The use of reflectance meters to screen for gestational diabetes has been advocated because of convenience and apparent accuracy. The present study addresses the possibility that imprecision of reflectance meters may affect the reliability of the blood sugar values obtained with their use. We tested four reflectance meters (Glucometer, Glucometer II, Glucoscan 3000, and Accuchek II) on 84 gravidas undergoing 1-hour, 50-g glucose screening, and calculated average coefficients of variation (9.6 +/- 10.9, 8.2 +/- 4.7, 6.8 +/- 4.5, and 7.4 +/- 4.4%, respectively). One-way analysis of variance testing indicated no significant difference between these values (P = .11). However, comparison with the coefficients of variation from standard laboratory technology (YSI Model 23A Glucose Analyzer) applied to simultaneously obtained venous samples (2.3 +/- 1.5, 1.5 +/- 0.6, 1.9 +/- 1.0, and 1.0 +/- 0.5%) revealed that the meters all had coefficients of variation significantly higher (P less than .0001). Use of reflectance meters would have resulted in 45.2% (38 of 84) of the subjects undergoing 3-hour oral glucose tolerance tests. Use of standard laboratory values would have indicated these tests in only 16% (14 of 84). The inconvenience and increased cost of these oral glucose tolerance tests suggest that meters may not be appropriate for screening.