema. In the DRS and ETDRS there was a lack of concordance between the professional graders and clinicians in determining macular edema with the two groups agreeing only 55% of the time on the diagnosis after taking into account the agreement due to chance.12 This subjective variability in diagnosing CSME may have practical implications in reproducing the beneficial results of laser therapy demonstrated in the ETDRS.

The DRS and the ETDRS demonstrated two thresholds of diabetic retinopathy severity for recommending laser therapy to help prevent vision loss: (1) proliferative diabetic retinopathy with high-risk characteristics and (2) clinically significant macular edema. The Global Diabetic Retinopathy Project Group has attempted to simplify the classification scheme to facilitate early detection of these threshold levels.6 They have proposed an International Classification Scale to help communication between comprehensive ophthalmologists and primary care physicians to hopefully allow these threshold levels to be detected earlier.

Rather than utilizing 12 levels of severity (level 10–85) with 12 corresponding descriptive terms (mild NPDR, moderate NPDR, moderately severe NPDR, mild PDR, high-risk PDR, etc.), this international panel of experts proposed reducing the number of levels. The distinguishing feature of this classification scheme is an attempt to distinguish patients with severe NPDR or worse from patients with less than severe NPDR. The rationale proposed lies in using the 4:2:1 rule as a simplified method to identify severe NPDR. As previously stated these definitions of severe NPDR approximate each other but are not equivalent.

In reviewing the ETDRS data the Global Diabetic Retinopathy Project Group realized that the presence of IRMA and venous beading was more predictive of the risk of progressing to high-risk PDR, whereas retinal hemorrhages and microaneurysms alone were poorly predictive. For this reason their classification scheme deemphasizes the recognition of hemorrhages and microaneurysms (see Table 5.7).

A separate severity scale was proposed for diabetic macular edema (see Table 5.8).

The global issues prompting a separate severity scale for diabetic macular edema were purported to

Table 5.7 Proposed International Clinical Diabetic Retinopathy Severity Scale

Reproduced with permission from Wilkinson et al.6

be the use of direct ophthalmoscopy by primary care physicians to examine the retina. The ETDRS standard of diagnosing ‘‘clinically significant macular edema’’ as the indication to benefit from laser treatment would therefore not be applicable to practitioners untrained in the use of slit lamp biomicroscopy. This proposed International Classification Scheme merits consideration in facilitating communication but has not been validated with regard to its ability to overdiagnose or underdiagnose clinically significant macular edema.

The advent of optical coherence tomogram technology to quantitate retinal thickness along with its universal acceptance has further complicated the issue of classifications schemes to classify diabetic macular edema. CSME as the indication for benefiting from laser therapy is either present or absent. The quantitative analyses provided by OCT allow the possibility of further splitting diabetic macular edema into subclassifications analogous to the progressive splitting of classifications between the Airlie House classification, modified Airlie House classification, and ETDRS classification schemes. It is not unreasonable that varying degrees of diabetic macular edema may respond differently to varying approaches of laser combined with pharmacologic therapy. However, in the absence of studies demonstrating such benefit, the ETDRS currently provides the most complete data set and ‘‘clinically significant macular edema’’ (either present or absent) remains the gold standard classification scheme.

The varying definitions of diabetic ‘‘macular edema’’ and ‘‘clinically significant macular edema’’ lend itself to difficulties in communication among clinicians. Although the ETDRS provided distinctly different definitions to ‘‘diabetic macular edema’’ and ‘‘clinically significant macular edema’’ some clinicians merge the two definitions into an undefined term ‘‘clinically significant diabetic macular edema.’’ Further vagaries in undefined terminology are ‘‘diffuse’’ and ‘‘focal’’ diabetic macular edema. The varying definitions for these terms are based on using different modalities (clinical examination, fundus photography, optical coherence tomography, and fluorescein angiography) to describe them. Although commonly used, these terms currently are poorly defined without evidence that additional subclassification of diabetic macular edema into ‘‘focal’’ and ‘‘diffuse’’ changes clinical outcome or treatment decisions.13

A DME severity scale using professional grading of the ETDRS stereo photographs has recently been proposed.14 This scale categorized edema based on the grader-determined area of edema within the macula and also the degree of retinal thickening at the foveal center relative to a reference thickness. The objective of this severity scale was to determine whether splitting macular edema into more levels than ‘‘present’’ or ‘‘absent’’ might

be beneficial. Not surprisingly this study did find an association. Greater severity of both diabetic macular edema and edema that had been present for a longer duration was associated with vision loss. Given the ability of OCT to accurately and reproducibly quantify macular thickness, the results of this DME severity scale study imply that accurate subclassification of diabetic macular edema may be helpful.

The beneficial role of fluorescein angiography in classifying diabetic retinopathy severity is unclear. The ETDRS did perform and evaluate fluorescein angiograms in a systematic fashion. It graded the size of the foveal avascular zone (FAZ), FAZ outline, capillary loss, capillary dilatation, focal narrowing, vessel wall staining, severity of leakage, source of leakage, cystoid spaces, and RPE defects.15 These characteristics were graded typically on a 3- to 5-step scale. Graders varied in their assessment of the degree of severity. With regard to grading severity of features of ischemia, there was complete agreement in grading FAZ size in 74%; complete agreement in grading FAZ outline in 47%; and complete agreement in grading capillary loss in 47%. With regard to fluorescein angiographic signs of leakage there was complete agreement in grading severity of leakage in 63%; complete agreement in grading source of leakage in 60%; and complete agreement in grading cystoid spaces in 92%. It seems that if professional graders disagreed on classifying the severity level of fluorescein angiographic findings 40–50% of the time, consistent and reproducible grading of these lesions by clinicians may be a challenge. Reproducibility of grading fluorescein angiograms in diabetic retinopathy has not improved in the 20 years since the ETDRS, and although their use was employed in the first clinical trial of the Diabetic Retinopathy Clinical Research Network (DRCR.net), these gradings have since been abandoned as insufficiently reproducible and a poor allocation of resources in Network studies.

Besides being difficult to reproduce the gradings, it is not certain that the additional information gained from fluorescein angiographic classification helps in predicting risk of progression or for recommending therapy. In the DRS, the indication for benefiting from laser photocoagulation was the presence of high-risk characteristics