[Parental cigarette smoking and childhood risks of hepatoblastoma: OSCC data]

Historically health care data and especially health care cost data were not collected by disease. This is due on one side to a lack of readily available information on the diagnosis that led to the health care given and on the other side to the diversity of actors and organizations taking charge of patients. This results in a major difficulty of linking together incompatible and separate data. A few countries have tried to desagregate their global health cost data in a top-down approach. The limits of such undertakings appeared however quickly, not the least because of the difficulties in allocating the expenditures by sector and disease. Other approaches, probably more fecund in the long run, such as the one used by the SEER-Medicare database in the U.S. favour the linkage of individual patient clinical and cost data in a bottom-up approach. However one should not ignore the potential bias problems raised by the use of even large databases such as this one. A major advantage of clinical trials for costing purposes is that they include homogeneous groups of patients randomly allocated to several treatments for comparative purposes. They are therefore potentially better adapted for comparing new treatments with standard reference care.