T reatment has one of two fundamental goals: the relief of symptoms or a change in natural history. In general, a treatment goal of symptom relief is easier to achieve. One treats each day for events and symptoms that day, and patient compliance is maintained when symptoms are relieved. Treatment that has as its goal a change in natural history, without providing symptom relief, is fundamentally more difficult: one treats today for events that might occur tomorrow-or months or years later. A major element in the revolution in therapeutics that began several decades ago, and continues, has been the increasing importance of our attempts to change the natural history of chronic processes. The revolution in what we have available for therapy has been accompanied by an evolution in expectations. What has been the change in expectations? For the physician, the key words are “reliable” and “predictable”: the expectations now include that there be reliable evidence that a therapeutic agent matches the claims made for it, and that the results will be predictable. Patients, vaguely aware of the enormous advances that medicine has made, expect treatment to be effective and to provide minimal side effects. Regulatory agencies also have their expectations, reflecting both the physicians’ and the patients’ needs. In the area of hypertension the expectations are especially stringent. The physician cannot provide the patient with symptom relief, because there are no symptoms. The short-term goal is to control the high blood pressure at a minimal cost in terms of adverse reactions and dollars; the long-term goal is to change natural history. The essence of treatment of hypertension, therefore, has been and will be prevention of events related to the high blood pressure. Diuretic agents, introduced more than two decades ago, represented the first major advance in the drug treatment of hypertension. They long have been a mainstay in treatment, and remain so. They are still among the most widely prescribed class of agents in this country and elsewhere. Recent emphasis has been on reduced thiazide doses. The importance of a decrease in serum potassium levels, the most common and predictable side effect of thiazide diuretic use, has remained controversial. Indeed, the debate is often passionate. As is often the case when controversy exists, there is some truth on both sides. On the one hand, all physicians would agree that patients specifically at risk because of a decrease in serum potassium concentration-those who use digitalis, show signs of ventricular irritability, or have shown evidence of severe hypokalemia while receiving a thiazide diuretic-should have prophylactic treatment. At the other end of the spectrum, there are patients who do well with low doses of thiazide diuretics, have no specific predisposition to risk because of hypokalemia, and maintain a normal serum potassium concentration. In general, these people are relatively young, female, and enjoy a healthy, well-balanced diet. Between those two poles is the domain of the controversy. Physicians treat hypertension today because of what tomorrow will bring. Among the common events that might arise tomorrow is myocardial infarction. Thus, another element in the debate has emerged. It has become clear that during the evolution of a myocardial infarct, patients with a serum potassium concentration below 3.5 meq/liter show a sharp increase in the risk of clinically important ventricular arrhythmias [l-6]. Whether or not the low serum potassium level is provoked by diuretic use seems to matter little. Why should the serum potassium concen-