An incremental exercise test that linearly spans the tolerable work rate range has become the recommended procedure for initial assessment in clinical exercise testing (11,13,29). In these tests, work rate maybe advanced either as a ramp function (i.e., continuous increase) or using frequent stepwise increases in work rate. These tests are normally performed using either a cycle or treadmill ergometer. It has been argued that the treadmill has the advantage of educing a higher maximum metabolic rate and of using a mode of exercise that more closely approximates aspects of activities of daily living (5,11,13). Others prefer the cycle ergometer for this task, citing safety aspects of the test performance, ease of both blood sampling, and auscultation of blood pressure as well as the ability to precisely quantify work rate throughout the test (9,16,29). Because the linearity of the pattern of the oxygen uptake response is a major discriminating feature for assessing a cardiovascular basis for exercise intolerance (3,17,20,22, 28,30), it is naturally important to be confident of the linearity of the work rate profile that yielded the response.
For severely impaired patients, the metabolic cost of even normal walking or unloaded cycling can be a significant proportion of their maximum metabolic rate, which in turn limits the duration of the test and, consequently, the available range of data for interpretation. An incremental treadmill test starting at a slow and constant walking speed would provide a sufficiently low initial metabolic rate (12), but keeping the speed constant as the test progresses results in an inappropriately long test duration. Conversely, a faster and constant walking speed, although shortening the test, has the disadvantage of a high initial metabolic cost (5,34). We reasoned that the combination of an initially slow walking speed that is increased progressively in concert with an increasing grade would meet the demands of both an initially low exercise metabolic rate and optimum test duration. The simultaneous combination of change in speed and grade, however, does not directly provide a linear increase in work rate, thereby complicating the interpretation of the oxygen uptake profile, unless more complex methods of deconvoluting the oxygen uptake response are applied (14).
It was therefore the purpose of this study to design and implement a treadmill exercise test that a) requires a low initial metabolic demand; b) provides a constant rate of change in work rate (i.e., linear increase as a function of time) utilizing continuous changes in both speed and grade; and c) brings subjects to the limit of tolerance in approximately 10 min, i.e., meeting the requirements recommended for clinical exercise testing by the relevant major international organizations (1,2,9,15). Although this new protocol was thought to have its most valuable application in subjects with severely limited exercise tolerance (e.g., those with severe heart or lung disease), we chose to perform our initial validation in a group of healthy subjects with widely differing exercise tolerances, so that the robustness of the method might be evaluated. We evaluated this new treadmill protocol by comparing the cardiopulmonary and metabolic response profiles of healthy sedentary individuals with those from a cycle ramp test with an equivalent work rate incrementation profile.
The rate of work done against gravity while walking up an incline depends on the subject’s weight, the walking speed and the angle of inclination itself (21, p. 271). Specifically:MATH
where WR(t) is the time course of work rate in watts, m is body mass in kilograms, g is the gravitational acceleration (9.81 m·s−2), v(t) is the time course of velocity in meters per second, and α is the angle of inclination. [Although the expressions “inclination,” “angle,” and “grade” all refer to the slope of the treadmill belt, the terms are not synonymous. Inclination/grade refers to the “rise/run” of the treadmill. It is, therefore, the tangent of the angle of elevation and is commonly expressed as percentage (i.e., units of rise per 100 units of run). It is important to recognize that the work rate equation considers only that done against gravity (calculated work rate will be zero, when the grade is zero) (18)].
Straightforward ways to achieve a linear increase in work rate are 1) to maintain a constant inclination while increasing the speed linearly or 2) to maintain a constant speed while increasing the sine of the angle of inclination linearly. However, neither option is totally satisfactory. With the first option, a choice of a low inclination means that the speed will be increased to such a fast pace that tolerance may be dictated by the ability to move the legs quickly and/or efficiently enough rather than metabolic factors. Alternatively, the choice of a steeper grade yields a high initial metabolic cost. With the second option, the choice of a low speed means that treadmill inclination will increase to a very steep grade before tolerance is reached, whereas the choice of a higher initial speed will yield a high initial metabolic cost.
We considered that it should be possible to design a protocol featuring a linear increase in work rate that starts with a slow speed, which is advanced linearly such that the subject fatigues at a moderate walking speed. To accomplish this, the treadmill slope must be incremented in a pattern predetermined to yield a linear work rate profile. Thus, if:MATH
where S is the work rate slope (W·min−1) and WR0 is the initial work rate, and MATH
where Vs is the rate of change of velocity (m·s−2) and V0 is initial treadmill velocity, then MATH
where f(t) is the time course of sin(α).
Certain considerations lead to a highly useful computational form of this equation. First, it is desirable that the incremental phase of the exercise test be of approximately 10-min duration.
where WRmax is the projected maximum end-test work rate.
Second, if the desired end exercise treadmill velocity is specified as Vmax, then:MATH
Finally, for most treadmills, slope of the treadmill is programmed as the grade, which is the tangent of the angle of inclination (18). Therefore:MATH
By trigonometric identity:MATH
Importantly, for the range of treadmill grades typically encountered during exercise testing, the denominator of this equation is very near 1, so that f(t) and the grade are very nearly equal (e.g., for a grade of 0.05 (a 5% grade), f(t) is 0.499; for a grade of 0.20, f(t) is 0.196).
The time course of the inclination predicted to achieve a linear increase in work rate is:MATH
Therefore, the time course of inclination can be calculated based on the patient’s body weight, the desired initial and final treadmill speeds (V0 and Vmax), the initial grade (grade0), and the predicted peak work rate. Although this procedure provides a pure ramp of work rate, with appropriate electronic control of the treadmill, its functional equivalent can also be generated, as was done in this experiment. We used small increments of work rate expressed over time intervals that are short relative to the response kinetics of the variables of interest to yield a response that is indistinguishable from that of a “pure” ramp (e.g., Fig. 1).
To design an individual test, first, the predicted peak work rate is obtained from a published formula (e.g., 21, p. 306), as modified by an a priori estimate of the patient’s level of disability. The average maximum work rate attained on the treadmill was not significantly different from those provided by the above formula (21, p. 306), although there was marked variability (i.e., the actual value averaging 105% of that predicted, but with a large standard deviation of 25%). It has been our practice, therefore, to set the peak work rate to a somewhat higher value (approximately 30%) in order to ensure that subjects who have a higher exercise tolerance than that of the predicted mean are also accommodated by the test. Were the test still to exceed the 10-min target, the work rate would, however, still increase effectively linearly. This is because the grade continues to increase slowly as it approaches its asymptotic value and the treadmill speed, continues to increase linearly, allowing the linearity of the work rate to be sustained. Next, the treadmill velocity range is determined. It is useful to keep the initial velocity slow (e.g., 0.5 mph, i.e., 0.25 m·s−1, or the lowest value achievable (e.g., 1.0 mph) with “older” treadmills) to provide the desirably low initial metabolic rate (Fig. 1). The top speed can be chosen according to the subject’s ability to walk comfortably which, for a debilitated subject, might be as low as 1–1.5 mph, and for a healthy subject might be up to 3.5–4.0 mph (1.6–1.8 m·s−1). Higher speeds can induce substantial change in the gait pattern, leading to a steeper rise in the metabolic rate per calculated work rate increase (i.e., ΔV̇O2/ΔW) than established over the slower speed range (6). The changes in work rate and speed range are then spread over 21 stages, of which the first is maintained for 3 min (to give an equivalent of the unloaded cycling) and the remaining 20 stages, each lasting for 30 s to result a ramp duration of 10 min. The initial grade should be low, but greater than “0,” say 1%. The inclination for each of the remainder of the steps is calculated as detailed above.
Subjects and experimental design.
Sedentary healthy subjects (9 females and 13 males) gave written informed consent for their participation in this study as approved by the Harbor-UCLA Research and Education Institute’s Institutional Review Board. Their age range was 18–54 yr (mean: 37.6 ± 12.3 yr). The basic demographic characteristics are presented in Table 1. Each performed the treadmill protocol (Type 2000, SensorMedics, Yorba Linda, CA; the speed of the treadmill is continuously adjustable from 0.0 to 13.5 mph and the grade from 0.0 to 25.0%, respectively) featuring 3 min at the initial velocity and grade followed by the incremental phase as described above; they were not allowed to grasp the handrails or employ any other form of partial support during the test. Each subject also performed a ramp test on an electromagnetically braked cycle ergometer (Ergoline 800, SensorMedics), at a constant cycling frequency of approximately 60 rpm; 3 min of unloaded pedaling was followed by a ramp-wise increase in work rate. Subjects were divided into two subgroups. To facilitate testing of the linearity of the responses, in 8 of 22 subjects (sample of convenience), the work rate slopes of the cycle and treadmill tests were chosen to be identical (15 W·min−1), and the test was stopped at 150 W in each case. This was done in order to have a reasonably wide work rate range over which to directly compare the responses to the two tests. As both the work rate slope and the ramp duration were identical in these tests, the difference between the physiologic responses (in particular, V̇O2) at equivalent work rates could be compared. The remaining 14 subjects performed ramp exercise tests to the limit of tolerance; subjects were verbally encouraged to continue exercise for as long as possible. In these subjects, the treadmill protocol was designed as indicated above; the cycle ergometer protocol utilized a rate of work rate increment designed to result in tests with an incremental phase of 10-min duration. As previously recommended (29), the slope of the work rate increase was altered (e.g., 10, 15, or 20 W·min−1), depending on the subject’s estimated exercise tolerance (21). Comparison of these tests allowed determination of differences in the overall pattern of response as well as peak exercise responses.
Pulmonary gas exchange (minute ventilation (V̇E), oxygen uptake (V̇O2) and carbon dioxide output (V̇CO2)) was determined breath by breath with a commercial metabolic measurement system (Vmax 229, SensorMedics). This was calibrated for airflow and gas concentrations before each test and the system accuracy was checked periodically using a metabolic simulator (19). The 12-lead electrocardiogram was recorded electronically (Corina, Cardiosoft, SensorMedics), and the heart rate was derived from the R-R interval. The breath-by-breath data were interpolated to 1-s intervals, and the averages of uniform 10-s bins were used for graphical purposes and further calculations.
Sigma Plot 2000 (SPSS Science, Chicago, IL) was used for graphical display and calculating basic statistics (mean, SD) and linear regression analysis; Microsoft Excel 2000 was used for conducting paired t-tests. Differences were deemed to be significant at P < 0.05.
As shown in Figure 1, for a representative subject who performed maximal tests, our proposed protocol, incorporating a linear increase in walking speed with a treadmill grade that increases curvilinearly, resulted in linear increases in both calculated work rate and in the consequent oxygen uptake response during an exercise test that lasted approximately 10 min. This was consistently the case in all subjects. The walking speed for this test began at 0.7 mph and a grade of 2.7% producing an initial control-phase V̇O2 of less than 0.5 L·min−1. This was very similar to that achieved on the cycle ergometer in this subject during unloaded cycling. For the group as a whole, the oxygen cost of the “slow-walking” control phase was slightly (though significantly) higher than that achieved during the unloaded cycling phase of the cycle ergometer test: 0.54 ± 0.16 compared with 0.46 ± 0.18 L·min−1, respectively (P < 0.001).
The middle panel of Figure 1 shows an example of the response to a treadmill test also designed to last approximately 10 min but employing a fixed treadmill velocity (3.1 mph) and a linearly increasing grade, which began at 0.6%. Note that the V̇O2 during the initial phase of exercise was approximately 1 L·min−1, substantially more than that obtained either on the cycle ergometer or, importantly in the present context, on the new treadmill protocol. This initial requirement of 1 L·min−1 would, of course, have been close to the functional limit for a subject significantly impaired by cardiovascular or pulmonary disease, for example.
Figure 2 presents the responses of two representative subjects who performed incremental exercise to 150 W on both treadmill and cycle ergometer. Note that, although the V̇O2 response during the treadmill test was a linear function of the work rate, the slope of the V̇O2–work-rate relationship was higher than that achieved on the cycle ergometer. For the group as a whole (Table 2), the slope of the V̇O2 response in the treadmill test averaged 11.4 ± 2.4 compared with 9.6 ± 2.0 mL·min−1·W−1 for the cycle tests (P < 0.001). This figure also demonstrates a utilitarian feature of the proposed treadmill test: the range of both work rate and oxygen uptake attained on the test were effectively the same in two subjects differing considerably in body weights; these responses were achieved despite identical treadmill velocity ranges because the inclination profiles were considerably different. In the lighter subject, the grade increased to approximately 20% whereas in the heavier subject the inclination increased only up to approximately 10%.
Figure 3 presents the linearity of the V̇O2–work-rate relationship in treadmill and cycle ergometer tests. Figure 3 panels A and B present a subject with a high exercise tolerance, and Figure 3 panels C and D depict the responses of a subject with a low exercise tolerance. Figure 3 panels E and F show the ensemble-average data of the eight subjects in which the incremental test was done both on the cycle and on the treadmill at 15 W·min−1, and the incremental test was stopped at 150 W. In all three sets of panels, the very high regression coefficients and the minimal variation of residuals around zero suggest that the response is, in fact, linear. However, the slope of V̇O2–work-rate relationship is consistently steeper on treadmill than on the cycle (10.62 mL O2·min−1·W−1 vs 9.27 mL O2·min−1·W−1;Fig. 3, panels F and E, respectively).
Using the proposed treadmill protocol, it was possible to discriminate, as clearly as for the cycle test, the ventilatory and gas exchange responses that allow the lactate threshold to be estimated, as depicted for a typical subject by the asterisks in each of the panels in Figure 4. For example, as shown in the lower panels, there was a clear increase in the ventilatory equivalent for oxygen (V̇E/V̇O2) without a simultaneous increase in the ventilatory equivalent for CO2 (i.e., confirming the V-slope representation of the threshold depicted in the upper panels). Both the peak oxygen uptake, as shown by the solid vertical line at the end of the data, and the estimated lactate threshold, depicted by the asterisks, were higher on the treadmill than on the cycle ergometer: in this subject 2.18 versus 1.96 and 1.5 versus 1.1 L·min−1, respectively. This was true also for the group as a whole: 2.69 ± 0.74 versus 2.18 ± 0.60 and 1.37 ± 0.42 versus 1.06 ± 0.32 L·min−1, respectively.
In those subjects in whom the ramp slopes were identical on both the cycle ergometer and the treadmill, it was possible to calculate the difference in V̇O2 responses at equivalent work rates. The differences were calculated for the responses at the end of the control period and for each half minute of the incremental and plotted as a function of the treadmill speed (Fig. 5). At ostensibly equivalent work rates, the difference between the treadmill and the cycle ergometer responses was linearly related to the treadmill speed. This suggests that the factors other than the work against gravity are contributing more to the overall metabolic response at higher treadmill speeds.
There are a large number of treadmill exercise protocols that are useful for particular purposes (5,7,23,25,26). The Bruce test for example, still widely used in clinical (especially cardiac) exercise testing, proceeds with relatively large and uneven increments in work rate leading to rapid exhaustion in impaired patients. The nonlinear work rate profile also limits its utility for readily interpreting the pattern of the oxygen uptake response so pivotal in discriminating a cardiovascular source of impaired exercise tolerance (2, pp. 228–229 and 251–252;17,20,22,28). In fact, the American Thoracic Society and the American College of Chest Physicians in their joint statement (2, p. 225), the American College of Cardiology/American Heart Association (15), as well as the American College of Sports Medicine Guidelines (1) recommend alternative protocols. In contrast to the Bruce protocol, the standard Balke test does provide a linear increase in work rate. However, it utilizes a constant speed (i.e., 3.4 mph), which is relatively high for clinical exercise testing and proceeds with grade incrementation that is constant, regardless of the subject’s exercise capacity. Consequently, unless the treadmill speed is reduced and/or the grade increments are reduced (resulting in a long test duration), subjects with poor exercise tolerance will rapidly reach exhaustion. There have been recent attempts to obviate these concerns. Will and Walter (33), for example, have modified the Bruce protocol to provide a quasi-linear ramp format over the Bruce work rate regions. The set work rate profile for these tests, however, results in widely varying test durations in subjects who vary in exercise tolerance. Consequently, such protocols with their predetermined work rates do not optimize the test design with respect to the subject’s known or predicted functional capacity, as recommended (1,2,9,15).
Myers and his associates (25), however, designed a test that was promising in this regard: the work rate incrementation, achieved by both speed and grade changes, was designed to produce a linear increase in oxygen uptake (not work rate per se, however) with the test lasting approximately 10 min. In their large-cohort study, the measured V̇O2, though approximately linear, differed markedly from that predicted. This, presumably, is consequent in part to the nonsteady state ramp predictions being based on “steady-state” predictions of O2 uptake, the response of which is known to become nonlinear at work rates above the lactate threshold (32,35). Northridge et al. (27) devised a test in which the work rate is increased exponentially in order to provide test durations that do not appreciably differ among subjects varying widely in exercise tolerance. This work rate profile, however, requires more complex techniques for analyzing the responses.
We report here a novel application of the ramp profile exercise test designed for treadmill use that, we believe, will be especially useful in patient populations or others with relatively low exercise tolerance. The protocol uses a linear increase in belt speed that can be adjusted to be within the subject’s comfortable walking range, usually from about 0.5 mph up to a maximum of about 3.0 mph, although both the initial speed and the projected maximum can be adjusted according to the subject’s ability to walk. However, in order to achieve the linear increase in work rate, it is necessary that the treadmill inclination be incremented in a curvilinear fashion (Fig. 1), as described in “Methods.” The comparison of the physiological responses to this treadmill test with those of the more-standard cycle ergometer ramp is important in this regard. The cycle ergometer ramp yields a lagged-linear V̇O2 response in normal subjects (31), deviations from which maybe interpreted with respect to the normalcy of cardiac, peripheral circulatory, and muscular function. For example, with cycle ergometer exercise, patients with coronary artery disease often display a response slope that begins to decrease at work rates associated with ST segment changes (20), whereas patients with peripheral vascular disease (28), or hypertrophic cardiomyopathy (22) often display a V̇O2 response slope, which although linear, is low compared with the normal or expected response to the work rate profile. A similarly linear treadmill protocol allows the responses to be interpreted in a similar manner. However, an additional feature of our protocol is that it establishes the “work rate” with units equivalent to that used on the cycle ergometer, i.e., “watts.” This allows direct comparison of responses at particular work rates.
The profile of the metabolic rate change using this treadmill protocol is, in fact, similar to that seen on the cycle ergometer. As previously demonstrated, the subjects’ lactate threshold and peak V̇O2 on the treadmill were significantly higher than on the cycle ergometer. For example, in the paper of Buchfuhrer et al. (8), the differences averaged 13% and 6%, respectively. This is likely to be due, in part, to the larger total muscle mass over which metabolic cost of exercise is distributed in treadmill exercise and also that the muscle groups involved in treadmill walking are those more commonly used during every day activity, in our subjects, and consequently are more “trained” (10).
The steeper slope of the oxygen uptake to work-rate relationship on the treadmill relative to the cycle ergometer is likely to reflect the influence of unmeasured work being performed, for example, by swinging the legs and arms more frequently as the speed increases and also to the higher inherent work against friction as speed increases. And so, although the linearity of the work rate change on the treadmill overcomes an important concern for treadmill use in clinical exercise testing, the fact that the slope was consistently and significantly greater would require new normal values for the slope to be considered. It would not, of course, interfere with the interpretation of deviations in the linearity during the test (24,28). Due to the steeper slope, the V̇O2 at equivalent work rates were higher on the treadmill than on the cycle. As shown in Figure 5, this difference increases apparently linearly as a function of speed. In walking or running at much higher speeds, this relationship has been found to be a “tear-away-like” exponential (6). Our results may therefore be considered to reflect a “piece-wise linear” region covering the low speeds appropriate to this protocol. The additional kinetic energy of moving the limbs and arms at higher speeds, in addition to the increasing frictional losses, is likely to contribute to this difference (4,6).
It is important to recognize that the linearity of the work rate incrementation on the treadmill (as shown in Fig. 1 and implicit in Fig. 2, for example) is directly related to the ergometer itself; it would be distinctly altered by factors such as a variably inefficient gait, or the subject being supported by holding on to the handrails, or being partially supported by the investigator (e.g., during blood pressure auscultation). In order that the V̇O2 response profile be appropriately interpreted, we believe that such external subject support during the test should be discouraged, and with the initial walking speed in our test being so slow and the upper speed having been determined by the “comfortable walking” range for the particular subject, it should be unnecessary.
In summary, in this study we present a novel approach for applying a ramp-wise increase in work rate on the treadmill to yield a protocol equivalent to that which continues to be widely used for clinical-diagnostic purposes on the cycle ergometer. The linearly increasing work rate profile is achieved by a linear increase in walking speed, starting with slow walking, which yields a metabolic rate comparable to that of “unloaded” cycle ergometry. Using the relation we derived, the time course of treadmill inclination is calculated to provide a linear increase in work rate. This equation is suitable to be programmed into contemporary computerized systems, so that this treadmill protocol can be implemented under computer control. The O2 uptake response elicited by this protocol was proven to be linear in sedentary subjects with a substantially wide range of exercise tolerance. Naturally, patients with even lower exercise tolerance would simply operate over a lower region of the work rate profile for the same projected duration. Further studies will be required, however, to demonstrate its advantages in patient populations.
1. American College of Sports Medicine. Guidelines for Exercise Testing and Prescription (Exercise Protocols). E. P. Johnson (Ed.). Philadelphia: Lippincott Williams & Wilkins, 2000, pp. 97–99.
2. American Thoracic Society/American College of Chest Physicians. Statement on cardiopulmonary exercise testing. Am. J. Respir. Crit. Care Med. 167: 211–277, 2003.
3. Ashley, E. A., J. Myers, and V. Froelicher. Exercise testing in clinical medicine. Lancet 356: 1592–1597, 2000.
4. Asmussen, E. Muscular exercise. In: Handbook of Physiology. Section 3: Respiration, W. O. Fenn and H. Rahn (Eds.). Washington, DC: American Physiological Society, 1965, p. 955.
5. Balke, B., and R. Ware. An experimental study of physical fitness of Air Force personnel. U. S. Armed Forces Med. J. 10: 675, 1959.
6. Bøje, O. Energy production, pulmonary ventilation, and length of steps in well-trained runners working on a treadmill. Acta. Physiol. Scand. 7: 362–375, 1944.
7. Bruce, R. A. Exercise testing of patients with coronary artery disease. Ann. Clin. Res. 3: 323–332, 1971.
8. Buchfuhrer, M. J., J. E. Hansen, T. E. Robinson, D. Y. Sue, K. Wasserman, and B. J. Whipp. Optimizing the exercise protocol for cardiopulmonary assessment. J. Appl. Physiol. 55: 1558–1564, 1983.
9. Casaburi, R., C. Prefaut, and J. E. Cotes. Equipment, measurements and quality control in clinical exercise testing. In: Clinical Exercise Testing: European Respiratory Monograph (6), J. Roca and B. J. Whipp (Eds.). Sheffield, UK: European Respiratory Society Journals Ltd., 1997, p. 73.
10. Chilibeck, P. D., D. H. Paterson, W. D. Smith, and D. A. Cunningham. Cardiorespiratory kinetics during exercise of different muscle groups and mass in old and young. J. Appl. Physiol. 81: 1388–1394, 1996.
11. Ellestad, M. H. Stress testing protocol. In: Stress Testing: Principles and Practice. Philadelphia: F. A. Davis, 1996, pp. 169–198.
12. Franklin, B. A., A. Pamatmat, S. Johnson, J. Scherf, M. Mitchell, and M. Rubenfire. Metabolic cost
of extremely slow walking
in cardiac patients: implications for exercise testing and training. Arch. Phys. Med. Rehabil. 64: 564–565, 1983.
13. Froelicher, V. F., and J. N. Myers. Exercise testing methodology. In: Exercise and the Heart. Philadelphia: W.B. Saunders, 2000, pp. 11–37.
14. Fukuba, Y., K. Hara, Y. Kimura, A. Takahashi, S. A. Ward, and B. J. Whipp. Estimating the parameters of aerobic function during exercise using an exponentially increasing work rate protocol. Med. Biol. Eng. Comput. 38: 433–437, 2000.
15. Gibbons, R. J., G. J. Balady, J. W. Beasley, et al. Guidelines for exercise testing: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines. J. Am. Coll Cardiol. 30: 260–311, 1997.
16. Hansen, J. E. Exercise instruments, schemes, and protocols for evaluating the dyspneic patient. Am. Rev. Respir. Dis. 129: S25–S27, 1984.
17. Hansen, J. E., D. Y. Sue, A. Oren, and K. Wasserman. Relation of oxygen uptake to work rate in normal men and men with circulatory disorders. Am. J. Cardiol. 59: 669–674, 1987.
18. Howley, E. T. The exercise testing laboratory. In: Resource Manual for Guidelines for Exercise Testing and Prescription, S. N. Blair, P. Painter, R. R. Pate, L. K. Smith, and C. B. Taylor (Eds.). Philadelphia: Lea & Febiger, 1988, pp. 406–413.
19. Huszczuk, A., B. J. Whipp, and K. Wasserman. A respiratory gas exchange simulator for routine calibration in metabolic studies. Eur. Respir. J. 3: 465–468, 1990.
20. Itoh, H., A. Tajima, A. Koike, et al. Oxygen uptake abnormalities during exercise in coronary artery disease. In: Cardiopulmonary Exercise Testing and Cardiovascular Health, K. Wasserman (Ed.). Armonk, NY: Futura, 2002, pp. 165–172.
21. Jones, N. L. Clinical Exercise Testing. Philadelphia: W.B. Saunders, 1988.
22. Jones, S., P. M. Elliott, S. Sharma, W. J. McKenna, and B. J. Whipp. Cardiopulmonary responses to exercise in patients with hypertrophic cardiomyopathy. Heart 80: 60–67, 1998.
23. Kaminsky, L. A., and M. H. Whaley. Evaluation of a new standardized ramp protocol: the BSU/Bruce ramp protocol. J. Cardiopulm. Rehabil. 18: 438–444, 1998.
24. Koike, A., M. Hiroe, H. Adachi, et al. Anaerobic metabolism as an indicator of aerobic function during exercise in cardiac patients. J. Am. Coll. Cardiol. 20: 120–126, 1992.
25. Myers, J., N. Buchanan, D. Smith, et al. Individualized ramp treadmill: observations on a new protocol. Chest 101: 236S–241S, 1992.
26. Naughton, J., J. Patterson, and S. M. Fox III. Exercise tests in patients with chronic disease. J. Chronic Dis. 24: 519–522, 1971.
27. Northridge, D. B., J. Christie, J. McLenachan, et al. Novel exercise protocol suitable for use on a treadmill or a bicycle ergometer. Br. Heart J. 64: 313–316, 1990.
28. Wasserman, K., J. Hansen, D. Y. Sue, R. Casaburi, and B. J. Whipp. Clinical applications of cardiopulmonary exercise testing. In: Principles of Exercise Testing and Interpretation: Including Pathophysiology and Clinical Applications. Philadelphia: Lippincott Williams & Wilkins, 1999, p. 180.
29. Wasserman, K., J. Hansen, D. Y. Sue, R. Casaburi, and B. J. Whipp. Clinical exercise testing. In: Principles of Exercise Testing and Interpretation: Including Pathophysiology and Clinical Applications. Philadelphia: Lippincott Williams & Wilkins, 1999, pp. 131–132.
30. Wasserman, K., and B. J. Whipp. Exercise physiology in health and disease. Am. Rev. Respir. Dis. 112: 219–249, 1975.
31. Whipp, B. J., J. A. Davis, F. Torres, and K. Wasserman. A test to determine parameters of aerobic function during exercise. J. Appl. Physiol. 50: 217–221, 1981.
32. Whipp, B. J., and M. Mahler. Dynamics of pulmonary gas exchange during exercise. In: Pulmonary Gas Exchange: Organism and Environment, J. B. West (Ed.). New York: Academic Press, 1980, pp. 33–96.
33. Will, P. M., and J. D. Walter. Exercise testing: improving performance with a ramped Bruce protocol. Am. Heart J. 138: 1033–1037, 1999.
34. Wolthuis, R. A., V. F. Froelicher, Jr., J. Fischer, et al. New practical treadmill protocol for clinical use. Am. J. Cardiol. 39: 697–700, 1977.
35. Zoladz, J. A., K. Duda, and J. Majerczak. Oxygen uptake does not increase linearly at high power outputs during incremental exercise test in humans. Eur. J. Appl. Physiol. 77: 445–451, 1998.