. . . "Several relevant sources were searched and attempts were made to minimise publication bias by including studies reported only as abstracts." .