The standard fracture toughness test is usually limited in practical engineering projects due to the expensive and sophisticated experimental procedure or limited specimens. The traditional method to evaluate the structural integrity is to estimate the ductile-brittle transition region by Charpy tests. However, this fuzzy estimation is sometimes too conservative and cannot obtain quantitative fracture toughness data. In recent years, many achievements have been made in the study of the relations between Charpy impact test and fracture toughness test. Although acceptable, the errors caused by these formulas are rarely compared in detail formulas with stable deviation will be more suitable for research or engineering application. Therefore, this paper compares the data of four different ferritic steels to test the error stability of each formula, so as to provide reference for experimental research and engineering facilities. The reliability of each correlation is checked from the view of accuracy or appropriate conservatism. In this research, We found that the T 28J-T 0 correlations, the empirical formula proposed by Rolfe, Novak and Barsom (abbreviated as RNB) and the Mean-4 Procedure obtained by IGC-parameter give the reference temperatures more accurately or appropriately conservatively. These formulas should be preferred in practical application.