Bridging differences in outcomes of pharmacoepidemiological studies: design and first results of the PROTECT project