Role of Diabetologist in Evaluating Diabetic Retinopathy
Objective: To evaluate the ability of diabetologists to screen diabetic patients for diabetic retinopathy. Research Design and Methods: Comparison of eye examination performed by diabetologists with direct ophthalmoscopy through an undilated pupil and by ophthalmologists through a dilated pupil with sevenfield stereoscopic fundus photography (gold standard). The study consisted of 67 insulin-dependent and noninsulin-dependent diabetic outpatients attending a diabetes clinic. Results: On the basis of fundus photography, patients were classified as having no or insignificant (30%), minimal (31%), moderate (24%), or severe (15%) retinopathy. The diabetologists and ophthalmologists performed similarly in their ability to classify severity of diabetic retinopathy accurately. When no or insignificant retinopathy (isolated microaneurysms only) was detected by examination, clinically significant retinopathy detected by fundus photography was highly unlikely (8 yr of clinical practice experience in diabetes and who routinely examine patients with direct ophthalmoscopy through undilated pupils. The endocrine fellows were either board certified or board eligible in internal medicine and were completing a clinical year of training in endocrinology. They had received both didactic and practical instruction in the examination and recognition of diabetic retinopathy. The ophthalmologists who performed the clinical examination (P.L.L., D.J.D., C.D.J.R.) are board-certified ophthalmologists who specialize in diseases of the retina with a special interest in diabetic retinopathy. Both ophthalmologists (P.L.L., T.M.T.) who graded the fundus photographs have extensive experience in research studies such as the ETDRS (8) and Diabetes Control and Complications Trial (14), which used similar stereoscopic fundus photography. In addition to the demographic data collected, each examiner completed a questionnaire noting the presence and number of microaneurysms and the presence of hemorrhages, hard exudates, circinate exudates, cotton-wool spots, and cataracts in each eye. In addition, the presence of macular or perimacular microaneurysms, hemorrhages or exudates (defined as >1 lesion within 1 -disk diam of the maculae), macular edema, vitreous hemorrhage, or retinal detachment and the presence and location of proliferative retinopathy were determined. The examiners were also asked whether they had achieved satisfactory visualization of the fundus, including the maculae, and the number of minutes that the examination required. Finally, the examiners were asked to summarize their findings into one of four possible diagnostic categories and recommend a time interval when ophthalmologic consultation would be appropriate based on the respective examiner's clinical judgment. The summary assessment for each patient was based on the grading of the more severely affected eye. The four diagnostic categories included 7) no or no significant retinopathy (few microaneurysms only); 2) minimal nonproliterative retinopathy; 3) moderately severe retinopathy including extensive nonproliferative retinopathy, preproliferative retinopathy, or macular edema; and 4) severe retinopathy including proliferative
27
SCREENING FOR DIABETIC RETINOPATHY
retinopathy, vitreous hemorrhage, or traction retinal detachment. The choices of referral included 7) no ophthalmologic examination required for at least 1 yr, 2) ophthalmologic consultation appropriate in 6-9 mo, 3) ophthalmologic examination necessary in next 2-3 mo, and 4) immediate ophthalmologic referral when appropriate. The questionnaires were completed by each examiner and each grader of the stereoscopic fundus photographs and nonmydriatic fundus photographs. Statistics. Data were entered and analyzed on the CLINFO system. Spearman's rank-correlation coefficients, x2-ana'yses, and the calculation of sensitivities and specificities were performed as indicated. Sensitivity was defined as true-positive readings divided by truepositive plus false-negative readings, and specificity was defined as true-negative readings divided by true-negative plus false-positive readings. Overall error rates were calculated as the number of cases in which an examiner missed the correct diagnosis as a percentage of the total number of patients examined. The serious error rates were calculated as the percentage of patients with moderate or severe retinopathy in whom the examiner underestimated the degree of retinopathy by two or more grades. These are the patients who are most likely to have an inappropriate delay in referral for proper treatment.
RESULTS Sixty-seven diabetic (38 type I, 29 type II) patients who fulfilled the eligibility criteria were examined by one of the staff diabetologists and had seven-field stereoscopic fundus photography completed. Sixty-five of the patients were examined by at least one ophthalmologist. The endocrine fellows examined 20 of the patients, and nonmydriatic photographs were obtained on 28 patients. Six of the patients were treated previously with laser therapy. Because such patients should be followed routinely by an ophthalmologist and are not suitable candidates for screening, their retinopathy examination results were analyzed and presented separately. Mean ages for type I and type II patients were 30.8 ±10.7 and 58.2 ± 10.9 yr, respectively. Most (24 of 29) of the type II patients were insulin treated. The duration of diabetes was 16.7 ± 9.6 and 11.8 ± 6.6 yr, respectively, for type I and type II patients. Hypertension was present in 15.8% of type I diabetic patients and 58.6% of type II diabetic patients. Stereoscopic fundus photograph findings. Sixtyseven sets of stereoscopic fundus photographs (61 without laser) were graded. The two fundus photography graders agreed on the category of retinopathy (based on the more severely involved eye) in 58 of 67 (86.6%) cases and disagreed by one grade in 9 of 67 (13.4%) cases. On the identification of specific lesions (10 different lesions graded in 130 eyes), the graders agreed on 1240 of 1300 items (95.4%). The frequency of disagreements for specific lesions ranged from 0 to 10%.
28
The most common disagreements involved microaneurysm count (10% disagreements), presence or absence of perimacular and macular involvement (9%), differentiation of cotton-wool spots as representing diabetic or hypertensive changes (9%), and differentiation of hard exudates from drusen (8%). All disagreements were resolved by consensus regrading of photographs by the two ophthalmologists. In three patients (including 1 with previous laser treatment), there were disagreements between the diagnosis of proliferative lesions and intraretinal microvascular abnormalities. Two of these cases were eventually classified as proliferative and one as intraretinal microvascular abnormalities with the aid of fluorescein angiography, which was available in two of three cases. The distribution of the category of retinopathy was 30% of patients with no or insignificant retinopathy, 31.3% with minimal nonproliferative retinopathy, 23.9% with extensive nonproliferative or preproliferative retinopathy or macular edema, and 14.9% with proliferative retinopathy. There was no significant difference in the distribution of retinopathy beween type I and type II diabetic patients. The frequencies of individual lesions in our patients are shown in Table 1. Diabetologists' examination results. The correlation (Spearman) of the diabetologists' classification of the 61 nonlaser-treated patients, with the classification based on the fundus photography, was r = 0.798 (P < 0.001). The diabetologists' classification agreed with the photographs in 42 of 61 (69%) patients; classifications were underestimated by one grade in 12 of 61 (20%), overestimated by one grade in 6 of 61 (10%), and underestimated by two grades in 1 of 61 (2%). Thus, the diabetologists' overall error rate was 19 of 61 (31%), and the serious error rate was 1 of 20 (5%). The sensitivity and specificity of the diabetologists' examination in detecting individual eye lesions are shown in Figs. 1 and 2. Most errors involved the inability to detect or quantitate the eye lesions (false-neg-
TABLE 1 Frequency of specific eye lesions Lesion
n eyes with finding
%*
Microaneurysms Hard exudates Circinate exudates Cotton-wool spots Perimacular plus macular involvement Macular edema Neovascular disk Neovascular elsewhere Vitreous hemorrhage Retinal detachment
98 44 13 31 57 25 3 15 4 0
73.7 33.1 9.8 23.3 42.9 18.8 2.3 11.3 3.0 0
Based on 7-field stereoscopic fundus photographs. *n = 133 for eyes evaluated; 1 eye not graded because of vitreous hemorrhage.
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
D.M. NATHAN AND ASSOCIATES
FIG. 1. Sensitivity of diabetologists (open bars), ophthalmologists (hatched bars), and nonmydriatic photography (stippled bars) in detecting individual eye lesions compared with 7-field stereoscopic fundus photography. NVD, neovascularization (proliferative retinopathy) within 1 disk-diam of disk; NVE, neovascularization disease elsewhere (>1 disk-diam from disk). See Table 1 for denominators for each eye lesion. *P < 0.05, ophthalmologists vs. diabetologists; +P < 0.001, ophthalmologists vs. diabetologists.
ative). The more severe retinal lesions went undetected more frequently. Neovascularization was missed in five of eight cases, and macular edema was not detected in any of the 22 eyes in which it was present. False-positive findings were less common, with specificity for the detection of all lesions except microaneurysms being >90%. The recommendation for ophthalmologic referral by the diabetologists strongly correlated with their classification of degree of retinopathy (r = 0.915, P < 0.001) and the recommendation based on grading of fundus photographs (r = 0.71, P < 0.001). The recommendation agreed with that derived from the fundus photograph results in 39 of 61 (64%) patients and disagreed in 22 of 61 (36%) patients. The diabetologist suggested earlier referral in 7 of 61 (11 %) patients and later referral in 13 of 61 (21 %) (Table 2). Only two recommendations (3%) differed by more than one grade. The amount of time required for the examination, including visual acuity testing, was ~3 min. Endocrinology fellows' examination results. The endocrinology fellows examined 18 of 61 nonlaser-treated patients. The correlation (Spearman) of their classification with that from the fundus photographs was r = 0.547 (P < 0.05). The fellows' classifications agreed with the stereoscopic photographs in 10 of 18 (56%) patients and were underestimated by one grade in 3 of 18 (1 7%), overestimated by one grade in 3 of 18 (17%), and underestimated by two grades in 2 of 18 (11 %) (an overall error rate of 8 of 18 [44%] and a serious error rate of 2 of 5 [40%]). In detecting individual eye lesions, the fellows' examination had lower sensitivity than the internists' ex-
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
amination for all lesions (detection of hard exudates and perimacular lesions different from diabetologists by x2analysis [P < 0.01], others not significantly different) and lower specificity for microaneurysms and hard exudates. The fellows detected none of the seven cases of cotton-wool spots, four cases of circinate exudates, five cases of macular edema, and five cases of neovascularization. The sensitivity and specificity of the fellows' classification of degree of severity of retinopathy were less than those of the diabetologists' examination, but this difference did not reach statistical significance (data not shown). The recommendation for ophthalmologic referral by the fellows correlated strongly with their classification of degree of retinopathy (r = 0.733, P < 0.001). The fellows' recommendation agreed with that made by the fundus photography reading in 8 of 18 (44%) patients. They suggested earlier referral in 6 of 18 (33%) patients and later referral in 4 of 18 (22%; Table 2). Ophthalmologists' examination results. The ophthalmologists' examination provided information similar to the diabetologists' examination. There were no significant differences in classification of severity of retinopathy compared with the diabetologists' assessment (P > 0.10). On the other hand, the ophthalmologists' examination was significantly more sensitive than the diabetologists' for the detection of specific eye lesions such as perimacular lesions (P < 0.05) and macular edema (P < 0.001). Their sensitivity was not significantly different from the diabetologists' for other lesions (Fig. 1). The ophthalmologists had fewer false-positive readings than other examiners, with a specificity of 5:97% for all lesions (Fig. 2). As with the other examiners, more severe lesions went undetected more frequently. The recommendation for referral by the ophthalmologists strongly correlated with their classification of de-
FIG. 2. Specificity of diabetologists, ophthalmologists, and nonmydriatic photographs compared with results from 7-field stereoscopic fundus photography. NVD, neovascularization (proliferative retinopathy) within 1 diskdiam of disk; NVE, neovascularization disease elsewhere (>1 disk-diam from disk).
29
SCREENING FOR DIABETIC RETINOPATHY
TABLE 2 Agreement of examiner's recommendations for referral with stereoscopic fundus photography results in nonlasertreated patients Disagreement
Examiner
Total
Agreement (%)
Later by 1 grade (%)
Earlier by 1 grade (%)
Later by >1 grade (%)
Earlier by >1 grade (%)
Diabetologist Fellow Ophthalmologist Nonmydriatic photograph
61 18 87 26
39 (64) 8(44) 49 (56) 18 (69)
13 (21) 2(11) 28 (32) 6(23)
7(11) 4(22) 2(2) 2(8)
2(3) 2(11) 8(9) 0
0 2(11) 0 0
More than 1 ophthalmologist examined 33 patients (see METHODS).
gree of retinopathy (r = 0.909, P < 0.001). The ophthalmologists' recommendation agreed with that made by reading of the fundus photographs in 49 of 87 (56%) patients. The ophthalmologists suggested earlier referral in 2 of 87 (2%) patients and later referral in 36 of 87 (41%; Jable 2). Nonmydriatic photography results. The nonmydriatic photograph readings were similar to the diabetologists' examination results. The correlation (Spearman) of the nonmydriatic photograph classification was r = 0.810 (P < 0.001) with the classification by stereoscopic fundus photography and r = 0.895 (P < 0.001) with the diabetologists' classification. The classification by nonmydriatic photograph agreed with that by stereoscopic photograph in 15 of 26 (58%) patients and was underestimated by one grade in 9 of 26 (35%) and overestimated by one grade in 2 of 26 (8%). The classification by nonmydriatic photographs agreed with the diabetologists' classification in 21 of 26 patients, was more accurate (based on stereoscopic fundus photographs as gold standard) in 1 of 26, and was less accurate in 4 of 26. The sensitivity and specificity of the nonmydriatic photographs in detecting individual eye lesions were again similar to those of the diabetologists' examination (Figs. 1 and 2). Severe lesions again went frequently undetected. Neovascularization was missed in three of four cases, and macular edema was not detected in any of the 12 eyes in which it was present. False-positive readings were less common than false-negative readings. The recommendation for ophthalmologic referral
based on readings of the nonmydriatic photographs strongly correlated with the classification grading of these photographs {r = 0.981, P < 0.001). The recommendation based on the readings of nonmydriatic photographs agreed with that based on stereoscopic photography in 18 of 26 (69%) patients and suggested earlier referral in 2 of 26 (8%) and later referral in 6 of 26 (23%; Table 2). Visual acuity. Thirty-four patients (including 4 lasertreated patients) had visual acuity evaluated by a handheld or wall-mounted Snellen chart. Acuity was recorded as either 20/30 or better in both eyes or worse than 20/30 in either eye (corrected). Seven of 34 patients who were tested had corrected acuity poorer than 20/30. Of these, 5 had severe retinopathy and 2 had moderate retinopathy. Also, 5 of the 7 had macular edema in one or both eyes. The sensitivity, specificity, and predictive value of abnormal visual acuity for moderate and severe retinopathy and macular edema are shown in Table 3. Laser-treated patients. The six patients with previous laser treatment all had severe retinopathy as classified by fundus photography. Both nonophthalmologists and ophthalmologists had difficulty in classifying these patients because of the inability to distinguish between regressed versus active proliferative disease and to identify lesions in the midst of laser scars. Note that the recommendations for ophthalmologic follow-up by all examiners would have led to a delay in referral compared with the recommendation based on fundus photography reading in - 5 0 % of the laser-treated patients.
TABLE 3 Sensitivity and specificity of visual acuity
For diagnosis of moderate and severe retinopathy (grades C and D) For diagnosis of macular edema
Sensitivity (%)
Specificity (%)
Positive predictive value (%)
Negative predictive value (%)
43.8
100.0
100.0
66.7
38.5
90.5
71.4
70.4
Based on 34 examinations (corrected visual acuity worse than 20/30, see RESULTS).
30
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
D.M. NATHAN AND ASSOCIATES
CONCLUSIONS We evaluated the ability of diabetologists to screen for diabetic retinopathy and make appropriate decisions with regard to the need for ophthalmologist consultation. Current American Diabetes Association (ADA) guidelines advise annual referral for routine eye evaluation from time of diagnosis of diabetes in patients >30 yr of age and after a 5-yr duration of diabetes in patients aged 12-30 yr (9). Although the level of training of the eye specialist is not stipulated, the ADA suggests that in referring patients, the primary-care physician should be guided by the expertise and qualifications of the eye doctor to perform the examinations described. Implicit in these guidelines is the assumption that nonophthalmologists cannot adequately screen diabetic patients; therefore, all patients require automatic ophthalmologic referral. Similar guidelines have been adopted by other agencies (15). With an estimated population of 6 million diagnosed diabetic individuals in the U.S., most of whom are >30 yr of age, the cost of full compliance with these guidelines (at an estimated $75/ophthalmologic examination) would be $450 million/yr (16). Adequate and timely retinal examinations are made essential by the risk of visual loss from proliferative retinopathy and macular edema and the availability of efficacious therapy. An accurate and cost-effective method of screening is desirable. In this study, we compared the accuracy of diabetologists, endocrinology fellows, ophthalmologists, and nonmydriatic photographs in the diagnosis of diabetic retinopathy compared with the gold standard of sevenfield stereoscopic fundus photography. The two highly trained diabetologists were able to identify accurately patients without significant retinopathy. Of 23 patients classified by the diabetologists as having no or insignificant retinopathy, only 1 of 23 (4.3%) patients had moderate retinopathy according to seven-field stereoscopic fundus photography, and none had severe retinopathy. When the diabetologist detected any lesions, the lesions were often accompanied by other more severe lesions that went undetected. Of the 24 patients classified by the diabetologists as having minimal retinopathy, 7 of 24 (29.2%) had moderate retinopathy noted on sevenfield stereoscopic fundus photography. Although the diabetologists missed all cases of macular edema and most cases of proliferative retinopathy, they did detect other accompanying lesions in all but 1 of these patients. The ability of nonophthalmologists to screen effectively and safely for severe retinopathy apparently is predicated on their ability to detect other lesions that accompany severe retinopathy, rather than an ability to detect the severe lesions themselves. Therefore, the detection of any retinopathy by a nonophthalmologist should prompt referral. The overall accuracy of classification by the ophthalmologists was similar to that of the diabetologists. Not
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
unexpectedly, the dilated examination allowed the ophthalmologists to visualize the macula better and detect more perimacular disease and macular edema. But, unexpectedly, like the diabetologists, the ophthalmologists missed many cases of proliferative retinopathy. Of 10 patients with proliferative retinopathy on stereoscopic photographs, 5 had previous laser photocoagulation therapy and 3 had neovascularization located in peripheral regions of the retina. The difficulties in evaluating peripheral and previously treated retina, in addition to the low magnification of indirect ophthalmoscopy, may have contributed to some of the errors in diagnosis. The diagnostic accuracy of the ophthalmologists in this study is similar to results from the San Antonio Heart Study (17); both indicate that ophthalmologists specializing in retinal diseases may miss a substantial number of cases of diabetic retinopathy compared with sevenfield fundus photography. Other studies (11,12) have indicated greater accuracy on the part of ophthalmologists. The explanation of differences between the results in these studies and ours are conjectural. The use of direct plus indirect ophthalmoscopy may improve accuracy. The results of the only previous study to compare the accuracy of nonophthalmologists and ophthalmologists in the diagnosis of diabetic retinopathy differed from this study (12). The most marked difference was a substantially lower error rate by ophthalmologists in the previous study. There are several differences in methodology that may have contributed to this discrepancy. The previous study selected 11 patients with eye findings representative of common types of diabetic retinal disease and asked the physicians to examine each patient and identify the eye lesions. It was conducted in a prepared setting that the authors described as optimal. This study was more realistic in that the patients were not selected and it was conducted in physicians' offices during regular office hours. The endocrine fellows were less accurate in the classification of retinopathy than the diabetologists. The small number of cases examined by the fellows limits the conclusions that can be drawn, but we suggest that the results reflect improvement in skills with greater experience. Nonmydriatic fundus photographs read by the diabetologists had similar sensitivity to direct ophthalmoscopy by a nonophthalmologist in the classification of retinopathy. These photographs detected few lesions not noted by the diabetologists and missed a similarly small number of lesions detected by the diabetologists on examination. This differs from two previous studies that suggested that nonmydriatic photographs improve detection of diabetic retinopathy compared to ophthalmoscopy (10,18). The reasons for this discrepancy are not clear but may relate to the variation among the small number of examiners performing the ophthalmoscopy in these studies. Based on our results, we suggest that
31
SCREENING FOR DIABETIC RETINOPATHY
nonmydriatic photographs are not a useful addition to ophthalmoscopic examination by a highly trained diabetologist. They may be useful as an alternative screening method in the absence of a properly trained examiner. This possibility warrants further study. Finally, we found that visual acuity, tested with a reading card, is a useful addition to ophthalmoscopic examination. Although normal acuity did not rule out diabetic retinopathy, including proliferative disease, reduced corrected acuity was associated with moderate to severe retinopathy in all cases in this study. We recommend that corrected visual acuity be checked in all diabetic patients and that abnormal corrected visual acuity be an indication for referral to an ophthalmologist for additional evaluation. Based on our findings, we suggest that patients with no retinopathy noted on direct ophthalmoscopic examination by a properly trained diabetologist and with corrected visual acuity better than 20/30 do not require ophthalmologic referral for at least the next year. Patients with more than insignificant retinopathy on direct ophthalmoscopy should be referred for seven-field stereoscopic fundus photographs, because additional potentially serious lesions are likely to be missed by both nonophthalmologists and ophthalmologists. Our study design does not allow us to determine whether a subset of patients with minimal retinopathy on direct ophthalmoscopy can also have ophthalmology referral deferred. These recommendations are similar to those provided by the National Diabetes Data Group (19). Our recommendations are based on nondilated examinations by two diabetologists experienced in direct ophthalmoscopy. Their results may not be applicable to other nonophthalmologists. The lower accuracy of the endocrine fellows suggests that greater experience may be required to ensure acceptable results. To implement these recommendations successfully, internists and pediatricians caring for patients with diabetes need to be trained in the evaluation of diabetic retinopathy by direct ophthalmoscopy and need to maintain these skills. A recent study has demonstrated the feasibility of training primary-care physicians to perform direct ophthalmoscopy through a dilated pupil with a high degree of accuracy (20). In addition to providing safe and effective screening of patients at risk and limiting the number of unnecessary referrals to ophthalmologists, these recommendations should lower the cost of evaluation for diabetic retinopathy. In our study population, 66 of 67 patients would have been referred under the ADA guidelines to an ophthalmologist at an estimated cost of $75/examination for a total cost of $4950. With our proposal, 23 of 67 patients would have been spared an ophthalmologic examination, and 44 would have been referred for fundus photographs (estimated cost of $100/examination) for a savings of —10%. In the U.S., with —5.4 million type II diabetic patients, a population-based study revealed that 53.4% of type II diabetic patients had no retinopathy (2). The cost of ophthalmologic ex-
32
aminations with automatic referral for all of these patients would be $405 million. We would recommend fundus photographs for —50% of these patients (most of those with retinopathy plus - 1 0 % false-positive readings) at a cost of $270 million or a savings of $135 million or 33%. The potential reduction in the cost of evaluation of retinopathy is substantial. If the cost of fundus photography could be reduced, savings would be even greater. We conclude that the recent recommendations by the ADA and other groups be reexamined. Although the applicability of our findings to the general medical community needs to be studied, we suggest that an alternate strategy for screening for diabetic retinopathy might be the following. First, appropriately trained diabetologists can make ophthalmology referral decisions for their patients with an acceptably low risk of error. Patients with no detectable retinopathy need not be referred to an ophthalmologist, whereas patients with any retinopathy should be referred for further evaluation. Second, patients with retinopathy detected by a diabetologist should be referred to an ophthalmologist for seven-field stereoscopic fundus photographs, because ophthalmologists may also miss significant lesions. Third, nonmydriatic fundus photographs do not improve the accuracy of classification of retinopathy over examination by an appropriately trained diabetologist. Their use as an alternative screening method warrants further study. Fourth, corrected visual acuity should be tested in all diabetic patients. Those with reduced acuity should be referred for further evaluation. Fifth, because the detection of eye lesions is more problematic in the setting of previous laser therapy, such patients should be followed by an ophthalmologist. Sixth, all practitioners who care for patients with diabetes should receive training in the detection of diabetic retinopathy. Finally, these recommendations may result in significant cost savings over current guidelines.
ACKNOWLEDGMENTS This work was supported by a grant from the Diabetes Association of Greater Fall River, Inc., and by the donation of equipment and supplies by Canon U.S.A., Inc., Lake Success, New York, and Polaroid Corporation, Cambridge, Massachusetts. We thank Kathy Hurxthal, RN, for technical assistance and Carol Bovest for secretarial assistance in preparing the manuscript. Parts of this study were presented at the 49th annual meeting of the American Diabetes Association, Detroit, Michigan, 3-6 June 1989.
REFERENCES 1. Kahn HA, Bradley RF: Prevalence of diabetic retinopathy: age, sex, and duration of diabetes. Br J Ophthalmol
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
D.M. NATHAN AND ASSOCIATES
59:345-49, 1975 Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. III. Prevalence and risk of diabetic retinopathy when age at diagnosis is 30 or more years. Arch Ophthalmol 102: 527-32, 1984 Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. III. Prevalence and risk of diabetic retinopathy when age at diagnosis is less than 30 years. Arch Ophthalmol 102:520-26, 1984 Palmberg PK, Smith M, Waltman S, Krupin T, Singer P, Burgess D, Wendtlant T, Achtenberg J, Cryer P, Santiago J, White N, Kilo C, Daughaday W: The natural history of retinopathy in insulin-dependent juvenile-onset diabetes. Ophthalmology 88:613-18, 1981 Frank RN, Hoffman WH, Podgor MJ, Joondeph HC, Lewis RA, Margherio RR, Nachagel DP Jr, Weiss H, Christopherson KW, Cronin MA: Retinopathy in juvenile-onset type I diabetes of short duration. Diabetes 31:874-82, 1982 Klein R, Klein BEK: Vision disorders in diabetes. In Diabetes in America. Bethesda, Maryland, U.S. Dept. of Health and Human Services, 1985 (NIH publ. no. 851468) The Diabetic Retinopathy Study Research Group: Preliminary report on effects of photocoagulation therapy. Am J Ophthalmol 81:383-96, 1976 Early Treatment Diabetic Retinopathy Study Research Group: Photocoagulation for diabetic macular edema: Early Treatment Diabetic Retinopathy Study Report Number 1. Arch Ophthalmol 103:1796-806, 1985 American Diabetes Association: Position statement: eye care guidelines for patients with diabetes mellitus. Diabetes Care 11:745-46, 1988 10. Klein R, Klein BEK, NeiderMW, Hubbard LD, MeuerSM, Brothers RJ: Diabetic retinopathy as detected using ophthalmology, a nonmydriatic camera and a standard fun-
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
dus camera. Ophthalmology 92:485-91, 1985 11. Moss SE, Klein R, Kessler SD, Richie KA: Comparison between ophthalmoscopy and fundus photography in determining severity of diabetic retinopathy. Ophthalmology 92:62-67, 1985 12. Sussman EJ, Tsiasas WG, Soper KA: Diagnosis of diabetic eye disease. JAMA 247:3231-34, 1982 13. Diabetic Retinopathy Study Research Group: Report no. 6: design, methods, and baseline results. Invest Ophthalmol Visual Sci 21:1-209, 1981 14. The Diabetes Control and Complications Trial Research Group: Color photography vs. fluorescein angiography in the detection of diabetic retinopathy in the Diabetes Control and Complications Trial. Arch Ophthalmol 105: 1344-51, 1987 15. Kentucky Diabetic Retinopathy Group: Guidelines for eye care in patients with diabetes mellitus. Arch Intern Med 149:769-70, 1989 16. Harris Ml: Prevalence of noninsulin-dependent diabetes and impaired glucose tolerance. In Diabetes In America. Bethesda, Maryland, U.S. Dept. of Health and Human Services, 1985 (NIH publ. no. 85-1468) 17. Valez R, Haffner S, Stern MP, Van Heuven WAJ: Ophthalmologist vs. retinal photographs in screening for diabetic retinopathy (Abstract). Clin Res 35:363A, 1987 18. Ryder REJ, Vora JP, Atiea JA, Owens DR, Hayes TM, Young S: Possible, new method to improve detection of diabetic retinopathy: Polaroid non-mydriatic retinal photography. Br Med I 291:1256-57, 1985 19. National Diabetes Advisory Board: Detection and Prevention of Visual Impairment in the Prevention and Treatment of Five Complications of Diabetes. Washington, DC, U.S. Dept. of Health and Human Services, 1983 (publ. no. 83-8392) 20. Baker S, Vallbona C, Hamill MB, Goetz B, West MS: A study of the proficiency of diabetic eye disease detection by primary care physicians (Abstract). Diabetes 39 (Suppl. 1):126A, 1990
33
Objective: To evaluate the ability of diabetologists to screen diabetic patients for diabetic retinopathy. Research Design and Methods: Comparison of eye examination performed by diabetologists with direct ophthalmoscopy through an undilated pupil and by ophthalmologists through a dilated pupil with sevenfield stereoscopic fundus photography (gold standard). The study consisted of 67 insulin-dependent and noninsulin-dependent diabetic outpatients attending a diabetes clinic. Results: On the basis of fundus photography, patients were classified as having no or insignificant (30%), minimal (31%), moderate (24%), or severe (15%) retinopathy. The diabetologists and ophthalmologists performed similarly in their ability to classify severity of diabetic retinopathy accurately. When no or insignificant retinopathy (isolated microaneurysms only) was detected by examination, clinically significant retinopathy detected by fundus photography was highly unlikely (8 yr of clinical practice experience in diabetes and who routinely examine patients with direct ophthalmoscopy through undilated pupils. The endocrine fellows were either board certified or board eligible in internal medicine and were completing a clinical year of training in endocrinology. They had received both didactic and practical instruction in the examination and recognition of diabetic retinopathy. The ophthalmologists who performed the clinical examination (P.L.L., D.J.D., C.D.J.R.) are board-certified ophthalmologists who specialize in diseases of the retina with a special interest in diabetic retinopathy. Both ophthalmologists (P.L.L., T.M.T.) who graded the fundus photographs have extensive experience in research studies such as the ETDRS (8) and Diabetes Control and Complications Trial (14), which used similar stereoscopic fundus photography. In addition to the demographic data collected, each examiner completed a questionnaire noting the presence and number of microaneurysms and the presence of hemorrhages, hard exudates, circinate exudates, cotton-wool spots, and cataracts in each eye. In addition, the presence of macular or perimacular microaneurysms, hemorrhages or exudates (defined as >1 lesion within 1 -disk diam of the maculae), macular edema, vitreous hemorrhage, or retinal detachment and the presence and location of proliferative retinopathy were determined. The examiners were also asked whether they had achieved satisfactory visualization of the fundus, including the maculae, and the number of minutes that the examination required. Finally, the examiners were asked to summarize their findings into one of four possible diagnostic categories and recommend a time interval when ophthalmologic consultation would be appropriate based on the respective examiner's clinical judgment. The summary assessment for each patient was based on the grading of the more severely affected eye. The four diagnostic categories included 7) no or no significant retinopathy (few microaneurysms only); 2) minimal nonproliterative retinopathy; 3) moderately severe retinopathy including extensive nonproliferative retinopathy, preproliferative retinopathy, or macular edema; and 4) severe retinopathy including proliferative
27
SCREENING FOR DIABETIC RETINOPATHY
retinopathy, vitreous hemorrhage, or traction retinal detachment. The choices of referral included 7) no ophthalmologic examination required for at least 1 yr, 2) ophthalmologic consultation appropriate in 6-9 mo, 3) ophthalmologic examination necessary in next 2-3 mo, and 4) immediate ophthalmologic referral when appropriate. The questionnaires were completed by each examiner and each grader of the stereoscopic fundus photographs and nonmydriatic fundus photographs. Statistics. Data were entered and analyzed on the CLINFO system. Spearman's rank-correlation coefficients, x2-ana'yses, and the calculation of sensitivities and specificities were performed as indicated. Sensitivity was defined as true-positive readings divided by truepositive plus false-negative readings, and specificity was defined as true-negative readings divided by true-negative plus false-positive readings. Overall error rates were calculated as the number of cases in which an examiner missed the correct diagnosis as a percentage of the total number of patients examined. The serious error rates were calculated as the percentage of patients with moderate or severe retinopathy in whom the examiner underestimated the degree of retinopathy by two or more grades. These are the patients who are most likely to have an inappropriate delay in referral for proper treatment.
RESULTS Sixty-seven diabetic (38 type I, 29 type II) patients who fulfilled the eligibility criteria were examined by one of the staff diabetologists and had seven-field stereoscopic fundus photography completed. Sixty-five of the patients were examined by at least one ophthalmologist. The endocrine fellows examined 20 of the patients, and nonmydriatic photographs were obtained on 28 patients. Six of the patients were treated previously with laser therapy. Because such patients should be followed routinely by an ophthalmologist and are not suitable candidates for screening, their retinopathy examination results were analyzed and presented separately. Mean ages for type I and type II patients were 30.8 ±10.7 and 58.2 ± 10.9 yr, respectively. Most (24 of 29) of the type II patients were insulin treated. The duration of diabetes was 16.7 ± 9.6 and 11.8 ± 6.6 yr, respectively, for type I and type II patients. Hypertension was present in 15.8% of type I diabetic patients and 58.6% of type II diabetic patients. Stereoscopic fundus photograph findings. Sixtyseven sets of stereoscopic fundus photographs (61 without laser) were graded. The two fundus photography graders agreed on the category of retinopathy (based on the more severely involved eye) in 58 of 67 (86.6%) cases and disagreed by one grade in 9 of 67 (13.4%) cases. On the identification of specific lesions (10 different lesions graded in 130 eyes), the graders agreed on 1240 of 1300 items (95.4%). The frequency of disagreements for specific lesions ranged from 0 to 10%.
28
The most common disagreements involved microaneurysm count (10% disagreements), presence or absence of perimacular and macular involvement (9%), differentiation of cotton-wool spots as representing diabetic or hypertensive changes (9%), and differentiation of hard exudates from drusen (8%). All disagreements were resolved by consensus regrading of photographs by the two ophthalmologists. In three patients (including 1 with previous laser treatment), there were disagreements between the diagnosis of proliferative lesions and intraretinal microvascular abnormalities. Two of these cases were eventually classified as proliferative and one as intraretinal microvascular abnormalities with the aid of fluorescein angiography, which was available in two of three cases. The distribution of the category of retinopathy was 30% of patients with no or insignificant retinopathy, 31.3% with minimal nonproliferative retinopathy, 23.9% with extensive nonproliferative or preproliferative retinopathy or macular edema, and 14.9% with proliferative retinopathy. There was no significant difference in the distribution of retinopathy beween type I and type II diabetic patients. The frequencies of individual lesions in our patients are shown in Table 1. Diabetologists' examination results. The correlation (Spearman) of the diabetologists' classification of the 61 nonlaser-treated patients, with the classification based on the fundus photography, was r = 0.798 (P < 0.001). The diabetologists' classification agreed with the photographs in 42 of 61 (69%) patients; classifications were underestimated by one grade in 12 of 61 (20%), overestimated by one grade in 6 of 61 (10%), and underestimated by two grades in 1 of 61 (2%). Thus, the diabetologists' overall error rate was 19 of 61 (31%), and the serious error rate was 1 of 20 (5%). The sensitivity and specificity of the diabetologists' examination in detecting individual eye lesions are shown in Figs. 1 and 2. Most errors involved the inability to detect or quantitate the eye lesions (false-neg-
TABLE 1 Frequency of specific eye lesions Lesion
n eyes with finding
%*
Microaneurysms Hard exudates Circinate exudates Cotton-wool spots Perimacular plus macular involvement Macular edema Neovascular disk Neovascular elsewhere Vitreous hemorrhage Retinal detachment
98 44 13 31 57 25 3 15 4 0
73.7 33.1 9.8 23.3 42.9 18.8 2.3 11.3 3.0 0
Based on 7-field stereoscopic fundus photographs. *n = 133 for eyes evaluated; 1 eye not graded because of vitreous hemorrhage.
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
D.M. NATHAN AND ASSOCIATES
FIG. 1. Sensitivity of diabetologists (open bars), ophthalmologists (hatched bars), and nonmydriatic photography (stippled bars) in detecting individual eye lesions compared with 7-field stereoscopic fundus photography. NVD, neovascularization (proliferative retinopathy) within 1 disk-diam of disk; NVE, neovascularization disease elsewhere (>1 disk-diam from disk). See Table 1 for denominators for each eye lesion. *P < 0.05, ophthalmologists vs. diabetologists; +P < 0.001, ophthalmologists vs. diabetologists.
ative). The more severe retinal lesions went undetected more frequently. Neovascularization was missed in five of eight cases, and macular edema was not detected in any of the 22 eyes in which it was present. False-positive findings were less common, with specificity for the detection of all lesions except microaneurysms being >90%. The recommendation for ophthalmologic referral by the diabetologists strongly correlated with their classification of degree of retinopathy (r = 0.915, P < 0.001) and the recommendation based on grading of fundus photographs (r = 0.71, P < 0.001). The recommendation agreed with that derived from the fundus photograph results in 39 of 61 (64%) patients and disagreed in 22 of 61 (36%) patients. The diabetologist suggested earlier referral in 7 of 61 (11 %) patients and later referral in 13 of 61 (21 %) (Table 2). Only two recommendations (3%) differed by more than one grade. The amount of time required for the examination, including visual acuity testing, was ~3 min. Endocrinology fellows' examination results. The endocrinology fellows examined 18 of 61 nonlaser-treated patients. The correlation (Spearman) of their classification with that from the fundus photographs was r = 0.547 (P < 0.05). The fellows' classifications agreed with the stereoscopic photographs in 10 of 18 (56%) patients and were underestimated by one grade in 3 of 18 (1 7%), overestimated by one grade in 3 of 18 (17%), and underestimated by two grades in 2 of 18 (11 %) (an overall error rate of 8 of 18 [44%] and a serious error rate of 2 of 5 [40%]). In detecting individual eye lesions, the fellows' examination had lower sensitivity than the internists' ex-
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
amination for all lesions (detection of hard exudates and perimacular lesions different from diabetologists by x2analysis [P < 0.01], others not significantly different) and lower specificity for microaneurysms and hard exudates. The fellows detected none of the seven cases of cotton-wool spots, four cases of circinate exudates, five cases of macular edema, and five cases of neovascularization. The sensitivity and specificity of the fellows' classification of degree of severity of retinopathy were less than those of the diabetologists' examination, but this difference did not reach statistical significance (data not shown). The recommendation for ophthalmologic referral by the fellows correlated strongly with their classification of degree of retinopathy (r = 0.733, P < 0.001). The fellows' recommendation agreed with that made by the fundus photography reading in 8 of 18 (44%) patients. They suggested earlier referral in 6 of 18 (33%) patients and later referral in 4 of 18 (22%; Table 2). Ophthalmologists' examination results. The ophthalmologists' examination provided information similar to the diabetologists' examination. There were no significant differences in classification of severity of retinopathy compared with the diabetologists' assessment (P > 0.10). On the other hand, the ophthalmologists' examination was significantly more sensitive than the diabetologists' for the detection of specific eye lesions such as perimacular lesions (P < 0.05) and macular edema (P < 0.001). Their sensitivity was not significantly different from the diabetologists' for other lesions (Fig. 1). The ophthalmologists had fewer false-positive readings than other examiners, with a specificity of 5:97% for all lesions (Fig. 2). As with the other examiners, more severe lesions went undetected more frequently. The recommendation for referral by the ophthalmologists strongly correlated with their classification of de-
FIG. 2. Specificity of diabetologists, ophthalmologists, and nonmydriatic photographs compared with results from 7-field stereoscopic fundus photography. NVD, neovascularization (proliferative retinopathy) within 1 diskdiam of disk; NVE, neovascularization disease elsewhere (>1 disk-diam from disk).
29
SCREENING FOR DIABETIC RETINOPATHY
TABLE 2 Agreement of examiner's recommendations for referral with stereoscopic fundus photography results in nonlasertreated patients Disagreement
Examiner
Total
Agreement (%)
Later by 1 grade (%)
Earlier by 1 grade (%)
Later by >1 grade (%)
Earlier by >1 grade (%)
Diabetologist Fellow Ophthalmologist Nonmydriatic photograph
61 18 87 26
39 (64) 8(44) 49 (56) 18 (69)
13 (21) 2(11) 28 (32) 6(23)
7(11) 4(22) 2(2) 2(8)
2(3) 2(11) 8(9) 0
0 2(11) 0 0
More than 1 ophthalmologist examined 33 patients (see METHODS).
gree of retinopathy (r = 0.909, P < 0.001). The ophthalmologists' recommendation agreed with that made by reading of the fundus photographs in 49 of 87 (56%) patients. The ophthalmologists suggested earlier referral in 2 of 87 (2%) patients and later referral in 36 of 87 (41%; Jable 2). Nonmydriatic photography results. The nonmydriatic photograph readings were similar to the diabetologists' examination results. The correlation (Spearman) of the nonmydriatic photograph classification was r = 0.810 (P < 0.001) with the classification by stereoscopic fundus photography and r = 0.895 (P < 0.001) with the diabetologists' classification. The classification by nonmydriatic photograph agreed with that by stereoscopic photograph in 15 of 26 (58%) patients and was underestimated by one grade in 9 of 26 (35%) and overestimated by one grade in 2 of 26 (8%). The classification by nonmydriatic photographs agreed with the diabetologists' classification in 21 of 26 patients, was more accurate (based on stereoscopic fundus photographs as gold standard) in 1 of 26, and was less accurate in 4 of 26. The sensitivity and specificity of the nonmydriatic photographs in detecting individual eye lesions were again similar to those of the diabetologists' examination (Figs. 1 and 2). Severe lesions again went frequently undetected. Neovascularization was missed in three of four cases, and macular edema was not detected in any of the 12 eyes in which it was present. False-positive readings were less common than false-negative readings. The recommendation for ophthalmologic referral
based on readings of the nonmydriatic photographs strongly correlated with the classification grading of these photographs {r = 0.981, P < 0.001). The recommendation based on the readings of nonmydriatic photographs agreed with that based on stereoscopic photography in 18 of 26 (69%) patients and suggested earlier referral in 2 of 26 (8%) and later referral in 6 of 26 (23%; Table 2). Visual acuity. Thirty-four patients (including 4 lasertreated patients) had visual acuity evaluated by a handheld or wall-mounted Snellen chart. Acuity was recorded as either 20/30 or better in both eyes or worse than 20/30 in either eye (corrected). Seven of 34 patients who were tested had corrected acuity poorer than 20/30. Of these, 5 had severe retinopathy and 2 had moderate retinopathy. Also, 5 of the 7 had macular edema in one or both eyes. The sensitivity, specificity, and predictive value of abnormal visual acuity for moderate and severe retinopathy and macular edema are shown in Table 3. Laser-treated patients. The six patients with previous laser treatment all had severe retinopathy as classified by fundus photography. Both nonophthalmologists and ophthalmologists had difficulty in classifying these patients because of the inability to distinguish between regressed versus active proliferative disease and to identify lesions in the midst of laser scars. Note that the recommendations for ophthalmologic follow-up by all examiners would have led to a delay in referral compared with the recommendation based on fundus photography reading in - 5 0 % of the laser-treated patients.
TABLE 3 Sensitivity and specificity of visual acuity
For diagnosis of moderate and severe retinopathy (grades C and D) For diagnosis of macular edema
Sensitivity (%)
Specificity (%)
Positive predictive value (%)
Negative predictive value (%)
43.8
100.0
100.0
66.7
38.5
90.5
71.4
70.4
Based on 34 examinations (corrected visual acuity worse than 20/30, see RESULTS).
30
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
D.M. NATHAN AND ASSOCIATES
CONCLUSIONS We evaluated the ability of diabetologists to screen for diabetic retinopathy and make appropriate decisions with regard to the need for ophthalmologist consultation. Current American Diabetes Association (ADA) guidelines advise annual referral for routine eye evaluation from time of diagnosis of diabetes in patients >30 yr of age and after a 5-yr duration of diabetes in patients aged 12-30 yr (9). Although the level of training of the eye specialist is not stipulated, the ADA suggests that in referring patients, the primary-care physician should be guided by the expertise and qualifications of the eye doctor to perform the examinations described. Implicit in these guidelines is the assumption that nonophthalmologists cannot adequately screen diabetic patients; therefore, all patients require automatic ophthalmologic referral. Similar guidelines have been adopted by other agencies (15). With an estimated population of 6 million diagnosed diabetic individuals in the U.S., most of whom are >30 yr of age, the cost of full compliance with these guidelines (at an estimated $75/ophthalmologic examination) would be $450 million/yr (16). Adequate and timely retinal examinations are made essential by the risk of visual loss from proliferative retinopathy and macular edema and the availability of efficacious therapy. An accurate and cost-effective method of screening is desirable. In this study, we compared the accuracy of diabetologists, endocrinology fellows, ophthalmologists, and nonmydriatic photographs in the diagnosis of diabetic retinopathy compared with the gold standard of sevenfield stereoscopic fundus photography. The two highly trained diabetologists were able to identify accurately patients without significant retinopathy. Of 23 patients classified by the diabetologists as having no or insignificant retinopathy, only 1 of 23 (4.3%) patients had moderate retinopathy according to seven-field stereoscopic fundus photography, and none had severe retinopathy. When the diabetologist detected any lesions, the lesions were often accompanied by other more severe lesions that went undetected. Of the 24 patients classified by the diabetologists as having minimal retinopathy, 7 of 24 (29.2%) had moderate retinopathy noted on sevenfield stereoscopic fundus photography. Although the diabetologists missed all cases of macular edema and most cases of proliferative retinopathy, they did detect other accompanying lesions in all but 1 of these patients. The ability of nonophthalmologists to screen effectively and safely for severe retinopathy apparently is predicated on their ability to detect other lesions that accompany severe retinopathy, rather than an ability to detect the severe lesions themselves. Therefore, the detection of any retinopathy by a nonophthalmologist should prompt referral. The overall accuracy of classification by the ophthalmologists was similar to that of the diabetologists. Not
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
unexpectedly, the dilated examination allowed the ophthalmologists to visualize the macula better and detect more perimacular disease and macular edema. But, unexpectedly, like the diabetologists, the ophthalmologists missed many cases of proliferative retinopathy. Of 10 patients with proliferative retinopathy on stereoscopic photographs, 5 had previous laser photocoagulation therapy and 3 had neovascularization located in peripheral regions of the retina. The difficulties in evaluating peripheral and previously treated retina, in addition to the low magnification of indirect ophthalmoscopy, may have contributed to some of the errors in diagnosis. The diagnostic accuracy of the ophthalmologists in this study is similar to results from the San Antonio Heart Study (17); both indicate that ophthalmologists specializing in retinal diseases may miss a substantial number of cases of diabetic retinopathy compared with sevenfield fundus photography. Other studies (11,12) have indicated greater accuracy on the part of ophthalmologists. The explanation of differences between the results in these studies and ours are conjectural. The use of direct plus indirect ophthalmoscopy may improve accuracy. The results of the only previous study to compare the accuracy of nonophthalmologists and ophthalmologists in the diagnosis of diabetic retinopathy differed from this study (12). The most marked difference was a substantially lower error rate by ophthalmologists in the previous study. There are several differences in methodology that may have contributed to this discrepancy. The previous study selected 11 patients with eye findings representative of common types of diabetic retinal disease and asked the physicians to examine each patient and identify the eye lesions. It was conducted in a prepared setting that the authors described as optimal. This study was more realistic in that the patients were not selected and it was conducted in physicians' offices during regular office hours. The endocrine fellows were less accurate in the classification of retinopathy than the diabetologists. The small number of cases examined by the fellows limits the conclusions that can be drawn, but we suggest that the results reflect improvement in skills with greater experience. Nonmydriatic fundus photographs read by the diabetologists had similar sensitivity to direct ophthalmoscopy by a nonophthalmologist in the classification of retinopathy. These photographs detected few lesions not noted by the diabetologists and missed a similarly small number of lesions detected by the diabetologists on examination. This differs from two previous studies that suggested that nonmydriatic photographs improve detection of diabetic retinopathy compared to ophthalmoscopy (10,18). The reasons for this discrepancy are not clear but may relate to the variation among the small number of examiners performing the ophthalmoscopy in these studies. Based on our results, we suggest that
31
SCREENING FOR DIABETIC RETINOPATHY
nonmydriatic photographs are not a useful addition to ophthalmoscopic examination by a highly trained diabetologist. They may be useful as an alternative screening method in the absence of a properly trained examiner. This possibility warrants further study. Finally, we found that visual acuity, tested with a reading card, is a useful addition to ophthalmoscopic examination. Although normal acuity did not rule out diabetic retinopathy, including proliferative disease, reduced corrected acuity was associated with moderate to severe retinopathy in all cases in this study. We recommend that corrected visual acuity be checked in all diabetic patients and that abnormal corrected visual acuity be an indication for referral to an ophthalmologist for additional evaluation. Based on our findings, we suggest that patients with no retinopathy noted on direct ophthalmoscopic examination by a properly trained diabetologist and with corrected visual acuity better than 20/30 do not require ophthalmologic referral for at least the next year. Patients with more than insignificant retinopathy on direct ophthalmoscopy should be referred for seven-field stereoscopic fundus photographs, because additional potentially serious lesions are likely to be missed by both nonophthalmologists and ophthalmologists. Our study design does not allow us to determine whether a subset of patients with minimal retinopathy on direct ophthalmoscopy can also have ophthalmology referral deferred. These recommendations are similar to those provided by the National Diabetes Data Group (19). Our recommendations are based on nondilated examinations by two diabetologists experienced in direct ophthalmoscopy. Their results may not be applicable to other nonophthalmologists. The lower accuracy of the endocrine fellows suggests that greater experience may be required to ensure acceptable results. To implement these recommendations successfully, internists and pediatricians caring for patients with diabetes need to be trained in the evaluation of diabetic retinopathy by direct ophthalmoscopy and need to maintain these skills. A recent study has demonstrated the feasibility of training primary-care physicians to perform direct ophthalmoscopy through a dilated pupil with a high degree of accuracy (20). In addition to providing safe and effective screening of patients at risk and limiting the number of unnecessary referrals to ophthalmologists, these recommendations should lower the cost of evaluation for diabetic retinopathy. In our study population, 66 of 67 patients would have been referred under the ADA guidelines to an ophthalmologist at an estimated cost of $75/examination for a total cost of $4950. With our proposal, 23 of 67 patients would have been spared an ophthalmologic examination, and 44 would have been referred for fundus photographs (estimated cost of $100/examination) for a savings of —10%. In the U.S., with —5.4 million type II diabetic patients, a population-based study revealed that 53.4% of type II diabetic patients had no retinopathy (2). The cost of ophthalmologic ex-
32
aminations with automatic referral for all of these patients would be $405 million. We would recommend fundus photographs for —50% of these patients (most of those with retinopathy plus - 1 0 % false-positive readings) at a cost of $270 million or a savings of $135 million or 33%. The potential reduction in the cost of evaluation of retinopathy is substantial. If the cost of fundus photography could be reduced, savings would be even greater. We conclude that the recent recommendations by the ADA and other groups be reexamined. Although the applicability of our findings to the general medical community needs to be studied, we suggest that an alternate strategy for screening for diabetic retinopathy might be the following. First, appropriately trained diabetologists can make ophthalmology referral decisions for their patients with an acceptably low risk of error. Patients with no detectable retinopathy need not be referred to an ophthalmologist, whereas patients with any retinopathy should be referred for further evaluation. Second, patients with retinopathy detected by a diabetologist should be referred to an ophthalmologist for seven-field stereoscopic fundus photographs, because ophthalmologists may also miss significant lesions. Third, nonmydriatic fundus photographs do not improve the accuracy of classification of retinopathy over examination by an appropriately trained diabetologist. Their use as an alternative screening method warrants further study. Fourth, corrected visual acuity should be tested in all diabetic patients. Those with reduced acuity should be referred for further evaluation. Fifth, because the detection of eye lesions is more problematic in the setting of previous laser therapy, such patients should be followed by an ophthalmologist. Sixth, all practitioners who care for patients with diabetes should receive training in the detection of diabetic retinopathy. Finally, these recommendations may result in significant cost savings over current guidelines.
ACKNOWLEDGMENTS This work was supported by a grant from the Diabetes Association of Greater Fall River, Inc., and by the donation of equipment and supplies by Canon U.S.A., Inc., Lake Success, New York, and Polaroid Corporation, Cambridge, Massachusetts. We thank Kathy Hurxthal, RN, for technical assistance and Carol Bovest for secretarial assistance in preparing the manuscript. Parts of this study were presented at the 49th annual meeting of the American Diabetes Association, Detroit, Michigan, 3-6 June 1989.
REFERENCES 1. Kahn HA, Bradley RF: Prevalence of diabetic retinopathy: age, sex, and duration of diabetes. Br J Ophthalmol
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
D.M. NATHAN AND ASSOCIATES
59:345-49, 1975 Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. III. Prevalence and risk of diabetic retinopathy when age at diagnosis is 30 or more years. Arch Ophthalmol 102: 527-32, 1984 Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. III. Prevalence and risk of diabetic retinopathy when age at diagnosis is less than 30 years. Arch Ophthalmol 102:520-26, 1984 Palmberg PK, Smith M, Waltman S, Krupin T, Singer P, Burgess D, Wendtlant T, Achtenberg J, Cryer P, Santiago J, White N, Kilo C, Daughaday W: The natural history of retinopathy in insulin-dependent juvenile-onset diabetes. Ophthalmology 88:613-18, 1981 Frank RN, Hoffman WH, Podgor MJ, Joondeph HC, Lewis RA, Margherio RR, Nachagel DP Jr, Weiss H, Christopherson KW, Cronin MA: Retinopathy in juvenile-onset type I diabetes of short duration. Diabetes 31:874-82, 1982 Klein R, Klein BEK: Vision disorders in diabetes. In Diabetes in America. Bethesda, Maryland, U.S. Dept. of Health and Human Services, 1985 (NIH publ. no. 851468) The Diabetic Retinopathy Study Research Group: Preliminary report on effects of photocoagulation therapy. Am J Ophthalmol 81:383-96, 1976 Early Treatment Diabetic Retinopathy Study Research Group: Photocoagulation for diabetic macular edema: Early Treatment Diabetic Retinopathy Study Report Number 1. Arch Ophthalmol 103:1796-806, 1985 American Diabetes Association: Position statement: eye care guidelines for patients with diabetes mellitus. Diabetes Care 11:745-46, 1988 10. Klein R, Klein BEK, NeiderMW, Hubbard LD, MeuerSM, Brothers RJ: Diabetic retinopathy as detected using ophthalmology, a nonmydriatic camera and a standard fun-
DIABETES CARE, VOL. 14, NO. 1, JANUARY 1991
dus camera. Ophthalmology 92:485-91, 1985 11. Moss SE, Klein R, Kessler SD, Richie KA: Comparison between ophthalmoscopy and fundus photography in determining severity of diabetic retinopathy. Ophthalmology 92:62-67, 1985 12. Sussman EJ, Tsiasas WG, Soper KA: Diagnosis of diabetic eye disease. JAMA 247:3231-34, 1982 13. Diabetic Retinopathy Study Research Group: Report no. 6: design, methods, and baseline results. Invest Ophthalmol Visual Sci 21:1-209, 1981 14. The Diabetes Control and Complications Trial Research Group: Color photography vs. fluorescein angiography in the detection of diabetic retinopathy in the Diabetes Control and Complications Trial. Arch Ophthalmol 105: 1344-51, 1987 15. Kentucky Diabetic Retinopathy Group: Guidelines for eye care in patients with diabetes mellitus. Arch Intern Med 149:769-70, 1989 16. Harris Ml: Prevalence of noninsulin-dependent diabetes and impaired glucose tolerance. In Diabetes In America. Bethesda, Maryland, U.S. Dept. of Health and Human Services, 1985 (NIH publ. no. 85-1468) 17. Valez R, Haffner S, Stern MP, Van Heuven WAJ: Ophthalmologist vs. retinal photographs in screening for diabetic retinopathy (Abstract). Clin Res 35:363A, 1987 18. Ryder REJ, Vora JP, Atiea JA, Owens DR, Hayes TM, Young S: Possible, new method to improve detection of diabetic retinopathy: Polaroid non-mydriatic retinal photography. Br Med I 291:1256-57, 1985 19. National Diabetes Advisory Board: Detection and Prevention of Visual Impairment in the Prevention and Treatment of Five Complications of Diabetes. Washington, DC, U.S. Dept. of Health and Human Services, 1983 (publ. no. 83-8392) 20. Baker S, Vallbona C, Hamill MB, Goetz B, West MS: A study of the proficiency of diabetic eye disease detection by primary care physicians (Abstract). Diabetes 39 (Suppl. 1):126A, 1990
33
