Grading Protocol for CMV, FPRC

Protocol for Grading Cytomegalovirus (CMV) Retinitis from Fundus Photographs

The Studies of the Ocular Complications of AIDS (SOCA) Research Group developed a protocol for assessment of cytomegalovirus (CMV) retinitis and other abnormalities from fundus photographs. Using wide-angle cameras, standardized fundus photography is performed to document the post-equatorial fundus in a set of color slides. For each eye, the array of nine photographic fields (a non-simultaneous stereoscopic slide pair of the posterior pole of the retina surrounded by eight slightly overlapping non-stereoscopic slides) is mounted in a plastic sheet and examined with a 5X-magnifying viewer. The principal abnormalities evaluated are the white retinal opacities typical of active CMV retinitis and the subsequent retinal atrophy (usually accompanied by pigment abnormalities). Chief goals of the grading protocol are: (1) to evaluate features of CMV retinitis, such as presence of lesions, their location, and stage of border activity; (2) to assess progression (either movement of existing lesions or appearance of new ones) of retinitis lesions between visits; and (3) to estimate the area of retina involved by retinitis via imposition of a grid, so that change in area can be calculated over time. The set of standard photographs used to classify CMV retinitis and other abnormalities is available from the University of Wisconsin Fundus Photograph Reading Center (610 Walnut St., Room 450, Madison, WI 53705) at cost of reproduction.

Protocol for Grading Cytomegalovirus (CMV) Retinitis from Fundus Photographs

This chapter begins with the standard protocol for 9-field wide-angle photography (summarized in section 2, and reproduced in its entirety as Appendix A), states the assumptions made in measuring distances and areas (section 3), and describes the division of the retina into conventional zones (section 4). General procedures for grading, such as how to view the photographs on the light box, are described in section 5. Section 6 introduces the use of the grading forms (reproduced as exhibits 1 and 2), and stipulates rules used in determining grades. Next, three key sections of the protocol describe the grading forms in detail - section 7 for the primary form regarding status of retinitis and progression from baseline, section 8 for the primary form regarding other abnormalities (including some appearances related to retinitis), and section 9 for the supplementary form recording the estimate of retinitis area. These sections describe the abnormalities, present the questions contained on the grading forms, list the codes for answering them, and discuss how the appropriate answers are determined. To provide background for grading CMV retinitis, section 7 opens with discussions of the various appearances of retinitis lesions and of abnormalities that might be confused with retinitis. Finally, the chapter documents the algorithms for summarizing estimates of retinitis area into analysis variables (section 10) and describes the editing of completed gradings (section 11). The detailed rules used to check gradings for completeness and internal consistency are listed in Appendix B.

A set of standard photographs used to classify CMV retinitis and other abnormalities is available from the University of Wisconsin Fundus Photograph Reading Center (610 Walnut St., Room 450; Madison, WI 53705) at cost of reproduction. (A more extensive set of example photographs is used internally to train graders in the classification.)

This grading protocol assumes that a wide-angle fundus camera, preferably the Canon 60^o but alternatively the Topcon 50^o or similar camera, has been used to take a standardized set of retinal photographs according to the SOCA manual of operations (the photographic protocol is attached as Appendix A). The complete set consists of a stereoscopic pair centered midway between the center of the macula and the temporal edge of the disc (field 1-2), and eight non-stereoscopic photographs arrayed around the central photograph and overlapping slightly with it superiorly (field 9), supero-nasally (field 6), nasally (field 8), infero-nasally (field 7), inferiorly (field 10), infero-temporally (field 5), temporally (field 3), and supero-temporally, (field 4). A fundus reflex photograph is also included to show the degree of pupillary dilation and to document media opacities that might blur the retinal image, unless its presence might unmask the grader to the patient's clinical trial treatment assignment (e.g., by inadvertently depicting an intraocular ganciclovir implant). Since the Canon and Topcon photographs have different magnifications, it is necessary to differentiate between them. Both cameras produce notches on the edge of the photographic image at about 2 o'clock - the Canon a semicircular notch and the Topcon a square notch.

The linear unit utilized in this grading protocol is the disc diameter (DD), represented here as 1,500 µm in accordance with convention, although 1,800 to 1,900 may be a better estimate. ^,

The area unit is the disc area (DA), defined as the area of a circle with a diameter of 1 DD.

The approach used to establish the correspondence between distance on the retina and distance on the film was adapted from the Diabetic Retinopathy Study/Early Treatment Diabetic Retinopathy Study. In those studies, 4.7 mm on the film was selected for the diameter of the average disc based on measurements from Zeiss 30^o fundus photographs of the horizontal and vertical disc diameters of 400 eyes. The relationship between the two wide-angle cameras utilized in SOCA studies was established pragmatically by photographing eyes of an emmetropic individual with the Zeiss camera and with both the Canon 60^o and the Topcon 50^ocameras. The distance between various pairs of retinal features was measured across the photographic fields in each type of photograph, and the ratios calculated for difference in magnification between the Zeiss camera and the two wide-angle cameras. The following distances on the film are assumed to correspond to 1.0 DD on the retina: for the Canon 60^o camera, 3.0 mm; and for the Topcon 50^o camera, 3.4 mm.

For measurements performed directly upon the slide transparencies, the grader uses either of two devices: (a) a measuring magnifier or "reticle" (7X magnification, Bausch and Lomb catalog number 81-34-35) with a millimeter scale, or (b) a selection of graduated standard circles printed on a transparent overlay (reproduced as Figure 1), with diameters of 63, 125, 250, 375, 500, 750, 1,000, and 1,500 µm. There are two versions of the standard circles, appropriately scaled for the Canon 60^o and Topcon 50^o cameras. When recording distances, the grader rounds the result to the nearest tenth of a DD.

To indicate the location of CMV retinitis lesions, the retina is conventionally divided into three zones as illustrated in figure 2: zone 1 includes the posterior pole, zone 2 extends from zone 1 to a circle through the ampullae of the vortex veins, and zone 3 extends from zone 2 to the ora serrata. For application in grading, the boundaries between these zones are imprinted on transparency stock, hereafter termed the SOCA grid (reproduced as figure 3), which can be placed over or under the retinal photographs to determine the zone boundaries. Two variants of this grid are available, sized for the Canon 60^o camera and Topcon 50^o camera respectively. (the Topcon grid is larger than the Canon grid; both are labeled as to camera type.)

Zone 1 includes all of the retinal area within 2 DD of the center of the macula or within 1 DD of the margin of the optic disc. Because of anatomic variations and differences in refractive error, the distance on the film between the center of the macula and the temporal disc margin varies somewhat between patients. For the SOCA grid, a disc-to-macula distance of 2.00 DD (3,000 µm) was chosen. When the grader encounters an eye with a different disc-to-macula distance, the grid must be repositioned slightly to superimpose it on either the disc or the macula as examination shifts from one region to the other.

To allow more precise description of the location and extent of lesions, the SOCA grid further divides zone 1 (see Figure 3). Two "safety zones" are defined, one surrounding the disc and one the central part of the fovea that typically has no retinal capillaries and is referred to as the foveal avascular zone (FAZ). The term safety zone was adopted to reflect the provision in some CMV retinitis treatment protocols that persistent or recurrent retinitis in these zones constitutes an imminent threat to vision, and may trigger a change in treatment without waiting for border movement of 750 µm (see section 7.8). The macular safety zone includes the area within 1,000 µm of the center point - assuming that the average FAZ is about 500 µm in diameter, this provides a margin around the FAZ of 750 µm. The disc safety zone includes the area of retina within 750 µm of the disc margin.

For estimation of the area of retina involved by CMV retinitis, the portion of zone 1 outside the safety zones is further divided into subfields. The area surrounding the macular safety zone (out to 3,000 µm from the center of the macula, i.e., the zone 1 boundary) is divided into four subfields: superior, nasal, inferior, and temporal. The area surrounding the disc safety zone (out to 1,500 µm from the disc margin, i.e., the zone 1 boundary), excluding the area which falls within the macular nasal subfield, is designated as the "disc other" subfield (see Figure 3).

The subfields of zone 1 have the following areas, as established via planimetry on an image processor:

Subfield	Area in DA's
Disc Safety Zone	3.0
Disc Outer Subfield	3.5
Macular Safety Zone	1.8
Superior Subfield	3.6
Nasal Subfield	2.9
Inferior Subfield	3.6
Temporal Subfield	3.6

Zone 2 includes all of the retina anterior to zone 1 and posterior to a circle intersecting the ampullae of the vortex veins. Because of variations in anatomy and refractive error the location of the zone 2/3 boundary varies somewhat among patients. This grading protocol assumes the outer boundary of zone 2 to be a circle of 9 DD radius centered midway between the center of the macula and the temporal margin of the disc. (This locus in the papillomacular bundle is the central point around which the photographic protocol and the grid construction are organized.) Given this definition, the anterior boundary of zone 2 is assumed to be located approximately 1 DD posterior to the equator.

The SOCA grid (Figure 3) can be applied to each peripheral photograph to approximate the boundary between zones 2 and 3. The grader first positions the zone 1 portion of the grid to correspond with the disc and macula in field 1-2. After examining the relationship between retinal landmarks in the part of field 1-2 overlapping the peripheral field of interest, the grader adjusts the peripheral field over the grid so that the relationship of the retinal landmarks to the grid is maintained.

The part of the grid applicable to zone 2 was constructed to create equal-area blocks that could be counted to estimate the total area of retinal involved. Given the circular shape of zone 2, a radially symmetrical grid seemed most appropriate. However, since zone 1 is irregularly shaped (i.e., not circular) and zone 2 surrounds zone 1, it was necessary to divide zone 2 into a small irregular portion and a large regular portion. A circle of 3 DD radius centered on the central point between the disc and the macula separates the regular from the irregular portion of zone 2.

The regular part of the zone 2 grid consists of 120 blocks formed by the intersections of meridians and concentric circles, each containing an equivalent area of retina. There are 24 meridians emanating from the central photographic point, radially oriented so that the wedges they create are centered on each hour and half-hour. There are six concentric circles centered upon the photographic central point, with radii of 3.00 DD, 4.84 DD, 6.15 DD, 7.23 DD, 8.16 DD, and 9.00 DD. As a result of this spacing, the blocks formed by the meridians and circles each have an area of 2.22 DA. The grid is structured so that each peripheral photographic field has a correspondingly numbered sector of the grid, consisting of three wedges with five blocks in each wedge. These sectors are divided from each other by broader meridian lines. If the photographic field is properly placed, the corresponding sector of the grid falls completely within the photographic field.

The irregular position of zone 2 immediately surrounding zone 1 is divided into six subfields by extending the major meridian lines for the superior and inferior peripheral fields (fields 9 and 10) to the zone 1 boundary. These subfields have an irregular shape similar to that left over after cookies have been stamped out of a sheet of dough (zone 1 and the regular part of zone 2 correspond to the cookies), and are referred to informally as the "cookie cutter" part of zone 2. The irregular subfields of zone 2 are designated by whether they are superior (S) or inferior (I) and numbered according to the peripheral photographic field they adjoin. According to planimetry performed on an image processor, these subfields have the following areas:

Subfield	Area in DA's
Superior temporal (S4)	1.6
Superior (S9)	3.1
Superior nasal (S6)	1.6
Inferior nasal (I7)	3.2
Inferior (I10)	2.9
Inferior temporal (I5)	1.1

Zone 3 includes all of the retinal area anterior to the zone 2 boundary, i.e., between it and the ora serrata. Only a small portion of zone 3 can be readily photographed, and there can be substantial inconsistency in the area captured from visit to visit. Zone 3 can be divided into eight sectors by mentally extending the major meridians of the zone 2 grid.

In order to become thoroughly familiar with the CMV retinitis and other abnormalities documented by the photographs, the grader begins by examining them upon a light box (equipped with"daylight" fluorescent tubes) using a Donaldson stereoscopic viewer (5X magnification, manufactured by George Davco, Holbrook, MA). Decisions about the presence and extent of lesions are based upon careful study of the appearances seen in the photographs. Once the grader has made basic decisions about the status of CMV retinitis and other abnormalities, the templates for the boundaries of retinal zones and the measuring devices are utilized as necessary.

To compare follow-up visits with the relevant baseline, the grader remounts the individual frames in the plastic mounting sheet if necessary so that every slide from the follow-up visit can be juxtaposed with the corresponding slide from the baseline visit. Such juxtaposition allows the grader to use the Donaldson stereo viewer to look at the baseline with one eye and the follow-up with the other. By alternating the view from one side to the other, the grader is able to compare individual features and their location closely. This stratagem takes advantage of short-term visual memory to enhance the ability to detect subtle differences.

With minor remounting of the individual slides (diagramed in figure 4), it is possible to obtain clear columns in both the baseline and follow-up sets so that corresponding slides can be juxtaposed without removing any from their original mounting sheets. To accomplish this, one member of the stereoscopic pair of field 1-2 must be transferred to the bottom of the sheet temporarily - for baseline sets, the right member is shifted, and for follow-up sets, the left member. The central column of photographs (including fields 9 and 10) is shifted if necessary to the left for baseline photographs and to the right for follow-up photographs. Once these configurations are attained, one mounting sheet can be placed on top the other and manipulated until the desired juxtaposition of photographs is attained.

In some cases, the grader might find it helpful to relate the location of a lesion boundary to a fixed retinal landmark, such as a vessel fork. This approach is sometimes necessary when aberrant field definition makes it difficult to fuse two fields being compared across visits, or when photographs being compared were taken at different magnifications.

The protocol employs two grading forms, a primary form for recording the status of CMV retinitis and other features (exhibit 1) and a supplementary form for estimating the proportion of retina involved by retinitis (exhibit 2).

The main grading form, titled "Assessment of Ocular Photographs," is used to describe the current status of CMV retinitis and other abnormalities, and to record the comparison of follow-up visits to the baseline and to the previous visit regarding the expansion of existing retinitis lesions and development of new retinitis lesions. Separate but identical forms are used for the right and left eyes.

An addendum to the primary grading form, the form titled "Estimate of Retinal Area Involved by CMVR" is known informally as the "gridwork" form. Gridwork forms are used to record the area involved by CMV retinitis in regions of the SOCA grid (zone 1, and the irregular and regular parts of zone 2) and the presence of retinitis in zone 3. To help orient the grader, an enlarged copy of the grid is printed on the grading form. Consequently, this form has variants for the right and left eyes (labeled "RE" and "LE"), which are mirror images of each other. The gridwork forms for right and left eyes are completed at baseline plus selected follow-up visits, depending upon the visit schedule of the particular trial.

Both forms include questions referred to as "gatekeepers" or "stop conditions," which control whether the remainder of a section or form must be completed. Several examples illustrate how such questions operate. (1) On the primary assessment form the absence of any retinal photographs is a stop condition. (2) If photographs are present but CMV retinitis is absent, the grader is not required to complete the sections regarding its characteristics and whether it has progressed. (3) On the gridwork form, absence of retinitis in a region of the grid (e.g., in all of zone 1) releases the grader from having to record answers for each subfield of zone 1.

The grade "questionable" is chosen if the grader thinks that the lesion or condition under consideration is probably present but is not sure that all other experienced graders would agree. Expressed another way, if the grader is at least 50% but less than 90% certain that the lesion is present then "questionable" is marked. If the appearance is more definite than this, then a grade denoting definite presence is selected; if the appearance is less definite than this, then the "absent" grade must be chosen.

If the grader is unable to answer a question with reasonable assurance, due to inadequate photographic quality or to lack of comparability between serial photographs, "cannot grade" (code 8) is marked. The threshold at which the grader decides to use "cannot grade" varies with the subtlety of the lesion being assessed. CMV retinitis lesions themselves (at least in the fully active stage or those of substantial size) tend to be prominent enough that they can be seen relatively well in spite of moderate photographic problems. However, more subtle features such as the degree of activity of treated CMV retinitis may not be gradable in compromised photographs.

7.0 Evaluation of CMV Retinitis Status and Progression, Primary Grading Form

The primary grading form, the "Assessment of Ocular Photographs," is used to evaluate the status of CMV retinitis and other abnormalities, and to assess whether retinitis has progressed.

Description of CMV retinitis status includes various features: lesion presence, location of lesions by retinal zone, stage of activity (degree of opacification) of lesion margins, proportion of lesion margins which are active, proximity of lesions to the disc and the center of macula, involvement of the disc and the foveal avascular zone by retinitis, characterization of the active margin (baseline only), and presence of retinal detachment. Evaluation of progression is based upon comparison of follow-up photographs with those taken at baseline and the previous visit to detect movement of the lesion edge or occurrence of a new lesion.

The grader begins by recording identifying information, including clinical center, patient ID number, namecode, visit, and date of photographs. Following are the items contained on the form.

The grader completes two items recording the status of photographs, the first pertaining to their presence and the second to the type of camera used to take them.

Acceptable values are code 0 for no photographs, code 1 for lens photograph only (no retinal photographs), and code 2 for retinal photographs present. The grader is not required to complete the remainder of the grading form if there are no retinal photographs. If a lens photograph is present, the grader is allowed to complete the comment bloc to describe its appearance. The lens photo may provide information explaining the absence of retinal photographs, such as media opacities (e.g., vitreous hemorrhage) or a small bound-down pupil.

The type of fundus camera used to take the photographs is recorded: code 0 for the Canon 60^ocamera, code 1 for the Topcon 50^o, code 2 for the Zeiss 30^o, and code 3 for any other camera. Type of fundus camera can usually be determined from the characteristics of the image produced on the film. Both of the most common wide-angle cameras, the Canon 60^o and the Topcon 50^o, produce a round image cropped across the top and bottom, the Canon more elongated horizontally than the Topcon. In contrast, the Zeiss camera produces an uncropped round image. Both the Canon and Topcon cameras produce orientation notches on the edge of the photographic image at about 2 o'clock - the Canon a semicircular notch and the Topcon a square notch. (Although not authorized for SOCA studies, photographs may sometimes be seen from a Kowa camera, which produces a triangular notch on the photograph edge at 12 o'clock.) In contrast, the Zeiss camera produces an image without any orientation notch.

For grading purposes, the most critical differences among these cameras are degree of field and magnification. Obviously cameras with a smaller degree of field document a lesser extent of retina, which in the case of the Zeiss may be insufficient for confident assessment of CMV retinitis status. It is imperative that the grader consider the degree of magnification when selecting grids and measurement devices or conversion factors. The scaling factors used to construct grids and to convert measurements are based on the assumption that 1.0 DD on the retina equals the following distances on film: 3.0 mm for the Canon 60^o, 3.4 mm for the Topcon 50^o, and 4.7 mm for the Zeiss 30^o. If the grader is uncertain about the degree of magnification, measurement of the disc with the measuring magnifier or trial with the 1,500 µm circle from the two sets of measurement circles (60^oor 50^o) provide an approximate gauge of scale. Since some wide-angle cameras are variable, i.e., can also be used to take photographs of narrower field and higher magnification (both the Canon at its 40^o setting and the Topcon at its 35^o setting approximate the Zeiss 30^o magnification), the grader needs to be vigilant for wide-angle photographs taken at the wrong setting.

As background for the grading form questions regarding presence and location of CMV retinitis (presented in section 7.2.4), sections 7.2.1 through 7.2.3 discuss the various appearances of retinitis lesions, discrimination of retinitis lesions from similar appearances, and conventions adopted in classification of lesions.

Typical CMV retinitis first becomes visible ophthalmoscopically as an area of white opacification of the retina, sometimes accompanied by retinal hemorrhages. Although early retinitis lesions are often observed straddling blood vessels, they may appear anywhere in the retina. Small isolated foci of retinitis may be difficult to distinguish from cotton-wool spots, but large patches have the appearance characteristic of viral retinitis, i.e., opaque, white, full-thickness retinal necrosis (without the subtle striations typical of cotton wool spots) that usually obscures the underlying retinal pigment epithelium (RPE). Small white spots called "satellites" (typically 63 to125µm in diameter, the latter dimension being about the caliber of a large vein at the edge of an average optic disc - assumed here to be 1,500 µm in diameter) are often seen in otherwise normal-appearing retina adjacent to larger lesions. The features of CMV retinitis have been described previously by others (DeVenecia et al., Holland et al.^,, Freeman et al., Palestine et al., Rosenberg et al., Jabs et al. ), as summarized in a review by Holland and coworkers.

Untreated the infection gradually advances, evidenced by the spread of opacification into adjacent retina. Concurrently the initially infected retina, now the interior of the lesion, atrophies to become thin and transparent. Once again visible, the underlying RPE usually develops some degree of pigment mottling over time. In some areas the RPE may be obscured by residual deposits resembling lipid, or by permanent sheets or strands of white (fibrous/glial) tissue.

Example photographs 1A-1D [NEI 01-012] show the natural history course of a CMV retinitis lesion located nasal to the disc in a left eye. This patient was assigned to deferral of treatment until retinitis progression in an early clinical trial investigating an anti-CMV drug. Photographs were taken shortly after diagnosis at study baseline (1A) and weekly thereafter until three weeks (1B through 1D), at which point the patient began treatment.

Assessment of the degree of active inflammation and of progression of retinitis into previously uninvolved areas is critical to the clinical management of patients and to the evaluation of treatment efficacy. Thus, particular attention is focused upon the lesion "margin" - the zone of transition from infected to normal-appearing retina. In active (particularly untreated) lesions, the margin usually includes two components: a "solid" part within which the majority of retina is opacified, and an outermost "satellite" part (not always observed) in which small foci of CMV infection are interspersed with apparently uninvolved retina. By convention, the margin is considered to be a band approximately 1,000 microns wide that begins at the outermost satellites and extends toward the interior of the lesion.

At presentation (i.e., prior to treatment), the typical CMV retinitis lesion margin has a solid portion that is opaquely white, with such density that the pattern of the underlying RPE cannot be seen. (For examples, examine the lesion margins shown in standard photographs 7A and 7B.) The retinal vessels coursing through this area are often difficult to see because they are obscured, narrowed, or obliterated.

Some investigators have described a spectrum of CMV retinitis appearance ranging from "fulminant/edematous" to "indolent/granular."¹¹ Toward the fulminant end of the spectrum, the active margin is densely opacified, often with hemorrhages and accompanied by inflammatory perivascular cuffing. In such cases this part of the lesion margin is usually relatively broad, extending 1 disc diameter (DD) or more into the lesion and often accompanied by a substantial band of satellites. For illustrations of fulminant margins, see example photographs 8 and 35A. Toward the indolent end of the spectrum, the solid part of the margin is more granular and less opaque, usually without hemorrhages and unaccompanied by perivascular cuffing. In such cases this part of the margin is often relatively shallow, at points almost disappearing, and usually lacks a broad band of satellites (although the granular texture of the solid margin itself somewhat resembles satellites). For an illustration of an indolent margin, see example photograph 9. In many cases the appearance of the active margin falls between the two ends of the fulminent/indolent spectrum, and these lesions cannot readily be classed as either.

When CMV retinitis is reactivating (i.e., treatment is not suppressing CMV sufficiently), the solid portion of the margin usually does not regain the dense, opaquely white appearance observed upon presentation. Rather, the margin characteristic of lesion reactivation often features localized areas of subtle opacification, even though other areas (perhaps adjacent to the active areas) are atrophic or show partially-resolved opacity. Example photograph 35E shows reactivation of a lesion margin despite treatment (compare with example photograph 35A, which shows the higher degree of activity prior to treatment, and with example photograph 35D, before reactivation).

Upon presentation (and when activity is persistent or recurrent) many CMV retinitis lesions display satellites, small grayish-white dots, typically between 63 and 125?m in diameter (see example photograph 1). These spots have fuzzy edges and may appear faint, with only the larger ones achieving notably greater opacity at their centers. Often satellites occur in clusters, with the density of these clusters decreasing farther from the solid edge. Most lesions appear to have at least a few areas with satellites in their margins, although in some cases no satellites are evident (i.e., there appears to be no transitional zone between solid edge and entirely uninvolved retina). Typically, the satellite band is restricted to about 1/2 DD from the solid edge, although there are cases with satellite margins as wide as 1 DD or more.

When the active stage of CMV retinitis has run its course in a given region or has been suppressed by treatment, the whitish opacification clears and the retina becomes atrophic. (Some lesions may display areas of atrophy within their active margins even when untreated, particular those toward the indolent end of the spectrum.) Usually atrophic retina becomes transparent, revealing once again the RPE (often mottled by clumps of hyperpigmentation) and the underlying pattern of the choroidal vasculature (most clearly in individuals with lighter pigmentation). Vessels coursing through the atrophic part of the lesion often become narrowed and irregular in contour, and eventually may disappear or persist as white threads. Standard photograph 5C illustrates the atrophy of the interior of a retinitis lesion - examine the inferior part of the photograph below the active lesion margin.

Portions of atrophic retina (often well back from the active margin if there is one) may remain opaque, either temporarily or permanently, after other parts nearer the edge become clear, particularly if the lesion has been fulminant. In some cases this opacification is composed of dense, confluent, homogenous plaques (appearing variously "dry," "hard" or "lipid-like"). Over time such residual opacities tend to decrease in size and eventually disappear. Example photograph 10A shows this appearance in a resolving lesion margin; six weeks later most of this material has been absorbed, as revealed by photograph 10B. In other cases more delicate, often fibrillar opacification persists or arises in the interior of the lesion, presumably consisting of fibrous or glial tissue. If the lesion has an active margin for comparison, the color of this fibrosis within the lesion is usually more bluish-white than the yellowish-white of the border. Example photograph 11 shows this appearance within a lesion with resolved borders.

There may be an intermediate zone with a translucent or milky appearance separating the active margin from the atrophic core of the lesion. It is neither so dense that the underlying RPE is completely obscured nor so transparent that the RPE can be seen with complete clarity. After suppression of retinitis by treatment, the entire margin may be seen to pass through an intermediate stage prior to becoming atrophic. For example, examine the lesion margin shown in standard photograph 4. Superior to the section with subtle but fairly definite opacification (the portion located straight nasally from the disc), the margin has a faint milky appearance - the arteriole exiting the field at 11:30 courses through this area.

Example photograph 15 presents the resolution over a six-week period of an active CMV retinitis margin, located in Field 8 (the area nasal to the disc in this right eye). Photograph 15A shows the untreated lesion at diagnosis, with definite opacification of the margin at its juncture with normal-appearing retina. In photograph 15B, taken two weeks after initiation of treatment, previously clear areas of retina within the margin have become opacified (the margin has ?filled in?), but there has been no increase in degree of opacification and no actual expansion of the lesion. In photograph 15C, taken at four weeks, the opacification of the lesion margin has mostly cleared, leaving a line of hyper pigmentation at its juncture with normal-appearing retina. In photograph 15D, taken at six weeks, the opacification of the lesion margin has cleared completely except for a persistent area of faint whitening (thought to be fibrous tissue, and not representative of activity) in the temporal part of the field.

[Describe Exp 36A-D as an example of persistent fibrosis within a resolved lesion.]

Two common appearances are sufficiently similar to CMV retinitis that they may be confused with it: cotton-wool spots (soft exudates) and vascular cuffing associated with retinitis.

Cotton wool spots are sometimes termed soft exudates, although they are not composed of exudative material. Since cotton wool spots appear as white or grayish-white spots surrounded by normal retina, they may look similar to small active CMV retinitis lesions. However, cotton wool spots are different from retinitis lesions in several ways. (a) Typically, cotton wool spots have feathery edges while retinitis lesions have granular edges. (b) In some cases cotton wool spots display striations in the direction of the nerve fibers, while CMV retinitis lesions are unstriated. (c) Cotton wool spots are often bounded by the forks of small retinal vessels and seldom straddle them, while retinitis lesions often have vessel segments running through their centers and appear to be spreading from them. (d) Since cotton wool spots are usually concentrated around the posterior pole, the stereoscopic photograph of this region helps the grader to determine if the lesion is superficial, suggesting a cotton wool spot, or involves the full thickness of the retina, suggesting CMV retinitis. (e) Finally, cotton wool spots and retinitis lesions follow different courses: individual cotton wool spots tend to wax and wane somewhat independently of each other, whereas retinitis lesion margins usually resolve with treatment or remain opacified (as the lesion enlarges) if untreated. Unfortunately this last point is not helpful at grading time, when the grader is required to discriminate on the basis of current appearance.

Standard photograph 7A provides an example of the differing appearances of active retinitis and cotton wool spots. The field displays one retinitis lesion, with an active margin above the disc, and four cotton prominent wool spots (one about 1.5 DD supero-nasal to the center of the macula, two just infero-nasal and infero-temporal to the disc, and one about 2 DD infero-nasal to the disc just next to a small hemorrhage). There are three or four fainter cotton wool spots located near the temporal disc margin and in the infero-temporal quadrant. In standard photograph 6A, which shows the same eye as 7A after initiation of treatment, the active retinitis margin and most of the cotton wool spots are resolving at the same time - however, a prominent new cotton wool spot has arisen slightly more than 1 DD infero-temporal to the disc.

While CMV retinitis lesions and cotton wool spots are distinguishable from each other most of the time, there are instances in which they are very difficult to differentiate until later in the case series. (The grader is required to record his/her best judgement at the time a visit is graded, and is not allowed to change the grade based upon later information.)

Example photograph 15A shows an active retinitis lesion adjacent to the disc supero-nasally and at least partially involving it. Only two spots can be identified unequivocally as cotton wool spots, one in the superior arcade and one inferior to the center of the macula. White spots adjacent to the disc margin from 3 o'clock clockwise to 11 o'clock could be cotton wool spots or active retinitis. In photographs 15B, C, and D the active retinitis margins are resolving, while most of the white spots adjacent to the disc become more prominent and then resolve, suggesting that they are (were) cotton wool spots.

Example photograph 13A shows a white spot just temporal to the disc during follow-up. This abnormality is larger than most cotton wool spots and there is subtle swelling or detachment of the retina adjacent to it, strongly suggesting that it is retinitis. Further follow-up revealed that this spot was indeed a retinitis lesion (see photograph 13B, taken four weeks later).

Vascular cuffing is a striking feature seen in some cases of active CMV retinitis, particularly the fulminant variety, at presentation or when reactivated. Dense white sheathing surrounds vessel segments, usually localized near retinitis lesions but sometimes distant from them, and occasionally distributed throughout the fundus. This appearance is similar to the "frosted branch angiitis" sometimes seen in immunocompetent patients. It is thought that vascular cuffing consists of the aggregation of neutrophils along the vessel wall, stimulated by the challenge CMV poses to the immune system. Whatever its cause, vascular cuffing appears to be an acute inflammatory response to CMV retinitis rather than retinal necrosis.

Vascular cuffing associated with active CMV retinitis may be difficult to differentiate from retinitis lesions proper for two reasons. First, the opaque white appearance of vascular cuffing resembles the whitish opacification of small retinitis lesions (which often appear to emanate from around vessels as well). Second, vascular cuffing often is located immediately contiguous to retinitis lesions.

Several features allow discrimination between vascular cuffing and perivascular CMV retinitis. (a) Vascular cuffing has a limited appearance, displaying a uniform and narrow breadth (rather like perivenous exudate in diabetic retinopathy) and terminating abruptly. In contrast, retinitis along vessels usually has variable breadth, widening in places to several times the caliber of the vessel and terminating gradually. (b) The color of opacification in vascular cuffing is generally whitish, while that of retinitis is yellowish-white. (c) Opacification in vascular cuffing often obscures the blood column itself, while that from retinitis usually does not.

When CMV retinitis is treated, cuffing (even when prominent) disappears more rapidly than retinitis lesions can resolve. As retinitis lesions enlarge and the active margins spread away from the vessels, the retina adjacent to the vessels becomes atrophic and loses its whitish appearance. While resolution of vascular cuffing leaves behind little or no trace, the atrophy of retinitis almost always results in pigment disturbance in the RPE. (These natural history differences are not helpful at grading time, however, since the grader is required to discriminate based upon current appearance.)

Example photograph 14A shows prominent vascular cuffing which has arisen not just in proximity to retinitis lesions (nasal to the disc and inferior to the inferior vascular arcade) but also at considerable distance from these lesions (for example, in the supero-temporal quadrant). One month later (after initiation of treatment) the acute appearance of cuffing has disappeared, as shown in photograph 14B, leaving a few isolated loci of what looks like perivenous exudate.

This section presents conventions for determining the presence of CMV retinitis lesions, demarcating their edges, and deciding whether areas of retinitis constitute one or multiple lesions.

The following considerations are applied in determining whether CMV retinitis is present. The protocol specifies a minimum size criterion: a lesion must be at least 1/4 DA in extent to be considered definitely present. For a roundish lesion, this requires a diameter of at least 1/2 DD. Any lesion totaling less than 1/4 DA is considered a special case of questionable retinitis, even if the grader is confident that the lesion is indeed retinitis. Sometimes a coherent locus of retinitis spots, no one of which is 1/4 DA by itself, may be observed apart from any larger lesion. Such clusters may display subtle whitening of the retina between the spots. If the locus in its aggregate involves at least 1/4 DA of retina, it should be considered a definite retinitis lesion (this appearance presumably represents lesion onset).

In preparation for the following section on edge determination, it is useful to specify criteria for presence of satellites within the margin of an area of definite CMV retinitis. Any spots smaller than 63 µm in diameter are disregarded, since they are not sufficiently large to judge whether they truly constitute retinitis. Spots of opacification at least 63 but less than 125µm are included as satellites if located within 1/2 DD of the solid edge of a definite lesion. If they are farther away from the solid edge, they can be classed as satellites only if they are part of a cluster consisting of two or more spots. Spots of opacification at least 125µm in diameter are included as satellites if located within 1 DD of the solid edge of a definite lesion. Spots located farther than 1 DD from the solid edge of a definite lesion may be considered satellites if the gestalt of the lesion warrants. In equivocal cases, the general line of the satellite edge serves as the determining factor.

It is necessary to demarcate the edge of the CMV retinitis lesion so that the grader can decide whether a lesion projects into a retinal zone, can measure the distance from the nearest lesion margin to the disc and the center of the macula, and can assess whether the edge has moved between visits. The following considerations are applied in establishing the edge of the retinitis lesion. Section 7.2.1.1 introduced the concepts of the solid and satellite margins. If a satellite margin is present, its outer edge takes precedence as the boundary of the lesion. Otherwise if a solid edge is present its outer edge defines the lesion boundary. If neither is present because opacification has resolved, the edge of discernible atrophy serves as the boundary of the lesion.

The solid edge is traced by mentally circumscribing the portion of the lesion in which at least half of the retina has become opacified. If there are satellites, the solid edge is located where the satellites become sufficiently dense to involve the majority of retina. (One way for an observer to approximate this edge is to squint until specific features disappear, leaving only the general outline.) In drawing the solid edge the grader preserves within reason the presence of bays or promontories in the line. Any spot separated from a confluent patch of retinitis by normal-appearing retina is included in the solid edge only if it is at least 1/4 DA in extent and part of it is within 1/2 DD of the general line of the solid edge.

The satellite edge is traced by mentally drawing a line through the satellites furthermost from the solid edge, i.e., at the juncture with normal retina. To decide whether to include particular spots, the grader considers their size and their location relative to the solid edge. If the outermost spots are within 1/2 DD of the solid edge, the line is determined any spots that are at least 63 µm in diameter. If the outermost spots are more than 1/2 DD from the solid edge, they must be at least 125 µm in diameter to serve as points in the line unless there is a cluster of smaller spots.

The grader is bound by guidelines regarding how to shape the line connecting satellites to each other or to the solid edge. When connecting satellites to form the edge the grader is restricted to a maximum distance between spots of less than 1/2 DD. Otherwise the line must return toward the solid edge, either to the solid edge itself or to another satellite closer to the solid edge. When projecting the line to include an outlying satellite, the grader must draw a narrow rectangular promontory or salient rather than 'tenting' the line excessively. The purpose of these restrictions is to avoid including overly large areas of apparently uninvolved retina within the lesion margin. In equivocal cases, the grader takes into account the general line of the satellite edge as the determining factor.

Sometimes the margins of CMV retinitis are marked by retinal hemorrhages, particularly in fulminant lesions. These hemorrhages are often sufficiently dense to obscure the opaque white appearance characteristic of active CMV retinitis. (See example photograph 2, which shows the extremely hemorrhagic margin of a lesion located in the supero-nasal quadrant of a right eye.) Upon occasion hemorrhages may appear to become more prominent even as retinitis activity is resolving. These hemorrhages are not considered to be an integral part of the margin, and are disregarded when the edge is traced. If there is sufficient hemorrhage to obscure the edge for a portion of the lesion, the grader bisects the hemorrhage unless the general line of the edge strongly suggests another solution to be more appropriate.

To be classified as a separate lesion, an area of retinitis must be located at least 1/2 DD from the outermost edge of another definite lesion and must be at least 1/4 DA in extent. If an area of definite retinitis is at least 1 DD away from another definite lesion, it is assumed to be a new lesion without further question. However if an area of definite retinitis appears in the region between 1/2 and 1 DD from the edge of another definite (typically larger) lesion, the decision whether to designate it as a new lesion rather than an outpost of the adjacent lesion depends upon several considerations. Evidence supporting classification as a new lesion includes appearance in a separate vascular "watershed" from the previously existing lesion, attainment of a size sufficient to develop a center in which activity appears to be subsiding, and movement of the lesion margin toward the previous lesion as well as in other directions. Evidence against designation as a new lesion consists primarily of a trail of satellites or a zone of milky retina bridging the gap between the lesion in question and the previous lesion.

In practice, whether an area of retinitis constitutes a separate lesion is important primarily for determining whether progression has occurred. When a new area of retinitis appears in a previously uninvolved eye or in a distant area of a previously involved eye, the grader has little difficulty concluding that progression has occurred by virtue of that new lesion. When a sizable new area of retinitis appears close to an existing lesion, the decision may be more difficult - however the ultimate conclusion regarding progression will probably be the same. If the grader decides that the equivocal lesion does not qualify as separate, the conclusion will likely be that progression has occurred through movement of an existing lesion border; if the grader decides that the equivocal lesion is in fact separate, the conclusion will likely be that progression has occurred through appearance of a new lesion.

The presence of CMV retinitis is assessed separately for zone 1 and zones 2/3, with a "gate-keeper" question presented for each. If retinitis is questionably or definitely present, the remaining questions for that zone must be completed. If retinitis is absent in one zone, the grader skips the remaining questions for that zone but completes those for the other zone. If retinitis is absent in both zones, the grader does not complete the remainder of the grading form.

The grader indicates whether CMV retinitis is questionably or definitely present using the criteria described in sections 7.2.1 through 7.2.3, first in zone 1 and then in zone 2/3. The grader first determines the presence of the lesion in question without regard for its location by zone, and then determines whether any part of that lesion is within a given zone. The protocol specifies a minimum size criterion: a lesion must be at least 1/4 DA in extent to be considered present. Any lesion totaling less than 1/4 DA is considered a special case of questionable retinitis, even if the grader is confident that the lesion is indeed CMV retinitis. The grader marks code 0 for no retinitis, code 1 for questionable retinitis, code 2 for presence in the zone of any part of a definite retinitis lesion totaling less than 1/4 DA, code 3 for presence in the zone of any part of a definite retinitis totaling at least 1/4 DA, and code 8 for cannot grade. In other words the grader marks code 3 for presence of definite retinitis in a zone even if the portion of the lesion located within that zone is less than 1/4 DA, provided that the total extent of the lesion over all zones is at least 1/4 DA. This situation sometimes occurs when a lesion is located mostly in one zone but extends slightly into another.

As described in section 4.1, the SOCA grid demarcates two subfields in zone 1 as "safety zones," a term adopted to reflect the provision in some CMV retinitis treatment protocols that persistent or recurrent retinitis in these zones constitutes an imminent threat to vision, and may trigger a change in treatment. The grader examines the disc safety zone (the area within 750 µm of the disc margin) and the macular safety zone (the area within 1,000 µm of the center of the macula) for presence of retinitis. Assuming that the average FAZ is about 500 µm in diameter, this provides a margin around the FAZ of 750 µm. The results are recorded separately for the two zones, using the same codes defined above for presence of retinitis in zones 1 and 2/3: code 0 for no retinitis, code 1 for questionable retinitis, code 2 for part of a retinitis lesion totaling less than 1/4 DA, code 3 for part of a retinitis lesion totaling at least 1/4 DA, and code 8 for cannot grade.

The stage of the most active lesion margin is determined for zone 1 and for zones 2/3. The proportion of the CMV retinitis margins exhibiting definite activity and definite active satellites is estimated. As described in section 7.2.1, active margins are typified by their opaque and densely white appearance, such that underlying detail of the retinal pigment epithelium is obscured. Presence of active satellite lesions is interpreted as a sign of active margins. Atrophic margins are characterized by the resolution of opacity to the point that the detail of the underlying RPE is clearly visible and virtually no trace of whiteness remains. Intermediate margins are those between these two states, i.e., not easily identifiable as either one or the other. A translucent but milky appearance is the hallmark of the intermediate stage. (The use of the term intermediate does not necessarily imply that the change in stage is from active to atrophic; it may be from atrophic to active when a patient is suffering reactivation of retinitis.)

In determining maximum activity, the grader carefully reviews all CMV retinitis margins. If all margins are atrophic or only questionably active, the grader marks the maximum activity as such. If any portion of the lesion margins exhibits opacity suggesting activity, the grader chooses the most active portion of the retinitis margins and compares that area to reference areas in standard photographs. (In each standard photograph, the reference area is the most active portion of the margin shown.) The area of maximum activity under consideration must be a segment of border at least 750 µm or 1/2 DD in length.

Standard photograph 4 provides a reference area showing the most subtle opacification that qualifies as active according to this protocol. Within the lesion margin that extends from 8 to 12 o'clock, the reference area is the small segment of lesion margin (about 1,000 µm in length) near the vertical center of the photographic field and about 3 DD from its left margin. Standard photograph 4 also exhibits subtle white opacity along the more superior portion of the margin pictured. Some experienced graders think this second area is definitely active, while others would evaluate it as only questionably active - thus it is most appropriately graded questionably active.

Standard photographs 5A, 5B and 5C provide reference areas for the next higher stage of activity. These are considered typical of treated but moderately active CMV retinitis. All three standards exhibit activity which is partially resolved, with a solid margin that is narrow and an appearance that is congealed in some portions while other portions (the reference areas) show the slightly fluffier appearance of definite activity. The three standards provide alternate views of this stage of activity, with photograph 5B showing its appearance in zone 1. In photograph 5A, the reference area consists of the inferior 1/3 of the lesion margin. Photographs 5B and 5C have several reference areas, each about 1 DD long. In photograph 5B, there are two reference areas along the lesion margin superior to the disc (one just above the disc itself and the other at the superior margin of the field) and another reference area located about 1.5 DD temporal to the center of the macula. In photograph 5C, one reference area is located at the left margin of the field and another is located about 3 DD from the right margin of the field.

Standard photographs 6A, 6B (the same eye used for 5B but at another visit with more activity), and 6C (the same eye used for 5C but at another visit with more activity) provide reference areas showing greater activity than standard photographs 5A, 5B and 5C. Standard photographs 6A, 6B and 6C are typical of very active treated CMV retinitis, but are less active than most newly-presenting, untreated retinitis. Standards 6B and 6C show this stage of activity in zone 1. In photograph 6A, there are three reference areas each about 1 DD long - one 2 DD superior to the center of the macula, one about 2 DD supero-nasal to the disc (visible in both field 1-2 and field 6), and one near the center of field 6 (the supero-nasal field). In photograph 6B, the reference area consists of the entire margin of the lesion located in the superior temporal vascular arcade - except for obvious gaps supero-temporal to the disc and on the temporal aspect of the lesion. In photograph 6C, the reference area is restricted to a segment of the margin about 2 DD long located about 3 DD from the right margin of the photographic field (just inferior to a major bend in the inferior temporal vein).

Standard photographs 7A (the same eye used for 6A but at another visit with more activity) and 7B exhibit activity characteristic of moderately active, newly-presenting CMV retinitis. In addition to a solid active border of moderate depth along all visible margins, these lesions display a characteristically active satellite margin. Standard photographs 7A and 7B were chosen so that about half of newly-presenting, untreated eyes would be less active than these standards and about half would be as or more active. In photograph 7A, the reference areas consist of the entire lesion margin in field 1-2 and the inferior half of the margin in field 6 (the supero-nasal field). In photograph 7B, the reference areas include the entire lesion margin in field 1-2 and the inferior third of the margin in field 8 (the nasal field).

The code values for maximum activity are: code 0 for no CMV retinitis; code 1 for atrophic retinitis; code 2 for questionably active retinitis; code 3 for activity as much as the reference area in standard photograph 4 but less than that in standard photographs 5A, 5B or 5C; code 4 for activity as much as that in standards 5A, 5B and 5C, but less than that in standard photographs 6A, 6B or 6C; code 5 for activity as much as that in standards 6A, 6B and 6C, but less than that in standard photographs 7A or 7B; code 6 for as much as or more than that in standards 7A and 7B; and code 8 for cannot grade.

The grader considers all of zone 1 including the safety zones to determine the region of maximum activity. The segment of maximum activity compared to the standard photographs must extend at least 750 µm or 1/2 DD along a lesion margin.

The grader considers all CMV retinitis margins in zones 2 and 3, and chooses the most active appearance for comparison with the activity standards. The section of maximum activity compared to the standards must extend at least 750µm or 1/2 DD along the border of a lesion.

The grader estimates the percentage of all CMV retinitis margins in the eye which equal or exceed the activity of the reference area in standard photograph 4. The choices available are code 0 for no activity equal to the reference area, code 5 for approximately 5% equal in activity to the reference area, code 10 for approximately 10%, code 20 for approximately 20%, and so on by increments of 10 up to 100% of the margin. If the grader is unable to estimate the edge proportion active because of photographic quality, he/she chooses code 888 for cannot grade. All retinitis margins from zone 1 and zones 2/3 are considered together. In an eye with extensive or complex margins, the grader may find it necessary to use a measuring magnifier to measure active and inactive margins so as to arrive at the correct proportion.

The grader estimates the proportion of the CMV retinitis margins with active satellites to the nearest per cent, if the maximum activity of retinitis in the eye is at least as great as that in standard photograph 4. As with "Edge Proportion Active," the choices are code 0 for none, code 5 for 5%, code 10 for 10%, code 20 for 20%, and so on up to 100%. The grader chooses code 888 for cannot grade. The grader may use a measuring magnifier to measure the extent of the margins with and without satellites so as to calculate the correct proportion.

The grader uses a measuring magnifier to measure the distance from the disc or macula to the nearest CMV retinitis lesion. The satellite edge of the lesion is used in performing these measurements. When the most posterior lesion occurs in zones 2 or 3 it may be necessary to measure across photographic fields, using match points in the overlap of field 1-2 and the field exhibiting retinitis. If there is no overlap, the grader may be unable to measure and may choose code 88.0 for cannot grade. After measuring with the measuring magnifier to the nearest tenth of a millimeter, the grader divides the raw measurement by 3.0 for Canon 60^o photographs or 3.4 for Topcon 50^o photographs to convert the measurement to disc diameters, and records the result rounded to the nearest tenth.

The grader measures from the edge of the optic nerve head to the nearest CMV retinitis margin.

The grader measures from the center of the macula to the satellite margin of the nearest CMV retinitis lesion.

When the appearance of the disc is abnormal, the grader must determine whether the disc is actually involved by CMV retinitis or by serous fluid in advance of the edge of retinitis. Involvement by serous fluid results in mild blurring of the disc margin and pale discoloration of the disc. (In such cases, the presence of serous fluid may mislead the grader into thinking that retinitis is immediately contiguous to the disc when in fact there may be an intervening strip of uninvolved retina.) In contrast, frank involvement of the disc by retinitis is apparent by the whitening characteristic of active CMV retinitis elsewhere, often accompanied in this location by linear retinal hemorrhages. Typically, swelling of the disc is evident in the stereoscopic photograph (appearing somewhat similar to papillary swelling in diabetic retinopathy) and at least part of the disc margin is obscured.

Example photograph 34 illustrates the appearance over time of disc involvement by CMV retinitis. In photograph 34A, taken one month after initiation of treatment, retinitis adjacent to the disc margin infero-nasally is largely atrophic - a few small areas of mild opacification remain. In photograph 34B, taken one month later, CMV retinitis has reactivated and expanded dramatically, completely involving the disc and extending along the inferior temporal vascular arcade. In photograph 34C, taken one month later, the appearance of active disc involvement has resolved (the disc is beginning to pale), although the lesion margins temporal and immediately superior to the disc remain active.

The following codes are used to describe the status of the disc: uninvolved, code 0; questionably involved, code 1; edematous in appearance but not involved by CMV retinitis, code 2; involved by active retinitis over less than 50% of the disc, code 3; involved by active retinitis over at least 50% but less than 100% of the disc, code 4; involved by active retinitis over 100% of the disc, code 5; cannot grade, code 8.

The foveal avascular zone (FAZ) is assumed to be the area within 1/6 DD or 250 µm of the center of the macula. For recording status of the FAZ, the grader chooses from the following codes: CMV retinitis absent, code 0; questionable retinitis, code 1; retinitis present but satellites only, code 2; retinitis present but involving less than 50% of the FAZ, code 3; retinitis present involving at least 50% but less than 100% of the FAZ, code 4; retinitis involving 100% of the FAZ, code 5; and cannot grade, code 8.

In an attempt to describe the fulminant or indolent nature of the CMV retinitis (discussed in section 7.2.1.1), the grader evaluates the depth of the solid and satellite borders or margins, and their hemorrhagic content. This assessment is performed only at baseline, when lesion margins usually active.

The depth of the active margin is evaluated for both the solid and satellite margins. For the solid margin, this depth is defined as the perpendicular distance from the outer edge of the solid margin to the inner boundary between clearly active retinitis and a less active stage, either intermediate or atrophic. The depth of the margin is measured along a front at least 750µm or 1/2 DD in length.

To record depth of the solid margin, the grader utilizes the following categories: no activity (code 0), questionably active solid border (code 1), definitely active solid border < 0.25 DD deep (code 2), solid border < 0.50 DD deep (code 3), solid border < 0.75 DD deep (code 4), solid border <1.00 DD deep (code 5), or solid border > 1.00 DD deep (code 6).

If the lesion is of sufficiently small size, there may be no discernible core of atrophy. When no atrophic or intermediate area is visible, the measurement of the depth of the active solid border is obtained by measuring across the round or oval lesion in its shortest dimension. If such a lesion does not measure at least 2 DD across, the grader indicates that the depth cannot be determined (code 7). If the depth of the solid border cannot be determined because of photographic quality, the grader employs code 8 for cannot grade.

For the satellite margin, this depth is the perpendicular distance from the outer edge of the active satellite margin to the outer edge of the solid margin. To make this measurement more reproducible, the depth recorded must occur along a breadth of at least 750µm or 1/2 DD. For depth of the satellite margin, the following categories are used: no active satellites (code 0), questionable active satellites (code 1), definitely active satellite border < 0.25 DD deep (code 2), satellite border < 0.5 DD deep (code 3), satellite border < 0.75 DD deep (code 4), satellite border < 1.00 DD deep (code 5), or satellite border > 1.00 DD deep (code 6). Code 8 is used for cannot grade.

As described above, the solid margin of CMV retinitis lesions is sometimes marked by intraretinal hemorrhages. When these are prominent (typically in an eye with fulminant retinitis), they tend to be sufficiently dense and superficial that they obscure other features within the retina. The purpose of this item is to identify eyes that display mild, moderate or severe hemorrhages within the retinitis margin. The grader compares the most hemorrhagic lesion border to those in standard photographs. Standard photograph 9 is used to differentiate mild from moderate hemorrhage, while standard photographs 10, 11, and 12 are used to distinguish moderate from severe hemorrhage. The latter group of standards shows retinitis lesions in which about 30% of the active margins have retinal hemorrhage, illustrating different patterns of hemorrhagic margins that may be encountered. Factors considered in determining the grade are both the density and the extent of hemorrhages.

The grader utilizes the following categories to describe the status of hemorrhage in the CMV retinitis lesion margin: no hemorrhage (code 0), questionable hemorrhage (code 1), definite hemorrhage but less severe than that in standard photograph 9 (code 2), hemorrhage as much as that in standard photograph 9 but less than that in any of standard photographs 10, 11 or 12 (code 3), hemorrhage as much as or more than that in any of standard photographs 10, 11 or 12 (code 4), or cannot grade (code 8).

After the grader has assessed the current status of CMV retinitis as described above, follow-up photographs are compared with those from the baseline, and then with those from the previous visit, to determine if progression of retinitis has occurred. Progression as conventionally defined consists of either movement of the border of an existing lesion by a specified distance or occurrence of a new retinitis lesion. Direct comparison of slides is facilitated by temporarily shifting the position of the slides in their mounting sheets, as described in section 5.

For zones 1 and 2/3 separately, the grader determines whether there has been expansion of retinitis (movement of the outermost lesion edge) into previously normal-appearing retina, and if definite movement is detected records its extent. In order of precedence, the grader considers the satellite edge, the solid edge, or the atrophic edge. The movement observed must have occurred along a front at least 750 µm in length to qualify for a given category of extent. Movement of at least 750 µm has conventionally been defined as disease progression for many studies.

The grader proceeds systematically through each photographic field showing lesion margins to compare their positions between visits. The corresponding fields are juxtaposed to each other as described in section 5. Most graders find it helpful to alternate their view between baseline and follow-up very frequently in order to be sure movement is not overlooked. The grader must remember that movement of the lesion edge and degree of activity have been separated in this protocol. While movement is typically accompanied by persistent or increased activity, there are cases in which perceptible movement of the edge has been detected unaccompanied by notable activity. Consequently, the grader is required to examine the lesion margins carefully for movement even if no activity is apparent. If definite movement is detected, the grader formally measures the greatest distance moved. To perform the measurement, the grader uses either the measuring magnifier or the graduated standard circles to confirm that the advance meets the criteria specified. In particular, the circle of 750 µm diameter can be moved through the area of interest to check both dimensions simultaneously.

Example photograph 35 provides a longitudinal view of a case in which CMV retinitis is treated, resolves without expanding significantly, reactivates, and then demonstrates border movement exceeding 750 µm (i.e., progresses according to convention). In photograph 35A, taken before initiation of treatment, fulminantly active retinitis is present just temporal to the center of the macula. In photograph 35B, taken two weeks after start of treatment, some areas previously involved with satellites have become more uniformly opacified. Beyond this 'filling in' of the margin there has been some expansion of the lesion into previously normal-appearing retina, most notably towards the disc in the superior part of the papillo-macular bundle. However, the distance moved has been less than 750 µm - the conventional criterion for progression. In photograph 35C, taken at four weeks, the opacification of the lesion margins has resolved somewhat and there has been no further expansion. In photograph 35D, taken at six weeks, the opacification has almost completely disappeared (except for a small area superior to the papillo-macular bundle) and the margins appear static. However in photograph 35E, taken at 12 weeks, there has been subtle reactivation of the lesion margin. Around most of the lesion perimeter the apparent change does not qualify as definite progression because the movement of the lesion margin has been less than 750 µm. However, the tongue of retinitis that has expanded nasally from the 7'oclock aspect of the lesion is more than 750 µm broad and has moved as much as 1,500 µm (an intraretinal hemorrhage marks the center of the newly-involved area). In photograph 35F, taken at 16 weeks just before a change in treatment, the margin has become frankly active and has expanded around its entire perimeter. In the area of greatest expansion, inferior to the papillo-macular papillo-macular bundle, the distance the edge has moved from baseline exceeds 2 DD.

Several other cases of progression by movement of the lesion border are discussed in SOCA report number 8 from the Foscarnet Ganciclovir Cytomegalovirus Retinitis Trial (FGCRT).

Results are recorded using the following codes: no evident movement of the lesion edge, code 0; questionable edge movement, code 1; definite edge movement but less than 750 µm over a 750 µm front, code 2; movement at least 750 µm but less than 1,500 µm along a 750 µm front, code 3; movement at least 1,500 µm but less than 3,000 µm along a 750 µm front, code 4; movement equaling or exceeding 3,000 µm along a 750 µm front, code 5; cannot grade, code 8.

If movement of a lesion border crosses the boundary between zone 1 and zones 2/3, it is recorded to its full extent for both zones (i.e., the grader does not distribute the total distance moved between the two zones).

When a lesion previously less than 1/4 DA in extent has expanded sufficiently to satisfy the 1/4 DA criterion, it is graded solely as a new lesion and its border movement is excluded. However if there is further movement at the next visit that is included when assessing border movement compared to the last visit (but still excluded from movement compared to the baseline).

It is assumed in this protocol that CMV retinitis lesions essentially do not retract from their farthest advanced positions. Thus if a grader indicates that movement of a specified distance has occurred from baseline to a given follow-up visit, the grader is constrained at successive visits to select either that code for movement or a higher one (if further movement has occurred).

Sometimes either the baseline or the follow-up set of photographs may not capture all the lesion margins documented in the other set, due to missing or shifted fields, poor image quality, or presence of photographic artifacts. In such cases, the grader must decide whether to record a substantive grade or to indicate that the comparison cannot be made. If movement satisfying a criterion for progression is observed the grader records it, even if only a minor portion of the lesion edges can be compared. On the other hand if movement qualifying as progression is not observed, the grader must be able to compare satisfactorily at least half the known margins in order to record a substantive result rather than cannot grade.

If photographs were taken with different magnifications at the two visits, it may be necessary to measure from the lesion edges to retinal landmarks within each set, convert the raw measurements to the common unit of disc diameters, and calculate the difference to see if movement occurred.

The grader marks all photographic fields exhibiting movement greater than 750µm along a 750µm front (either code 3, 4, or 5). This item serves as a scratch pad for the convenience of the grader to record the photographic fields in which movement qualifying as progression was observed.

Separately for zones 1 and 2/3, the grader assesses eyes for the occurrence of new CMV retinitis lesions (?spots?) using criteria presented in section 7.2. Occurrence of a new lesion at least 1/4 DA in extent is conventionally defined as retinitis progression.

The grader scans each photographic field systematically for new CMV retinitis lesions, paying particular attention to the peripheral parts of photographic fields where the view of the retina might be compromised by artifacts.

Example photograph 33 illustrates appearance of a new CMV retinitis lesion during follow-up. Photograph 33A (a left eye), taken before initiation of treatment, displays no definite retinitis lesions in field 1-2. Definite retinitis was not detected anywhere in this eye; the patient was entered into a treatment trial for retinitis in the contra-lateral eye. In photograph 33B, taken six months after initiation of treatment, an area of subtle whitening has appeared at 7 o'clock about one DD from the disc margin. At this point the grade would be borderline between questionable and definite presence of a new retinitis lesion, for two reasons. First, the appearance is still somewhat equivocal (reexamination of the baseline photograph reveals that this area did not appear entirely normal, perhaps due to heightened reflex or a subtle cotton wool spot). Second, even if the abnormality is considered to be definite retinitis it may not quite fill the 1/4 DA measuring circle, depending upon exactly where the lesion edge is determined. In photograph 33C, taken at eight months (the seven month photograph was not taken), the abnormality is clearly definite retinitis, based upon its characteristic opacification, full thickness involvement, and unmistakable satellites.

If no new CMV retinitis lesion is present the grader marks code 0. If the grader thinks that there is a new lesion but that its appearance is too equivocal to be called definite, code 1 is selected. If the grader is confident that the new lesion is retinitis but the total area of the lesion is less than 1/4 DA, the grader selects code 2. If the new lesion totals at least 1/4 DA but less than 1 DA, the grader marks code 3. If the new lesion equals or exceeds 1 DA in extent, the grader chooses code 4.

If a new retinitis lesion straddles the boundary between zone 1 and zones 2/3, it is recorded for both regions.

It is assumed in this protocol that definite CMV retinitis lesions, once established, do not disappear. Thus if the grader has detected the appearance of a new lesion from the baseline to a given follow-up visit, the grader is constrained at successive visits to choose the same code for new lesion or a higher code (if the lesion has increased in size).

When a questionable or definite CMV retinitis lesion is detected in an area not included in the relevant baseline photographs, or if that area was included but photographic quality is inadequate to document retinitis, the grader marks cannot grade (code 8).

The grader marks all photographic fields exhibiting a definite new spot of at least 1/4 DA (code 3 or 4). The grader utilizes this item as a scratch pad to note the photographic fields in which new lesions were found.

Individual studies may have their own windows as to when photographs are considered to serve as the baseline - ideally within two or three days of randomization. When a series of patient photographs lacks baseline photographs, evaluation of follow-up photographs is problematic. The goal is to provide useful information if possible so that the case is not lost to analysis, but to return an indeterminate result when necessary. Lacking baseline photographs, the grader is faced with the following choices.

(1) Eyes free of CMV retinitis at the follow-up visit have not progressed by definition, despite the lack of baseline photographs.

(2) Eyes with CMV retinitis at the follow-up visit fall into one of several categories.

(a) If photographs taken prior to the allowed baseline window are available, the grader compares the follow-up position of CMV retinitis lesions to the pre-baseline photograph. If retinitis has not progressed compared to the pre-baseline photographs, it can be assumed to have not progressed compared to the undocumented baseline. This assertion is based on the assumptions that retinitis lesion margins never retract and new retinitis lesions never disappear.

(b) If CMV retinitis has progressed compared to any pre-baseline photographs, or if such are not available, then the grader compares the follow-up photographs to those from the first documented follow-up visit in lieu of baseline photographs. The first follow-up visit itself is necessarily indeterminate ("cannot grade") regarding progression. If retinitis has not progressed compared to the first follow-up visit, the grader is constrained to select "cannot grade" for the conclusion regarding progression. If retinitis has progressed compared to the first follow-up visit, it has necessarily progressed compared to baseline. The grader records the degree of progression relative to the first follow-up visit, and records the caveat that retinitis may have progressed sooner and more than can be ascertained from the documentation available.

Overall, the grader is advised to be cautious about providing a progression conclusion, under the policy that an indeterminate result is preferable to an erroneous result.

8.0 Evaluation of Abnormalities Other than CMV Retinitis, Primary Grading Form

Remaining sections of the grading form deal with the non-infectious retinopathy seen in persons with AIDS, and with other abnormalities often seen in persons with CMV retinitis - many directly related to presence of retinitis.

Rhegmatogenous retinal detachment is a major vision-threatening sequela of CMV retinitis. The grader examines the entire region of retina documented in the photographs for detachment, although only field 1-2 provides stereoscopic effect to aid this determination. Retinal detachment is more easily and accurately identified when viewed stereoscopically, so that elevation of the detached area relative to adjacent flat retina can be perceived directly.

Another important clue to retinal detachment is blurring of the RPE/choroidal pattern beneath the translucent (but not completely transparent) detached retina. When shallow retinal detachment is viewed in non-stereoscopic photographs, this blurring may be the only sign of detachment. The contrast of appearances between detached and attached areas of retina depends upon the plane of focus of the camera. If the camera is focused on detached retina, in detached areas only retinal details (not the choroidal pattern visible through the RPE) will be clear while in adjacent attached areas both the retinal details and the RPE/choroidal pattern will be blurred. If the camera is focused on attached retina, in attached areas both the retinal details and the RPE/choroidal pattern will be clear while in adjacent detached areas both will be blurred.

Care must be taken to distinguish between blurring from shallow retinal detachment and from other causes, such as vitreous or lens opacities or from improper focus of the fundus camera. In the latter cases both the retinal details and the RPE/choroidal pattern are likely to be blurred, either throughout the photographic field or in areas of localized vitreous opacity.

Rhegmatogenous detachment is characterized by the presence of holes or tears in the sensory retina that have allowed fluid vitreous to flow beneath it, separating the sensory retina from the retinal pigment epithelium. Retinal tears are discussed in section 8.3.8.1 below, and illustrated in example photograph 23. (Although visibility of retinal holes can help confirm presence of detachments, they are not required to consider detachment present.) Rhegmatogenous detachments are usually larger than serous detachments and sometimes have folds within them. They may occur more commonly when large areas of retina have become atrophic and hence fragile. Thinning of the retina following CMV retinitis infection leaves the retina vulnerable to tearing along the juncture of normal and atrophic retina or within the atrophic area.

Serous retinal detachment (caused not by fluid from the vitreous cavity but by fluid coming from retinal or choroidal blood vessels) may also occur, usually adjacent to areas of active retinitis and often involving the macula. When the grader is confident that all detachment seen is strictly serous, which may be the case when a shallow detachment is restricted to the posterior pole and the entire detached area and its margin is visible, it is recorded separately from rhegmatogenous detachment under its own item as discussed in section 8.3.9 below. Otherwise all detachments are recorded under the primary retinal detachment item presented in this section, without further attempt to differentiate between rhegmatogenous and serous detachments.

In example 23 (a right eye), a large CMV retinitis lesion extends across the inferior quadrants of the retina. Retinal breaks are visible in field 8 (nasal), and perhaps inferior to the disc in field 1-2. The retina appears detached nasal to the disc (between it and the retinal break), and may be detached through the posterior pole, and into field 3 (temporal), where it is limited by the horse-shoe shaped margin of the retinitis lesion. In example photograph 24 (a right eye), a large retinal tear with a rolled edge is visible inferiorly. The retina has become detached throughout much of the eye, including the posterior pole, except for areas in field 9 (superior) and field 3 (temporal). Inferior to the tear (looking through the break) the choroidal pattern is clearly visible; superior to the tear the choroidal pattern can no longer be detected through the detached retina. In example photograph 26 (a right eye), there was a CMV retinitis lesion in the far inferior nasal quadrant (not shown). An area of detachment extends from there towards the disc, displaying an obvious demarcation between attached and detached retina.

Findings are recorded for zone 1 (separately for the FAZ and for the zone overall) and for zones 2/3 (by quadrant). For each region, the grader chooses from among the following codes: no detachment, code 0; questionable detachment, code 1; definite detachment, code 2; cannot grade, code 8.

Some lesions are observed with or without CMV retinitis, and are perhaps independently associated with HIV infection. These include hemorrhages/microaneurysms (H/Ma) and cotton wool spots, each assessed separately for zones 1 and zones 2-3.

Both microaneurysms and hemorrhages are dark red spots (dots and blots) within the retina. A red punctate spot that is less than 125 µm in its longest dimension (approximately the width of a vein at the disc margin) and has sharp margins and even density is considered a microaneurysm. However, a red spot of this size that has irregular margins and/or uneven density is considered to be hemorrhage. A red spot that is at least 125 µm in its longest dimension is considered a hemorrhage, unless features such as round shape, smooth margins, and a central light reflex suggest that it is a microaneurysm. Although this distinction is sometimes difficult (especially in wide-angle photographs) it is not critical in this classification because these two types of red spots are graded as a single lesion (H/Ma). Areas of hemorrhage located within the margin of a CMV retinitis lesion are considered separately.

The amount of H/Ma is compared to standard photographs 1 and 2. Standard photograph 1 shows a retinal hemorrhage in the nerve fiber layer infero-nasal to the disc and overlapping the boundary between zones 1 and 2. Standard photograph 2 shows a blot retinal hemorrhage and several microaneurysms in the far periphery of zone 2, the total area being roughly equal to the area of the hemorrhage in standard photograph 1. For zone 1, the total area of retinal hemorrhages and microaneurysms within the zone 1 boundary is compared with the total area in either standard photograph 1 or 2 (not just the area covered by retinal hemorrhage within zone 1 in standard photograph 1). For zones 2/3, the amount of retinal hemorrhages and microaneurysms in any one photographic field must be compared with the amount of H/Ma in all of standard photograph 1 or 2.

The grading codes used are: code 0 for no H/Ma, code 1 for questionable H/Ma, code 2 for definite H/Ma less than those in standard photographs 1 or 2, code 3 for definite H/Ma equal to or greater than standard photographs 1 or 2, and code 8 for cannot grade.

Cotton wool spots (sometimes termed soft exudates by convention) are superficial abnormalities that are white, pale yellowish-white, or grayish-white with ill-defined or feathery edges, frequently showing striations parallel to the nerve fibers. These areas of swelling of the nerve fiber layer should not be confused with the normal pattern of nerve fiber striation often seen along the superior and inferior vascular arcades. Differentiation of cotton wool spots from CMV retinitis lesions is discussed in section 7.2.2.1, and an example of cotton wool spots presented in standard photograph 7A.

The presence and number of cotton wool spots is recorded on an ascending scale with a doubling in number at each step of the scale. The number of cotton wool spots in zone 1 and in zones 2/3 is recorded separately. Each distinct, non-contiguous cotton wool spot is counted as one locus. The code values for cotton-wool spots, are: code 0 for none, code 1 for questionable cotton wool spots, code 2 for one definite cotton wool spot, code 3 for two or three definite cotton wool spots (less than four loci), code 4 for four to seven definite cotton wool spots (less than eight loci), code 5 for eight or more definite cotton wool spots (greater than or equal to eight loci), and code 8 for cannot grade.

A variety of other features is assessed in addition to the evaluation of CMV retinitis itself. While some of these are presumably linked to retinitis, others may not be (they are included because they might confound assessment of retinitis or cause visual impairment). These other features are assessed for the eye as a whole, under the control of an item which asks if any of them are questionably or definitely present. If none are present, the grader marks code 0 and skips the remainder of the section. If any are present, the grader marks code 1 for those judged questionable and code 2 for those thought definite. Any features not specifically listed are marked as "other" and described in the comment section.

Vascular cuffing is an opaque white appearance bordering blood vessels and located outside of the lesion proper. (Vascular sheathing within the lesions themselves, often observed during the atrophic stage, is disregarded.) Although cuffing typically occurs in the vicinity of CMV retinitis lesions, sometimes it occurs elsewhere in the retina, e.g., near the posterior pole when retinitis is limited to the periphery. Cuffing has fairly uniform and narrow breadth, somewhat similar to perivenous exudate in diabetic retinopathy, and terminates relatively abruptly. Differentiation of this appearance from CMV retinitis lesions is discussed in section 7.2.2.2. Vascular cuffing can be observed throughout example photograph 14A, even at some distance from the retinitis.

Vitreous or preretinal precipitates are small whitish deposits just anterior to the retinal surface or floating in the vitreous, typically appearing in clusters. Such precipitates are assumed to be deposits from the diffuse vitritis sometimes associated with CMV retinitis, especially the active stage. The grader needs to be careful to distinguish such precipitates from the photographic artifacts sometimes seen due to debris on the camera lens. Example photograph 12 shows a stereoscopic view of vitreous precipitates located in the bottom part of field 10 (inferior to the posterior pole). These appear as clusters of out-of-focus whitish dots, located anterior to the plane of the retina (perhaps on the posterior face of the detached vitreous).

Vitreous haze is a partial loss of transparency of the vitreous causing the appearance of the underlying fundus (both the retina and the underlying pattern of the RPE and choroid) to be fuzzy. Presumably this obscuration is a result of inflammatory cells infiltrating the vitreous. Although vitreous haze is usually generalized, such clouding may be restricted to discernible areas surrounded by others where the fundus is clearly visible. Haziness from vitreous hemorrhage is not included in this category, but recorded separately (see section 8.3.6 below). Retinal detachment may also result in decreased visibility of the underlying pattern of RPE and choroid; its differentiation from vitreous haze is discussed in section 8.1 above. Example photograph 17A shows an eye with diffuse vitreous haze prior to initiation of treatment for retinitis (an active lesion margin is visible superior to the disc). The fundus reflex photograph confirms that the hazy appearance of the posterior pole in field 1-2 is accounted for by opacities in the vitreous rather than on the cornea or in the lens. In photograph 17B, taken four weeks later, the vitreous haze has mostly cleared (and the lesion margin superior to the disc has become atrophic).

Occasionally eyes with more peripheral involvement by CMV retinitis develop abnormalities in the macular region. They include diffuse macular edema and/or shallow serous sensory retinal detachment, which are difficult to distinguish from each other A yellowish appearance near the center of the macula is often manifest in such cases. Given that only shallow stereoscopic effect is attainable with a wide-angle camera, the grader may find it difficult to detect subtle retinal thickening without relying upon other features such as a partial loss of retinal transparency.

Over time, macular edema may appear to become cystoid. Cystoid spaces are vacuolar areas of increased retinal transparency (typically round or oval with diameters one to four times that of a major retinal vein at the disc margin) occurring at or near the center of the macula, usually in regions that appear somewhat thickened. Reflexes from the anterior surface of the retina and/or spots of mild depigmentation may be mistaken for cystoid changes. For cysts to be considered definitely present, a distinct edge bounding an area of increased retinal transparency must be traceable around at least half the circumference of one cyst.

Macular abnormalities in an eye with CMV retinitis outside of zone are shown in example photograph 20, which includes stereoscopic 30^o Zeiss photographs of the disc and macula in addition to the standard wide-angle photographs.

The category for disturbances of the retinal surface includes several appearances. Initially, such disturbances may appear as regions of "cellophane" reflex and moderate loss of transparency. More severe occurrences may display gliotic or fibrotic tags or sheets, and may be perceived as distinct epiretinal membranes. Adhesions appear as areas in which the posterior face of the vitreous seems to be attached to the internal limiting membrane, with a clear boundary circumscribing the affected area. Sometimes this process may result in surface wrinkling or tension lines (small folds in the retinal surface, oriented parallel to an assumed vector of tension within the retina).

Surface abnormalities within 2 DD of the center of the macula are illustrated in example photograph 19. In this right eye with CMV retinitis located supero-nasal to the disc, a sheet of cellophane reflex developed by three weeks after entry (19A) that progressed by week 19 to a more fibrotic appearance with accompanying tension lines (19B).

Sometimes the optic nervehead may be surrounded by visible fibrotic thickening. At first this phenomenon may look like a "halo" circling the disc, but later it may take on a more dense and strand-like appearance characteristic of fibrosis. This feature may be accompanied by vascular anomalies, such as dilation of capillaries, avulsion of retinal vessels, or formation of new vessels. (Such features are recorded under the 'Other Ocular Abnormalities,' as discussed in section 8.3.13 below.)

Example photographs 18 and 20 illustrate these peripapillary changes in an eye with CMV retinitis located some distance away from the disc. In both cases, a narrow fibrotic process with a vascular component appears to be pulled anteriorly from its point of origin on the disc.

Preretinal and/or vitreous hemorrhages may occur in eyes with CMV retinitis. Preretinal hemorrhage is located just anterior to the retina or under the internal limiting membrane, while vitreous hemorrhage is located in the vitreous cavity (i.e., anterior to preretinal hemorrhage, including hemorrhage on or within fibrovascular proliferations). Occasionally, preretinal or vitreous hemorrhages are seen in newly presenting eyes with retinitis. More often, they are seen in treated eyes subsequent to surgical intervention such as retinal detachment repair or insertion of an intra-ocular ganciclovir implant.

Preretinal hemorrhages are depicted in example photograph 6 (the boat-shaped hemorrhage superior to the disc) and example photograph 36 (the large, dense hemorrhage on the inferior margin of the retinitis lesion).

Vitreous hemorrhage can appear in several manifestations, as the following example photographs illustrate. Example photograph 3 shows rather discrete hemorrhage in a right eye subsequent to surgical implantation of a ganciclovir device - the hemorrhage is located infero-temporally near the point of insertion of the implant. Example photograph 4 displays ?old white blood? in field 10 (inferior region) of a left eye several weeks after ganciclovir implant surgery. Example photograph 5 shows hemorrhage diffused throughout the vitreous in the inferior part of a right eye, but confined so that the retina is clearly visible supero-nasal to the disc. Example photograph 37A shows a clear view of the retina in a right eye prior to implantation of a ganciclovir device. Three weeks after surgery (photograph 37B) the view of the retina is hazy, presumably due to vitreous hemorrhage, and one heavier blob is visible high in the vitreous cavity. Seven months after surgery (photograph 37C) the haziness has cleared and the retina is once again clearly visible.

Subretinal hemorrhage is sometimes seen within or adjacent to CMV retinitis lesions, particularly in eyes with very fulminant retinitis. Subretinal hemorrhage is located deep in the retina, in the potential space above the RPE but beneath the rods and cones. Example photograph 7 shows a subretinal hemorrhage beneath the superior vascular arcade of a right eye; example photograph 36B shows a subretinal hemorrhage located about 2 DD inferior to the disc (just inferior to the crossing of an arteriole and a venule).

Since retinal detachment is a major vision-threatening sequela of CMV retinitis, it is discussed separately in section 8.1 above and recorded under its own questions within the retinitis portion of the form. However, several features listed under "Other Lesions" are related to retinal detachments and their repair.

Retinal tears within or along the edge of rhegmatogenous detachments may sometimes be visible in photographs. The adjacent detached retina is typically translucent rather than completely transparent, obscuring the pattern of the underlying structures, but the RPE and choroidal detail may be clearly visible through the retinal tear. This feature is illustrated in the inferior portion of field 8 of example photograph 23, where there is an area of detachment nasal to the disc in a right eye. Within the detached region is a squarish area of greater visibility of the RPE/choroidal pattern, probably indicating a break in the retina. Example photograph 24 presents a more dramatic example in which there is a large tear with a rolled edge extending across much of the inferior aspect of the retina. Retinal detachment has involved most or all of the retina in this eye. There appears to be a thin remnant of retina across the nasal part of the tear.

Retinal reattachment surgery may entail the insertion of silicon oil or a gas bubble to push detached retina back against the RPE. This tamponade can often be detected in the photographs, either by a fluid level between silicon or gas and vitreous, or in the case of silicon oil by prominent characteristic light reflexes in the posterior pole. Because the broad whitish appearance of silicon oil reflex could be mistaken for active CMV retinitis, the grader must be skeptical of this appearance in eyes with silicon oil tamponades. Since the appearance at issue is typically in the posterior pole, examination of the overlapping region from adjacent photographs often resolves the matter. Seen from another angle the problematic reflex usually changes location or disappears, allowing a clear view of the retina. Presence of a tamponade usually alters the refractive index of the eye, possibly resulting in a discernible change of magnification compared with photographs taken before surgery.

Example photograph 30 shows a right eye with appearances typical of a silicon oil tamponade - the silicon reflex is prominent over the center of the macula (here uninvolved by retinitis) in field 1-2, and the edge of the bubble is visible in field 5 (infero-temporal to the posterior pole). Example photograph 31 displays a case in which silicon reflex partially obscures the appearance of active retinitis. In the field 3 photograph of this left eye, marked reflexes are seen to overlay several areas of retina temporal to the macula. Examination of this same area using the field 1-2 photograph reveals that a small, active area of retinitis (located at 2 o'clock from the center of the macula and about 3 DD away) was partially obscured by silicon reflex. If only the field 3 photograph were available, this appearance might well be overlooked or graded questionable.

Scleral buckling operations are sometimes performed to repair retinal detachments subsequent to CMV retinitis. Application of the buckle results in band-like elevations of the retina and choroid visible in retinal photographs, usually at or near the equator of the globe and parallel to it - and thus near the peripheral edge of one or more photographic fields. Example photograph 32 illustrates the appearance of a buckle in a left eye, with the buckle oriented circumferentially in field 6 (supero-nasal), field 8 (superior), and field 4 (supero-temporal).

Photocoagulation treatment may be applied adjacent to retinitis lesions in the hope of either preventing retinal detachment or walling off a localized area of detachment to prevent its spread. Photocoagulation scars can be differentiated from the RPE disturbance associated with atrophic CMV retinitis by the greater regularity of photocoagulation scars. When laser burns are applied confluently, they typically result in a distinct band. When isolated individual burns are visible outside of any confluent treatment, their roundish shape and uniform size (often about 750 µm diameter) contrasts distinctly with the irregular shapes and sizes of retinitis lesions.

Example photograph 28 illustrates the appearance of fresh photocoagulation burns, which appear as localized whitening of the retina. In field 1-2 of this left eye, the burns wall off an area of detached retina temporal to the macula; in field 4 (further temporal to the detachment) they are scattered over an area that appears to have been involved by retinitis previously. (It is important not to mistake fresh photocoagulation burns adjacent to retinitis lesions for active retinitis margins.) More typically seen are old photocoagulation scars, as displayed in example photograph 29. This right eye has a shallow detachment located infero-nasal to the disc, with a cordon of scars flanking the detachment in field 8 (nasal to the disc) and field 10 (inferior to the posterior pole).

Serous detachment and macular edema may be difficult to differentiate from each other, and thus are considered together. They are sometimes observed near the margins of active CMV retinitis within the posterior pole and near the macula or (for serous detachment only) around the disc.

Serous detachment is a separation of the sensory retina from the retinal pigment epithelium, caused by accumulation between these two layers of fluid arising presumably from the retinal or choroidal vasculature (rather than from the vitreous cavity via a tear in the sensory retina). Such detachments are usually smooth in contour and localized. Macular edema is a thickening of the retina in the macular region caused by an accumulation of fluid within the retina. There may be partial loss of transparency, and sometimes the edematous retina contains or is bounded by hard exudates. Shallow serous detachments and macular edema are difficult to detect in non-stereoscopic photographs, i.e., outside of field 1-2.

Serous detachment/macular edema is recorded both for the eye as a whole and for the center of the macula.

Example photograph 22 shows macular edema (and perhaps serous detachment) accompanying active CMV retinitis. In this right eye, an active retinitis lesion arches superiorly above the macula, approaching to within about 1/2 DD of the center. The stereoscopic photograph shows retinal thickening extending from the edge of the retinitis lesion towards the center and just involving it.

Example photograph 36 provides an example of macular edema or serous detachment resolving over time. In photograph 36A, prior to initiation of treatment, an area of thickening or detachment extends from the active retinitis lesion located in the inferior temporal vascular arcade toward the center of the macula. In photograph 36B, taken two weeks after treatment began, retinal edema thickening is still present - although it is decreasing as fluid is reabsorbed. At this stage, hard exudates have become prominent in the macular region, and retinal tension lines are visible. In photograph 36C, taken at 24 weeks, the retinal thickening and hard exudates have completely disappeared and the retinitis lesion remains atrophic. Photograph 36D shows reactivation of the retinitis lesion margin with expansion superiorly. Now the granular whitish spots approaching the macula are part of the active retinitis margin - not hard exudates.

Example photograph 13B illustrates serous retinal detachment or edema involving the center of the macula. Fluid presumably arising from the active CMV retinitis lesion located in the papillo-macular bundle and arcing above the macula has penetrated under the sensory retina, detaching it so that the yellow xanthophyll pigment in the fovea has become apparent. Example photograph 21 was taken of a patient assigned to deferral of treatment in a clinical trial of one of the first anti-CMV drugs available. It demonstrates that serous detachments can be large, and can affect the macular region even when the retinitis lesion is somewhat distant. In photograph 21A, an active retinitis lesion located in field 10 (inferior) has resulted in an extensive serous detachment involving the entire inferior temporal vascular arcade and extending to within 1 DD of the center of the macula. Subtle foci of vascular cuffing can be detected in the vascular arcade. In photograph 21B, taken one week later and just before initiation of treatment, serous detachment has involved more of the area surrounding the macula. (Retinal elevation is less apparent because the stereoscopic effect is less - compare the degree of optic disc cupping in photographs 21A and 21B.) Vascular cuffing is now extremely prominent, and involves several vessels in the macular region.

In eyes with CMV retinitis, hard exudates frequently appear concurrently with or subsequent to serous detachments or retinal edema adjacent to retinitis lesions. Usually hard exudates are restricted to the posterior pole, around the disc and macular regions. Hard exudates are small white or yellowish-white deposits of lipid with sharp margins, often appearing waxy, shiny or glistening. They are typically located in the outer layers of the retina, but may be more superficial, particularly when retinal edema is present. Hard exudates may be arranged as individual dots or confluent patches. Care must be taken to differentiate them from drusen. Although somewhat similar in appearance, drusen appear deeper in the retina, are less shiny, and usually have less distinct margins.

Introduced above in section 8.3.9, example photograph 22 illustrates hard exudate accompanying macular edema (and perhaps serous detachment). Small hard exudates are interspersed within an area of retinal edema that extends inferiorly from the retinitis lesion toward the center of the macula. (Hard exudates are not considered retinitis satellites, and in this case are located outside of the lesion margin.) Example photograph 36B (also discussed above in section 8.3.9) provides an additional example of hard exudates, which often become more prominent as retinal edema/serous detachment are reabsorbed.

Chorioretinal lesions from inflammations other than CMV retinitis may appear either active or atrophic.

Active chorioretinal infections occasionally occur along with CMV retinitis, and in some cases the former may be confused with or confound assessment of the latter. The differential diagnosis of CMV retinitis versus other conditions is discussed in the review by Holland and coworkers.¹¹ Probably the most commonly seen active chorioretinal lesions in patients with AIDS other than CMV retinitis are infiltrates from <

Grading Protocol - CMVR