Importance of standardization of hemoglobin A(1c) in the analysis of factors that predict hemoglobin A(1c) levels in non-diabetic residents of three distinct areas of Japan

We performed a statistical analysis to elucidate effects of standardized me asurement of hemoglobin A(1c) (HbA(1c)) on analysis of factors that affect HbA(1c) values. Subjects were participants in the Japan Public Health Cente r-based Prospective Study on Cancer and Cardiovascular Diseases, and a tota l of 1789 men and 3150 women in three distinct areas who did not have overt diabetes (HbA(1c)greater than or equal to6.1% or prior diagnosis) were ana lyzed. A different method of HbA(1c) assay was used in each area: high-perf ormance liquid chromatography in one area and a different immunochemical me thod in each of the other two areas. Then, calibration of HbA(1c) was perfo rmed using two HbA(1c) standards (5.5 and 10.5%,) provided by the Japan Dia betes Society. Analysis of co-variance was performed separately in men and women. When raw HbA(1c) data were used as the outcome,'area', which represe nts differences in assay systems, lifestyles, etc, had a significant effect on HbA(1c) levels. When calibrated HbA(1c) data were used, however, 'area' was no longer a significant factor. In the latter analysis, age and BMI we re the principal contributors to HbA(1c), and parental history of diabetes had a weak effect in women. Thus, standardization of HbA(1c) reduced the di fference between assay systems, and uncovered two common factors to determi ne HbA(1c) levels. (C) 2001 Elsevier Science Ireland Ltd. All rights reserv ed.