This study was conducted (a) to establish a normal cut-off value for glycos
ylated haemoglobin measured as HbA(1a) in South Indian subjects, and (b) to
evaluate its usefulness in demarcating different categories of glucose int
olerance. HbA(1c) measurement was carried out in 1261 cases with no known h
istory of diabetes, while being tested by oral glucose tolerance test (M:F
850:411, mean age 40 +/- 12 years). An immunoturbidimetric procedure for Hb
A(1c) assay (Tina-Quant, Boehringer Mannheim, Germany) was used. The specif
icity and sensitivity of HbA(1c) in demarcating normal glucose tolerance (N
GT) from abnormal tolerance were calculated using the ROC procedure. By the
ROC analysis, a cut-off Value of HbA(1c) greater than or equal to 6.0% gav
e a sensitivity of 88.5% and specificity of 62.8% using the WHO criteria (2
-h plasma glucose greater than or equal to 200 mg/dl). Using the ADA criter
ion (fasting plasma glucose > 125 mg/dl) the sensitivity and specificity fo
r the same cut-off value were 85.2 and 61.2%. In NGT, only a small percenta
ge of the variance in HbA(1c) was explained by the fasting plasma glucose (
FPG) values. The overall correlation coefficient between the fasting plasma
glucose and HbA(1c) was r = 0.8, r(2) = 0.64 and, in the case of 2-h post
glucose, r = 0.82, r(2) = 0.67. This showed that more than 35% of the varia
tions in HbA,, were not explained by the plasma glucose values. The study s
howed that HbA(1c) values of 2 6.0% gave a reasonably high sensitivity and
specificity for diagnosis using the WHO or ADA criteria. However, nearly 35
% of the variations in HbA(1c) were not explained by the variations in plas
ma glucose. Wide inter-individual variations even in the normoglycaemic ran
ge make the test unsuitable for diagnostic purpose. (C) 2000 Elsevier Scien
ce Ireland Ltd. All rights reserved.