The empiric administration of 50% dextrose to all patients presenting to the ED with altered mental status is a standard of care predicated on the assumption that glucose administration is harmless to non-hypoglycemic patients. Considerable evidence now disputes this assumption. Glucose administration before complete cerebral ischemia in experimental animals worsens neurologic and histologic outcome. Administration of glucose during severe incomplete ischemia has a similar detrimental effect. The translation of these experimental findings into clinical practice has been slow, perhaps hindered by the frequent use of rodent models and large bolus doses of glucose. However, evidence is now provided by primate and human studies and by experimental designs using clinically relevant doses of glucose. These clinical and experimental findings in conjunction with the wide availability of a rapid bedside screen for hypoglycemia provide the rationale for an alteration in the standard of care. The empiric administration of glucose should be avoided in patients at risk for cerebral ischemia, such as those with acute stroke, impending cardiac arrest, or severe hypotension or receiving CPR. A bedside fingerstick blood glucose estimation should be performed immediately on all patients presenting with altered mental status. The administration of 50% dextrose should be reserved for those patients in whom hypoglycemia is demonstrated; this practice will uphold Hippocrates' most basic principle of clinical medicine, "The physician must ... do no harm."