Two-Field Photography Can Identify Patients with Vision-Threatening Diabetic Retinopathy a Screening Approach in the Primary Care Setting

Emerging Treatments and Technologies ORIGINAL ARTICLE

Two-Field Photography Can Identify Patients With Vision-Threatening Diabetic Retinopathy A screening approach in the primary care setting

CATHARINA STELLINGWERF, MD ina is photographed through a dilated PETER L.L.J. HARDUS, PHD pupil, and the photographs are evaluated JOHANNA M.M. HOOYMANS, PHD by an expert. This method of screening is very suitable for use in a primary care setting. Screening in a primary care setting could dramatically reduce the number of premature referrals to an ophthalmolo- OBJECTIVE — To compare the effectiveness of two 45° photographic fields per eye in the gist. It will favor accessibility of the screening for diabetic retinopathy with the routine ophthalmologist’s examination and to study screening and increase compliance. the effectiveness of visual acuity measurement in the detection of diabetic macular edema. The gold standard for evaluating dia- RESEARCH DESIGN AND METHODS — Type 1 and 2 diabetic patients without a betic retinopathy is the grading of stereo- known history of more than minimal retinopathy (n ϭ 469) had a routine examination by an scopic photographs of seven fields (6). ophthalmologist, including visual acuity measurement, indirect retinoscopy, and biomicros- Several studies have demonstrated that a copy. At the same time, two-field nonstereoscopic retinal photographs were made of both eyes reduced number of photographic fields and assessed in a masked fashion by a retinal specialist. The results were compared. can provide sufficient information in clinical trials and epidemiological studies RESULTS — The prevalence was 4.3% for vision-threatening retinopathy and 24% for any (7,8). However, for screening purposes, retinopathy. The sensitivity of two-field photography in identifying diabetic patients with sight- this was not evaluated (9). threatening retinopathy was 95% (specificity 99%) and sensitivity for detecting any retinopathy In the present study, the grading re- was 83% (specificity 88%). The percentage of referrals to an ophthalmologist was 6.2%. All sults of two central nonstereoscopic 45° patients with macular edema detected by biomicroscopy were classified as having vision- threatening retinopathy on the photographs. photographic fields were compared with the findings from indirect retinoscopy CONCLUSIONS — Two-field retinal photography is a promising alternative to the routine and slitlamp biomicroscopy performed ophthalmologist’s examination in the screening for diabetic retinopathy. Visual acuity measure- by an ophthalmologist. In a small study, ment is not a sensitive tool for the detection of macular edema. Screening for diabetic retinopathy the agreement of this technique and sev- using two-field retinal photography is feasible in a primary care setting and can substantially en-field stereo photography was satisfac- lower the number of ophthalmic referrals. tory (8). Standard photography does not pro- Diabetes Care 24:2086–2090, 2001 vide stereopsis. Therefore, macular thickening, which is important for the diagnosis of clinically significant macular iabetic retinopathy is currently the restoring vision (3,4). Therefore, it is vital edema, will be missed. Visual acuity is leading cause of blindness in the to detect retinopathy in time to prevent considered as an aid for detecting macular D western world in people aged the permanent and destroying effects of edema (10), in addition to the retinal pho- 20–65 years (1). Several studies indicate this disease. Regular fundus examinations tographs. We evaluated the effectiveness that severe visual loss by diabetic retinop- in mydriasis are recommended for all pa- of visual acuity measurements in the di- athy is caused to a large extent by lack of tients with diabetes (5). agnosis of macular edema. adequate screening (2). When not discov- Screening in the Netherlands is typi- ered before vision loss occurs, the disease cally retinoscopy through dilated pupils usually progresses severely, and at that performed by an ophthalmologist. An al- RESEARCH DESIGN AND point, laser treatment is rarely effective in ternative is retinal photography. The ret- METHODS — Consecutive patients ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● were recruited from a university hospital From the Department of Ophthalmology, University Hospital of Groningen, Groningen, the Netherlands. practice and a general practitioner labora- Address correspondence and reprint requests to C. Stellingwerf, MD, Department of Ophthalmology, tory over a 2-year period. Patients were University Hospital of Groningen, Postbus 30001, 9700 RB Groningen, The Netherlands. E-mail address: eligible for inclusion if they had type 1 [email protected]. diabetes (based on autoimmunity) for at Received for publication 1 June 2001 and accepted in revised form 6 September 2001. Abbreviations: ETDRS, Early Treatment Diabetic Retinopathy Study. least 5 years or type 2 diabetes (based on A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion insulin resistance) regardless of the dura- factors for many substances. tion of the diabetes. For the diagnosis of

2086 DIABETES CARE, VOLUME 24, NUMBER 12, DECEMBER 2001 Stellingwerf, Hardus, and Hooymans diabetes, the following World Health Or- Table 1—Definitions of retinopathy grades ganization criteria were used: fasting Ͼ blood glucose of 6.1 mmol/l or blood Grade Definition glucose of Ն11.1 mmol/l 2 h after a glucose load. If patients had been examined 1 No retinopathy previously, they were excluded if they 1.5 Only intraretinal hemorrhages had grade 3 retinopathy (see Table 1) or 2 Only microaneurysms more retinopathy in one or both eyes on 3 Microaneurysms and intraretinal hemorrhages, maximum 20 red dots their previous visit. In the general practi- 3.5 Grade 3 plus hard exudates or Ͼ20 red dots tioner laboratory, only type 2 diabetic pa- 4 Grade 3 plus Ͼ1 cotton wool spot or intraretinal microangiopathy or venous tients were screened. beading 5 Neovascularization or fibrosis in the optic disc (and/or elsewhere in the retina), Visual acuity preretinal hemorrhage, and/or vitreous hemorrhage Best visual acuity was measured on both 6 Not gradable eyes using a Snellen chart. In the university practice, visual acuity was measured after refraction; in the general practitioner the optic disc at the edge of the photo- system (12) (Table 1). A level of severity laboratory, distance correction glasses graph and a nasal field with 1 disc diam- for each person was assigned according to were worn for testing if the patient nor- eter at the temporal edge of the optic disc. the degree of retinopathy present in the mally used them, and a pinhole was also An ophthalmic photographer or a doc- more affected eye. We modified the used with the glasses. tor’s assistant performed the photogra- method for the use of retinoscopy. We phy; the doctor’s assistants were trained considered retinopathy grade 3.5 or Examination for 1 week by an ophthalmic photogra- higher in at least one eye as a reason for Both eyes were inspected using bio- pher. We used a Canon CF-60 UV retinal referral to an ophthalmologist, and we microscopy of the anterior segment to camera with Ektachrome professional defined this retinopathy as vision- determine the presence of iris neovascu- ISO 64 film (Kodak) and had them pro- threatening. Ungradable photographs of larization or significant cataract. Mydria- cessed in a photo laboratory. one or both eyes were also considered as a sis was achieved by 0.5% tropicamide and reason for referral. 2.5% phenylephrine; if the pupil diame- Grading ter was still Ͻ6 mm after 20 min, another The images were graded by a retinal spe- RESULTS — Of the 469 diabetic pa- drop of each medicine was instilled. The cialist in a masked fashion using an up- tients included, 231 were men and 238 grade of diabetic retinopathy was deter- right slide magnifier. The quality of the were women. Their age (mean Ϯ SD) was mined by indirect retinoscopy and by photographs was assessed as “good” (le- 51 Ϯ 17.7 years. Of these subjects, 311 slitlamp biomicroscopy. In addition, sions well discernible as present or not), had type 2 diabetes, and 158 had type 1 when clinically significant macular edema “fair” (lesions discernible with difficulty), diabetes; 47% used only insulin, 43% was present on biomicroscopy, it was or “insufficient” (presence of lesions not used only oral antidiabetic medication, noted. Macular edema was defined ac- discernible). Grades 1–5 were allocated 1% used both, and 9% were treated only cording to the Early Treatment Diabetic only if the quality was at least “fair” for with diet. More patients had findings of Retinopathy Study (ETDRS) (11): 1) both eyes and sufficient retinal area could no retinopathy with retinoscopy than thickening of the retina within 500 ␮mof be assessed on the photographs; if this with photographs (76 vs. 70%) (Table 2). the fovea and/or 2) hard exudates within was not the case, grade 6 was allocated. From six (1.3%) patients, the photo- 500 ␮m of the fovea if associated with Retinopathy was assessed according graphs of one eye (two patients) or both thickening of the adjacent retina and/or 3) to a modified Airlie House Classification (four patients) were not gradable. In the zone(s) of retinal thickening of one disc area or larger, any part of which is within Table 2—Grade of diabetic retinopathy with retinoscopy and retinal photographs one disc diameter of the fovea. This examination was done by one of six graduated ophthalmologists from the University Retinoscopy Retinal photography Hospital of Groningen. The patients in Grade Type 1 Type 2 Total Type 1 Type 2 Total the general practitioner laboratory were examined by one of the authors (C.S.). 1 97 (61) 258 (83) 355 (76) 84 (53) 242 (78) 326 (70) This examination, which includes indi- 1, 5 5 (3.2) 19 (6.1) 24 (5.1) 4 (2.5) 20 (6.4) 24 (5.1) rect retinoscopy and biomicroscopy, will 2 27 (17.1) 16 (5.1) 43 (9.2) 39 (25) 23 (7.4) 62 (13) henceforth be called retinoscopy. 3 16 (10.1) 10 (3.2) 26 (5.5) 17 (10.8) 11 (3.5) 28 (6.0) 3, 5 6 (3.8) 5 (1.6) 11 (2.3) 10 (6.3) 6 (1.9) 16 (3.4) Photography 4 5 (3.2) 2 (0.64) 7 (1.5) 2 (1.3) 2 (0.64) 4 (0.85) Two nonstereoscopic 45° photographs of 5 2 (1.3) 1 (0.32) 3 (0.64) 2 (1.3) 1 (0.32) 3 (0.64) each eye were made through dilated pu- 600006(1.9) 6 (1.3) pils as follows: the photograph was cen- Total 158 (100) 311 (100) 469 (100) 158 (100) 311 (100) 469 (100) tered on the fovea, with the nasal edge of Data are n (%).

DIABETES CARE, VOLUME 24, NUMBER 12, DECEMBER 2001 2087 Two-ﬁeld photography in diabetic retinopathy screening

Table 3—Test statistics for the detection of retinopathy by retinal photography compared on the photographs was 8.9% for the type with retinoscopy 1 diabetic patients and 4.8% for the type 2 diabetic patients; the overall referral rate Vision-threatening was 6.2%. retinopathy Any retinopathy CONCLUSIONS — The minimum Type 1 Type 2 Overall Type 1 Type 2 Overall sensitivity required for screening meth- Sensitivity 92.3 100 95.2 83.6 81.1 82.5 ods for diabetic retinopathy is often con- Specificity 98.6 99.3 99.1 76.3 92.1 87.7 sidered to be 80% (13). For both type 1 Positive predictive value 85.7 77.8 82.6 68.9 68.3 68.8 and type 2 diabetic patients, the sensitiv- Negative predictive value 99.3 100 99.8 88.1 96.0 93.9 ity of two-field retinal photography was Agreement (weighted ␬) 0.70 0.71 0.71 0.69 0.71 0.71 far above 80%; in fact, the sensitivity was 95% for the detection of sight-threatening Data are %, unless otherwise indicated. retinopathy in this group of patients with a low prevalence of vision-threatening following analysis, they were not in- photographs was good, with a weighted ␬ retinopathy. Few photographs were of included. Of the photographs, 80% were of 0.71. Exact agreement between assess- sufficient quality for grading, both in the assessed as “good” and 19% were “fair.” ment using retinoscopy and retinal pho- university hospital and in the general Altogether, 81% of the photographs made tographs was 80.8%. For 13.8% of the practitioner laboratory, where doctor’s at the general practitioner laboratory— patients, a higher grade of retinopathy assistants who had had only a short train- where only type 2 diabetic patients are was found with the photographs, and in ing in retinal photography made the screened—were assessed as “good,” 5.8% of the patients, more abnormalities photographs. In this predominantly Cau- whereas 2% were “insufficient.” In the were seen with retinoscopy. We checked casian population, the acceptable photo- university hospital, 77% of the type 2 di- whether the agreement between the grad- graphic rate was better than in other abetic patients had good photographs, ings with retinoscopy and retinal photo- studies using 45° photography. It may be and 0.6% of the patients had photographs graphs would be better when not taking that in a Northern European population, of insufficient quality. This shows that the into account the distinction made be- ocular characteristics favor this tech- photographs made by the doctor’s assis- tween intraretinal hemorrhages and micro- nique, whereas other ethnic groups and tants at the general practitioners’ labora- aneurysms. The weighted ␬ for retinoscopy populations may have larger numbers of tory were not of inferior quality compared compared with retinal photography was patients with poor dilation or lens or me- with those made at the hospital. 0.71 in the original grading and 0.73 for dia opacity. Forty-five degree photographs The prevalence of vision-threatening the grading without this distinction. are more susceptible than 30° photo- retinopathy (see RESEARCH DESIGN AND METH- Visual acuity was Ͻ0.8 in one or both graphs to anything that degrades image ODS for the definition) as determined by eyes in 120 (25.6%) patients. Macular quality because of the smaller details and retinoscopy was 4.5% (21 patients); using edema was detected with slitlamp biomi- difficulty in detecting lesions at the edges retinal photography, it was 4.9% (23 pa- croscopy in only nine of these patients. of the fields. tients). Table 3 lists test statistics for the The sensitivity of visual acuity measure- In their review of the effectiveness of detection of vision-threatening retinopa- ment for the detection of macular edema screening tests for diabetic retinopathy, thy and any retinopathy as determined by is therefore quite low (7.5%). Lowering Hutchinson et al. (14) showed that myd- retinal photographs compared with reti- the visual acuity limit to values Ͻ0.8 did riatic 45° retinal photographs read by dif- noscopy. not improve sensitivity. However, speci- ferent health care professionals mostly In addition, we analyzed the groups ficity was high (99.0%). In the other pa- reach a sensitivity Ͼ80%. This could not of types 1 and 2 diabetic patients sepa- tients, cataracts, macular degeneration, be reached by using direct or indirect rately. The prevalence of vision-threaten- glaucoma, or amblyopia could explain the ophthalmoscopy, even when used by ex- ing retinopathy was 2.3% in the group of subnormal visual acuity. All of the pa- perienced ophthalmologists. The number type 2 diabetic patients. The sensitivity of tients with macular edema determined by of photographs per eye was not evaluated. the retinal photograph grading was 100% retinoscopy had diabetic retinopathy We chose mydriatic photography because (specificity 99%). In the group of type 1 grade 3.5 or higher on the retinal photo- it has higher sensitivity in the detection of diabetic patients, the prevalence of vision- graphs. One patient had iris neovascular- vision-threatening retinopathy than the threatening retinopathy was higher, i.e., ization in one eye; he had grade 5 nonmydriatic method (14). 7.6%. The specificity for detecting these retinopathy determined by both retinos- We compared two-field 45° retinal pho- patients using retinal photographs was copy and photographs. tography to slitlamp biomicroscopy and 92% (specificity 98%). Vision-threaten- indirect ophthalmoscopy as a standard. ing retinopathy was not detected with ret- Referrals In clinical trials, seven-field stereopho- inal photographs in one patient, and he A total of 23 subjects (4.9%) with retinop- tography graded by more than one inde- was assessed as having grade 3 retinopa- athy grade 3.5 or higher in at least one eye pendent grader is the most reliable thy; peripheral exudates were missed on and 6 subjects (1.3%) with ungradable noninvasive gold standard (15). How- the photographs. photographs for at least one eye needed (if ever, slitlamp biomicroscopy in combina- The agreement between the gradings not included in this study) referral to an tion with indirect ophthalmoscopy is the with retinoscopy and those with retinal ophthalmologist. The referral rate based preferred and widely accepted method

2088 DIABETES CARE, VOLUME 24, NUMBER 12, DECEMBER 2001 Stellingwerf, Hardus, and Hooymans used by ophthalmologists when screen- care setting, however, best-corrected vi- is not unimportant. Most guidelines rec- ing for diabetic retinopathy. It is a sensi- sual acuity can be difficult and time- ommend screening patients without any tive method for detecting vision- consuming to measure. In addition, it is retinopathy twice yearly, whereas pa- threatening diabetic retinopathy (11). not very specific; it will detect cataracts tients with slight signs of retinopathy are Thus, it is a logical reference standard for and age-related macular degeneration, screened yearly. Like other studies (6,19), a screening procedure to be evaluated. which is not the purpose of the screening. we found a higher prevalence of back- Harding et al. (16) photographed three Finally, macular edema is preferably de- ground retinopathy with retinal photo- overlapping 45°fields per eye and used tected and treated before visual acuity loss graphs than with retinoscopy. Probably, slitlamp biomicroscopy as a reference occurs (19). In our study, visual acuity small abnormalities like one or two mi- standard as well. They found an 89% sen- measurement had low sensitivity in the croaneurysms are more easily detected on sitivity for the detection of vision- detection of macular edema. retinal photographs than with retinos- threatening diabetic retinopathy. All of the techniques using nonstereo- copy. It is likely that the retinal photo- The agreement in our study between retinal imaging have limitations in the de- graphs have a greater sensitivity in the two-field photography and slitlamp tection of macular edema. Besides detection of any retinopathy. Thus, reti- biomicroscopy is good, and it is better slitlamp biomicroscopy, oral fluorescein nal photography is an even more power- than that found by Verhoeven et al. (17), could be an alternative, because this ful tool for the screening of diabetic who made one photograph per eye (␬ screening method has good sensitivity in retinopathy. 0.41), but was comparable with results detecting macular edema. However, it is Conventional retinal photography is by Schachat et al. (18), who graded the much more time-consuming than retinal partly being replaced by digital imaging. individual characteristics of diabetic reti- photography and has a small risk of aller- The instantaneous availability and the nopathy using photography and biomi- gic reaction, which is a drawback when rapid transfer of the images are some of croscopy (␬ 0.56–0.79). The agreement using the technique on a large scale. the advantages of digital processing over between two-field and seven-field stereo- An alternative method for the detec- 35-mm film processing. At present, how- photography was reasonable to good in tion of macular edema is to use the pres- ever, 35-mm photographs reach a higher the study of Aldington et al. (8) (␬ varied ence of hard exudates on the photographs sensitivity in the detection of any retinop- from 0.44 to 0.77). within one disc diameter of the fovea as a athy and vision-threatening retinopathy Unfortunately, there is little unifor- parameter. This parameter was shown to because they have a higher resolution mity in the grading systems used when be sufficiently sensitive for the detection (20–22). In addition, the equipment nec- screening for diabetic retinopathy. The of macular edema because 98% of the pa- essary for digital imaging and processing definition of vision-threatening retinopa- tients with reduced vision (Ͻ6/9) due to is not yet available to most ophthalmol- thy often differs as well in the studies pub- macular edema have an exudate in at least ogy or diabetes practices, whereas 35-mm lished. The grading we chose was based one eye (13). Bresnick et al. (9), who used cameras are readily available. on the grading used in the larger studies, the ETDRS database (with a high preva- Screening for diabetic retinopathy like the ETDRS and the U.K. Prospective lence of vision-threatening retinopathy), can be done effectively with two non- Diabetes Study (UKPDS). We followed showed that the sensitivity for detecting stereoscopic retinal photographs per eye. the suggestions for modification for clinically significant macular edema is It is a very sensitive tool for detecting ret- screening purposes made by Klein et al. 94% for the criterion of any hard exudate inopathy in both type 1 and type 2 dia- (12), which does not grade the severity of within one disc diameter of the center of betic patients. When applied in the specific lesions separately. This grading the macula. To make the screening as sen- primary care setting, it can substantially differentiates between intraretinal hemor- sitive as possible without losing too much lower the number of ophthalmic referrals. rhages and microaneurysms, which can specificity, we propose referring to an However, before widespread application be difficult to distinguish with retinos- ophthalmologist those patients with at can be recommended, a larger multi- copy or on retinal photographs. The least one hard exudate in at least one eye center study should be conducted. Using agreement we found when not taking into (grade 3.5 and higher). In our series, we the presence of hard exudates as a param- account this distinction was almost the can conclude that although retinal thick- eter for the presence of macular edema is same, which means that in our series, the ening caused by macular edema could be more effective than using visual acuity confusion of hemorrhages and microan- overlooked on retinal photographs, all of screening. Further study is required on eurysms was small. the patients with macular edema had this issue. Harding et al. (16) noted that macular other characteristics of vision-threatening edema in particular was missed on the retinopathy as well and were referred to photographs. In our series, all of the pa- an ophthalmologist. With this referral Acknowledgments— This work was sup- tients with macular edema diagnosed on policy, visual acuity measurement may ported by grants from the University Hospital retinoscopy were graded on retinal pho- therefore not be indicated. Because this of Groningen. tographs as having vision-threatening study had only 23 patients with vision- retinopathy and were referred to an oph- threatening retinopathy, a larger study References thalmologist. Because macular edema can will be necessary on this issue. 1. Moss SE, Klein R, Klein BEK: The 14-year be missed on photographs, visual acuity Although the purpose of screening for incidence of visual loss in a diabetic pop- measurement is often recommended in diabetic retinopathy is the detection of vi- ulation. Ophthalmology 105:998–1003, addition to photography in the screen- sion-threatening retinopathy, the detec- 1998 ing for diabetic retinopathy. In a primary tion of minimal background retinopathy 2. Klein R: Barriers to prevention of vision

DIABETES CARE, VOLUME 24, NUMBER 12, DECEMBER 2001 2089 Two-ﬁeld photography in diabetic retinopathy screening

loss caused by diabetic retinopathy. Arch JC, Cole DR: A screening approach to the ficity of photography and direct ophthal- Ophthalmol 115:1073–1074, 1997 surveillance of patients with diabetes for moscopy in screening for sight threatening 3. Beetham WP: Visual prognosis of prolif- the presence of vision-threatening reti- eye disease: the Liverpool Diabetic Eye erating diabetic retinopathy. Br J Ophthal- nopathy. Ophthalmology 107:19–24, 2000 Study. BMJ 311:1131–1135, 1995 mol 47:611–619, 1963 10. Lee CM, Olk RJ: Modified grid laser pho- 17. Verhoeven S, Ballegooie E van, Crijns H, 4. Early Treatment Diabetic Retinopathy tocoagulation for diffuse diabetic macular Hylkema HA, Loeve AA, Casparie AF: Is Study (ETDRS) Research Group: Early edema: long term visual results. Ophthal- fundusfotografie zinvol bij de screening photocoagulation for diabetic retinopa- mology 98:1594–1602, 1991 op diabetische retinopathie bij patie¨nten thy: ETDRS report number 9. Ophthalmol- 11. Early Treatment Retinopathy Study met diabetes mellitus type II? Ned Tijdschr ogy 98 (Suppl.):766–785, 1991 Group: Photocoagulation for diabetic Geneesk 137:1713–1717, 1993 5. American College of Physicians, Ameri- macular edema: ETDRS report number 1. 18. Schachat AP, Hyman L, Leske MC, Con- can Diabetes Association and American Arch Ophthalmol 103:1796–1806, 1985 nell AM, Hiner C, Alexander J: Compari- Academy of Ophthalmology: Screening 12. Klein R, Klein BEK, Magli YL, Brothers RJ, son of diabetic retinopathy detection by guidelines for diabetic retinopathy. Ann Meuer SM, Moss SE, Davis MD: An alter- clinical examinations and photograph Intern Med 116:683–685, 1992 native method of grading diabetic reti- gradings. Arch Ophthalmol 111:1064– 6. Early Treatment Retinopathy Study Re- nopathy. Ophthalmology 93:1183–1187, 1070, 1993 search Group: Grading diabetic retinopa- 1986 19. Newsom R, Moate B, Casswell T: Screen- thy from stereoscopic color fundus 13. Moss SE, Klein R, Kessler SD, Richie KA: ing for diabetic retinopathy using digital photographs—an extension of the modi- Comparison between ophthalmoscopy fied Airlie House classification: ETDRS and fundus photography in determining colour photography and oral fluorescein report number 10. Ophthalmology 98 severity of diabetic retinopathy. Ophthal- angiography. Eye 14:579–582, 2000 (Suppl.):786–806, 1991 mology 92:62–67, 1985 20. George LD, Halliwell M, Hill R, Aldington 7. Lee VS, Kingsley RM, Lee ET, Lu M, Rus- 14. Hutchinson A, McIntosh A, Peters J, SJ, Lusty J, Dunstan F, Owens DR: A com- sell D, Asal NR, Bradford RH Jr, Wilkinsin O’Keefe C, Khunti K, Baker R, Booth A: parison of digital retinal images and 35 CP: The diagnosis of diabetic retinopathy: Effectiveness of screening and monitoring mm colour transparanies in detecting and ophthalmoscopy versus photography. tests for diabetic retinopathy—a system- grading diabetic retinopathy. Diabet Med Ophthalmology 10:1504–1512, 1993 atic review. Diabet Med 17:495–506, 15:250–253, 1998 8. Aldington SJ, Kohner EM, Meuer S, Klein 2000 21. Henricsson M, Karlsson C, Ekholm L, R, Sjolie AK, the EURODIAB IDDM Com- 15. Hooymans JMM: The course of diabetic ret- Kaikkonen P, Sellman A, Steffert E, Tyr- plications Study Group: Methodology for inopathy during treatment with continuous berg M: Colour slides or digital photogra- retinal photography and assessment of di- subcutaneous insulin infusion. Thesis. Uni- phy in diabetes screening—a comparison. abetic retinopathy: the EURODIAB IDDM versity of Groningen, the Netherlands, Acta Ophthalmol Scand 78:164–168, 2000 Complications Study. Diabetologia 38: Van Denderen, 1986 22. Koch Jensen P, Scherfig E: Resolution of 437–444, 1995 16. Harding SP, Broadbent DM, Neoh C, retinal digital colour images. Acta Oph- 9. Bresnick GH, Mukamel DB, Dickinson White MC, Vora J: Sensitivity and speci- thalmol Scand 77:526–529, 1999

2090 DIABETES CARE, VOLUME 24, NUMBER 12, DECEMBER 2001