We evaluated the use of adenosine, dobutamine and arbutamine with (99m)Tc-tetrofosmin myocardial perfusion imaging. Forty patients under investigation for suspected coronary artery disease were recruited. Each had a resting scan and two separate stress scans on different days, in a randomized cross-over study. Resultant images were blindly reported in 13 segments per scan as normal, reversible or fixed defects. A score was given (0-3) for segmental defect severity. Haemodynamic responses were as expected for each agent. Subjective side effect scores did not differ overall between agents. Adenosine caused a significantly higher incidence of abnormal taste (54%) than dobutamine and arbutamine (both 23%) and a lower incidence of palpitations (25% vs 69% and 54%, respectively), all P<0.05. Arbutamine caused significantly more chest pain than adenosine (77% vs 46%) though less flushing (35% vs 68%), both P<0.05. Comparison of the results obtained showed highly significant levels of segmental agreement for visual and semi-quantitative analysis between adenosine and arbutamine, kappa value and correlation coefficient of 0.78 and 0.86, respectively, dobutamine and adenosine 0.69 and 0.78, and arbutamine and dobutamine 0.75 and 0.78, all P<0.0001. Adenosine, arbutamine and dobutamine differ in their haemodynamic response and side effect profile but provide highly comparable results during (99m)Tc SPECT imaging.