Does exclusion of extreme reporters of energy intake (the "Goldberg cutoffs") reliably reduce or eliminate bias in nutrition studies? Analysis with illustrative associations of energy intake with health outcomes.