PubMed1997
To assess the performance of different cutoff points for fasting plasma glucose (FPG) including the World Health Organization (WHO) cutoff of FPG > and = 7.8 mmol l-1 for the screening of diabetes mellitus, data from oral glucose tolerance tests done on 865 Singaporean adults over a 4-year period were analysed. The relationship between FPG and 2 h post-load plasma glucose (2H-PG) was determined using regression analysis and receiver operating characteristic (ROC) analysis. Out of 220 subjects with 2H-PG exceeding 11.1 mmol l-1, only 90 had FPG exceeding 7.8 mmol l-1, yielding a sensitivity of only 40.9%. The relationship between FPG and 2H-PG was non-linear and conformed to a cubic equation [FPG = 3.0890 + 0.0272 (2H-PG)2 - 0.0006 (2H-PG)3]. ROC analysis determined an FPG of 5.7 mmol l-1 to yield optimal sensitivity and specificity. Although in the clinical setting, FPG is an extremely useful screening test for diabetes, it is clear, as has been shown in other studies, that the traditional WHO criterion of FPG > and = 7.8 mmol l-1 is highly insensitive for detecting diabetes. Lower FPG cutoffs achieve greater sensitivity albeit at the expense of reduced specificity. Although ROC analysis in this population determines the optimally sensitive and specific FPG cutoff to be 5.7 mmol l-1, we suggest that the FPG cutoff should be adjusted to 7.0 mmol l-1 for preservation of specificity.