This paper describes the content of an H-mode confinement database that has been assembled for the ITER project. Data are collected from six machines of very different sizes ASDEX, DIII-D, JET, JFT2M, PBX-M and PDX. A detailed description of the criteria that are used in the selection of data and the definition of each of the variables are given. After an analysis of the condition of the database, scalings of the energy confinement time with both the engineering and physics variables are determined. Both power law and offset linear models are used to fit the data and the predictions of the confinement time in ITER using the different formulations are presented.