Derivation & validation of glycosylated haemoglobin (HbA 1c ) cut-off value as a diagnostic test for type 2 diabetes in south Indian population

Abstract

Background & objectives: Glycosylated haemoglobin (HbA 1c ) has been in use for more than a decade, as a diagnostic test for type 2 diabetes. Validity of HbA 1c needs to be established in the ethnic population in which it is intended to be used. The objective of this study was to derive and validate a HbA 1c cut-off value for the diagnosis of type 2 diabetes in the ethnic population of Rayalaseema area of south India.

Methods: In this cross-sectional study, consecutive patients suspected to have type 2 diabetes underwent fasting plasma glucose (FPG) and 2 h post-load plasma glucose (2 h-PG) measurements after a 75 g glucose load and HbA 1c estimation. They were classified as having diabetes as per the American Diabetes Association criteria [(FPG ≥7 mmol/l (≥126 mg/dl) and/or 2 h-PG ≥11.1 mmol/l (≥200 mg/dl)]. In the training data set (n = 342), optimum cut-off value of HbA 1c for defining type 2 diabetes was derived by receiver-operator characteristic (ROC) curve method using oral glucose tolerance test results as gold standard. This cut-off was validated in a validation data set (n = 341).

Results: On applying HbA 1c cut-off value of >6.3 per cent (45 mmol/mol) to the training data set,sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for diagnosing type 2 diabetes were calculated to be 90.6, 85.2, 80.8 and 93.0 per cent, respectively. When the same cut-off value was applied to the validation data set, sensitivity, specificity, PPV and NPV were 88.8 , 81.9, 74.0 and 92.7 per cent, respectively, although the latter were consistently smaller than the proportions for the training data set, the differences being not significant.

Interpretation & conclusions: HbA 1c >6.3 per cent (45 mmol/mol) appears to be the optimal cut-off value for the diagnosis of type 2 diabetes applicable to the ethnic population of Rayalaseema area of Andhra Pradesh state in south India.

Fig. 1

Flow chart showing study design. OGTT = Oral glucose tolerance test; FPG = Fasting plasma glucose; HbA1c = Glycosylated haemoglobin.

Fig. 2

(A) Receiver-operator characteristic (ROC) curve along with 95 per cent confidence bounds for calculating the cut-off value for glycosylated haemoglobin for the diagnosis of type 2 diabetes using oral glucose tolerance test as gold standard. The area under the ROC curve = 0.930; standard error = 0.0143; 95 per cent confidence interval = 0.898-0.955; Z-statistic = 30.180; significance level P (area = 0.5) = 0.0001. (B) Interactive dot diagram for glycosylated haemoglobin (HbA1c) values observed in training data set patients with newly diagnosed type 2 diabetes and those without type 2 diabetes, diagnosed based on oral glucose tolerance test. The horizontal line depicts the cut-off value. PPV = Positive predictive value; NPV = Negative predictive value.

Figure 3

Graph plotted with specificity (%) on the X-axis and sensitivity (%) on the Y-axis along with 95 per cent confidence bands. The ideal sensitivity and specificity are indicated by the line that would begin in the lower left corner, go straight up to the upper left corner and then to the upper right corner (i.e., with 100% sensitivity and 100% specificity). The diagonal line, which bisects the graph with the area under this curve of 0.5, indicates that the test is uninformative/worthless. The line connecting the derived optimal glycosylated haemoglobin cut-off (glycosylated haemoglobin >6.3; sensitivity 90.6%, specificity 85.2%) and the ideal cut-off (sensitivity 100%, specificity 100%) is also shown. The inset shows the graph with the glycosylated haemoglobin (%) on the X-axis and error in judgement (%E) (with 95% confidence bands) on the Y-axis.