Geographic variations in sperm counts: a potential cause of bias in studies of semen quality