Author + information
- Received January 30, 2003
- Revision received February 10, 2004
- Accepted February 11, 2004
- Published online July 21, 2004.
- Babak Azarbal, MD*,
- Sean W. Hayes, MD*,†,‡,
- Howard C. Lewin, MD*,†,‡,
- Rory Hachamovitch, MD, MSc, FACC†,‡,
- Ishac Cohen, PhD§ and
- Daniel S. Berman, MD, FACC*,†,‡,§,* ()
- ↵*Reprint requests and correspondence:
Dr. Daniel S. Berman, Cedars-Sinai Medical Center, Taper 1258, 8700 Beverly Boulevard, Los Angeles, California 90048, USA.
Objectives We sought to determine whether chronotropic incompetence (CI) adds incremental value in predicting cardiac death (CD) and all-cause mortality and to determine which marker of CI is superior.
Background Chronotropic incompetence, defined by either a low percent heart rate (HR) reserve achieved or failure to achieve 85% maximal age-predicted heart rate (MA-PHR), is a predictor of mortality. These variables have not been examined together in a comprehensive myocardial perfusion single-photon emission computed tomographic (SPECT), or MPS, model.
Methods A total of 10,021 patients who underwent exercise MPS, evaluated by a summed stress score (SSS), were followed up for 719 ± 252 days. Percent HR reserve = (peak HR − rest HR)/(220 − age − rest HR) × 100, with <80% considered abnormal.
Results A total of 2,956 patients (29.5%) had low %HR reserve; 1,331 (13.3%) achieved <85% MA-PHR; and 1,296 (13.0%) had both. There were 234 deaths (93 CDs). On multivariate analysis, the SSS, %HR reserve, and inability to achieve 85% MA-PHR were predictors of all-cause mortality and CD (all p < 0.01). Myocardial perfusion SPECT was the most powerful predictor of CD (chi-square = 50). When the %HR reserve and ability to achieve 85% MA-PHR were considered, only the former remained a predictor of CD (p = 0.006 vs. p = 0.59).
Conclusions In a comprehensive MPS model, CI was an important predictor of CD and all-cause mortality. Percent HR reserve was superior to the ability to achieve 85% MA-PHR in predicting CD; MPS was superior to both. Combined with previous studies, the findings suggest that %HR reserve should become the standard for assessing the adequacy of HR response during exercise testing, and that it should be routinely incorporated in risk stratification algorithms.
Myocardial perfusion single-photon emission computed tomography (SPECT), or MPS, is a well-established prognostic marker of cardiac death (CD) (1,2). Recent publications suggest that chronotropic incompetence (CI), defined by either a low percent heart rate (HR) reserve achieved, a measure of HR response to exercise that accounts for age and resting HR, or failure to achieve 85% of the maximal age-predicted heart rate (MA-PHR), is an independent predictor of mortality after adjustment for clinical and standard exercise variables (3–8).
One study suggested that CI might be as important as MPS in the prediction of death and that the prognostic information gained from %HR reserve was additive to information obtained by MPS (6). However, this work considered MPS as simply normal or abnormal. To date, CI has not been compared with a comprehensive model of MPS variables.
The purpose of this study was two-fold: first, to determine whether CI adds incremental value to a comprehensive model of MPS in the prediction of CD and all-cause mortality in a large cohort of patients undergoing exercise MPS; and second, to determine whether %HR reserve is superior to the ability to achieve 85% MA-PHR in the prediction of CD.
We prospectively identified 10,793 consecutive patients who underwent exercise MPS and who had no history of valvular heart disease and who were not under the influence of beta-blockers at the time of the exercise study. Of this initial population, 772 patients underwent early revascularization, defined as revascularization <60 days after exercise testing, and were censored from prognostic analysis. Institutional review board approval was obtained for the performance of this research at our institution, and patients gave informed consent for participating in the research protocol.
Data on the presence of hypertension, diabetes mellitus, hypercholesterolemia, smoking, and family history of premature coronary artery disease (CAD) were collected prospectively by questioning the patients or from review of medical records at the time of stress testing.
Rest thallium imaging
All patients underwent stress dual-isotope MPS, as previously described (1,2,9). Initially, thallium-201 (3.0 to 4.5 mCi) was injected intravenously at rest, with dose variance based on the patient's weight. Rest thallium imaging was initiated 10 min after injection of the isotope. For patients with extensive rest perfusion defects, additional SPECT thallium-201 images were obtained 24 h after injection.
Exercise myocardial perfusion protocol
All patients performed a symptom-limited treadmill exercise test, using the standard Bruce protocol, with a 12-lead electrocardiographic (ECG) recording each minute of exercise and continuous monitoring of leads aVF, V1, and V5. At near maximal exercise, a 25- to 40-mCi weight-adjusted dose of technetium-99m sestamibi was injected, and exercise was continued for 1 min after injection. Whenever possible, patients were taken off calcium channel blockers for 48 h and nitrate compounds for at least 6 h before testing.
The rest ECG was considered abnormal if there were any ECG abnormalities other than sinus tachycardia, sinus bradycardia, first-degree atrioventricular block, premature atrial contractions, premature ventricular contractions, mild intraventricular conduction delay, or early repolarization abnormality.
During the treadmill exercise test, HR and blood pressure were measured and recorded at rest, at the end of each stress stage, and at peak stress. Exercise end points included physical exhaustion, severe angina, sustained ventricular tachycardia, hemodynamically significant supraventricular dysrhythmias, or significant exertional hypotension. The maximal degree of ST-segment change at 80 ms after the J point on the ECG was measured and assessed as horizontal, upsloping, or downsloping.
Estimated exercise capacity in metabolic equivalents (METs) was obtained using previously published tables (10). Subjects were subclassified into two groups according to their exercise duration (≤6 min [7 METs] and >6 min [7 METs]) to broadly classify patients with poor and moderate to superior exercise tolerance, respectively (11).
SPECT acquisition protocol
The SPECT studies were performed, as previously described, with a circular or elliptical 180° acquisition for 64 projections at 20 to 25 s/projection for thallium-201 and at 15 to 25 s/projection for technetium-99m sestamibi (2). All images were subject to quality-control measures, as previously described (12), including cinematic display for assessment of patient motion (13)and corrections for field nonuniformity and center of rotation. The projection data were reconstructed into tomographic transaxial images using filtered backprojection and automatic reorientation. No attenuation or scatter correction was used.
The MPS images were visually assessed using a 20-segment model of the left ventricle and a 5-point scoring scale (0 = normal; 4 = absent uptake) (2). The summed stress score (SSS) and summed rest score (SRS) were calculated as previously described (1). The summed difference score (SDS), representing the amount of reversible perfusion defect, was calculated by subtracting SRS from SSS. A negative SDS (rest score > stress score) was set at 0. The MPS results were subdivided by SSS as follows: <4 (normal); 4 to 8 (mildly abnormal); 9 to 13 (moderately abnormal); and >13 (severely abnormal).
Calculation of %HR reserve and ability to achieve 85% MA-PHR
Chronotropic incompetence was defined by either the failure to achieve 85% MA-PHR (MA-PHR defined as 220 − age) or as low %HR reserve. The following equation was used: %HR reserve = ([peak HR − rest HR]/[MA-PHR − rest HR])× 100 (14). A low %HR reserve was defined as <80% (3,6).
Patient follow-up was performed by a scripted telephone interview by research staff blinded to the patients' test results. Events were defined as either CD, as noted and confirmed by a review of the death certificate and hospital chart or physician's records, or nonfatal myocardial infarction, as evidenced by the appropriate combination of symptoms, ECG, and enzyme changes. All-cause mortality was defined by any death occurring during follow-up. Patients were followed up for 719 ± 252 days (median 677 days, range 355 to 3,297 days; median 682 days for survivors; 96.4% complete).
Comparisons between patient groups were performed by use of one-way analysis of variance (with Bonferroni correction where appropriate) for continuous variables and the chi-square test for categorical variables. All continuous variables are expressed as the mean value ± SD. A value p < 0.05 was considered significant.
The Cox proportional hazards model (SPSS for Windows, Version 11.0, SPSS Inc., Chicago, Illinois) identified univariate and multivariate predictors of CD and overall mortality and was used to construct survival curves. Selection of variables for entry consideration was based on both univariate statistical significance and clinical judgment. Variables initially included in the model were age, gender, family history of premature CAD, hypertension, smoking, hypercholesterolemia, diabetes mellitus, previous myocardial infarction, previous angioplasty, previous bypass surgery, calcium channel blocker use, resting ECG abnormalities, resting HR, exercise duration, exercise-induced chest pain, and exercise-induced ischemic ST-segment depression.
After the Cox proportional hazards model most predictive of CD was determined, MPS data, %HR reserve, and ability to reach 85% MA-PHR were added to each Cox proportional hazards model to determine the incremental value of each variable. A conservative value of p < 0.01 was considered significant for entry into the final model to minimize spurious covariates. Using this vigorous threshold for entry into the model, no interactions were statistically significant to enter the final model. All model assumptions were tested, including those of proportional hazards, linearity, additivity, multiplicity, and collinearity. Predicted events rates were based on hazard scores generated for each patient, based on the Cox proportional hazards model presented in the current report.
The baseline and stress test (including exercise and MPS) characteristics of the 10,021 patients enrolled in this study are summarized in Tables 1, 2, 3, and 4. ⇓⇓⇓Patients with either a low %HR reserve or <85% MA-PHR were older and had a higher prevalence of hypertension, diabetes, smoking, previous MI, and abnormal rest ECG (all p < 0.001). Regarding stress test characteristics (Tables 3 and 4), they had a higher resting HR, lower peak HR, shorter exercise duration, and higher incidence of exertional hypotension, as compared with patients with normal HR responses (all p < 0.001). Regarding MPS variables, patients with abnormal HR responses had a higher SSS, SRS, and SDS (all p < 0.001). The baseline and stress test characteristics of patients with an inability to reach 85% MA-PHR were similar to those with a low %HR reserve, with the exception of gender, a family history of CAD, and a gender-related difference in exercise duration.
Table 5demonstrates the relationship between %HR reserve and the ability to reach 85% MA-PHR. There were 2,956 patients with a low %HR reserve and 1,331 patients with an inability to reach 85% MA-PHR. Of those patients who were unable to reach 85% MA-PHR, 97.4% (1,296 of 1,331) also had a low %HR reserve. Conversely, of those patients with a low %HR reserve, only 43.8% (1,296 of 2,956) were unable to reach 85% MA-PHR (p < 0.001).
There were 234 deaths, of which 93 were cardiac. There were 145 myocardial infarctions. Figures 1 and 2⇓⇓demonstrate overall survival and survival free of CD as a function of MPS (when considered as normal vs. abnormal) and either %HR reserve (Fig. 1) or ability to achieve 85% MA-PHR (Fig. 2).
For the end point of CD, when considered separately, a low %HR reserve and inability to achieve 85% MA-PHR were independent predictors of CD, even after adjustment for SSS (p < 0.001 and p = 0.006, respectively). However, when these variables were both included in the model, the inability to reach 85% MA-PHR was no longer an independent predictor of CD (p = 0.59), whereas a low %HR reserve remained an independent predictor of CD (p = 0.006). In fact, of the 35 patients with an inability to reach 85% MA-PHR but with a normal %HR reserve, there were no CDs.
In the final model, SSS was the most powerful predictor of CD (chi-square = 50.0), followed by a low %HR reserve (chi-square = 11.7), resting ECG abnormalities (chi-square = 10.5), age (chi-square = 10.0), exercise duration (chi-square = 9.7), and resting HR (chi-square = 6.3). When SDS and SRS were used instead of SSS, the significant predictors were SDS (chi-square = 33.7), followed by SRS (chi-square = 28.6), low %HR reserve (chi-square = 11.3), resting ECG abnormalities (chi-square = 10.6), exercise duration (chi-square = 9.4), age (chi-square = 9.4), and resting HR (chi-square = 6.9).
The prognostic information gained from MPS and %HR reserve was additive when MPS was considered as either normal or abnormal (Fig. 1). Patients with a normal %HR reserve and normal MPS had a very low risk of CD (0.2%). Patients with either a low %HR reserve or abnormal MPS were at increased risk of CD, although having an abnormal MPS alone carried a greater risk of CD at two years, as compared with having a low %HR reserve alone (0.7% vs. 1.4%, p < 0.001). Patients with both an abnormal MPS and low %HR reserve had the highest risk of CD (3.5%, p < 0.001).
Adjusted (predicted) risk of CD and all-cause mortality, as a function of %HR reserve and MPS categories, is shown in Figure 3. There was an increasing incidence of CD as a function of worsening MPS category (p < 0.001), with further stratification within SSS categories by %HR reserve (p < 0.05). Patients with a normal %HR reserve had a <1% CD rate over the follow-up period, even with a moderately abnormal SSS; patients achieving ≥85% MA-PHR with a moderately abnormal SSS category did not reach this very low risk and had a >1% CD rate.
Figure 4shows adjusted (predicted) survival free of CD, as well as overall survival, as a function of exercise capacity and %HR reserve. Patients who were able to exercise >6 min of the standard Bruce protocol (7 METs) and also had a normal %HR reserve were at extremely low risk of CD (0.3%), whereas those who exercised ≤6 min (7 METs) and had a low %HR reserve were at highest risk of CD (2.6%).
For the outcome of all-cause mortality, a low %HR reserve and inability to achieve 85% MA-PHR were independent predictors (p < 0.001). Four models were evaluated. In the model including SSS and %HR reserve, age was the most powerful predictor of all-cause mortality (chi-square = 43.1), followed by SSS (chi-square = 23.7), exercise duration (chi-square = 21.1), low %HR reserve (chi-square = 20.0), resting HR (chi-square = 19.6), resting ECG abnormalities (chi-square = 11.7), and gender (chi-square = 6.2). In the model using SDS and SRS (instead of SSS) and %HR reserve, the significant predictors of all-cause mortality were age (chi-square = 42.7), followed by exercise duration (chi-square = 21.1), low %HR reserve (chi-square = 19.8), resting HR (chi-square = 19.8), SRS (chi-square = 14.2), SDS (chi-square = 12.2), resting ECG abnormalities (chi-square = 11.6), and gender (chi-square = 6.8).
In the model considering SSS and inability to achieve 85% MA-PHR (instead of %HR reserve), age was the most powerful predictor of all-cause mortality (chi-square = 45.7), followed by resting HR (chi-square = 25.5), exercise duration (chi-square = 22.0), SSS (chi-square = 22.0), inability to achieve 85% MA-PHR (chi-square = 21.5), resting ECG abnormalities (chi-square = 13.0), and gender (chi-square = 5.9). Finally, when an inability to achieve 85% MA-PHR was considered along with SDS and SRS (instead of SSS), the significant predictors were age (chi-square = 45.3), followed by resting HR (chi-square = 25.7), exercise duration (chi-square = 21.9), inability to achieve 85% MA-PHR (chi-square = 21.1), SRS (chi-square = 13.4), resting ECG abnormalities (chi-square = 12.8), SDS (chi-square = 11.1), and gender (chi-square = 5.7).
Although a low %HR reserve and inability to achieve 85% MA-PHR were univariate predictors of myocardial infarction (p < 0.01 and p < 0.02, respectively), neither marker was an independent predictor of myocardial infarction in multivariate modeling. The final predictors of myocardial infarction in the final model were SSS (chi-square = 68.8), diabetes mellitus (chi-square = 12.3), exercise duration (chi-square = 11.0), chest pain during exercise stress (chi-square = 10.0), age (chi-square = 8.0), and family history of CAD (chi-square = 5.7).
Calcium channel blockers
For the end points of CD, overall mortality, and myocardial infarction, the results remained unchanged for a low %HR reserve and inability to achieve 85% MA-PHR, when adjusted for the use of calcium channel blockers.
In a large cohort of patients undergoing exercise MPS, we have demonstrated that CI, as defined by either a low %HR reserve or inability to achieve 85% MA-PHR, was an independent predictor of CD and all-cause mortality, even after adjustment for the univariate predictors of these events, including SSS. Additionally, while SSS was a more powerful predictor, %HR reserve added incremental value in the prediction of CD and all-cause mortality when combined with SSS. Myocardial perfusion SPECT was superior to %HR reserve in the prediction of CD, even when MPS was interpreted as either normal or abnormal.
Differentiating between all-cause mortality and cardiac death, as we have done in this study, adds clinically relevant information. In our population, 40% of deaths were attributable to cardiac death. Although all-cause mortality is considered the most unbiased measure of outcome, patients are generally referred for exercise testing to assess cardiac symptoms and risk of cardiac events. Further work-up and treatment of patients undergoing exercise testing, including choice of medical therapy or need for coronary angiography/revascularization, are often predicated on the patient's estimated risk of adverse cardiovascular outcomes.
The %HR reserve further risk-stratified patients within each SSS category. Patients who had a normal SSS were at low risk of CD irrespective of %HR reserve; however, if the %HR reserve was normal, the risk of CD was extremely low (0.1%/year). In those with an abnormal %HR reserve, although the risk was higher (p < 0.01), it was still less than 0.5%/year. Patients in the mild to moderate SSS categories who had a normal %HR reserve had a very low CD risk that was comparable to the incidence of CD observed in patients with a normal MPS study. As demonstrated in Figure 3A, based on a CD risk threshold of 1% over two years, where patients would be at very low risk of CD and would not be expected to benefit from referral for coronary angiography, 1,169 patients in the mild and moderately abnormal MPS categories (11.7% of total population, 62.0% of those in these two MPS categories) are reclassified as very low risk if the %HR reserve is normal. It is important to note that this additional clinical information is derived at no additional cost or effort, and all the variables needed for the derivation of %HR reserve are routinely measured during standard exercise testing.
Although patients with severely abnormal SSS and normal %HR reserve were at lower risk of CD compared with patients with severely abnormal SSS and a low %HR reserve (2.4% vs. 5.1%, p < 0.05), they remained at intermediate to high risk of CD and should be considered for coronary angiography/revascularization.
Low %HR reserve versus inability to achieve 85% MA-PHR as a measure of CI
Multivariable analysis showed that %HR reserve was superior to the ability to reach 85% MA-PHR in the identification of patients at risk of CD. Patients with a normal %HR reserve had a <1% CD rate over the follow-up period, even with a moderately abnormal SSS; in contrast, if achieving ≥85% MA-PHR were used as the low-risk exercise criterion, patients in the moderately abnormal SSS category demonstrated a >1% CD rate. There were 2.2 times as many higher risk individuals identified using the %HR reserve variable. Whereas 97.4% of patients with an inability to reach 85% MA-PHR had a low %HR reserve, only 43.8% of patients with a low %HR reserve achieved <85% MA-PHR. In addition, of the 2.6% (35 of 1,331) of patients with an inability to reach 85% MA-PHR but with a normal %HR reserve, there were no CDs, thus underscoring the superior prognostic information gained from %HR reserve in this population. The reason that the %HR reserve identifies a larger subset of patients at risk of CD who would not be identified by simple assessment of the ability to achieve 85% MA-PHR is that it accounts for resting HR in addition to age and peak HR. For example, a 70-year-old patient with a resting HR of 90 beats/min, who exercises only 2 min of a modified Bruce protocol limited by shortness of breath and who achieves a peak HR of 130 beats/min, would have reached 85% MA-PHR. However, common clinical sense would suggest that the ability to achieve 85% MA-PHR does not appropriately identify this patient as being at low risk. The %HR reserve in this same patient would be: (130 − 90)/(220 − 70 − 90) = 0.67. This patient would be appropriately identified as high risk by the %HR reserve variable and not by the ability to reach 85% MA-PHR variable. Conversely, a young 20-year-old athlete with a resting HR of 50 beats/min, who exercises 15 min of a Bruce protocol and achieves a peak HR of 160 beats/min, would not have achieved his or her 85% MA-PHR of 170 beats/min. However, this same patient would have a %HR reserve of: (160 − 50)/(220 − 20 − 50) = 0.9, which is well within normal limits and would be classified appropriately as low risk.
Several potential mechanisms could account for the association between CI and CD. First, it has been demonstrated in this study as well as previous studies that CI is a marker for CAD (5,6,15–18). Patients with CI had significantly higher SSS, SDS, and SRS values as compared with patients with normal HR responses. However, even after adjustment for the perfusion defects on MPS, CI was still an independent predictor of CD. It is possible that CI is still a marker for CAD that is not detected by either ST-T segment abnormalities on exercise testing or MPS. Concordant with these findings, Chin et al. (15)evaluated 53 patients with CI and no ST-segment abnormalities after maximal treadmill testing with coronary angiography and found that 72% of these patients had significant angiographic CAD. Also, a recent study by Brener et al. (18)demonstrated that CI is predictive of the severity of angiographic CAD.
Additionally, patients with CI may have an increased mortality independent of the risk associated with CAD (5,6,19). This finding is supported by a recent study by Dresing et al. (19), who demonstrated that CI was predictive of all-cause mortality in patients without severe angiographic CAD, as defined by the Duke CAD Prognostic Weight Score.
However, the specific mechanisms leading to an increased incidence of mortality in patients with CI, independent of the risk associated with CAD, have not been elucidated. It is possible that CI may be a marker for autonomic dysfunction, which in turn leads to an increase in mortality similar to that observed in patients with congestive heart failure (20). Analogies have been drawn between the increased mortality observed with CI and that seen with decreased heart rate variability (3,21). It is possible that both of the mechanisms already outlined contribute to the observed relationship between CI and CD.
This study is a large observational study and not a randomized clinical trial. We did not adjust for late revascularizations in our study. Although data on the use of calcium channel blockers and beta-blockers are complete, we did not have such data on the use of aspirin, statins, or angiotensin-converting enzyme inhibitors; these medications are known to affect cardiovascular outcomes.
In our study, a low %HR reserve was superior to an inability to achieve 85% MA-PHR and identified 2.2 times as many individual at increased risk of CD. However, as a routine in our exercise laboratories, exercise is not considered adequate unless the patient achieves at least 85% MA-PHR. If this target heart rate is not achieved, patients are generally converted over to pharmacologic stress and thus are not part of the prognostic analysis in the exercise cohort. This could introduce a bias, where the number of individuals with an inability to achieve 85% MA-PHR would be reduced relative to those with low %HR reserve. However, in an earlier study by Lauer et al. (6)in which patients were exercised until exhaustion or limiting symptoms (fatigue was the reason for termination in 95% of patients without CI and in 83% of patients with CI), there were 2.4 times as many patients with CI identified using a low %HR reserve compared with an inability to achieve 85% MA-PHR. Also, as we have demonstrated in this report, of all patients with an inability to achieve 85% MA-PHR, >97% will also have a low HR reserve.
Chronotropic incompetence, as determined by either a low %HR reserve or an inability to achieve 85% MA-PHR, is an important predictor of CD and all-cause mortality. We have demonstrated, for the first time, that CI adds incremental value when combined with a comprehensive MPS model. The %HR reserve is superior to an inability to achieve 85% MA-PHR in identification and risk stratification of patients at increased risk of CD.
Submaximal tests (i.e., exercise tests where patients have achieved a low %HR reserve) are not merely submaximal but are actually quite meaningful. Patients with CI, as assessed by %HR reserve, are at substantially increased risk of CD and all-cause mortality. This finding underscores the importance of obtaining symptom-limited tests and not merely stopping the exercise test because the patient has achieved an arbitrary heart rate.
Our study suggests that the additional prognostic information gained from calculation of %HR reserve in patients undergoing exercise stress testing may have implications regarding further evaluation and management of these patients. Combined with previous studies (6), the findings suggest that %HR reserve should become the standard for assessing the adequacy of HR response during exercise testing and that it should be routinely incorporated in risk stratification algorithms.
☆ This work was supported in part by a grant from Bristol-Myers Squibb Medical Imaging, Inc., Billerica, Massachusetts. This study was presented in part at the 50th Annual Scientific Sessions of the American College of Cardiology, Orlando, Florida, March 2001. Dr. James Udelson acted as the Guest Editor for this paper.
- coronary artery disease
- cardiac death
- chronotropic incompetence
- heart rate
- maximal age-predicted heart rate
- MPS or SPECT
- myocardial perfusion single-photon emission computed tomography
- summed difference score
- summed rest score
- summed stress score
- Received January 30, 2003.
- Revision received February 10, 2004.
- Accepted February 11, 2004.
- American College of Cardiology Foundation
- Hachamovitch R.,
- Berman D.S.,
- Shaw L.J.,
- et al.
- Berman D.S.,
- Kiat H.,
- Friedman J.D.,
- et al.
- Lauer M.S.,
- Okin P.M.,
- Larson M.G.,
- Evans J.C.,
- Levy D.
- Lauer M.S.,
- Pashkow F.J.,
- Larson M.G.,
- Levy D.
- Lauer M.S.,
- Mehta R.,
- Pashkow F.J.,
- Okin P.M.,
- Lee K.,
- Marwick T.H.
- Robbins M.,
- Francis G.,
- Pashkow F.J.,
- et al.
- Berman D.S.,
- Kiat H.S.,
- Van Train K.F.,
- Germano G.,
- Maddahi J.,
- Friedman J.D.
- American College of Sports Medicine
- Thompson C.A.,
- Jabbour S.,
- Goldberg R.J.,
- et al.
- Colucci W.S.,
- Ribeiro J.P.,
- Rocco M.B.,
- et al.
- Tsuji H.,
- Venditti F.J. Jr..,
- Manders E.S.,
- et al.