Analysis of the method for conversion between levels of HbA 1c and glycated albumin by linear regression analysis using a measurement error model

Abstract

Aim To establish a method for conversion between HbA 1c and glycated albumin (GA) using a measurement error model (MEM). Methods Type 2 diabetic patients, without complications that might affect either HbA 1c or GA, were enrolled in the study ( n = 154, age 68.4 ± 9.9). HbA 1c, GA and postprandial plasma glucose (PPG) levels were measured simultaneously on ≥3 occasions. Results PPG showed a significant correlation with HbA 1c and GA ( p < 0.001 for both). Correlation between HbA 1c and GA was very high ( r = 0.747, p < 0.001). When the independent variable was assumed to be GA, common regression analysis yielded a regression line HbA 1c = 2.59 + 0.204 GA. When the independent variable was changed to HbA 1c, the regression line became GA = 2.26 + 2.74 HbA 1c. The y-intercept of the first line was significantly positive, whereas that of the second was not. The regression line using MEM was HbA 1c = 1.73 + 0.245 GA. The y-intercept was 1.73 ± 0.38 ( p < 0.001) and the slope was 0.245 ± 0.018 ( p < 0.001), showing that 1% increase in HbA 1c level corresponds to 4% increase in GA level. Conclusions The relationship between HbA 1c and GA was examined by regression analysis using MEM. HbA 1c levels in Japan appear to have a positive shift of approximately 1.7%. Incremental ratio 4 of GA vs. HbA 1c showed good consistency with values derived from in vitro data.