Utility of HbA_1c and fasting plasma glucose for screening of Type 2 diabetes: a meta-analysis of full ROC curves

There is still debate on the optimal threshold for population-based screening of diabetes (diagnosed by the oral glucose tolerance test) using tests like HbA_1c or fasting plasma glucose. Meta-analyses provide meaningful input in such situations. The aim of this article is to perform a meta-analysis that includes the complete information reported in the individual studies of two existing systematic reviews.

We screened the individual studies from two systematic reviews and reconstructed the full four-fold tables for every reported threshold. Using a recently proposed meta-analysis model for the comparison of two diagnostic tests, we compared HbA_1c with fasting plasma glucose, and estimated meta-analytic receiver operating characteristic curves for both tests using the 11.1 mmol/l threshold of the 2-h post-challenge glucose level (2 h-PG) as the gold standard.

We included nine studies from two existing systematic reviews in our analysis. Based on our data set, the optimal threshold lies between 42 and 44 mmol/mol (6.0-6.2%) for HbA_1c , and 6.2-6.4 mmol/l for fasting plasma glucose choosing the Youden index as the technical criterion. In addition, we found that there is no relevant difference in the performance of HbA_1c and fasting plasma glucose.

In our meta-analysis, we found that the optimal threshold with reference to the 2 h-PG should be chosen between 42 and 44 mmol/mol (6.0-6.2%) for HbA_1c , and 6.2-6.4 mmol/l for fasting plasma glucose on the basis of maximal sensitivity and specificity.