Purpose: The heart rate response (HRR, percentage change from baseline) to regadenoson during myocardial perfusion imaging (MPI) can provide incremental prognostic value in patients with known or suspected coronary artery disease. Our purpose was to evaluate the variability and prognostic value of HRR on serial measurements. Methods: We studied 648 consecutive patients (61 ± 11 years, 48 % with diabetes) who underwent two regadenoson MPI studies (16 ± 9 months between studies). HRR <30 % was defined as abnormal. All-cause mortality was determined by chart review and verified using the US Social Security Death Master File. Results: HRR was well correlated between the two studies (intraclass correlation coefficient 0.72, 95 % CI 0.67 – 0.76) with no systematic bias (mean difference 0.88 %, p = 0.2) or proportional bias (p = 0.5) by Bland-Altman analysis in all patients and in those with normal MPI on both studies. Of the 308 patients (48 %) with normal baseline HRR (HRR-1), 33 % had developed a blunted HRR on the second MPI study (HRR-2). Older age, male gender, end-stage renal disease, and abnormal baseline left ventricular ejection fraction were independent predictors of a new-onset abnormal HRR. During a mean follow-up of 2.4 ± 1.2 years, 55 patients (8.5 %) died. Patients with a blunted HRR-1 had increased mortality risk irrespective of their HRR-2 (p = 0.9, log-rank test). Among patients with normal HRR-1, a blunted HRR-2 was an independent predictor of all-cause mortality beyond clinical and traditional MPI data (hazard ratio 2.83, 95 % CI 1.14 – 7.03). Finally, patients with a normal HRR-1 and HRR-2 had the lowest event rate, while those with any abnormal HRR had an increased risk of death (hazard ratio 2.5, 95 % CI 1.2 – 5.4). Conclusion: There was good correlation in the HRR to regadenoson on serial measurements without systematic or proportional biases. Patients with consistently normal HRR had the best prognosis.