To evaluate the clinical efficacy of ambulatory electrocardiographic (ECG) monitoring and to develop guidelines for its use in clinical practice. Studies reported since January 1978 were identified both through computer searches using Index Medicus and extensive manual searching of bibliographies of identified articles. Only studies that fulfilled methodologic criteria designed to limit bias were reviewed. Information describing population and study results was assessed in four major categories (variability, diagnosis, prognosis, and therapy guidance) for both arrhythmia monitoring and ST-segment analysis. The day-to-day variability of arrhythmia and myocardial ischemia detected by ambulatory ECG monitoring may be considerable in an individual patient. Caution must therefore be used in interpreting serial tests. Ambulatory ECG monitoring with diary correlation permits documentation of cardiac arrhythmias causing symptoms, but the diagnostic yield is low unless symptoms are frequent. Such monitoring can provide information about prognosis in patients after acute myocardial infarction. The amount of prognostic information obtained is modest and is outweighed by other measures. There is insufficient information to make conclusions about such monitoring and prognosis in other conditions. Serial ambulatory ECG monitoring may be used to assess the effect of an antiarrhythmic drug in patients with frequent and reproducible ventricular ectopy. The effect of arrhythmia suppression on survival is uncertain. Because of its low sensitivity and specificity, analysis of ST-segment changes during ambulatory ECG monitoring is inaccurate in establishing or excluding the presence of coronary disease. Although anti-ischemic interventions reduce the frequency and duration of ST-segment changes on monitoring, there are no data on the utility of using reduction or elimination of the changes as the endpoint of therapy. Ambulatory ECG monitoring can provide diagnostic, prognostic, and therapeutic information in many situations, but similar information often may be better obtained in other ways.