The correlation of angina attacks with ST segment changes detected during ambulatory Holter monitoring was evaluated in patients with unstable angina. Forty hospitalized patients had one to three 24-hour Holter recordings each. Twenty-three patients had a cardiac catheterization, confirming significant coronary artery disease. The Holter recordings, scanned blindly by computer, were evaluated for ST segment shifts (defined as +/- 1.5 mm from baseline, lasting 60 seconds or longer). Angina attacks were carefully logged. Over the total forty patient experience, only 15 of 74 (20.3%) angina attacks had corresponding ST segment shifts on the Holter recordings. Nine of 34 (26.5%) angina attacks in the 23 patients who had a cardiac catheterization had corresponding ST segment shifts. A total of 159 ST segment shifts were recorded on these forty patients, but only 15 (9.4%) ST shifts corresponded to a time when the patients were actually experiencing angina attacks. The performance of the test procedure was quantified by use of Youden's J statistic. The aggregate J, over all patients, was 0.203 (J = 1.0 is perfect, J = 0.0 is useless). When consideration was restricted to patients with cardiac catheterization, the aggregate experience J was 0.263. Dealing with only the patients who had angina attacks during the monitoring, and computing the J statistic for each individual patient, the resulting mean J statistic was 0.146, with SEM = 0.0731. The Holter monitoring worked reasonably well in only 2 of the 14 patients who gave clear tests of the procedure. In an attempt to improve the performance of the procedure, 21 Holter recordings in eight patients were reread for ST segment shifts of only +/- 1 mm from baseline, lasting 30 seconds or longer. In these eight patients with rescanned Holter recordings, only five of 17 (29.4%) angina attacks resulted in an ST segment shift. In conclusion, ambulatory Holter recordings proved not to be a suitable method of documenting ST segment shifts during angina attacks in this study.