Positron emission tomography studies have shown that caffeine impairs myocardial blood flow response to physical exercise,10,11 dipyridamole,12,13 and adenosine triphosphate13 but not regadenoson.14 The effect of caffeine administration on adenosine-induced myocardial blood flow response has not been studied using positron emission tomography but intravenous caffeine (4 mg/kg resulting in serum caffeine range of 2-8 with a mean of 3.7 mg/L) infusion had no effect on adenosine-induced myocardial hyperemia as assessed by fractional flow reserve using a pressure wire in subjects with angiographic evidence of coronary stenosis.15 Although these agents all ultimately work by activating adenosine A_2A receptors in the coronary circulation, it cannot be stressed enough that they induce variable effects on myocardial blood flow depending on the interstitial availability of adenosine or its receptor agonist (regadenoson) locally. Further, it is unclear whether a modest decrease in the myocardial blood flow reserve with caffeine to the level detected in these studies will affect myocardial perfusion by MPI since nuclear tracer extraction plateaus at myocardial blood flow lower than that seen after caffeine supplementation in these studies. For example, in the study by Bottcher et al12 dipyridamole-induced myocardial blood flow reserve decreased from 3.35 ± 0.75 to 2.3 ± 0.75 after caffeine administration (one or two cups of coffee 1-4 hours prior to MPI). The authors concluded that since the flow reserve was above 2 in the majority of subjects even after caffeine, “the diagnostic accuracy of pharmacological stress testing might not be altered” by caffeine intake.

Smits et al16 were the first to evaluate the effect of caffeine on vasodilator MPI. In a study on 8 patients with perfusion abnormalities on dipyridamole MPI, intravenous caffeine administration (4 mg/kg, serum caffeine 9.7 ± 1.3 mg/L) resulted in negative tests in 6 of the 8 patients with the other 2 showing near-complete redistribution in more segments after caffeine but lower number of segments with mild redistribution. On a semi-quantitative score (15 segments; 0 = no, 1 = mild, and 2 = near complete redistribution), caffeine administration resulted in a decrease from 9.0 ± 0.9 to 2.0 ± 1.1. Zoghbi et al17 studied 30 subject with known or high likelihood of coronary artery disease who had evidence of reversible defect on a clinically indicated adenosine MPI. A second adenosine MPI was performed 1 hour after the subjects drank an 8 oz cup of brewed coffee prepared similarly for all the subjects who abstained from caffeine for 24 hours prior to both MPIs. The caffeine study was performed using the same protocol and tracer as the initial study. Caffeine administration (serum caffeine range of 1-7, mean 3.1 mg/L) did not affect automated quantitative perfusion defect size (12.6 ± 10.1 vs 12.4 ± 10.4, P = .6). The summed stress (SSS), summed rest (SRS), and summed difference (SDS) scores were similarly unaffected by caffeine administration on blinded side-by-side interpretation. The SDS with and without caffeine administration were highly correlated (r = 0.88, P < .0001) and there was no difference in the defect size with and without caffeine when the subjects were stratified by caffeine blood levels. Reyes et al18 also studied 30 subjects with known or suspected coronary artery disease who underwent clinically indicated adenosine MPI and had unequivocal reversible perfusion defect at baseline. Subjects were asked to abstain from caffeine for 12 hours before MPI. 1 hour after receiving 2 large shots of espresso coffee (~200 mg caffeine, serum caffeine range of 2.9-12.3 mg/L), 12 subjects underwent a repeat adenosine MPI using the standard adenosine dose of 140 μg/kg/min for 6 min), while 18 subjects received a higher adenosine dose of 210 μg/kg/min for 6 min. All the groups performed supplemental exercise in addition to adenosine. SSS, SRS, and SDS were determined by 2 blinded readers and a third reader was involved in case of disagreement. SDS decreased from 12.0 ± 4.4 at baseline to 4.1 ± 2.1 after caffeine when the standard adenosine dose was used but did not significantly change with the high adenosine dose, 7.7 ± 4.0 at baseline to 7.8 ± 4.2 with high-dose adenosine, P = .7. It is important to mention that even in this study, there was no relationship between the change in SDS from baseline to caffeine study and serum caffeine concentration (r = −0.26, P = .4). The higher dose of adenosine is not approved for use in the United States in imaging. Further, Wilson et al19 more than 2 decades ago, showed that the standard dose of adenosine produces near-maximal coronary hyperemia.

In an observational study of 86 patients who underwent adenosine or dipyridamole MPI and reported that they abstained from caffeine products for >24 hours, serum caffeine levels were detectable in 40% (range 0.1-5 mg/L) of subjects and the frequency of thallium redistribution was similar among those with undetectable (22 of 52 patients, 42%), low caffeine levels 0.1-0.9 mg/L (8 of 29 patients 28%), and high caffeine levels >1.0 mg/L (2 of 5 patients 40%).20

The study by Lee et al1 adds to this literature by assessing the extent and reversibility of the stress perfusion defect with and without caffeine supplementation in 30 patients who have inducible perfusion abnormalities on standard adenosine MPI. Consistent with previous studies, subjects had detectable levels of caffeine despite being instructed to abstain from any caffeine intake (range 0.025-1.790 with a mean of 0.268 mg/L). In addition to resumption of their usual caffeine consumption, subjects received an extra cup of coffee (containing 0.5, 1, or 2 teaspoons of instant coffee) 1 hour prior to the repeat adenosine MPI. Not surprisingly, caffeine levels increased in these subjects (range 0.706-10.430 with a mean of 3.374 mg/L, P < .001). There was no change in the stress percent defect or percent defect reversibility on automated or on visual semi-quantitative scoring between the 2 adenosine MPIs. Granted there is substantial variation between the 2 stress scans with regards to percent defect before and after caffeine supplementation (see Figure 2 in Lee et al1) but this has to be considered in context of the variability in reprocessing the same set of rest images twice (see Figure 1 in Lee et al1) and the lack of a clear association between the direction of change of percent defect size and caffeine supplementation. This study is also unique in supplementing subjects with a random dose of caffeine (in addition to the variable usual consumption) which resulted in a wide range of caffeine levels. The caffeine concentration (at baseline or after supplementation) was not associated with percent defect reversibility and the change in caffeine concentration from baseline to supplementation had no effect on percent defect reversibility (mean change −0.003 for every 100 μg/L, P = .97).

Adenosine, by activating A_2A receptors, is known to increase heart rate via a direct effect on the sympathetic nervous system.21,22 In the original case report on the effect of caffeine on dipyridamole MPI in 1989, Smits et al23 reported on a 46-year-old man who underwent dipyridamole MPI after sustaining a myocardial infarction. Dipyridamole infusion after a 36-hour caffeine abstinence period resulted in an impressive 6-mm ST depression on the electrocardiogram and a reversible perfusion defect on MPI. The heart rate increased from 74 to 102 beats/min. One week later the study was repeated 30 minutes after an infusion of 4 mg/kg intravenous caffeine. The heart rate increased from 79 to 88 beats/min consistent with the antagonistic effect of caffeine on the adenosine receptors. The ST changes were not seen on ECG and the reversible defect on MPI was much smaller. In the subsequent study by the same group on 8 patients described in the previous section, a lower increase in HR was again evident after caffeine administration (heart rate increased from a mean of 70 to 73 beats/min, P = NS after caffeine vs 71 to 86 beats/min with caffeine abstinence, P < .001).16 In the report by Bottcher et al,12 the heart rate response to dipyridamole was again blunted by caffeine and it was inversely related to serum caffeine levels. The studies by Zoghbi et al17 and Reyes et al18 showed that the heart rate response to adenosine was similar with and without caffeine. Since the effects of caffeine on the heart rate response to adenosine and dipyridamole are not consistent with its effect on MPI, and since the change in heart rate is a poor predictor of the change in myocardial blood flow,24 it is unlikely that the original thesis first proposed by Bottcher et al12 that the effect of caffeine on dipyridamole MPI is mediated via blunting of the heart rate response and thus the rate-pressure product is true. In a study on conscious dogs, increasing does of intravenous caffeine (1-10 mg/kg) significantly attenuated regadenoson-induced rise in heart rate but not the peak increase in coronary blood flow.7 In the current report by Lee et al1 the heart rate response to adenosine was lower after caffeine administration despite no change in the percent perfusion defect. The reason why the heart rate response was affected by caffeine in this study but not in the previous 2 studies is not clear. Since the change in heart rate response to adenosine carries prognostic significance independent from imaging,25 it will be important to settle this discrepancy and determine in future studies whether caffeine obscures the effect of the cardiac autonomic system on the heart rate response to adenosine (and regadenoson).

Although we believe that it is reasonable to ask patients reporting for adenosine MPI to abstain from caffeinated products the morning of the test, we anticipate that a significant proportion will not comply with this strict requirement. Rather than cancelling and/or rescheduling the test to another day which will place un-necessary burden on the nuclear laboratory and delay the diagnostic and prognostic information from the referring physician, we believe that the weight of the evidence presented here justifies the performance of adenosine MPI in patients who have ingested one cup of coffee (acknowledging the variability of caffeine concentration in each cup) more than 1 hour prior to the test (which we submit to be the usual scenario) without jeopardizing the test results. In patients suspected of ingesting larger quantities of caffeine (or within 1 hour of the test), a reasonable alternative is to proceed with the rest MPI to allow for serum caffeine levels to fall followed by the adenosine portion of the test. Our recommendation is consistent with that by Salcedo and Kern26 that “the adenosine study, however, should not be delayed or cancelled if the patient has ingested one cup of coffee more than 1 hr before the test. Additionally, several cups of coffee may even be permissible depending on how long prior to the scheduled start time the last cup was consumed, keeping in mind that the half-life of caffeine is around 2.5-4.5 hr.” It is time to revise the practice guidelines to allow for more flexibility in performing vasodilator stress tests in patients who have inadvertently ingested a reasonable quantity of caffeine.