The fructosamine assay is about to be used for the control of diabetic patients and for the diagnosis of diabetes mellitus. An analysis of the original literature is presented to answer the question whether the evaluation of the test is methodologically reliable. Altogether 65 studies are evaluated. Thirty of these studies deal with technical aspects of the test (phase 1). A comparison of the distributions of test results for diabetic and non-diabetic patients is presented in 28 studies (phase 2). However, serious shortcomings in methodology and presentation are apparent. "Significant" differences and correlation coefficients are the basis for optimistic conclusions. Only 7 studies are identified where fundamental test parameters such as sensitivity and specificity are given, or can be calculated (phase 3). None of them achieves a satisfactory standard of methodology (selection of patients unclear, assessment made not mutually blind). There is no reliable phase-3-study for the metabolic control of diabetes mellitus. It can be concluded that the fructosamine test is not evaluated sufficiently to allow its routine clinical use. In fact, contrary to a widely quoted opinion, the results presented so far do not suggest a reliable test.