Optimal HbA1c cutoff for detecting diabetic retinopathy

Transcription

1 DOI /s ORIGINAL ARTICLE Optimal HbA1c cutoff for detecting diabetic retinopathy Nam H. Cho Tae Hyuk Kim Se Joon Woo Kyu Hyung Park Soo Lim Young Min Cho Kyong Soo Park Hak C. Jang Sung Hee Choi Received: 8 December 2012 / Accepted: 7 January 2013 Ó Springer-Verlag Italia 2013 Abstract The associations between high glucose levels and diabetic retinopathy have been the basis for the diagnosis of diabetes. We aimed to provide updated data on the relationship between HbA1c and diabetic retinopathy, and to assess the diagnostic accuracy of the proposed HbA1c cutoff for detecting diabetic retinopathy. This cross-sectional study included 3,403 adults from the 2009 to 2010 Ansung Cohort Study. Retinopathy was assessed with single-field nonmydriatic fundus photography and graded according to the International Clinical Diabetic Retinopathy Disease Severity Scale. HbA1c was measured by standardized assay using high performance liquid chromatography. Based on deciles distribution, the prevalence of retinopathy was very low until the HbA1c range of mmol/mol ( %). The optimal HbA1c cutoff Communicated by Massimo Federici. Nam H. Cho and Tae Hyuk Kim are contributed equally to this work. N. H. Cho Department of Preventive Medicine, Ajou University School of Medicine, Suwon, South Korea T. H. Kim S. Lim Y. M. Cho K. S. Park H. C. Jang S. H. Choi Department of Internal Medicine, Seoul National University College of Medicine, Seoul, South Korea S. J. Woo K. H. Park Department of Ophthalmology, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Seongnam, South Korea S. Lim H. C. Jang S. H. Choi (&) Department of Internal Medicine, Seoul National University Bundang Hospital, 300, Gumi-dong Bundang-gu, Seongnam , South Korea drshchoi@snu.ac.kr for detecting any diabetic retinopathy was 49 mmol/mol (6.6 %), moderate or severer retinopathy was 52 mmol/mol from receiver operating characteristic curve analysis. The proposed HbA1c threshold of 48 mmol/mol (6.5 %) from American Diabetes Association produced comparable accuracy for identifying both any and moderate/severer retinopathy. This study confirmed that the proposed HbA1c threshold of 48 mmol/mol (6.5 %) allowed the proper detection of diabetic retinopathy. Our data support the judicious use of HbA1c for the diagnosis of diabetes and detecting diabetic retinopathy as well. Keywords Diabetes Epidemiology Diagnosis HbA1c Retinopathy Introduction Glycated hemoglobin (HbA1c), which reflects the chronic blood glucose concentration, is a reliable marker of future diabetes and predicts cardiovascular disease and mortality better than fasting plasma glucose (FPG) [1]. HbA1c is a more convenient and reproducible tool than glucose testing, and the standardized HbA1c assay is not inferior to the blood glucose assay [2]. In January 2010, the American Diabetes Association (ADA) published revised criteria for the diagnosis of diabetes mellitus, which included HbA1c C 48 mmol/mol (6.5 %) [3]. However, the recommendation to accept HbA1c as a diagnostic criterion has stimulated debate concerning the accuracy of HbA1c based diagnoses, particularly because it is less sensitive than glycemic measures [2]. The association between high glucose levels and longterm diabetic complications, particularly diabetic retinopathy, has been the basis for the diagnosis of diabetes [3, 4].

2 The presence of diabetic retinopathy itself is an independent predictor of other diabetic complications, such as nephropathy, cardiovascular disease, and mortality in patients with diabetes [5, 6]. Moreover, diabetic retinopathy is the leading cause of blindness in the working-age population and affects the life quality of diabetic patients [7]. In this study, we aimed to provide updated data on the relationship between HbA1c and retinopathy and assessed the accuracy of the proposed HbA1c cutoff from ADA for detecting prevalent diabetic retinopathy in a Korean community based cohort sample. Materials and methods The design and methods used in the Ansung Cohort Study have been described previously [8]. Briefly, it is an ongoing prospective, community-based, large cohort study to investigate the trends in diabetes and associated risk factors. The baseline examination was performed in , and biennial follow-up examinations will continue through The eligibility criteria included an age of years, residence within the borders of the survey area for at least 6 months before testing, and sufficient mental and physical ability to participate. Ansung is a representative rural farming community and had a population of 132,906 in 2000 [9]. Of the 7,192 eligible subjects in Ansung, 5,018 were investigated at baseline (70 % response rate) using a cluster sampling method according to age, sex, and residential district. In this study, we used cross-sectional data of fifth examination during the period (n = 3,403), because no ophthalmological evaluations were performed at baseline. Informed written consent was obtained from all participants. The study protocol was approved by the ethics committee of the Korean Center for Disease Control and Ajou University School of Medicine Institutional Review Board. Throughout the study, the same trained researchers and instruments were used to collect the data. Anthropometric parameters and blood pressure were assessed by standard methods. The plasma concentrations of glucose, total cholesterol, triglycerides, HDL cholesterol, and serum creatinine were measured in a central laboratory. After 8 14 h of overnight fasting, all subjects underwent a 2 h, 75 g oral glucose tolerance test (OGTT) at inclusion and biennially. HbA1c was analyzed centrally with high performance liquid chromatography (Variant II, BioRad, Hercules, CA), and the values were converted to the diabetes control and complications trial (DCCT) aligned reference HbA1c [10]. Diabetes was defined according to the ADA criteria, based on the 75 g OGTT: FPG concentration C7.0 mmol/l or 2 h plasma glucose (PG) C11.1 mmol/l or current treatment with oral antidiabetic drugs or insulin [3]. Hypertension was defined as a systolic blood pressure C140 mmhg or a diastolic blood pressure C90 mmhg or taking antihypertensive medication. Smoking status was defined as neversmoker and eversmoker. Retinopathy was assessed with single-field 45 nonmydriatic fundus photography of each eye using a digital fundus camera (CR-DGi, Cannon Inc., Japan) [11]. Retinopathy was classified as mild nonproliferative diabetic retinopathy (NPDR), moderate NPDR, severe NPDR, proliferative diabetic retinopathy (PDR), or prior panretinal photocoagulation (PRP) according to the International Clinical Diabetic Retinopathy Disease Severity Scale [12], based on the grading of the worse eye by one ophthalmologist specialized in retinal disease. The prevalence of any diabetic retinopathy and moderate/severer retinopathy was examined by deciles of the distribution for HbA1c as well as FPG. To derive threshold range of HbA1c for increasing prevalence of diabetic retinopathy, we used logistic regression models with the lowest deciles as the references [13]. The analyses were repeated after adjusting for age, gender, systolic blood pressure, and cigarette smoking [14, 15]. Also, the diagnostic properties of the specific cutpoints of HbA1c were evaluated by calculating the sensitivity and specificity by the receiver operating characteristic (ROC) curves. ROC analyses allowed the determination of the optimal HbA1c cutpoint to maximize accuracy. We compared the overall discriminatory power of HbA1c and FPG levels as continuous variables by calculating the area under the ROC curves (AUCs). MedCalc software was used to calculate the ROC curves; the significance of differences between areas under these curves was calculated as shown elsewhere [16]. All other analyses were performed using the SPSS 19.0 for Windows (SPSS, Chicago, IL). Significance was defined as P \ 0.05 for two sided tests. Results The characteristics of the 3,403 study participants are shown in Table 1. The mean HbA1c was 44 ± 8.7 mmol/mol (6.2 ± 0.8 %) and the mean FPG was 5.7 ± 1.5 mmol/l. The overall prevalence of diabetic retinopathy in total population was 1.9 %, and the prevalence of moderate or severer retinopathy was 1.0 %: 28 mild NPDR, 11 moderate NPDR, 8 severe NPDR, 16 PDR or prior PRP. Twenty-four percent had diabetes by ADA criteria with 75 g OGTT, and the mean of diabetes duration was 8.0 ± 6.6 years. Among those with diabetes defined by 75 g OGTT criteria, the prevalence of any diabetic retinopathy

3 Table 1 Clinical and biochemical characteristics of the study population ( , fifth examination of the Ansung Cohort) n 3,403 Age (years) 63.3 ± 8.6 Men (%) 43.2 Systolic blood pressure (mmhg) 121 ± 17 Diastolic blood pressure (mmhg) 75 ± 10 Body mass index (kg/m 2 ) 24.3 ± 3.3 FPG (mmol/l) 5.7 ± h PG (mmol/l) 6.9 ± 2.7 HbA1c (mmol/mol) 44 ± 8.7 HbA1c (%) 6.2 ± 0.8 Total cholesterol (mmol/l) 4.9 ± 0.9 HDL cholesterol (mmol/l) 1.1 ± 0.3 Triglyceride (mmol/l) 3.7 ± 2.4 Serum creatinine (mg/dl) 0.9 ± 0.2 Hypertension (%) 22.4 Smoker (%) 31.0 Diabetes duration (years) 8.0 ± 6.6 Current diabetic medication (%) 11.6 Diabetic retinopathy, n (%) Any retinopathy 63 (1.9) Moderate or severer retinopathy 35 (1.0) Data are mean ± SD or column percentage FPG fasting plasma glucose; 2h PG 2 h post oral glucose load plasma glucose and moderate/severer retinopathy was 7.1 % and 4.1 %, respectively. Figure 1 shows the cross-sectional frequency distributions of any diabetic retinopathy and moderate/severer retinopathy by HbA1c (Fig. 1a) and FPG (Fig. 1b) deciles. The frequency of any and moderate/severer retinopathy tended to increase, particularly with increasing ranges of HbA1c. The prevalence of retinopathy was very low until the 9th decile for HbA1c [48 51 mmol/mol ( %)]. Logistic regression analysis revealed statistically significant differences in the multivariate adjusted ORs compared with the first deciles that occurred from the 10th deciles of HbA1c for any diabetic retinopathy [C52 mmol/ mol ; OR 13.1 (95 % CI ), P \ 0.001] and for moderate or severer retinopathy [ ; OR 7.2 [95 % CI ], P = 0.01) (Table 2). Table 3 presents sensitivity, specificity as well as positive and negative predictive value for diabetic retinopathy at different HbA1c cutoffs. The optimal HbA1c cutoff derived from the ROC curve analysis for any diabetic retinopathy was 49 mmol/mol (6.6 %), with 76.2 % sensitivity and 84.2 % specificity (Fig. 2a). The positive and negative predictive values (PPV and NPV) of this cutpoint were 8.4 % and 99.5 %, respectively. For moderate or severer retinopathy, the HbA1c cutoff of 52 mmol/mol produced the highest sum of sensitivity (77.1 %) and specificity (88.7 %), with PPV of 6.6 % and NPV of Fig. 1 Relationship between HbA1c and diabetic retinopathy: prevalence of retinopathy according to a HbA1c and b fasting plasma glucose (FPG) deciles

4 Table 2 HbA1c cutoff above which the prevalence of diabetic retinopathy increases Any diabetic retinopathy Logistic regression (decile distribution) Unadjusted Multivariate adjusted a ROC curve analysis (continuous distribution) ROC receiver operating characteristic C49 mmol/mol (6.6 %) Moderate or severer retinopathy a Adjusted for age, gender, systolic blood pressure, and cigarette smoking 99.7 % (Fig. 2b). The ADA-proposed HbA1c threshold of 48 mmol/mol (6.5 %) produced similar accuracy for detecting moderate or severer retinopathy with 82.9 % sensitivity, 80.1 % specificity, 4.2 % PPV, and 99.8 % NPV. We did the subsidiary analyses to examine the relation between FPG and diabetic retinopathy. Statistically significant difference was observed in the adjusted ORs compared with the first deciles that occurred from the 10th deciles of FPG for any diabetic retinopathy [C6.9 mmol/l; OR 3.5 (95 % CI ), P = 0.004] and for moderate or severer retinopathy [OR 4.0 (95 % CI ], P = 0.015). The optimal FPG thresholds using ROC curve analysis for any diabetic retinopathy was 6.0 mmol/l, with 69.8 % sensitivity and 77.1 % specificity. For moderate or severer retinopathy, the FPG cutoff of 6.3 mmol/l produced the highest sum of sensitivity (71.4 %) and specificity (82.8 %). The overall discriminatory power determined by AUC of HbA1c was good and greater than that for FPG, although the differences did not reach statistical significance, for identification of any diabetic retinopathy (0.83 [95 % CI ] vs [95 % CI ], P = 0.07) (Fig. 2a) and moderate/severer retinopathy (0.84 [95 % CI ] vs 0.77 [95 % CI ], respectively; P = 0.28) (Fig. 2b). Discussion In this community-based cross-sectional study, an HbA1c cutoff of 49 mmol/mol (6.6 %) and 52 mmol/mol best detected the presence of any diabetic retinopathy and moderate/severer retinopathy, respectively. The proposed HbA1c cutoff C48 mmol/mol (6.5 %) produced similar accuracy compared to that of optimal cutoff derived from ROC curve analysis for detecting diabetic retinopathy. These results are consistent with a recent south Asian study that demonstrated the optimal HbA1c thresholds for detecting mild and moderate retinopathy were 49 mmol/ mol (6.6 %) and 53 mmol/mol (7.0 %), respectively [17]. Table 3 Sensitivity, specificity, and positive and negative predictive value for diabetic retinopathy at different HbA1c cutoffs HbA1c Number of retinopathy cases/total number of participants (%) Sensitivity (%) Cutoff Below the cutoff Above the cutoff Specificity (%) Positive Predictive value (%) Negative Predictive value (%) Any diabetic retinopathy C43 mmol/mol (6.1 %) 8/1,780 (0.4) 55/1,622 (3.4) C45 mmol/mol (6.3 %) 13/2,465 (0.5) 50/937 (5.3) C48 mmol/mol (6.5 %) 14/2,704 (0.5) 49/698 (7.0) mmol/mol (6.6 %) 15/2,828 (0.5) 48/574 (8.4) C50 mmol/mol (6.7 %) 18/2,890 (0.6) 45/512 (8.8) C51 mmol/mol (6.8 %) 20/2,951 (0.7) 43/451 (9.5) /2,993 (0.7) 43/409 (10.5) C53 mmol/mol (7.0 %) 21/3,024 (0.7) 42/378 (11.1) Moderate or severer retinopathy C43 mmol/mol (6.1 %) 4/1,780 (0.2) 31/1,622 (1.9) C45 mmol/mol (6.3 %) 5/2,465 (0.2) 30/937 (3.2) C48 mmol/mol (6.5 %) 6/2,704 (0.2) 29/698 (4.2) C49 mmol/mol (6.6 %) 7/2,828 (0.2) 28/574 (4.9) C50 mmol/mol (6.7 %) 8/2,890 (0.3) 27/512 (5.3) C51 mmol/mol (6.8 %) 8/2,951 (0.3) 27/451 (6.0) mmol/mol 8/2,993 (0.3) 27/409 (6.6) C53 mmol/mol (7.0 %) 9/3,024 (0.3) 26/378 (6.9) HbA1c cutoffs in bold indicate the optimal values derived from the receiver operating characteristic curve analysis

5 Fig. 2 Receiver operating characteristic curves for a any diabetic retinopathy and b moderate or severer retinopathy In a data pooling analysis of five studies using a DCCTaligned HbA1c assay, conducted in , the optimal HbA1c cutoff for moderate or severer retinopathy was mmol/mol ( %) [13]. Recent French cohort study further evaluated the predictive performance of HbA1c on diabetic retinopathy 10 years later, which increased sharply from 42 mmol/mol (6.0 %) of HbA1c level [18]. The main strengths of our study are its communitybased cohort sample in a homogeneous ethnic background, the precise interpretation of retinal photography by a trained eye care provider, stringent method for detecting diabetes using 75 g OGTT in central laboratory analysis, and the standardization of the HbA1c values to a DCCTaligned assay. Diagnostic cutpoints of HbA1c in diabetes have been changed on consensus judgments regarding optimal sensitivity and specificity for diabetic retinopathy [19]. In 1997, the ADA recommended that the FPG cutpoint be lowered to 7.0 mmol/l from 7.8 mmol/l because several studies had reported a linear increase in the prevalence of retinopathy above this level and former correlated better with a 2 h PG C 11.1 mmol/l [4, 20, 21]. In 2010, ADA proposed HbA1c C 48 mmol/mol (6.5 %) as the diagnostic cutpoint for diabetes based especially on a data pooling analysis reporting the close association of HbA1c with diabetes-specific retinopathy (defined as moderate or severer retinopathy) [3, 13, 22]. To our knowledge, diagnostic accuracy of this proposed HbA1c cutoff for detecting diabetic retinopathy has not yet been properly tested in an independent populationbased sample. Although HbA1c is a convenient and practical tool for diagnosing diabetes, it is important to note that the sensitivity of the proposed HbA1c threshold [C48 mmol/mol (6.5 %)] for detecting diabetes, defined by glycemic measures, has been less than 50 % in many reports [2, 23]. In our recent study, converted to a DCCT-aligned assay, the sensitivity of C48 mmol/mol (6.5 %) cutoff was only 52.3 % with 96.8 % specificity; rather, an HbA1c value of 44 mmol/mol (6.2 %) best detected patients with undiagnosed diabetes defined by OGTT criteria with 67.6 % sensitivity and 90.7 % specificity [8]. This raises concerns that an overreliance on the proposed HbA1c criterion for diagnosis of diabetes could lead to delayed diagnoses and therapeutic interventions, most probably in patients with early diabetes [19, 24]. Primary care physicians should consider the limited sensitivity of HbA1c when choosing the test as a diagnostic measure for diabetes. Furthermore, some reports suggested that higher postprandial glucose in patients with diabetes was associated with a greater risk of diabetic complications, even after adjustment for HbA1c levels [25]. Our study has some limitations. Due to including several thousand people in this large cohort study, we used single-field retinal photography for measuring diabetic retinopathy. Whereas multiple-field photography may more efficiently detect early-stage retinopathy [26], especially peripheral lesions. However, compared to 7-field standard stereoscopic fundus photographs, single-field nonmydriatic retinal photography showed perfect agreement for identifying people with and without diabetic retinopathy and moderate agreement for retinopathy grading [27]. A systemic review of past studies showed that although the single-field fundus photography cannot be a substitute for a comprehensive ophthalmic examination, there is level I evidence that it can serve as a screening tool for diabetic retinopathy to identify patients for referral for ophthalmic evaluation and management [28]. To maintain the

6 community-based nature of the study and to replicate the methods of previous studies, we included participants with diabetes who were taking antidiabetic medications [4, 13, 20]. This was based on cross-sectional data and we are thus unable to define the relation of HbA1c to incident diabetic retinopathy. In addition, the study was conducted in a homogenous Korean population. Caution is needed in extrapolating the study results to other ethnic groups. In conclusion, although the proposed HbA1c cutoff [C48 mmol/mol (6.5 %)] and 75 g OGTT criteria are frequently discordant for diagnosing diabetes, this study confirmed that the proposed HbA1c threshold of 48 mmol/ mol (6.5 %), nevertheless, allowed proper detection of both any diabetic retinopathy and moderate/severer retinopathy. Our data support the judicious use of HbA1c for the diagnosis of diabetes and detecting diabetic retinopathy as well. Acknowledgments This study was supported by the National Genome Research Institute, the Korean Center for Disease Control and Prevention (contracts # , # , and # ). The funding source had no role in the collection of the data or in the decision to submit the manuscript for publication. Conflict of interest References None. 1. Selvin E, Steffes MW, Zhu H et al (2010) Glycated hemoglobin, diabetes, and cardiovascular risk in nondiabetic adults. N Engl J Med 362: Bonora E, Tuomilehto J (2011) The pros and cons of diagnosing diabetes with A1C. Diabetes Care 34(Suppl 2):S American Diabetes Association (2010) Diagnosis and classification of diabetes mellitus. Diabetes Care 33(Suppl 1):S Expert Committee on the Diagnosis and Classification of Diabetes Mellitus (1997) Report of the expert committee on the diagnosis and classification of diabetes mellitus. Diabetes Care 20: Kramer CK, Rodrigues TC, Canani LH et al (2011) Diabetic retinopathy predicts all-cause mortality and cardiovascular events in both type 1 and 2 diabetes: meta-analysis of observational studies. Diabetes Care 34: Karlberg C, Falk C, Green A et al (2012) Proliferative retinopathy predicts nephropathy: a 25-year follow-up study of type 1 diabetic patients. Acta Diabetol 49: Mohamed Q, Gillies MC, Wong TY (2007) Management of diabetic retinopathy: a systematic review. JAMA 298: Choi SH, Kim TH, Lim S et al (2011) Hemoglobin A1c as a diagnostic tool for diabetes screening and new-onset diabetes prediction: a 6-year community-based prospective study. Diabetes Care 34: STAT-Korea (2003) Census. Daejeon, Korea, Korea National Statistical Office 10. Hanas R, John G, International HBA1c Concensus Committee (2010) 2010 consensus statement on the worldwide standardization of the hemoglobin A1C measurement. Diabetes Care 33: Lin DY, Blumenkranz MS, Brothers RJ et al (2002) The sensitivity and specificity of single-field nonmydriatic monochromatic digital fundus photography with remote image interpretation for diabetic retinopathy screening: a comparison with ophthalmoscopy and standardized mydriatic color photography. Am J Ophthalmol 134: American Academy of Ophthalmology (2002) International clinical diabetic retinopathy disease severity scale. Available at: Last Accessed 12 Oct Colagiuri S, Lee CM, Wong TY et al (2011) Glycemic thresholds for diabetes-specific retinopathy: implications for diagnostic criteria for diabetes. Diabetes Care 34: Stratton IM, Kohner EM, Aldington SJ et al (2001) UKPDS 50: risk factors for incidence and progression of retinopathy in Type II diabetes over 6 years from diagnosis. Diabetologia 44: Yu T, Mitchell P, Berry G et al (1998) Retinopathy in older persons without diabetes and its relationship to hypertension. Arch Ophthalmol 116: DeLong ER, DeLong DM, Clarke-Pearson DL (1988) Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44: Sabanayagam C, Liew G, Tai ES et al (2009) Relationship between glycated haemoglobin and microvascular complications: is there a natural cut-off point for the diagnosis of diabetes? Diabetologia 52: Massin P, Lange C, Tichet J et al (2011) Hemoglobin A1c and fasting plasma glucose levels as predictors of retinopathy at 10 years: the French DESIR study. Arch Ophthalmol 129: Malkani S, Mordes JP (2011) Implications of using hemoglobin A1C for diagnosing diabetes mellitus. Am J Med 124: McCance DR, Hanson RL, Charles MA et al (1994) Comparison of tests for glycated haemoglobin and fasting and two hour plasma glucose concentrations as diagnostic methods for diabetes. BMJ 308: Engelgau MM, Thompson TJ, Herman WH et al (1997) Comparison of fasting and 2-h glucose and HbA1c levels for diagnosing diabetes. Diagnostic criteria and performance revisited. Diabetes Care 20: International Expert Committee (2009) International Expert Committee report on the role of the A1C assay in the diagnosis of diabetes. Diabetes Care 32: Cowie CC, Rust KF, Byrd-Holt DD et al (2010) Prevalence of diabetes and high risk for diabetes using A1C criteria in the U.S. population in Diabetes Care 33: Kowall B, Rathmann W, Landgraf R (2011) Is HbA1c a valid and feasible tool for the diagnosis of diabetes? Diabetes Res Clin Pract 93: Mannucci E, Monami M, Lamanna C et al (2012) Post-prandial glucose and diabetic complications: systematic review of observational studies. Acta Diabetol 49: van Leiden HA, Moll AC, Dekker JM et al (2003) Photography or ophthalmoscopy for detection of diabetic retinopathy? Diabetes Care 26: Vujosevic S, Benetti E, Massignan F et al (2009) Screening for diabetic retinopathy: 1 and 3 nonmydriatic 45-degree digital fundus photographs vs 7 standard early treatment diabetic retinopathy study fields. Am J Ophthalmol 148: Williams GA, Scott IU, Haller JA et al (2004) Single-field fundus photography for diabetic retinopathy screening: a report by the American Academy of Ophthalmology. Ophthalmology 111: