Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Glycemic Thresholds for Diabetes-Specific Retinopathy

Glycemic Thresholds for Diabetes-Specific Retinopathy Pathophysiology/Complications ORIGINAL ARTICLE Glycemic Thresholds for Diabetes-Specific Retinopathy Implications for diagnostic criteria for diabetes 1 6 STEPHEN COLAGIURI, MBBS JONATHAN E. SHAW, MD glycemic levels. The datasets used for this 1 7,8 CRYSTAL M.Y. LEE, PHD KNUT BORCH-JOHNSEN, DMSC purpose were from Pima Indians, an 2,3 TIEN Y. WONG, PHD THE DETECT-2 COLLABORATION WRITING Egyptian study, and unpublished data 4,5 BEVERLEY BALKAU, PHD GROUP* from the Third National Health and Nu- trition Examination Survey (NHANES) (2). OBJECTIVE — To re-evaluate the relationship between glycemia and diabetic retinopathy. Other studies (3–5) also have exam- ined this relationship, but the results have RESEARCH DESIGN AND METHODS — We conducted a data-pooling analysis of been inconsistent. All studies reported to nine studies from five countries with 44,623 participants aged 20 –79 years with gradable retinal date have had limited statistical power to photographs. The relationship between diabetes-specific retinopathy (defined as moderate or examine this relationship in detail and more severe retinopathy) and three glycemic measures (fasting plasma glucose [FPG; n have adopted a very broad definition of 41,411], 2-h post oral glucose load plasma glucose [2-h PG; n 21,334], and A1C [n 28,010]) retinopathy that included many cases of was examined. mild retinopathy, now known to have causes other than hyperglycemia (6). A RESULTS — When diabetes-specific retinopathy was plotted against continuous glycemic measures, a curvilinear relationship was observed for FPG and A1C. Diabetes-specific retinop- more clinically relevant end point is dia- athy prevalence was low for FPG6.0 mmol/l and A1C6.0% but increased above these levels. betes-specific retinopathy (moderate or Based on vigintile (20 groups with equal numbers) distributions, glycemic thresholds for dia- more severe levels of retinopathy) that is betes-specific retinopathy were observed over the range of 6.4 – 6.8 mmol/l for FPG, 9.8 –10.6 invariably attributed to hyperglycemia. mmol/l for 2-h PG, and 6.3– 6.7% for A1C. Thresholds for diabetes-specific retinopathy from Also different statistical methods have receiver-operating characteristic curve analyses were 6.6 mmol/l for FPG, 13.0 mmol/l for 2-h been used in previous studies, which has PG, and 6.4% for A1C. an important effect on derived cut points (5,7). CONCLUSIONS — This study broadens the evidence based on diabetes diagnostic criteria. Several new datasets with retinopathy A narrow threshold range for diabetes-specific retinopathy was identified for FPG and A1C but data have become available since the orig- not for 2-h PG. The combined analyses suggest that the current diabetes diagnostic level for FPG could be lowered to 6.5 mmol/l and that an A1C of 6.5% is a suitable alternative diagnostic inal studies used to derive current diabe- criterion. tes diagnostic cut points (1,2). The DETECT-2 collaboration has pooled Diabetes Care 34:145–150, 2011 these datasets to examine and re-evaluate the relationship between retinopathy and he current diagnostic cut points for tially increased risk of diabetes-associated three glycemic measures: FPG, 2-h PG, diabetes (fasting plasma glucose microvascular complications, particularly and A1C. The size of the DETECT-2 data- [FPG] of 7.0 mmol/l and 2-h post retinopathy, above these levels (1,2). set has allowed us to focus on the relation- oral glucose load plasma glucose [2-h PG] These cut points were derived from cross- ship between measures of glycemia and of 11.1 mmol/l) are largely based on gly- sectional epidemiological studies that ex- diabetes-specific retinopathy (i.e., mod- cemic levels associated with a substan- amined retinopathy across a range of erate or more severe levels of retinopa- thy). These analyses were designed to ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● inform current deliberations on possible From the Boden Institute of Obesity, Nutrition, and Exercise, University of Sydney, Sydney, Australia; the 2 3 revisions to the diagnostic criteria for Center for Eye Research Australia, University of Melbourne, Melbourne, Australia; the Singapore Eye Research Institute, National University of Singapore, Singapore; the Institut National de la Sante´ etdela diabetes. Recherche Me ´ dicale, Centre de Recherche en Epide ´ miologie et Sante ´ des Populations, Epidemiology of Diabetes, Obesity, and Chronic Kidney Disease Over the Lifecourse, Villejuif, France; Centre de Recher- che en Epide ´ miologie et Sante ´ des Populations, Universite ´ Paris Sud, Villejuif, France; the Heart and RESEARCH DESIGNS AND Diabetes Institute, Baker International Diabetes Institute, Melbourne, Australia; the Steno Diabetes 8 METHODS — The DETECT-2 project Center, Gentofte, Denmark; and the Faculty of Health Science, University of Aarhus, Aarhus, Denmark. is an international data-pooling collabo- Corresponding author: Stephen Colagiuri, stephen.colagiuri@sydney.edu.au. Received 23 June 2010 and accepted 8 October 2010. Published ahead of print at http://care. ration. The primary objective of the col- diabetesjournals.org on 26 October 2010. DOI: 10.2337/dc10-1206. laboration was to examine aspects of *A complete list of the members of the DETECT-2 Collaboration Writing Group are listed in the online screening for type 2 diabetes and im- appendix available at http://care.diabetesjournals.org/cgi/content/full/dc10-1206/DC1. paired glucose tolerance across various This article was prepared using limited-access datasets obtained from the NHLBI and does not necessarily reflect the opinions or views of MESA or the NHLBI. populations and ethnic groups. Details of © 2011 by the American Diabetes Association. Readers may use this article as long as the work is properly the collaboration are reported elsewhere cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons. (8,9). For the current analysis, studies in- org/licenses/by-nc-nd/3.0/ for details. cluded in the DETECT-2 database, in The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. which retinopathy data had been col- care.diabetesjournals.org DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 145 Glycemic thresholds for diabetic retinopathy Table 1—Summary of studies included in these analyses Name of study Country Year Age range n* Measures available ARIC (10) U.S. 1993–1995 (visit 3) 49–73 10,873 FPG AusDiab (5) Australia 1999–2000 25–90 2,052 FPG, 2-h PG, A1C BMES (11) Australia 1992–1994 45–97 2,915 FPG CURES (12) India 2002–2004 (phase III) 20–85 2,200 FPG, 2-h PG, A1C Hiroshima study (4) Japan 1990–2004 17–99 12,873 FPG, 2-h PG, A1C MESA (14) U.S. 2002–2004 (second examination) 45–85 5,920 FPG, A1C NHANES III (15) U.S. 1988–1994 40–74 2,869 FPG, 2-h PG, A1C Pima Indian study (17) U.S. 1982 (first examination) 15–85 1,829 FPG, 2-h PG SiMES (18) Singapore 2004 40–79 3,170 A1C *Number of participants aged 20 –79 years included in the analysis. lected, were invited to provide these data thy, which we defined as moderate or of the authors (T.Y.W) was personally in- for this analysis. Additional studies with more severe levels of retinopathy. volved in the grading of retinopathy using retinopathy data identified by coinvesti- All nine studies measured plasma glu- the modified Early Treatment Diabetic gators through personal contact or litera- cose, and six studies that measured A1C Retinopathy Study (Atherosclerosis Risk ture search also were invited to contribute used high-performance liquid chroma- in Communities [ARIC], AusDiab, Blue datasets. Retinopathy data were available tography, of which five used a Diabetes Mountains Eye Study [BMES], MESA, and from 12 studies in eight countries Control and Complications Trial SiMES); 3) studies in which participants (4,5,7,10 –18). This analysis focuses on (DCCT)-aligned assay (5,12,14,15,18). were predominantly Caucasian (ARIC, nine studies from five countries that had AusDiab, BMES, MESA, and NHANES retinopathy data by grading. Participants Statistical analysis III); 4) studies in which participants were aged 20 –79 years, including those with Prevalence of diabetes-specific retinopa- Asian (CURES, Hiroshima study, and known diabetes and with gradable retinal thy was examined by 1) 0.5-unit intervals SiMES); and 5) studies in which partici- photographs and at least one measure of of glycemic measures and 2) vigintiles (di- pants had all three measures of glycemia. glycemia, were included. All studies were viding participants into 20 equally sized All statistical analyses were performed us- approved by respective institutional re- groups) of the distribution for each mea- ing SAS 9.1 for Windows (SAS Institute, view boards and were conducted accord- sure of glycemia. Logistic regression mod- Cary, NC) and SPSS 16.0 for Windows ing to the Declaration of Helsinki. els were applied to test the relationships (SPSS, Chicago, IL). between diabetes-specific retinopathy Classification of retinopathy and glycemia by 0.5-unit intervals and by RESULTS The retinal photograph grading was per- vigintiles of each glycemic measure, with formed by individual study centers. Reti- the lowest range as the reference. The Study participants nopathy was classified as present or analyses were repeated after adjusting for In total, 44,623 participants had informa- absent for initial analysis. Where data study center. tion on both the presence and severity of were available, those with retinopathy The discriminatory power of each retinopathy (Table 1). A total of 1,589 were further classified as those having measure of glycemia for retinopathy was participants had minimal NPDR, 762 had minimal nonproliferative diabetic reti- assessed as the area under the receiver- mild NPDR, 430 had moderate NPDR, 50 nopathy (NPDR), mild NPDR, moderate operating characteristic (ROC) curve had severe NPDR, and 171 had PDR. The NPDR, severe NPDR, or proliferative dia- (AUC). An AUC of 1 indicates perfect dis- number of participants available for each betic retinopathy (PDR) based on the in- criminatory power and an AUC of 0.5 in- measure of glycemia was 41,334 for FPG, formation provided by individual studies dicates that the discrimination is no better 21,334 for 2-h PG, and 27,933 for A1C. using the modified Airlie House classifi- than chance. ROC curve analyses were Of these, 27,445 participants had at least cation levels (19), modified Early Treat- used to examine thresholds based on op- two measures and 18,533 participants ment Diabetic Retinopathy Study levels timizing sensitivity and specificity. The had all three measures. The characteris- (20), or the Fukuda standard (21). Lev- impact of various thresholds on the prev- tics of participants by study are shown els 14 –20 indicate minimal NPDR, levels alence of diabetes was examined by ap- in supplementary Table 1 in the on- 30 –35 or the Fukuda standard A1 indi- plying these values to 16,381 participants line appendix (available at http://care. cate mild NPDR, levels 40 – 47 or the without known diabetes who had all three diabetesjournals.org/cgi/content/full/dc10- Fukuda standard A2 indicate moderate measures of glycemia. 1206/DC1). NPDR, levels 50 –53 or the Fukuda stan- Sensitivity analysis was performed on dard A3 indicate severe NPDR, and levels 1) studies in which a DCCT-aligned assay Prevalence of retinopathy 60 –90 or the Fukuda standards A4 and for A1C was used (AusDiab, Chennai Ur- The overall prevalence of any retinopathy B1–B4 indicate PDR. The final retinopa- ban Rural Epidemiological Study was 6.7% and 1.5% for diabetes-specific thy grading for each participant was based [CURES], Multi-Ethnic Study of Athero- retinopathy. In people with known diabe- on the diagnosis in the more severely af- sclerosis and Air Pollution [MESA], tes, the prevalence of diabetes-specific fected eye. The primary outcome used in NHANES III, and Singapore Malay Eye retinopathy was 9.4%, in newly diag- this study was diabetes-specific retinopa- Study [SiMES]); 2) studies in which one nosed diabetes 1.0%, in impaired glucose 146 DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 care.diabetesjournals.org Colagiuri and Associates Figure 1—Prevalence of diabetes-specific retinopathy (moderate or more severe retinopathy) with 95% confidence intervals, number of retinopathy cases, and participants within each interval by 0.5 unit intervals for FPG and 2-h PG, and A1C. tolerance (1) 0.1%, in impaired fasting retinopathy. These plots suggest a curvi- FPG category of 6.0 – 6.4 mmol/l and glucose (1) 0.1%, and with normal glu- linear relationship for FPG and A1C and from the A1C category of 6.0 – 6.4%. The cose tolerance 0.1%. retinopathy. Diabetes-specific retinopa- curve for 2-h PG was flatter than for FPG Figure 1 shows the prevalence of ret- thy was virtually absent (prevalence and A1C, and no definite interval of in- inopathy by 0.5-unit intervals for each 0.4%) at low levels for each glycemic crease for 2-h PG was obvious. measure of glycemia for diabetes-specific measure but began to increase from the Logistic regression adjusted for study care.diabetesjournals.org DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 147 Glycemic thresholds for diabetic retinopathy Figure 2—Prevalence of diabetes-specific retinopathy (moderate or more severe retinopathy) by vigintiles of the distribution of FPG, 2-h PG, and A1C. center showed that the first interval where 9.8 –10.6 mmol/l; OR 10.1 [95% CI 1.3– mmol/l for 2-h PG, and 6.4% for A1C (Ta- the odds ratio (OR) for diabetes-specific 79.4]; P  0.03), from the 17th vigintile ble 2). These thresholds gave similar val- retinopathy was significantly different for FPG (6.4 – 6.8 mmol/l; 2.5 [1.2–5.2]; ues for positive and negative predictive from the reference FPG level of 4.0 – 4.4 P  0.01), and from the 18th vigintile for values. If these thresholds were used for mmol/l was 6.5– 6.9 mmol/l (OR 6.0 A1C (6.3– 6.7%; 4.5 [1.4 –15.2]; P  diagnosing diabetes, the prevalence of [95% CI 2.1–17.1]; P  0.01). The cor- 0.01). newly diagnosed diabetes would be responding result for A1C was 6.5– 6.9% Supplementary Table 2 in the online 11.9, 8.0, and 6.3% according to FPG, (16.8 [2.3–123.7]; P  0.01) compared appendix shows the ROC curve analyses. 2-h PG, and A1C, respectively. The dif- with an A1C of 4.0 – 4.4%. The overall discriminatory power deter- ferences in performance based on ROC Figure 2 shows the prevalence of dia- mined by AUCs was uniformly high for curve statistics for the three measures of betes-specific retinopathy by vigintiles of diabetes-specific retinopathy for each glycemia were minor for threshold val- the glycemic distributions. The preva- measure of glycemia (0.87 [95% CI 0.85– ues around the above values (supple- lence of diabetes-specific retinopathy was 0.89] for FPG, 0.89 [0.87– 0.91] for 2-h mentary Table 2). very low until the 15th vigintile for 2-h PG, and 0.90 [0.88 – 0.92] for A1C). The Sensitivity analyses showed that the PG (vigintile range 9.8 –10.6 mmol/l) and overlapping CIs suggests that there is no five studies in which T.Y.W used the same until the 17th vigintile for FPG (6.4 – 6.8 statistical difference between the three retinopathy grading system or the five mmol/l) and for A1C (6.1– 6.2%). measures of glycemia. The performance studies that used DCCT-aligned assays for Logistic regression models adjusted of a wide range of thresholds was exam- A1C measurements provided similar re- for study center confirmed a statistically ined, with particular attention to those sults to the overall study. The optimal significant difference in the OR for diabe- that overlapped from the continuous and threshold for FPG was 6.4 – 6.5 mmol/l, tes-specific retinopathy compared with vigintile distribution plots. The thresh- for A1C 6.4 – 6.5%, and for 2-h PG 10.1– the first vigintile that occurred from the olds that optimized sensitivity and speci- 11.2 mmol/l. 15th vigintile for 2-h PG (vigintile range ficity were 6.6 mmol/l for FPG, 13.0 CONCLUSIONS — The current di- Table 2—Threshold ranges for diabetes-specific retinopathy (moderate NPDR or more severe agnostic criteria for diabetes were derived retinopathy) derived from logistic regression models (adjusted for center) of the glycemic from analyses of the relationship between measures by continuous distribution and vigintile distribution and ROC curve analysis retinopathy and measures of glycemia (1). Our study is the largest to examine this association, using data from 45,000 FPG (mmol/l) 2-h PG (mmol/l) A1C (%) participants from five countries, and pro- Logistic regression vides the statistical power for a more de- Continuous distribution 6.5–6.9 No threshold 6.5–6.9 tailed and precise examination of Vigintile distribution 6.4–6.8 9.8–10.6 6.3–6.7 glycemic thresholds for diabetes-specific ROC curve analysis 6.6 13.0 6.4 retinopathy (moderate nonproliferative 148 DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 care.diabetesjournals.org Colagiuri and Associates and more severe retinopathy). Previous the vigintile distribution, continuous itations. First, this study was based on studies (7,16) have only reported the as- plots, and ROC curve analyses suggest cut cross-sectional data, whereas diagnostic sociation of glycemic measures with any point values of 6.5 mmol/l for FPG and thresholds would ideally be informed by retinopathy, which is less specific for hy- 6.5% for A1C, which could be considered incidence data of diabetes complications. perglycemia and is very frequently de- in deliberations on modifying the current Second, the methods used to assess and tected in people without diabetes. diagnostic criteria for diabetes. The re- classify retinopathy differed between The association between glycemic sults for 2-h PG were too inconsistent to studies, and it was not possible to inde- measures and retinopathy has tradition- consider modifying the current diagnos- pendently review the grading of all ally been investigated by plotting the tic cut point of 11.1 mmol/l. photographs. Nevertheless, inter- and in- prevalence of retinopathy against the de- It should be noted that these values traobserver consistency for retinopathy in cile distribution (the population divided do not result in equivalent estimates for the different studies was of the order of into 10 equal groups) of each glycemic prevalent diabetes. This has been an on- 80 –98% (3,10,15) and misclassification, measure (1,2). Our large dataset allows going issue with the current diagnostic especially for moderate or more severe analysis using vigintile distributions (the criteria, whereby using FPG alone or an forms of retinopathy, is likely to be mini- population divided into 20 equal groups), oral glucose tolerance test to diagnose di- mal but cannot be entirely eliminated. which narrows the glycemic range of each abetes gives different diabetes prevalence Furthermore, analysis of the studies in group. Based on logistic regression analy- (23). From our data (supplementary Ta- which T.Y.W. was involved in the stan- sis of these vigintile distributions, glyce- ble 2), lowering the FPG to 6.5 mmol/l dardized grading of retinal photographs mic thresholds for diabetes-specific would result in a diabetes prevalence of showed cut points for FPG and A1C sim- retinopathy were observed in the range of 13.0% based on FPG alone and 18.6% ilar to our entire study cohort. Third, no 6.4 – 6.8 mmol/l for FPG, 9.8 –10.6 based on an oral glucose tolerance test quality assurance of measures of glycemia mmol/l for 2-h PG, and 6.3– 6.7% for using an FPG of 6.5 mmol/l or a 2-h PG of could be applied across the studies. Nev- A1C (Table 2). 11.1 mmol/l. The prevalence of diabetes ertheless, all studies measured A1C using The large size of this dataset enables defined by an A1C of 6.5% is consider- high-performance liquid chromatogra- diabetes-specific retinopathy to be plot- ably lower (5.7%). This discrepancy in phy, and analysis of the five studies that ted against measures of glycemia as a prevalence may be problematic for epide- used a DCCT-aligned assay showed an continuous variable. A curvilinear rela- miological studies but is not necessarily a A1C cut point of 6.4 – 6.5%. Fourth, the tionship was observed, especially for FPG disadvantage for individual patient care. Hiroshima study, with its large sample and A1C, as opposed to the linear associ- An A1C of 6.5% was associated with a size, and the Pima Indian study, with its ation observed between blood pressure higher sensitivity and specificity than an high prevalence of diabetes-specific reti- and cardiovascular disease. Diabetes- FPG of 6.5 mmol/l and a higher specificity nopathy, may have influenced the results. specific retinopathy was rare at low levels than a 2-h PG of 11.1 mmol/l. In other However, sensitivity analyses that ex- of glycemia but increased from a range of words, fewer people would be identified cluded these two studies did not alter the 6.0 – 6.4 mmol/l for FPG and 6.0 – 6.4% as having diabetes, but this would not overall results. Finally, not all included for A1C. A threshold for increasing reti- compromise the identification of people studies were randomly sampled popula- nopathy was less obvious for 2-h PG, with diabetes-specific retinopathy. tions (e.g., MESA) and some (e.g., Aus- probably related to the smaller number of Whether this would have any deleterious Diab) oversampled people with diabetes study participants with this measure and ramifications in relation to identifying in- and/or prediabetes. Common to all such diabetes-specific retinopathy. Change dividuals at increased risk of other micro- analyses is the issue of whether to include point analyses, which were used previ- vascular or macrovascular disease people with previously diagnosed diabe- ously in two population-based studies remains to be determined. tes. If people with known diabetes cur- (22), were applied to these curves in an This study necessarily included pop- rently receiving blood glucose–lowering attempt to identify statistically significant ulations from different countries with var- treatment are included, the population- thresholds, but we were unable to dem- ious racial/ethnic backgrounds. There based characteristics of the study sample onstrate a clear threshold for any glycemic have been reports of differences in A1C are maintained, but a bias associated with measure by this method. This could sug- levels independent of glucose between treatment-induced effects on glycemia is gest that within the ranges of visually de- black, white, and South Asian popula- introduced and the level of glycemia as- tected thresholds for the three measures, tions (24,25). In our study, subgroup sessed in each study may be lower than changes in the prevalence of diabetes- analysis by Asian and predominantly that which led to retinopathy. Excluding specific retinopathy remain somewhat Caucasian populations showed no differ- people with treated diabetes from the linear. ence in the optimal A1C threshold (6.4% analyses eliminates this bias but changes The continuous and vigintile plots for both). However, our study was not the characteristics of the population by provided a similar range of threshold val- designed to have and did not have suffi- eliminating many individuals with reti- ues for FPG and A1C. ROC curve analyses cient numbers to examine a potential nopathy, making it much more difficult to were then used to compare performance black/white difference. identify a threshold (2,7). Large incidence in relation to optimizing sensitivity and Strengths of this study include its studies are needed to resolve these issues specificity of glycemic values in the range large sample size, which was drawn from and determine the optimal levels of glyce- around these thresholds. These analyses populations across different countries mia that predict the development of dia- suggest thresholds of 6.6 mmol/l for FPG and racial/ethnic groups; the ability to fo- betes-specific retinopathy. and 6.4% for A1C. The corresponding cus on diabetes-specific retinopathy; and In summary, this pooled analysis of ROC value for 2-h PG was 13.0 mmol/l. availability of data to examine three gly- glycemia and diabetes-specific retinopa- Combining the results derived from cemic measures. Our study has some lim- thy among close to 45,000 participants care.diabetesjournals.org DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 149 Glycemic thresholds for diabetic retinopathy Report of the expert committee on the di- 2006;29:1619 –1625 substantially broadens the evidence based agnosis and classification of diabetes mel- 14. Wong TY, Klein R, Islam FM, Cotch MF, on glucose-specific and A1C diabetes di- litus. Diabetes Care 1997;20:1183–1197 Folsom AR, Klein BE, Sharrett AR, Shea S. agnostic thresholds. Our results demon- 3. Wong TY, Liew G, Tapp RJ, Schmidt MI, Diabetic retinopathy in a multi-ethnic co- strate narrow glycemic threshold ranges Wang JJ, Mitchell P, Klein R, Klein BE, Zim- hort in the United States. Am J Ophthal- for the presence of diabetes-specific reti- met P, Shaw J. Relation between fasting glu- mol 2006;141:446 – 455 nopathy and suggest that the current dia- cose and retinopathy for diagnosis of 15. Harris MI, Klein R, Cowie CC, Rowland betes diagnostic level for FPG should be diabetes: three population-based cross-sec- M, Byrd-Holt DD. Is the risk of diabetic lowered to 6.5 mmol/l and that an A1C of tional studies. Lancet 2008;371:736 –743 retinopathy greater in non-Hispanic 6.5% is a suitable alternative diagnostic 4. Ito C, Maeda R, Ishida S, Harada H, Inoue blacks and Mexican Americans than in criterion. N, Sasaki H. Importance of OGTT for di- non-Hispanic whites with type 2 diabe- agnosing diabetes mellitus based on prev- tes? A US population study. Diabetes Care alence and incidence of retinopathy. 1998;21:1230 –1235 Acknowledgments — This study was sup- Diabetes Res Clin Pract 2000;49:181–186 16. Rajala U, Laakso M, Qiao Q, Keinanen-Kiu- ported by a Diabetes Australia Research Trust 5. Tapp RJ, Zimmer PZ, Harper CA, de kaanniemi S. Prevalence of retinopathy in Grant. The ARIC Study is carried out as a col- Courten MP, McCarty DJ, Balkau B, Tay- people with diabetes, impaired glucose tol- laborative study supported by the National lor HR, Welbourn TA, Shaw JE; AusDiab erance, and normal glucose tolerance. Dia- Heart, Lung, and Blood Institute (NHLBI) Study Group. Diagnostic thresholds for betes Care 1998;21:1664 –1669 Contracts N01-HC-55015, N01-HC-55016, diabetes: the association of retinopathy 17. McCance DR, Hanson RL, Charles MA, Ja- N01-HC-55018, N01-HC-55019, N01-HC- and albuminuria with glycaemia. Diabe- cobsson LT, Pettitt DJ, Bennett PH, Knowler 55020, N01-HC-55021, and N01-HC-55022. tes Res Clin Pract 2006;73:315–321 WC. Comparison of tests for glycated hae- Diabetes Australia had no involvement in 6. van Leiden HA, Dekker JM, Moll AC, Ni- moglobin and fasting and two hour plasma the study design, data collection, and analysis, jpels G, Heine RJ, Bouter LM, Stehouwer glucose concentrations as diagnostic meth- interpretation, and writing of the manuscript. CD, Polak BC. Risk factors for incident ods for diabetes. BMJ 1994;308:1323–1328 K.B.-J. is the Managing Director and Profes- retinopathy in a diabetic and nondiabetic 18. Wong TY, Cheung N, Tay WT, Wang JJ, sor of the Steno Diabetes Center, which is population: the Hoorn study. Arch Oph- Aung T, Saw SM, Lim SC, Tai ES, Mitchell owned by Novo Nordisk, Bagsværd, Den- thalmol 2003;121:245–251 P. Prevalence and risk factors for diabetic mark. No other potential conflicts of interest 7. Engelgau MM, Thompson TJ, Herman retinopathy. The Singapore Malay Eye relevant to this article were reported. WH, Boyle JP, Aubert RE, Kenny SJ, Bad- Study. Ophthalmology 2008;115:1869 – S.C., K.B.-J., and J.S. contributed to the ran A, Sous ES, Ali MA. Comparison of study concept and design. S.C., K.B.-J., and fasting and 2-hour glucose and HbA1c 19. Diabetic Retinopathy Study Research T.W. contributed to the acquisition of data. levels for diagnosing diabetes. Diagnostic S.C. and C.L. contributed to the drafting of the criteria and performance revisited. Diabe- Group. Report 7. A modification of the manuscript. S.C., C.L., T.W., B.B., J.S., and tes Care 1997;20:785–791 Airlie House classification of diabetic ret- K.B.-J. contributed to the analysis and inter- 8. Colagiuri S, Borch-Johnsen K. DETECT-2: inopathy. Invest Ophthalmol Vis Sci pretation of data and critical revision of the early detection of type 2 diabetes and IGT. 1981;21:210 –226 manuscript for important intellectual content. Diabetes Voice 2003;48:11–13 20. Early Treatment Diabetic Retinopathy C.L. contributed to the statistical analysis. 9. Glumer C, Vistisen D, Borch-Johnsen K, Study Research Group. Grading diabetic Dorte Vistisen, Steno Diabetes Center, was Colagiuri S; DETECT-2 Collaboration. retinopathy from stereoscopic color fundus responsible for gathering and maintaining the Risk scores for type 2 diabetes can be ap- photographs – an extension of the modified original DETECT-2 dataset. Federica Barzi, plied in some populations but not all. Di- Airlie House classification. Ophthalmology the George Institute for International Health, abetes Care 2006;29:410 – 414 1991;98(Suppl.):786 – 806 and Gerald Liew, Center for Vision Research, 10. Wong TY, Klein R, Amirul Islam FM, Cotch 21. Fukuda M Classification and treatment of provided statistical support. The authors MF, Couper DJ, Klein BE, Hubbard LD, diabetic retinopathy. Diabetes Res Clin thank the staff and participants of the ARIC Sharrett AR. Three-year incidence and cu- Pract 1994;24 Suppl.:S171–S6 study for their important contributions. The mulative prevalence of retinopathy: The 22. Wang JJ, Liew G, Klein R, Rochtchina E, MESA was conducted and supported by the Atherosclerosis Risk In Communities Knudtson MD, Klein BE, Wong TY, Bur- NHLBI, National Institutes of Health. The Study. Am J Ophthalmol 2007;143:970 – lutsky G, Mitchell P. Retinal vessel diam- BMES was supported by the Australian 976 eter and cardiovascular mortality: pooled NHMRC. The authors thank the members of 11. Mitchell P, Smith W, Wang JJ, Attebo K. data analysis from two older populations. the Gila River Indian Community for collabo- Prevalence of diabetic retinopathy in an Eur Heart J 2007;28:1984 –1992 ration, the University of Wisconsin Ocular Ep- older community. The Blue Mountains 23. Cowie CC, Rust KF, Ford ES, Eberhardt idemiology Reading Center, and support from Eye Study. Ophthalmology 1998;105:406 – MS, Byrd-Holt DD, Li C, Williams DE, the Intramural Research Program of the Na- 411 Gregg EW, Bainbridge KE, Saydah SH, tional Institute of Diabetes and Digestive and 12. Mohan V, Sandeep S, Deepa M, Goku- Geiss LS. Full accounting of diabetes and Kidney Diseases and the National Eye lakrishnan K, Datta M, Deepa R. A diabe- pre-diabetes in the U.S. population in Institute. tes risk score helps identify metabolic 1988 –1994 and 2005–2006. Diabetes syndrome and cardiovascular risk in Indi- Care 2009;32:287–294 ans – the Chennai Urban Rural Epidemi- 24. Likhari T, Gama R. Glycaemia-indepen- References ology Study (CURES-38). Diabetes Obes dent ethnic differences in HbA1c in sub- 1. World Health Organization and Interna- Metab 2007;9:337–343 jects with impaired glucose tolerance. tional Diabetes Federation. Definition and 13. Droumaguet C, Balkau B, Simon D, Caces Diabet Med 2009;26:1068 –1069 diagnosis of diabetes mellitus and inter- E, Tichet J, Charles MA, Eschwege E; 25. Davidson MB, Schriger DL. Effect of age and mediate hyperglycemia: report of a DESIR Study Group. Use of HbA1c in race/ethnicity on HbA1c levels in people WHO/IDF consultation. World Health predicting progression to diabetes in without known diabetes mellitus: implica- Organization: Geneva, 2006 French men and women: Data from an 2. The Expert Committee on the Diagnosis Epidemiology Study on the Insulin Resis- tions for the diagnosis of diabetes. Diabetes and Classification of Diabetes Mellitus. tance Syndrome (DESIR). Diabetes Care Res Clin Pract 2010;87:415– 421 150 DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 care.diabetesjournals.org http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diabetes Care Pubmed Central

Glycemic Thresholds for Diabetes-Specific Retinopathy

Loading next page...
 
/lp/pubmed-central/glycemic-thresholds-for-diabetes-specific-retinopathy-1Z56A5V6tZ

References (50)

Publisher
Pubmed Central
Copyright
© 2011 by the American Diabetes Association.
ISSN
0149-5992
eISSN
1935-5548
DOI
10.2337/dc10-1206
Publisher site
See Article on Publisher Site

Abstract

Pathophysiology/Complications ORIGINAL ARTICLE Glycemic Thresholds for Diabetes-Specific Retinopathy Implications for diagnostic criteria for diabetes 1 6 STEPHEN COLAGIURI, MBBS JONATHAN E. SHAW, MD glycemic levels. The datasets used for this 1 7,8 CRYSTAL M.Y. LEE, PHD KNUT BORCH-JOHNSEN, DMSC purpose were from Pima Indians, an 2,3 TIEN Y. WONG, PHD THE DETECT-2 COLLABORATION WRITING Egyptian study, and unpublished data 4,5 BEVERLEY BALKAU, PHD GROUP* from the Third National Health and Nu- trition Examination Survey (NHANES) (2). OBJECTIVE — To re-evaluate the relationship between glycemia and diabetic retinopathy. Other studies (3–5) also have exam- ined this relationship, but the results have RESEARCH DESIGN AND METHODS — We conducted a data-pooling analysis of been inconsistent. All studies reported to nine studies from five countries with 44,623 participants aged 20 –79 years with gradable retinal date have had limited statistical power to photographs. The relationship between diabetes-specific retinopathy (defined as moderate or examine this relationship in detail and more severe retinopathy) and three glycemic measures (fasting plasma glucose [FPG; n have adopted a very broad definition of 41,411], 2-h post oral glucose load plasma glucose [2-h PG; n 21,334], and A1C [n 28,010]) retinopathy that included many cases of was examined. mild retinopathy, now known to have causes other than hyperglycemia (6). A RESULTS — When diabetes-specific retinopathy was plotted against continuous glycemic measures, a curvilinear relationship was observed for FPG and A1C. Diabetes-specific retinop- more clinically relevant end point is dia- athy prevalence was low for FPG6.0 mmol/l and A1C6.0% but increased above these levels. betes-specific retinopathy (moderate or Based on vigintile (20 groups with equal numbers) distributions, glycemic thresholds for dia- more severe levels of retinopathy) that is betes-specific retinopathy were observed over the range of 6.4 – 6.8 mmol/l for FPG, 9.8 –10.6 invariably attributed to hyperglycemia. mmol/l for 2-h PG, and 6.3– 6.7% for A1C. Thresholds for diabetes-specific retinopathy from Also different statistical methods have receiver-operating characteristic curve analyses were 6.6 mmol/l for FPG, 13.0 mmol/l for 2-h been used in previous studies, which has PG, and 6.4% for A1C. an important effect on derived cut points (5,7). CONCLUSIONS — This study broadens the evidence based on diabetes diagnostic criteria. Several new datasets with retinopathy A narrow threshold range for diabetes-specific retinopathy was identified for FPG and A1C but data have become available since the orig- not for 2-h PG. The combined analyses suggest that the current diabetes diagnostic level for FPG could be lowered to 6.5 mmol/l and that an A1C of 6.5% is a suitable alternative diagnostic inal studies used to derive current diabe- criterion. tes diagnostic cut points (1,2). The DETECT-2 collaboration has pooled Diabetes Care 34:145–150, 2011 these datasets to examine and re-evaluate the relationship between retinopathy and he current diagnostic cut points for tially increased risk of diabetes-associated three glycemic measures: FPG, 2-h PG, diabetes (fasting plasma glucose microvascular complications, particularly and A1C. The size of the DETECT-2 data- [FPG] of 7.0 mmol/l and 2-h post retinopathy, above these levels (1,2). set has allowed us to focus on the relation- oral glucose load plasma glucose [2-h PG] These cut points were derived from cross- ship between measures of glycemia and of 11.1 mmol/l) are largely based on gly- sectional epidemiological studies that ex- diabetes-specific retinopathy (i.e., mod- cemic levels associated with a substan- amined retinopathy across a range of erate or more severe levels of retinopa- thy). These analyses were designed to ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● inform current deliberations on possible From the Boden Institute of Obesity, Nutrition, and Exercise, University of Sydney, Sydney, Australia; the 2 3 revisions to the diagnostic criteria for Center for Eye Research Australia, University of Melbourne, Melbourne, Australia; the Singapore Eye Research Institute, National University of Singapore, Singapore; the Institut National de la Sante´ etdela diabetes. Recherche Me ´ dicale, Centre de Recherche en Epide ´ miologie et Sante ´ des Populations, Epidemiology of Diabetes, Obesity, and Chronic Kidney Disease Over the Lifecourse, Villejuif, France; Centre de Recher- che en Epide ´ miologie et Sante ´ des Populations, Universite ´ Paris Sud, Villejuif, France; the Heart and RESEARCH DESIGNS AND Diabetes Institute, Baker International Diabetes Institute, Melbourne, Australia; the Steno Diabetes 8 METHODS — The DETECT-2 project Center, Gentofte, Denmark; and the Faculty of Health Science, University of Aarhus, Aarhus, Denmark. is an international data-pooling collabo- Corresponding author: Stephen Colagiuri, stephen.colagiuri@sydney.edu.au. Received 23 June 2010 and accepted 8 October 2010. Published ahead of print at http://care. ration. The primary objective of the col- diabetesjournals.org on 26 October 2010. DOI: 10.2337/dc10-1206. laboration was to examine aspects of *A complete list of the members of the DETECT-2 Collaboration Writing Group are listed in the online screening for type 2 diabetes and im- appendix available at http://care.diabetesjournals.org/cgi/content/full/dc10-1206/DC1. paired glucose tolerance across various This article was prepared using limited-access datasets obtained from the NHLBI and does not necessarily reflect the opinions or views of MESA or the NHLBI. populations and ethnic groups. Details of © 2011 by the American Diabetes Association. Readers may use this article as long as the work is properly the collaboration are reported elsewhere cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons. (8,9). For the current analysis, studies in- org/licenses/by-nc-nd/3.0/ for details. cluded in the DETECT-2 database, in The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. which retinopathy data had been col- care.diabetesjournals.org DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 145 Glycemic thresholds for diabetic retinopathy Table 1—Summary of studies included in these analyses Name of study Country Year Age range n* Measures available ARIC (10) U.S. 1993–1995 (visit 3) 49–73 10,873 FPG AusDiab (5) Australia 1999–2000 25–90 2,052 FPG, 2-h PG, A1C BMES (11) Australia 1992–1994 45–97 2,915 FPG CURES (12) India 2002–2004 (phase III) 20–85 2,200 FPG, 2-h PG, A1C Hiroshima study (4) Japan 1990–2004 17–99 12,873 FPG, 2-h PG, A1C MESA (14) U.S. 2002–2004 (second examination) 45–85 5,920 FPG, A1C NHANES III (15) U.S. 1988–1994 40–74 2,869 FPG, 2-h PG, A1C Pima Indian study (17) U.S. 1982 (first examination) 15–85 1,829 FPG, 2-h PG SiMES (18) Singapore 2004 40–79 3,170 A1C *Number of participants aged 20 –79 years included in the analysis. lected, were invited to provide these data thy, which we defined as moderate or of the authors (T.Y.W) was personally in- for this analysis. Additional studies with more severe levels of retinopathy. volved in the grading of retinopathy using retinopathy data identified by coinvesti- All nine studies measured plasma glu- the modified Early Treatment Diabetic gators through personal contact or litera- cose, and six studies that measured A1C Retinopathy Study (Atherosclerosis Risk ture search also were invited to contribute used high-performance liquid chroma- in Communities [ARIC], AusDiab, Blue datasets. Retinopathy data were available tography, of which five used a Diabetes Mountains Eye Study [BMES], MESA, and from 12 studies in eight countries Control and Complications Trial SiMES); 3) studies in which participants (4,5,7,10 –18). This analysis focuses on (DCCT)-aligned assay (5,12,14,15,18). were predominantly Caucasian (ARIC, nine studies from five countries that had AusDiab, BMES, MESA, and NHANES retinopathy data by grading. Participants Statistical analysis III); 4) studies in which participants were aged 20 –79 years, including those with Prevalence of diabetes-specific retinopa- Asian (CURES, Hiroshima study, and known diabetes and with gradable retinal thy was examined by 1) 0.5-unit intervals SiMES); and 5) studies in which partici- photographs and at least one measure of of glycemic measures and 2) vigintiles (di- pants had all three measures of glycemia. glycemia, were included. All studies were viding participants into 20 equally sized All statistical analyses were performed us- approved by respective institutional re- groups) of the distribution for each mea- ing SAS 9.1 for Windows (SAS Institute, view boards and were conducted accord- sure of glycemia. Logistic regression mod- Cary, NC) and SPSS 16.0 for Windows ing to the Declaration of Helsinki. els were applied to test the relationships (SPSS, Chicago, IL). between diabetes-specific retinopathy Classification of retinopathy and glycemia by 0.5-unit intervals and by RESULTS The retinal photograph grading was per- vigintiles of each glycemic measure, with formed by individual study centers. Reti- the lowest range as the reference. The Study participants nopathy was classified as present or analyses were repeated after adjusting for In total, 44,623 participants had informa- absent for initial analysis. Where data study center. tion on both the presence and severity of were available, those with retinopathy The discriminatory power of each retinopathy (Table 1). A total of 1,589 were further classified as those having measure of glycemia for retinopathy was participants had minimal NPDR, 762 had minimal nonproliferative diabetic reti- assessed as the area under the receiver- mild NPDR, 430 had moderate NPDR, 50 nopathy (NPDR), mild NPDR, moderate operating characteristic (ROC) curve had severe NPDR, and 171 had PDR. The NPDR, severe NPDR, or proliferative dia- (AUC). An AUC of 1 indicates perfect dis- number of participants available for each betic retinopathy (PDR) based on the in- criminatory power and an AUC of 0.5 in- measure of glycemia was 41,334 for FPG, formation provided by individual studies dicates that the discrimination is no better 21,334 for 2-h PG, and 27,933 for A1C. using the modified Airlie House classifi- than chance. ROC curve analyses were Of these, 27,445 participants had at least cation levels (19), modified Early Treat- used to examine thresholds based on op- two measures and 18,533 participants ment Diabetic Retinopathy Study levels timizing sensitivity and specificity. The had all three measures. The characteris- (20), or the Fukuda standard (21). Lev- impact of various thresholds on the prev- tics of participants by study are shown els 14 –20 indicate minimal NPDR, levels alence of diabetes was examined by ap- in supplementary Table 1 in the on- 30 –35 or the Fukuda standard A1 indi- plying these values to 16,381 participants line appendix (available at http://care. cate mild NPDR, levels 40 – 47 or the without known diabetes who had all three diabetesjournals.org/cgi/content/full/dc10- Fukuda standard A2 indicate moderate measures of glycemia. 1206/DC1). NPDR, levels 50 –53 or the Fukuda stan- Sensitivity analysis was performed on dard A3 indicate severe NPDR, and levels 1) studies in which a DCCT-aligned assay Prevalence of retinopathy 60 –90 or the Fukuda standards A4 and for A1C was used (AusDiab, Chennai Ur- The overall prevalence of any retinopathy B1–B4 indicate PDR. The final retinopa- ban Rural Epidemiological Study was 6.7% and 1.5% for diabetes-specific thy grading for each participant was based [CURES], Multi-Ethnic Study of Athero- retinopathy. In people with known diabe- on the diagnosis in the more severely af- sclerosis and Air Pollution [MESA], tes, the prevalence of diabetes-specific fected eye. The primary outcome used in NHANES III, and Singapore Malay Eye retinopathy was 9.4%, in newly diag- this study was diabetes-specific retinopa- Study [SiMES]); 2) studies in which one nosed diabetes 1.0%, in impaired glucose 146 DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 care.diabetesjournals.org Colagiuri and Associates Figure 1—Prevalence of diabetes-specific retinopathy (moderate or more severe retinopathy) with 95% confidence intervals, number of retinopathy cases, and participants within each interval by 0.5 unit intervals for FPG and 2-h PG, and A1C. tolerance (1) 0.1%, in impaired fasting retinopathy. These plots suggest a curvi- FPG category of 6.0 – 6.4 mmol/l and glucose (1) 0.1%, and with normal glu- linear relationship for FPG and A1C and from the A1C category of 6.0 – 6.4%. The cose tolerance 0.1%. retinopathy. Diabetes-specific retinopa- curve for 2-h PG was flatter than for FPG Figure 1 shows the prevalence of ret- thy was virtually absent (prevalence and A1C, and no definite interval of in- inopathy by 0.5-unit intervals for each 0.4%) at low levels for each glycemic crease for 2-h PG was obvious. measure of glycemia for diabetes-specific measure but began to increase from the Logistic regression adjusted for study care.diabetesjournals.org DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 147 Glycemic thresholds for diabetic retinopathy Figure 2—Prevalence of diabetes-specific retinopathy (moderate or more severe retinopathy) by vigintiles of the distribution of FPG, 2-h PG, and A1C. center showed that the first interval where 9.8 –10.6 mmol/l; OR 10.1 [95% CI 1.3– mmol/l for 2-h PG, and 6.4% for A1C (Ta- the odds ratio (OR) for diabetes-specific 79.4]; P  0.03), from the 17th vigintile ble 2). These thresholds gave similar val- retinopathy was significantly different for FPG (6.4 – 6.8 mmol/l; 2.5 [1.2–5.2]; ues for positive and negative predictive from the reference FPG level of 4.0 – 4.4 P  0.01), and from the 18th vigintile for values. If these thresholds were used for mmol/l was 6.5– 6.9 mmol/l (OR 6.0 A1C (6.3– 6.7%; 4.5 [1.4 –15.2]; P  diagnosing diabetes, the prevalence of [95% CI 2.1–17.1]; P  0.01). The cor- 0.01). newly diagnosed diabetes would be responding result for A1C was 6.5– 6.9% Supplementary Table 2 in the online 11.9, 8.0, and 6.3% according to FPG, (16.8 [2.3–123.7]; P  0.01) compared appendix shows the ROC curve analyses. 2-h PG, and A1C, respectively. The dif- with an A1C of 4.0 – 4.4%. The overall discriminatory power deter- ferences in performance based on ROC Figure 2 shows the prevalence of dia- mined by AUCs was uniformly high for curve statistics for the three measures of betes-specific retinopathy by vigintiles of diabetes-specific retinopathy for each glycemia were minor for threshold val- the glycemic distributions. The preva- measure of glycemia (0.87 [95% CI 0.85– ues around the above values (supple- lence of diabetes-specific retinopathy was 0.89] for FPG, 0.89 [0.87– 0.91] for 2-h mentary Table 2). very low until the 15th vigintile for 2-h PG, and 0.90 [0.88 – 0.92] for A1C). The Sensitivity analyses showed that the PG (vigintile range 9.8 –10.6 mmol/l) and overlapping CIs suggests that there is no five studies in which T.Y.W used the same until the 17th vigintile for FPG (6.4 – 6.8 statistical difference between the three retinopathy grading system or the five mmol/l) and for A1C (6.1– 6.2%). measures of glycemia. The performance studies that used DCCT-aligned assays for Logistic regression models adjusted of a wide range of thresholds was exam- A1C measurements provided similar re- for study center confirmed a statistically ined, with particular attention to those sults to the overall study. The optimal significant difference in the OR for diabe- that overlapped from the continuous and threshold for FPG was 6.4 – 6.5 mmol/l, tes-specific retinopathy compared with vigintile distribution plots. The thresh- for A1C 6.4 – 6.5%, and for 2-h PG 10.1– the first vigintile that occurred from the olds that optimized sensitivity and speci- 11.2 mmol/l. 15th vigintile for 2-h PG (vigintile range ficity were 6.6 mmol/l for FPG, 13.0 CONCLUSIONS — The current di- Table 2—Threshold ranges for diabetes-specific retinopathy (moderate NPDR or more severe agnostic criteria for diabetes were derived retinopathy) derived from logistic regression models (adjusted for center) of the glycemic from analyses of the relationship between measures by continuous distribution and vigintile distribution and ROC curve analysis retinopathy and measures of glycemia (1). Our study is the largest to examine this association, using data from 45,000 FPG (mmol/l) 2-h PG (mmol/l) A1C (%) participants from five countries, and pro- Logistic regression vides the statistical power for a more de- Continuous distribution 6.5–6.9 No threshold 6.5–6.9 tailed and precise examination of Vigintile distribution 6.4–6.8 9.8–10.6 6.3–6.7 glycemic thresholds for diabetes-specific ROC curve analysis 6.6 13.0 6.4 retinopathy (moderate nonproliferative 148 DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 care.diabetesjournals.org Colagiuri and Associates and more severe retinopathy). Previous the vigintile distribution, continuous itations. First, this study was based on studies (7,16) have only reported the as- plots, and ROC curve analyses suggest cut cross-sectional data, whereas diagnostic sociation of glycemic measures with any point values of 6.5 mmol/l for FPG and thresholds would ideally be informed by retinopathy, which is less specific for hy- 6.5% for A1C, which could be considered incidence data of diabetes complications. perglycemia and is very frequently de- in deliberations on modifying the current Second, the methods used to assess and tected in people without diabetes. diagnostic criteria for diabetes. The re- classify retinopathy differed between The association between glycemic sults for 2-h PG were too inconsistent to studies, and it was not possible to inde- measures and retinopathy has tradition- consider modifying the current diagnos- pendently review the grading of all ally been investigated by plotting the tic cut point of 11.1 mmol/l. photographs. Nevertheless, inter- and in- prevalence of retinopathy against the de- It should be noted that these values traobserver consistency for retinopathy in cile distribution (the population divided do not result in equivalent estimates for the different studies was of the order of into 10 equal groups) of each glycemic prevalent diabetes. This has been an on- 80 –98% (3,10,15) and misclassification, measure (1,2). Our large dataset allows going issue with the current diagnostic especially for moderate or more severe analysis using vigintile distributions (the criteria, whereby using FPG alone or an forms of retinopathy, is likely to be mini- population divided into 20 equal groups), oral glucose tolerance test to diagnose di- mal but cannot be entirely eliminated. which narrows the glycemic range of each abetes gives different diabetes prevalence Furthermore, analysis of the studies in group. Based on logistic regression analy- (23). From our data (supplementary Ta- which T.Y.W. was involved in the stan- sis of these vigintile distributions, glyce- ble 2), lowering the FPG to 6.5 mmol/l dardized grading of retinal photographs mic thresholds for diabetes-specific would result in a diabetes prevalence of showed cut points for FPG and A1C sim- retinopathy were observed in the range of 13.0% based on FPG alone and 18.6% ilar to our entire study cohort. Third, no 6.4 – 6.8 mmol/l for FPG, 9.8 –10.6 based on an oral glucose tolerance test quality assurance of measures of glycemia mmol/l for 2-h PG, and 6.3– 6.7% for using an FPG of 6.5 mmol/l or a 2-h PG of could be applied across the studies. Nev- A1C (Table 2). 11.1 mmol/l. The prevalence of diabetes ertheless, all studies measured A1C using The large size of this dataset enables defined by an A1C of 6.5% is consider- high-performance liquid chromatogra- diabetes-specific retinopathy to be plot- ably lower (5.7%). This discrepancy in phy, and analysis of the five studies that ted against measures of glycemia as a prevalence may be problematic for epide- used a DCCT-aligned assay showed an continuous variable. A curvilinear rela- miological studies but is not necessarily a A1C cut point of 6.4 – 6.5%. Fourth, the tionship was observed, especially for FPG disadvantage for individual patient care. Hiroshima study, with its large sample and A1C, as opposed to the linear associ- An A1C of 6.5% was associated with a size, and the Pima Indian study, with its ation observed between blood pressure higher sensitivity and specificity than an high prevalence of diabetes-specific reti- and cardiovascular disease. Diabetes- FPG of 6.5 mmol/l and a higher specificity nopathy, may have influenced the results. specific retinopathy was rare at low levels than a 2-h PG of 11.1 mmol/l. In other However, sensitivity analyses that ex- of glycemia but increased from a range of words, fewer people would be identified cluded these two studies did not alter the 6.0 – 6.4 mmol/l for FPG and 6.0 – 6.4% as having diabetes, but this would not overall results. Finally, not all included for A1C. A threshold for increasing reti- compromise the identification of people studies were randomly sampled popula- nopathy was less obvious for 2-h PG, with diabetes-specific retinopathy. tions (e.g., MESA) and some (e.g., Aus- probably related to the smaller number of Whether this would have any deleterious Diab) oversampled people with diabetes study participants with this measure and ramifications in relation to identifying in- and/or prediabetes. Common to all such diabetes-specific retinopathy. Change dividuals at increased risk of other micro- analyses is the issue of whether to include point analyses, which were used previ- vascular or macrovascular disease people with previously diagnosed diabe- ously in two population-based studies remains to be determined. tes. If people with known diabetes cur- (22), were applied to these curves in an This study necessarily included pop- rently receiving blood glucose–lowering attempt to identify statistically significant ulations from different countries with var- treatment are included, the population- thresholds, but we were unable to dem- ious racial/ethnic backgrounds. There based characteristics of the study sample onstrate a clear threshold for any glycemic have been reports of differences in A1C are maintained, but a bias associated with measure by this method. This could sug- levels independent of glucose between treatment-induced effects on glycemia is gest that within the ranges of visually de- black, white, and South Asian popula- introduced and the level of glycemia as- tected thresholds for the three measures, tions (24,25). In our study, subgroup sessed in each study may be lower than changes in the prevalence of diabetes- analysis by Asian and predominantly that which led to retinopathy. Excluding specific retinopathy remain somewhat Caucasian populations showed no differ- people with treated diabetes from the linear. ence in the optimal A1C threshold (6.4% analyses eliminates this bias but changes The continuous and vigintile plots for both). However, our study was not the characteristics of the population by provided a similar range of threshold val- designed to have and did not have suffi- eliminating many individuals with reti- ues for FPG and A1C. ROC curve analyses cient numbers to examine a potential nopathy, making it much more difficult to were then used to compare performance black/white difference. identify a threshold (2,7). Large incidence in relation to optimizing sensitivity and Strengths of this study include its studies are needed to resolve these issues specificity of glycemic values in the range large sample size, which was drawn from and determine the optimal levels of glyce- around these thresholds. These analyses populations across different countries mia that predict the development of dia- suggest thresholds of 6.6 mmol/l for FPG and racial/ethnic groups; the ability to fo- betes-specific retinopathy. and 6.4% for A1C. The corresponding cus on diabetes-specific retinopathy; and In summary, this pooled analysis of ROC value for 2-h PG was 13.0 mmol/l. availability of data to examine three gly- glycemia and diabetes-specific retinopa- Combining the results derived from cemic measures. Our study has some lim- thy among close to 45,000 participants care.diabetesjournals.org DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 149 Glycemic thresholds for diabetic retinopathy Report of the expert committee on the di- 2006;29:1619 –1625 substantially broadens the evidence based agnosis and classification of diabetes mel- 14. Wong TY, Klein R, Islam FM, Cotch MF, on glucose-specific and A1C diabetes di- litus. Diabetes Care 1997;20:1183–1197 Folsom AR, Klein BE, Sharrett AR, Shea S. agnostic thresholds. Our results demon- 3. Wong TY, Liew G, Tapp RJ, Schmidt MI, Diabetic retinopathy in a multi-ethnic co- strate narrow glycemic threshold ranges Wang JJ, Mitchell P, Klein R, Klein BE, Zim- hort in the United States. Am J Ophthal- for the presence of diabetes-specific reti- met P, Shaw J. Relation between fasting glu- mol 2006;141:446 – 455 nopathy and suggest that the current dia- cose and retinopathy for diagnosis of 15. Harris MI, Klein R, Cowie CC, Rowland betes diagnostic level for FPG should be diabetes: three population-based cross-sec- M, Byrd-Holt DD. Is the risk of diabetic lowered to 6.5 mmol/l and that an A1C of tional studies. Lancet 2008;371:736 –743 retinopathy greater in non-Hispanic 6.5% is a suitable alternative diagnostic 4. Ito C, Maeda R, Ishida S, Harada H, Inoue blacks and Mexican Americans than in criterion. N, Sasaki H. Importance of OGTT for di- non-Hispanic whites with type 2 diabe- agnosing diabetes mellitus based on prev- tes? A US population study. Diabetes Care alence and incidence of retinopathy. 1998;21:1230 –1235 Acknowledgments — This study was sup- Diabetes Res Clin Pract 2000;49:181–186 16. Rajala U, Laakso M, Qiao Q, Keinanen-Kiu- ported by a Diabetes Australia Research Trust 5. Tapp RJ, Zimmer PZ, Harper CA, de kaanniemi S. Prevalence of retinopathy in Grant. The ARIC Study is carried out as a col- Courten MP, McCarty DJ, Balkau B, Tay- people with diabetes, impaired glucose tol- laborative study supported by the National lor HR, Welbourn TA, Shaw JE; AusDiab erance, and normal glucose tolerance. Dia- Heart, Lung, and Blood Institute (NHLBI) Study Group. Diagnostic thresholds for betes Care 1998;21:1664 –1669 Contracts N01-HC-55015, N01-HC-55016, diabetes: the association of retinopathy 17. McCance DR, Hanson RL, Charles MA, Ja- N01-HC-55018, N01-HC-55019, N01-HC- and albuminuria with glycaemia. Diabe- cobsson LT, Pettitt DJ, Bennett PH, Knowler 55020, N01-HC-55021, and N01-HC-55022. tes Res Clin Pract 2006;73:315–321 WC. Comparison of tests for glycated hae- Diabetes Australia had no involvement in 6. van Leiden HA, Dekker JM, Moll AC, Ni- moglobin and fasting and two hour plasma the study design, data collection, and analysis, jpels G, Heine RJ, Bouter LM, Stehouwer glucose concentrations as diagnostic meth- interpretation, and writing of the manuscript. CD, Polak BC. Risk factors for incident ods for diabetes. BMJ 1994;308:1323–1328 K.B.-J. is the Managing Director and Profes- retinopathy in a diabetic and nondiabetic 18. Wong TY, Cheung N, Tay WT, Wang JJ, sor of the Steno Diabetes Center, which is population: the Hoorn study. Arch Oph- Aung T, Saw SM, Lim SC, Tai ES, Mitchell owned by Novo Nordisk, Bagsværd, Den- thalmol 2003;121:245–251 P. Prevalence and risk factors for diabetic mark. No other potential conflicts of interest 7. Engelgau MM, Thompson TJ, Herman retinopathy. The Singapore Malay Eye relevant to this article were reported. WH, Boyle JP, Aubert RE, Kenny SJ, Bad- Study. Ophthalmology 2008;115:1869 – S.C., K.B.-J., and J.S. contributed to the ran A, Sous ES, Ali MA. Comparison of study concept and design. S.C., K.B.-J., and fasting and 2-hour glucose and HbA1c 19. Diabetic Retinopathy Study Research T.W. contributed to the acquisition of data. levels for diagnosing diabetes. Diagnostic S.C. and C.L. contributed to the drafting of the criteria and performance revisited. Diabe- Group. Report 7. A modification of the manuscript. S.C., C.L., T.W., B.B., J.S., and tes Care 1997;20:785–791 Airlie House classification of diabetic ret- K.B.-J. contributed to the analysis and inter- 8. Colagiuri S, Borch-Johnsen K. DETECT-2: inopathy. Invest Ophthalmol Vis Sci pretation of data and critical revision of the early detection of type 2 diabetes and IGT. 1981;21:210 –226 manuscript for important intellectual content. Diabetes Voice 2003;48:11–13 20. Early Treatment Diabetic Retinopathy C.L. contributed to the statistical analysis. 9. Glumer C, Vistisen D, Borch-Johnsen K, Study Research Group. Grading diabetic Dorte Vistisen, Steno Diabetes Center, was Colagiuri S; DETECT-2 Collaboration. retinopathy from stereoscopic color fundus responsible for gathering and maintaining the Risk scores for type 2 diabetes can be ap- photographs – an extension of the modified original DETECT-2 dataset. Federica Barzi, plied in some populations but not all. Di- Airlie House classification. Ophthalmology the George Institute for International Health, abetes Care 2006;29:410 – 414 1991;98(Suppl.):786 – 806 and Gerald Liew, Center for Vision Research, 10. Wong TY, Klein R, Amirul Islam FM, Cotch 21. Fukuda M Classification and treatment of provided statistical support. The authors MF, Couper DJ, Klein BE, Hubbard LD, diabetic retinopathy. Diabetes Res Clin thank the staff and participants of the ARIC Sharrett AR. Three-year incidence and cu- Pract 1994;24 Suppl.:S171–S6 study for their important contributions. The mulative prevalence of retinopathy: The 22. Wang JJ, Liew G, Klein R, Rochtchina E, MESA was conducted and supported by the Atherosclerosis Risk In Communities Knudtson MD, Klein BE, Wong TY, Bur- NHLBI, National Institutes of Health. The Study. Am J Ophthalmol 2007;143:970 – lutsky G, Mitchell P. Retinal vessel diam- BMES was supported by the Australian 976 eter and cardiovascular mortality: pooled NHMRC. The authors thank the members of 11. Mitchell P, Smith W, Wang JJ, Attebo K. data analysis from two older populations. the Gila River Indian Community for collabo- Prevalence of diabetic retinopathy in an Eur Heart J 2007;28:1984 –1992 ration, the University of Wisconsin Ocular Ep- older community. The Blue Mountains 23. Cowie CC, Rust KF, Ford ES, Eberhardt idemiology Reading Center, and support from Eye Study. Ophthalmology 1998;105:406 – MS, Byrd-Holt DD, Li C, Williams DE, the Intramural Research Program of the Na- 411 Gregg EW, Bainbridge KE, Saydah SH, tional Institute of Diabetes and Digestive and 12. Mohan V, Sandeep S, Deepa M, Goku- Geiss LS. Full accounting of diabetes and Kidney Diseases and the National Eye lakrishnan K, Datta M, Deepa R. A diabe- pre-diabetes in the U.S. population in Institute. tes risk score helps identify metabolic 1988 –1994 and 2005–2006. Diabetes syndrome and cardiovascular risk in Indi- Care 2009;32:287–294 ans – the Chennai Urban Rural Epidemi- 24. Likhari T, Gama R. Glycaemia-indepen- References ology Study (CURES-38). Diabetes Obes dent ethnic differences in HbA1c in sub- 1. World Health Organization and Interna- Metab 2007;9:337–343 jects with impaired glucose tolerance. tional Diabetes Federation. Definition and 13. Droumaguet C, Balkau B, Simon D, Caces Diabet Med 2009;26:1068 –1069 diagnosis of diabetes mellitus and inter- E, Tichet J, Charles MA, Eschwege E; 25. Davidson MB, Schriger DL. Effect of age and mediate hyperglycemia: report of a DESIR Study Group. Use of HbA1c in race/ethnicity on HbA1c levels in people WHO/IDF consultation. World Health predicting progression to diabetes in without known diabetes mellitus: implica- Organization: Geneva, 2006 French men and women: Data from an 2. The Expert Committee on the Diagnosis Epidemiology Study on the Insulin Resis- tions for the diagnosis of diabetes. Diabetes and Classification of Diabetes Mellitus. tance Syndrome (DESIR). Diabetes Care Res Clin Pract 2010;87:415– 421 150 DIABETES CARE, VOLUME 34, NUMBER 1, JANUARY 2011 care.diabetesjournals.org

Journal

Diabetes CarePubmed Central

Published: Oct 26, 2010

There are no references for this article.