Assessment of microvolt T-wave alternans (MTWA) by the spectral analysis (SA) method requires the heart rate to be stable within a certain range. This can be achieved by ventricular pacing (VP). We compared MTWA during short-term VP versus atrial pacing (AP). Patients presenting for evaluation of risk of sudden cardiac death underwent an invasive electrophysiologic study. The concordance of results of MTWA-VP with MTWA-AP was evaluated, as well as the specificity, sensitivity, negative, and positive predictive values of MTWA-VP versus MTWA-AP. The maximum recorded amplitude of MTWA (MValt) in concordant positive results, as well as noise levels in all, were compared in both pacing modes. We studied 42 consecutive patients, of whom 31 completed both tests (32 by AP, 40 by VP). Compared to AP, VP-MTWA SA had a sensitivity of 93%, specificity of 71%, negative predictive value of 92%, and positive predictive value of 72%. The results were concordant in 25 patients (80%, kappa = 0.62, P < 0.001). The noise level was significantly higher during VP than AP (1.4 +/- 0.8 vs 1 +/- 0.8, P < 0.01), and there was a trend toward a higher amplitude of TWA by VP (10.7 +/- 5.3 vs 7.8 +/- 3.9, P = 0.058). MTWA SA is more likely to be completed during VP than AP. Overall there was concordance between both tests. VP generates higher amplitudes and noise, and a higher percentage of nonnegative results.