Author + information
- Received January 29, 2004
- Revision received July 23, 2004
- Accepted August 2, 2004
- Published online November 16, 2004.
- Robert S. Foote, MD*,†,* (, )
- Justin D. Pearlman, MD, PhD, FACC*,†,‡,
- Alan H. Siegel, MD†,‡ and
- Kiang-Teck J. Yeo, PhD†,§
- ↵*Reprint requests and correspondence:
Dr. Robert S. Foote, Section of Cardiology, Dartmouth Hitchcock Medical Center, One Medical Center Drive, Lebanon, New Hampshire 03756
Objectives The purpose of this study was to examine the effect of exercise-induced ischemia on levels of B-type natriuretic peptide (BNP) and its inactive N-terminal fragment (NT-pro-BNP)and to determine whether measurement of these peptides can improve the diagnostic accuracy of exercise testing.
Background The ability of exercise testing to detect coronary artery disease (CAD) is limited by modest sensitivity and specificity. B-type natriuretic peptides (NT-pro-BNP and BNP) are released by ventricular myocytes in response to wall stress. We hypothesized that exercise-induced ischemia results in increased wall stress and triggers release of NT-pro-BNP and BNP.
Methods A total of 74 patients with known CAD, normal left ventricular function, and normal resting levels of NT-pro-BNP and BNP who were referred for exercise testing with radionuclide imaging, and 21 healthy volunteers, were enrolled. Blood was drawn before and after maximal exercise and analyzed for NT-pro-BNP and BNP.
Results Of the patients with CAD, 40 had ischemia on perfusion images and 34 did not. Median post-exercise increases in NT-pro-BNP and BNP (ΔNT-pro-BNP and ΔBNP) were approximately four-fold higher in the ischemic group than in the nonischemic group (ΔNT-pro-BNP 14.5 vs. 4 pg/ml, p < 0.0001; ΔBNP 36.5 vs. 7.5 pg/ml, p < 0.0001). In volunteers, median ΔNT-pro-BNP was almost identical to that of the nonischemic patient group. At equal specificity to the electrocardiogram (ECG) (58.8%), the sensitivities of ΔNT-pro-BNP and ΔBNP for detecting ischemia were 90% and 80%, respectively; in contrast, the sensitivity of the exercise ECG was 37.5%.
Conclusions Measurement of exercise-induced increases in BNPs more than doubles the sensitivity of the exercise test for detecting ischemia with no loss of specificity.
Exercise electrocardiography is the most widely used noninvasive method to detect the presence of coronary artery disease (CAD); however, its usefulness is limited by relatively modest sensitivity and specificity (1–3). Other more accurate noninvasive tests, such as exercise echocardiography and exercise testing with radionuclide imaging, are less widely available and considerably more expensive.
B-type natriuretic peptide (BNP) is a neurohormone with diuretic, vasodilatory, and renin-angiotensin-aldosterone antagonist effects. It is secreted primarily by cells in the ventricular wall in response to increases in wall stress (4–6). BNP and the inactive amino-terminal fragment of its prohormone (N-terminal fragment of B-type natriuretic peptide pro-hormone [NT-pro-BNP]) (7) have been shown to have diagnostic or prognostic value in a variety of left and right ventricular structural and functional abnormalities, particularly heart failure, (4,8,9) as well as in systolic (10,11) and diastolic (12,13) dysfunction, unstable angina (14,15), acute coronary syndromes (16–18), and myocardial infarction (19,20). In addition, two studies (21,22) have found evidence that tissue hypoxia alone may trigger release of BNP in the absence of left ventricular dysfunction.
Exercise-induced ischemia is known to produce wall-motion abnormalities in the affected area of the ventricle (23). We hypothesized that these wall motion abnormalities, or ischemia per se, might trigger release of NT-pro-BNP and BNP and that a post-exercise increase in peptide levels might serve as a marker of inducible ischemia in individual patients and thus improve the diagnostic accuracy of the exercise test.
This study was approved by the Committee for the Protection of Human Subjects of Dartmouth College and Dartmouth HitchcockMedical Center; written informed consent was obtained from all subjects.
Seventy-four consecutive patients with documented CAD (69 by coronary angiography, 5 by previous abnormal myocardial perfusion imaging study) and normal resting levels of NT-pro-BNP and BNP (based on 95th percentiles corrected for age and gender, provided by test manufacturers), who were referred for exercise stress testing with single-photon emission computed tomographic (SPECT) myocardial perfusion imaging were enrolled. Patients with a history of heart failure, atrial fibrillation, pacemakers, significant valvular disease (including replacement), age >80 years, echocardiographicleft ventricular ejection fraction <55%, or recent (<2 months) infarction or re-vascularization were excluded. Also excluded were patients taking digitalis, or whose resting electrocardiograms (ECGs) showed abnormalities that would preclude interpretation of exercise-induced changes, for example, left bundle-branch block, left ventricular hypertrophy, >1-mm ST-segment changes, or pre-excitation. Also enrolled were 21 healthy volunteers (mean age 21.1 years) with no history of cardiovascular disease or other significant illness.
Exercise testing/perfusion imaging
In the patients with CAD, exercise testing with myocardial perfusion imaging was performed using a dual isotope, rest-stress protocol. Four mCi 201thallous chloride were injected, and resting images were acquired using a Philips Irix three-headed gamma camera (Cleveland, Ohio). Patients then underwent symptom-limited exercise testing on a treadmill using a Bruce protocol. Exercise was terminated for fatigue, marked dyspnea, exercise-limiting angina, >20 mm Hg decrease in systolic blood pressure, or >3-mm ST-segmentdepression. No cases of serious arrhythmia or severe hypertension necessitating termination of exercise were observed. Ninety seconds before termination of exercise, 33 mCi of 99mtechnetium tetrofosmin (Amersham Healthcare, Arlington Heights, Illinois) were administered, and stress images were subsequently acquired with ECG gating. Healthy volunteers underwent symptom-limited exercise testing without myocardial perfusion imaging.
Before exercise and again at 1 min post-exercise, a venous blood sample was collected via an indwelling 20-gauge intravenous cannula. Samples were placed in ethylenediaminetetraacetic acid anticoagulated polyethylene tubes, and the plasma was separated, aliquoted, and frozen at −80°C until analysis.
Interpretation of test results
Exercise ECGs were interpreted by an experienced physician unaware of the interpretation of perfusion images and results of analysis of blood samples. The ECGs were interpreted as positive for ischemia if they showed ≥1 mm horizontal or downsloping ST-segment depression at 80 ms after the J-point during exercise or recovery. Those ECGs showing no significant ST-segment depression during exercise or recovery were interpreted as negative for ischemia at that level of exercise regardless of the maximal heart rate achieved.
Radionuclide SPECT images were interpreted by an experienced radiologist who had no knowledge of the clinical history, exercise test data, and the results of analysis of blood samples. Images were classified as having no perfusion defects, fixed defects only, fixed and reversible defects, or reversible defects only; the defects were also characterized by size, severity, and vascular territory. Images were also assessed independently of the radiologist's interpretation with a computer software program (QPS, Cedars Sinai, Los Angeles, California), using a 20-segment polar model which compares acquired photon counts in each segment to a gender-specific database of normal studies. Values from 0 to 4 are assigned to each segment, 0 being normal and 4 being no counts; the total is expressed as a summed stress score (SSS), a summed rest score (SRS), and a summed difference score (SDS), the latter indicating the degree of reversibility. Myocardial function was assessed using quantitated gated SPECT imaging.
Analysis of blood samples
Resting and post-exercise blood samples were analyzed in batches for NT-pro-BNP using an electrochemoluminescent immunoassay (Roche Diagnostics, Indianapolis, Indiana) on an Elecsys 1010 autoanalyzer, and for BNP using a fluorescent point-of-care immunoassay (Biosite, San Diego, California). Coefficients of variation for the assays were: NT-pro-BNP 1.3% to 2.4% and BNP 11.2% to 14.6 % (24). The NT-pro-BNP assays were repeated in a separate run at a later time.
SPSS (Chicago, Illinois), Microsoft Excel (Redmond, Washington), Analyse-IT (Analyse-IT Software Ltd., Leeds, England),and GraphPad Prism 4.0 (San Diego, California) statistical software were used in our analyses or construction of figures. For continuous variables, the Student ttest was used to compare means of values with parametric distributions and the Mann-Whitney rank sum test to compare medians of values with nonparametric distribution; the chi-square test was used to compare dichotomous variables. All tests were two-tailed. Logistic regression and linear binary correlations (Pearson's correlations) were performed with SPSS, which incorporated correction for multiple comparisons. Sensitivity, specificity, and other test characteristics for the exercise ECG and for the post-exercise change in NT-pro-BNP and BNP were calculated with Analyse-IT software, and receiver operator characteristic curves were constructed for the changes in peptide levels using the same program.
Based on the radiologist's blinded interpretation of radionuclide images, 40 of the 74 enrolled patients were classified as having perfusion defects on stress imaging that reversed at rest (“ischemic group”); 14 (35%) of these patients also had fixed defects. The remaining 34 patients had no fixed or reversible defects (“nonischemic group”). No patient had fixed defects only. Clinical characteristics of the two patient groups and healthy volunteers are shown in Table 1;the two patient groups were comparable in all respects except age (ischemic group mean 61.2 years, nonischemic group mean 55.9 years, p = 0.025) and a trend toward more frequent history of myocardial infarction in the ischemic group (55% vs. 23.5% in the nonischemic group, p = 0.056). (Student ttest was used for these and all other comparisons of mean values.) No other significant differences were found in clinical history, previous re-vascularization, or treatment with various commonly used medications.
Table 2summarizes the findings on exercise testing and gated imaging. Analysis of exercise test data showed no significant difference between the two patient groups in maximal exercise capacity, maximal systolic blood pressure, the presence of exertional chest pain, or Duke Treadmill Score. The percentage of patients who developed ECG changes characteristic of ischemia did not differ between the two groups (nonischemic 41.2%, ischemic 37.5%, p = 0.99). Maximal heart rate and rate-pressure product were higher in the nonischemic group, and post-exercise left ventricular ejection fraction by gated SPECT was lower in the ischemic group (51.8% vs. 57.8%, p = 0.001).
Table 3shows the results of analysis of blood samples. Volunteer blood was analyzed for NT-pro-BNP only, and pre-exercise (baseline) levels were normal in all subjects. In the ischemic and nonischemic patient groups, although baseline levels of NT-pro-BNP and BNP were within normal limits, median levels of both peptides were significantly higher in the ischemic group (NT-pro-BNP 120.5 vs. 53.5 pg/ml, p < 0.0001; BNP 40.5 vs. 16.5 pg/ml, p < 0.001). (Mann-Whitney rank sum test was used for comparison of peptide levels owing to their nonparametric distribution.) Interquartile ranges showed no overlap in NT-pro-BNP values, and only modest overlap in BNP values. Resting NT-pro-BNP values were lower in the healthy volunteers (median 25 pg/ml) than in the CAD patient groups (p = 0.0053 vs. nonischemic group), consistent with their much younger age.
NT-pro-BNP and BNP increased with exercise in both ischemic and nonischemic groups, and NT-pro-BNP increased in the healthy volunteers (BNP was not measured in this group). The median incremental rise of NT-pro-BNP (ΔNT-pro-BNP) was almost identical in the healthy volunteers and in the nonischemic patient group (5 vs. 4 pg/ml, p = NS). However, the incremental rise of both peptides in the ischemic group was significantly higher than in the nonischemic group (ΔNT-pro-BNP: 14.5 vs. 4 pg/ml, p < 0.0001; ΔBNP: 36.5 vs. 7.5 pg/ml, p < 0.0001). As with resting levels, there was no overlap in the interquartile ranges for NT-pro-BNP and modest overlap for BNP. The ΔNT-pro-BNP and ΔBNP values are shown graphically in Figures 1A and 1B.
Because 14 patients in the ischemic group were found to have fixed as well as reversible defects on radionuclide images, we conducted a subset analysis on the 26 ischemic patients with reversible defects only. Results of this analysis are shown in Table 4.Median resting levels of NT-pro-BNP and BNP for this subgroup were 118 pg/ml and 44 pg/ml, respectively, values that did not differ significantly from the values for the entire ischemic group. Similarly, median ΔNT-pro-BNP and ΔBNP for this subgroup were 16 and 36 pg/ml, respectively; as with resting levels, the Δ values were not significantly different from the ischemic group as a whole.
Detection of ischemia
To evaluate the ability of ΔNT-pro-BNP and ΔBNP levels to predict the presence or absence of ischemia in individual patients, we constructed receiver operator characteristic curves for each peptide (Fig. 2).The area under the curve in Figure 1for NT-pro-BNP is 0.836 (95% confidence interval 0.742 to 0.930), and for BNP is 0.811 (95% confidence interval 0.713 to 0.908, p < 0.0001 for both). Table 5shows the test characteristics at selected cutoff points for ΔNT-pro-BNP and ΔBNP. Sensitivities and specificities tended to be higher for ΔNT-pro-BNP than for ΔBNP at comparable cutoffpoints. In addition, we compared the specificity of ΔNT-pro-BNP in the healthy volunteer and nonischemic groups at various cutoff points (Table 6)and found the values to be comparable.
Analysis of the exercise ECG data showed that the sensitivity and specificity of 1-mm horizontal or downsloping ST-segment depression for the detection of ischemia were 37.5% and 58.8%, respectively.
Linear binary correlation analysis (Pearson's) found that baseline peptide levels correlated positively with age (r = 0.57, p < 0.0001), SSS (r = 0.56, p < 0.0001), SRS (r = 0.45, p = 0.0001), and SDS (r = 0.50, p = 0.0001), and negatively with maximal heart rate (r = −0.35, p = 0.002) and exercise capacity (r = −0.29, p = 0.01); by contrast, ΔNT-pro-BNP and ΔBNP correlated only with SSS (r = 0.35, p = 0.002) and SDS (r = 0.33, p = 0.004), and less strongly with SRS (r = 0.25, p = 0.03), but not with any other measured clinical or exercise test-derived variables. Logistic regression analysis showed that after correcting for other variables, ΔBNP and NT-pro-BNP were strongly predictive of ischemia (z score 12.8, p < 0.001). In a generalized linear model, Δpeptide levels accurately predicted SDS values (F ratio 10.4, p < 0.001).
Association of increased ΔBNPs with ischemia
We hypothesized that exercise-induced ischemia, either directly or by causing regional wall motion abnormalities, would result in detectable increases in levels of BNPs. Our finding of significantly larger post-exercise increments of NT-pro-BNP and BNP in the group of patients with inducible ischemia, compared with the group without, supports this hypothesis.
Because the exercise-induced increments in peptide levels in all groups were small compared with resting levels, we did not find percent change over baseline to be a useful diagnostic parameter. By contrast, the absolute increases demonstrated a strong ability to discriminate ischemic from nonischemic patients (Fig. 1, Table 5). For ΔNT-pro-BNP, median absolute increases in the nonischemic group and healthy volunteers were 4 and 5 pg/ml, respectively (p = NS), whereas in the ischemic group ΔNT-pro-BNP was over threefold higher (14.5 pg/ml, p < 0.0001 vs. nonischemic group). Similarly, absolute ΔBNP in the ischemic group was over four-fold higher than in the nonischemic group (36.5 vs. 7.5 pg/ml, p < 0.0001).
Of particular note is that the median absolute ΔNT-pro-BNP in the nonischemic patient group was almost identical to that of the healthy volunteers (4 vs. 5 pg/ml, p = NS), despite marked differences in age, exercise capacity, maximal heart rates, and resting NT-pro-BNP levels in the nonischemic patients that were more than twice as high as those of the volunteers. This suggests that the higher Δ peptide levels in ischemic patients compared with nonischemic patients were not due to the slightly higher age or to the higher baseline peptide levels in the ischemic group. The higher Δ peptide levels in the ischemic group are also unlikely to be the result of the presence within this group of patients with previous myocardial infarction, because our subset analysis in which we excluded results of all patients with fixed perfusion defects on perfusion imaging showed results no different from the group as a whole (Table 4). Reanalyzing blood samples of all subjects for baseline and ΔNT-pro-BNP values in a separate duplicate assay also did not significantly alter the results or the receiver operator characteristic curve of ΔNT-pro-BNP. This is consistent with the low coefficient of variation for this assay.
A recent study by Wu et al. (25) of a small number of apparently healthy subjects examined percent intra-individual biologic variation in natriuretic peptide levels over eight weeks and found changes of 33.3% for NT-pro-BNP and 43.6% for BNP. In our study, blood samples were drawn at most a few minutes apart, with interval exercise being the only intervention, thus it is unlikely that the observed differences are the result of intra-individual variation over time. McNairy et al. (26) measured BNP levels pre-exercise and immediately post-exercise in a small group of normal subjects and found a median 55% increase in post-exercise levels, with a return to baseline levels (+0.9%) by 60 min post-exercise. This is consistent with the changes in BNP observed in our nonischemic patients (median 45.4% increase). We also measured NT-pro-BNP levels at intervals up to 1 h post-exercise and found minimal differences from baseline levels by 30 to 60 min (data not shown). Furthermore, our ischemic group showed a median 1-min post-exercise increase in BNP almost twice as high as the nonischemic group (85.7% vs. 45.4%). Earlier studies (27–29) have also shown evidence of larger increases in BNP in patients with evidence of inducible ischemia, compared with nonischemic patients. As noted earlier, we found absolute increases, rather than percent increases, to have much higher discriminatory power.
Although there was no difference in ejection fraction by echocardiography before study entrance between the two groups (mean 64% in both), the post-exercise ejection fraction by gated imaging was lower in the ischemic than in the nonischemic group (51.8% vs. 57.8%). A plausible explanation for this observation is that the post-exercise ejection fractions are lower in the ischemic group precisely because of induced ischemia, because wall motion abnormalities are known to persist for a period of time after an episode of ischemia, a finding consistent with the study hypothesis.
Resting peptide levels
An unexpected finding was that median levels of both NT-pro-BNP and BNP were significantly higher at rest in the group of patients with inducible ischemia than in the group without (Table 2). However, a recent study (30) of over 300 patients measured BNP levels in patients with known CAD and found similar elevations in resting levels in those patients with inducible ischemia on imaging studies, compared with those without, corroborating our observation. If, as our findings suggest, ischemia triggers the release of these natriuretic peptides, patients with ischemia on scans may have higher baseline levels of NT-pro-BNP and BNP because of chronic or recurrent episodes of ischemia before testing. Our ischemic group contained patients with evidence of myocardial scar on perfusion imaging; however, it is not likely that this accounts for the increase in resting peptide levels in this group, because, as previously noted, a subset analysis that excluded these patients showed resting levels not significantly different from the ischemic group as a whole (Table 4). The higher age of the ischemic group compared with the nonischemic group also does not account for the differences in baseline levels, because both groups fall within the same decile of age for normal ranges.
An additional unexpected finding was that the ischemic and nonischemic groups did not differ in the percent of patients with abnormal exercise ECGs. The sensitivity of the ECG for detecting ischemia in our patients (37.5%) was in the lower range of those commonly reported for exercise testing. However, most studies of exercise testing have examined the ability of the exercise ECG to detect coronary stenosis on angiography and are affected by workup bias, which tends to overestimate sensitivity (1); our results are similar to those reported for the few published studies with reduced workup bias (2,3). Only a small number of studies have examined the ability of the exercise ECG to predict reversible defects on nuclear imaging; two representative studies (31,32) compared ECG findings with perfusion images and found ECG sensitivities of 45.5% and 42.8%, which are similar to the findings in our patients. In addition, a recent study (33) of over 700 patients found the sensitivity of the exercise ECG for detecting ischemia on myocardial imaging to be 36.6% to 40.0%, a range that includes the value found in our study.
In clinical practice, it is common to interpret as “indeterminate” exercise ECGs with no abnormal changes but a maximal heart rate <85% of predicted maximum, on the grounds that lower heart rates may not be adequate to induce ischemia, particularly in territories supplied by arteries with only moderate stenosis. However, because all patients in our study underwent myocardial perfusion imaging, we knew in all cases whether ischemia was present, and thus we considered ECGs with no diagnostic changes for ischemia to be negative regardless of the maximal heart rate achieved; if ischemia was present and the ECG failed to detect it, the ECG was falsely negative.
Use of BNPs to detect ischemia
Our findings suggest that an exercise-induced rise in BNP levels is a marker of inducible ischemia, and in our patients it was considerably more accurate in the detection of ischemia than was ST-segment depression on exercise electrocardiography. Comparative test characteristics of ECG findings and ΔNT-pro-BNP and ΔBNP levels for the detection of ischemia, set at equal specificities to the ECG, are shown in Table 7.Compared with the ECG, measurement of ΔNT-pro-BNP and ΔBNP more than doubled the sensitivity of the exercise test for ischemia (ΔNT-pro-BNP 90%, ΔBNP 80%; ECG 37.5%) with no loss of specificity. ΔNT-pro-BNP, in particular, correctly predicted the presence or absence of ischemia almost twice as frequently as the ECG (diagnostic accuracy 75.7% vs. 47.3%).
Additionally, the presence in our study of two groups without ischemia, the nonischemic patients and healthy volunteers, provided the opportunity to compare the specificity of ΔNT-pro-BNP values at various cutoff points in these groups (Table 6). Although the numbers are small, this comparison suggests that the rate of true negative ΔNT-pro-BNP results at any given cutoff point is consistent in two populations with markedly different ages, clinical characteristics, resting NT-pro-BNP levels, and exercise capacities.
Reduced regional myocardial blood flow results in a cascade of changes beginning with relaxation failure and progressing to contraction abnormalities, rise in filling pressure, ECG changes, and finally symptoms (34). Because ECG abnormalities occur later in this process than changes in ventricular wall function, this may explain why BNPs would rise before ECG abnormalities appear; in other words, measuring NT-pro-BNP or BNP rise may be more sensitive because it detects reduced myocardial blood flow at an earlier stage.
In summary, our findings suggest that exercise-induced ischemia or its associated regional wall motion abnormalities trigger release of BNPs and that measurement of plasma levels of NT-pro-BNP and BNP before and immediately after symptom-limited exercise can distinguish patients with and without ischemia, defined as reversible defects on radionuclide imaging, with a high degree of accuracy. In our group of patients with known CAD, the ability of ΔBNP and ΔNT-pro-BNP to detect ischemia was considerably better than that of the exercise ECG (90% vs. 37.5%) and distinguished between ischemic and nonischemic patients nearly twice as accurately as the exercise ECG. Assays for these peptides are relatively inexpensive and precise; in addition, NT-pro-BNP, although not BNP, is stable in whole blood for at least 48 h (24,35), and thus special handling is not required.
Measurement of these markers before and after exercise offers the potential for a substantial increase in the diagnostic accuracy of exercise testing.
The authors wish to thank the following individuals for their generous technical support: Kim Lavalley, BS, Rick Mazurek, CNMT, Jill Lewis, CNMT, Matthew Maddock, CNMT, Beryl McPhetres, CNMT, Robert Wulpern, CNMT, Ling Gao, ME, HjoJin Chung, BS, Kim McCullough, BS, and Kam Cheong Wong, MS.
Grant supportwas received from the Hitchcock Foundation (Lebanon, New Hampshire). Research support was received from Roche Diagnostics (Indianapolis, Indiana).
- Abbreviations and acronyms
- B-type natriuretic peptide
- coronary artery disease
- N-terminal fragment of B-type natriuretic peptide pro-hormone
- summed difference score
- single-photon emission computed tomography
- summed rest score
- summed stress score
- Received January 29, 2004.
- Revision received July 23, 2004.
- Accepted August 2, 2004.
- American College of Cardiology Foundation
- Gianrossi R.,
- Detrano R.,
- Mulvihill D,
- et al.
- Froelicher V.F.,
- Lehmann K.G.,
- Thomas R,
- et al.
- ↵Espiner EA, Richards AM, Yandle TG, Nicholls MG. Natriuretic hormones. Endocrinol Metab Clin North Am 1995;24:481–509.
- Yasue H.,
- Yoshimura M.,
- Sumida H,
- et al.
- Dao Q.,
- Krishnaswamy P.,
- Kazanegra R,
- et al.
- Lubien E.,
- DeMaria A.,
- Krishnaswamy P,
- et al.
- Talwar S.,
- Squire I.B.,
- Downie P.F.,
- Davies J.E.,
- Ng L.L.
- Jernberg T.,
- Stridsberg M.,
- Venge P.,
- Lindahl B.
- Morita E.,
- Yasue H.,
- Yoshimura M,
- et al.
- Toth M.,
- Vuorinen K.H.,
- Vuolteenaho O,
- et al.
- Goetze J.P.,
- Christoffersen C.,
- Perko M,
- et al.
- Bibbins-Domingo K.,
- Ansari M.,
- Schiller N.B.,
- Massie B.,
- Whooley M.A.
- Nallamothu N.,
- Ghods M.,
- Heo J.,
- Iskandrian A.S.
- ↵Sigwart UG, Payot J, Goy A, Essinger A. Ischemic events during coronary artery balloon occlusion. In: Rutishauser W, Roskamm H, editors. Silent Myocardial Ischemia. New York, NY: Springer-Verlag, 1984:29–36.